index
int64
repo_id
string
file_path
string
content
string
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/RadixOrder.java
package water.rapids; import water.*; import water.fvec.Chunk; import water.fvec.Frame; import water.fvec.Vec; import water.util.ArrayUtils; import water.util.Log; import water.util.MathUtils; import java.math.BigInteger; import static java.math.BigInteger.ONE; import static java.math.BigInteger.ZERO; import static water.rapids.Merge.MEM_MULTIPLIER; import static water.rapids.Merge.OPTIMAL_BATCHSIZE; // counted completer so that left and right index can run at the same time class RadixOrder extends H2O.H2OCountedCompleter<RadixOrder> { private final Frame _DF; private final boolean _isLeft; private final int _whichCols[], _id_maps[][]; final boolean _isInt[]; final boolean _isCategorical[]; final int _shift[]; final int _bytesUsed[]; final BigInteger _base[]; final int[] _ascending; // 0 to sort ASC, 1 to sort DESC final long _mergeId; RadixOrder(Frame DF, boolean isLeft, int whichCols[], int id_maps[][], int[] ascending, long mergeId) { _DF = DF; _isLeft = isLeft; _whichCols = whichCols; _id_maps = id_maps; _shift = new int[_whichCols.length]; // currently only _shift[0] is used _bytesUsed = new int[_whichCols.length]; //_base = new long[_whichCols.length]; _base = new BigInteger[_whichCols.length]; _isInt = new boolean[_whichCols.length]; _isCategorical = new boolean[_whichCols.length]; _ascending = ascending; _mergeId = mergeId; } @Override public void compute2() { long t0 = System.nanoTime(), t1; initBaseShift(); // The MSB is stored (seemingly wastefully on first glance) because we need // it when aligning two keys in Merge() int keySize = ArrayUtils.sum(_bytesUsed); // 256MB is the DKV limit. / 2 because we fit o and x together in one OXBatch. int batchSize = OPTIMAL_BATCHSIZE ; // larger, requires more memory with less remote row fetch and vice versa for smaller // go through all node memory and reduce batchSize if needed long minMem = Long.MAX_VALUE; // memory size of nodes with smallest memory for (H2ONode h2o : H2O.CLOUD._memary) { long mem = h2o._heartbeat.get_free_mem(); // in bytes if (mem < minMem) minMem = mem; } // at some point, a MSB worth of compare columns will be stored at any one node. Make sure we have enough memory // for that. long minSortMemory = _whichCols.length*Math.max(_DF.numRows(), batchSize)*8*MEM_MULTIPLIER; if (minMem < minSortMemory) // if not enough, just throw an error and get out throw new RuntimeException("The minimum memory per node needed is too small to accommodate the sorting/merging " + "operation. Make sure the smallest node has at least "+minSortMemory+" bytes of memory."); // an array of size batchsize by numCols will be created for each sorted chunk in the end. Memory is in bytes long dataSetMemoryPerRow = 8*((long) _DF.numCols())*MEM_MULTIPLIER; // 8 to translate 64 bits into 8 bytes, MEM_MULTIPLIER to scale up long batchMemory = Math.max((long) batchSize*dataSetMemoryPerRow, minSortMemory); // memory needed to store one chunk of dataset frame if (batchMemory > minMem) { // batchsize is too big for node with smallest memory, reduce it batchSize = (int) Math.floor(minMem/dataSetMemoryPerRow); if (batchSize == 0) throw new RuntimeException("The minimum memory per node needed is too small to accommodate the sorting/merging " + "operation. Make sure the smallest node has at least "+minMem*100+" bytes of memory."); } // The Math.max ensures that batches of o and x are aligned, even for wide // keys. To save % and / in deep iteration; e.g. in insert(). Log.debug("Time to use rollup stats to determine biggestBit: " + ((t1=System.nanoTime()) - t0) / 1e9+" seconds."); t0=t1; if( _whichCols.length > 0 ) // batchsize is not used here new RadixCount(_isLeft, _base[0], _shift[0], _whichCols[0], _id_maps, _ascending[0], _mergeId).doAll(_DF.vec(_whichCols[0])); Log.debug("Time of MSB count MRTask left local on each node (no reduce): " + ((t1=System.nanoTime()) - t0) / 1e9+" seconds."); t0=t1; // NOT TO DO: we do need the full allocation of x[] and o[]. We need o[] anyway. x[] will be compressed and dense. // o is the full ordering vector of the right size // x is the byte key aligned with o // o AND x are what bmerge() needs. Pushing x to each node as well as o avoids inter-node comms. // Workaround for incorrectly blocking closeLocal() in MRTask is to do a // double MRTask and pass a key between them to pass output from first on // that node to second on that node. // TODO: fix closeLocal() blocking issue and revert to simpler usage of closeLocal() Key linkTwoMRTask = Key.make(); if( _whichCols.length > 0 ) new SplitByMSBLocal(_isLeft, _base, _shift[0], keySize, batchSize, _bytesUsed, _whichCols, linkTwoMRTask, _id_maps, _ascending, _mergeId).doAll(_DF.vecs(_whichCols)); // postLocal needs DKV.put() Log.debug("SplitByMSBLocal MRTask (all local per node, no network) took : " + ((t1=System.nanoTime()) - t0) / 1e9+" seconds."); t0=t1; if( _whichCols.length > 0 ) new SendSplitMSB(linkTwoMRTask).doAllNodes(); Log.debug("SendSplitMSB across all nodes took : " + ((t1=System.nanoTime()) - t0) / 1e9+" seconds."); t0=t1; // dispatch in parallel RPC[] radixOrders = new RPC[256]; Log.info("Sending SingleThreadRadixOrder async RPC calls ... "); for (int i = 0; i < 256; i++) radixOrders[i] = new RPC<>(SplitByMSBLocal.ownerOfMSB(i), new SingleThreadRadixOrder(_DF, _isLeft, batchSize, keySize, /*nGroup,*/ i, _mergeId)).call(); Log.debug("took : " + ((t1=System.nanoTime()) - t0) / 1e9); t0=t1; Log.info("Waiting for RPC SingleThreadRadixOrder to finish ... "); for( RPC rpc : radixOrders ) rpc.get(); Log.debug("took " + (System.nanoTime() - t0) / 1e9+" seconds."); tryComplete(); // serial, do one at a time // for (int i = 0; i < 256; i++) { // H2ONode node = MoveByFirstByte.ownerOfMSB(i); // SingleThreadRadixOrder radixOrder = new RPC<>(node, new SingleThreadRadixOrder(DF, batchSize, keySize, nGroup, i)).call().get(); // _o[i] = radixOrder._o; // _x[i] = radixOrder._x; // } // If sum(nGroup) == nrow then the index is unique. // 1) useful to know if an index is unique or not (when joining to it we // know multiples can't be returned so can allocate more efficiently) // 2) If all groups are size 1 there's no need to actually allocate an // all-1 group size vector (perhaps user was checking for uniqueness by // counting group sizes) // 3) some nodes may have unique input and others may contain dups; e.g., // in the case of looking for rare dups. So only a few threads may have // found dups. // 4) can sweep again in parallel and cache-efficient finding the groups, // and allocate known size up front to hold the group sizes. // 5) can return to Flow early with the group count. User may now realise // they selected wrong columns and cancel early. } private void initBaseShift() { for (int i=0; i<_whichCols.length; i++) { Vec col = _DF.vec(_whichCols[i]); // TODO: strings that aren't already categoricals and fixed precision double. BigInteger max=ZERO; _isInt[i] = col.isCategorical() || col.isInt(); _isCategorical[i] = col.isCategorical(); if (col.isCategorical()) { // simpler and more robust for now for all categorical bases to be 0, // even though some subsets may be far above 0; i.e. forgo uncommon // efficiency savings for now _base[i] = ZERO; assert _id_maps[i] != null; max = _isLeft?BigInteger.valueOf(ArrayUtils.maxValue(_id_maps[i])):BigInteger.valueOf(col.domain().length); } else { double colMin = col.min(); double colMax = col.max(); if (col.isInt()) { GetLongStatsTask glst = GetLongStatsTask.getLongStats(col); long colMini = glst._colMin; long colMaxi = glst._colMax; _base[i] = BigInteger.valueOf(Math.min(colMini, colMaxi*(_ascending[i]))); max = BigInteger.valueOf(Math.max(colMaxi, colMini*(_ascending[i]))); } else{ _base[i] = MathUtils.convertDouble2BigInteger(Math.min(col.min(), colMax*(_ascending[i]))); max = MathUtils.convertDouble2BigInteger(Math.max(col.max(), colMin*(_ascending[i]))); } } // Compute the span or range between min and max. Compute a // shift amount to bring the high order bits of the range down // low for radix sorting. Lower the lower-bound to be an even // power of the shift. long chk = computeShift(max, i); // On rare occasions, lowering the lower-bound also increases // the span or range until another bit is needed in the sort. // In this case, we need to re-compute the shift amount and // perhaps use an even lower lower-bound. if( chk == 256 ) chk = computeShift(max, i); assert chk <= 255; assert chk >= 0; _bytesUsed[i] = Math.min(8, (_shift[i]+15) / 8); // should not go over 8 bytes //assert (biggestBit-1)/8 + 1 == _bytesUsed[i]; } } // TODO: push these into Rollups? private static class GetLongStatsTask extends MRTask<GetLongStatsTask> { long _colMin=Long.MAX_VALUE; long _colMax=Long.MIN_VALUE; static GetLongStatsTask getLongStats(Vec col) { return new GetLongStatsTask().doAll(col); } @Override public void map(Chunk c) { for(int i=0; i<c._len; ++i) { if( !c.isNA(i) ) { long l = c.at8(i); _colMin = Math.min(_colMin, l); _colMax = Math.max(_colMax, l); } } } @Override public void reduce(GetLongStatsTask that) { _colMin = Math.min(_colMin, that._colMin); _colMax = Math.max(_colMax, that._colMax); } } // Compute the span or range between min and max. Compute a // shift amount to bring the high order bits of the range down // low for radix sorting. Lower the lower-bound to be an even // power of the shift. private long computeShift( final BigInteger max, final int i ) { int biggestBit = 0; int rangeD = max.subtract(_base[i]).add(ONE).add(ONE).bitLength(); biggestBit = _isInt[i] ? rangeD : (rangeD == 64 ? 64 : rangeD + 1); // TODO: feed back to R warnings() if (biggestBit < 8) Log.warn("biggest bit should be >= 8 otherwise need to dip into next column (TODO)"); assert biggestBit >= 1; _shift[i] = Math.max(8, biggestBit)-8; long MSBwidth = 1L << _shift[i]; BigInteger msbWidth = BigInteger.valueOf(MSBwidth); if (_base[i].mod(msbWidth).compareTo(ZERO) != 0) { _base[i] = _isInt[i]? msbWidth.multiply(_base[i].divide(msbWidth).add(_base[i].signum()<0?BigInteger.valueOf(-1L):ZERO)) :msbWidth.multiply (_base[i].divide(msbWidth));; // dealing with unsigned integer here assert _base[i].mod(msbWidth).compareTo(ZERO) == 0; } return max.subtract(_base[i]).add(ONE).shiftRight(_shift[i]).intValue(); } private static class SendSplitMSB extends MRTask<SendSplitMSB> { final Key _linkTwoMRTask; SendSplitMSB(Key linkTwoMRTask) { _linkTwoMRTask = linkTwoMRTask; } @Override public void setupLocal() { SplitByMSBLocal.MOVESHASH.get(_linkTwoMRTask).sendSplitMSB(); SplitByMSBLocal.MOVESHASH.remove(_linkTwoMRTask); } } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/Rapids.java
package water.rapids; import org.apache.commons.lang.math.NumberUtils; import water.H2O; import water.fvec.Frame; import water.rapids.ast.AstExec; import water.rapids.ast.AstFunction; import water.rapids.ast.AstParameter; import water.rapids.ast.AstRoot; import water.rapids.ast.params.*; import water.util.CollectionUtils; import water.util.StringUtils; import java.util.ArrayList; import java.util.Map; import java.util.Set; /** * <p> Rapids is an interpreter of abstract syntax trees. * * <p> This file contains the AstRoot parser and parser helper functions. * AstRoot Execution starts in the AstExec file, but spreads throughout Rapids. * * <p> Trees have a Lisp-like structure with the following "reserved" special * characters: * <dl> * <dt> '(' <dd> a nested function application expression till ')' * <dt> '{' <dd> a nested function definition expression till '}' * <dt> '[' <dd> a numeric or string list expression, till ']' * <dt> '"' <dd> a String (double quote) * <dt> "'" <dd> a String (single quote) * <dt> digits: <dd> a number * <dt> letters or other specials: <dd> an ID * </dl> * * <p> Variables are lexically scoped inside 'let' expressions or at the top-level * looked-up in the DKV directly (and must refer to a known type that is valid * on the execution stack). */ public class Rapids { private final String _str; // Statement to parse and execute private int _x; // Parse pointer, points to the index of the next character to be consumed /** * Parse a Rapids expression string into an Abstract Syntax Tree object. * @param rapids expression to parse */ public static AstRoot parse(String rapids) { Rapids r = new Rapids(rapids); AstRoot res = r.parseNext(); if (r.skipWS() != ' ') throw new IllegalASTException("Syntax error: illegal Rapids expression `" + rapids + "`"); return res; } /** * Execute a single rapids call in a short-lived session * @param rapids expression to parse */ public static Val exec(String rapids) { Session session = new Session(); try { H2O.incrementActiveRapidsCounter(); AstRoot ast = Rapids.parse(rapids); Val val = session.exec(ast, null); // Any returned Frame has it's REFCNT raised by +1, and the end(val) call // will account for that, copying Vecs as needed so that the returned // Frame is independent of the Session (which is disappearing). return session.end(val); } catch (Throwable ex) { throw session.endQuietly(ex); } finally { H2O.decrementActiveRapidsCounter(); } } /** * Compute and return a value in this session. Any returned frame shares * Vecs with the session (is not deep copied), and so must be deleted by the * caller (with a Rapids "rm" call) or will disappear on session exit, or is * a normal global frame. * @param rapids expression to parse */ @SuppressWarnings("SynchronizationOnLocalVariableOrMethodParameter") public static Val exec(String rapids, Session session) { try { H2O.incrementActiveRapidsCounter(); AstRoot ast = Rapids.parse(rapids); // Synchronize the session, to stop back-to-back overlapping Rapids calls // on the same session, which Flow sometimes does synchronized (session) { Val val = session.exec(ast, null); // Any returned Frame has it's REFCNT raised by +1, but is exiting the // session. If it's a global, we simply need to lower the internal refcnts // (which won't delete on zero cnts because of the global). If it's a // named temp, the ref cnts are accounted for by being in the temp table. if (val.isFrame()) { Frame frame = val.getFrame(); assert frame._key != null : "Returned frame has no key"; session.addRefCnt(frame, -1); } return val; } } finally { H2O.decrementActiveRapidsCounter(); } } //-------------------------------------------------------------------------------------------------------------------- // Private //-------------------------------------------------------------------------------------------------------------------- // Set of characters that cannot appear inside a token private static Set<Character> invalidTokenCharacters = StringUtils.toCharacterSet("({[]}) \t\r\n\\\"\'"); // Set of characters that cannot appear inside a number. Note that "NaN" or "nan" is also a number. private static Set<Character> invalidNumberCharacters = StringUtils.toCharacterSet(":,({[]}) \t\r\n\\\"\'"); // List of all "simple" backslash-escape sequences (i.e. those that are only 2-characters long, i.e. '\n') private static Map<Character, Character> simpleEscapeSequences = CollectionUtils.createMap(StringUtils.toCharacterArray("ntrfb'\"\\"), StringUtils.toCharacterArray("\n\t\r\f\b'\"\\")); /** * The constructor is private: rapids expression can be parsed into an AST tree, or executed, but the "naked" Rapids * object has no external purpose. * @param rapidsStr String containing a Rapids expression. */ private Rapids(String rapidsStr) { _str = rapidsStr; _x = 0; } /** * Parse and return the next expression from the rapids string. * '(' a nested function application expression ') * '{' a nested function definition expression '}' * '[' a numeric list expression, till ']' * '"' a String (double quote): attached_token * "'" a String (single quote): attached_token * digits: a double * letters or other specials: an ID */ private AstRoot parseNext() { switch (skipWS()) { case '(': return parseFunctionApplication(); case '{': return parseFunctionDefinition(); case '[': return parseList(); case '\"': case '\'': return new AstStr(string()); case ' ': throw new IllegalASTException("Expected an expression but ran out of text"); default : return parseNumberOrId(); } } /** * Parse "function application" expression, i.e. pattern of the form "(func ...args)" */ private AstExec parseFunctionApplication() { eatChar('('); ArrayList<AstRoot> asts = new ArrayList<>(); while (skipWS() != ')') asts.add(parseNext()); eatChar(')'); AstExec res = new AstExec(asts); if (peek(0) == '-') { eatChar('-'); eatChar('>'); AstId tmpid = new AstId(token()); res = new AstExec(new AstRoot[]{new AstId("tmp="), tmpid, res}); } return res; } /** * Parse and return a user defined function of the form "{arg1 arg2 . (expr)}" */ private AstFunction parseFunctionDefinition() { eatChar('{'); // Parse the list of ids ArrayList<String> ids = new ArrayList<>(); ids.add(""); // 1-based ID list while (skipWS() != '.') { String id = token(); if (!Character.isJavaIdentifierStart(id.charAt(0))) throw new IllegalASTException("variable must be a valid Java identifier: " + id); for (char c : id.toCharArray()) if (!Character.isJavaIdentifierPart(c)) throw new IllegalASTException("variable must be a valid Java identifier: " + id); ids.add(id); } // Single dot separates the list of ids from the body of the function eatChar('.'); // Parse the body AstRoot body = parseNext(); if (skipWS() != '}') throw new IllegalASTException("Expected the end of the function, but found '" + peek(0) + "'"); eatChar('}'); return new AstFunction(ids, body); } /** * Parse and return a list of tokens: either a list of strings, or a list of numbers. * We do not support lists of mixed types, or lists containing variables (for now). */ private AstParameter parseList() { eatChar('['); char nextChar = skipWS(); AstParameter res = isQuote(nextChar)? parseStringList() : parseNumList(); eatChar(']'); return res; } /** * Parse a list of strings. Strings can be either in single- or in double quotes. */ private AstStrList parseStringList() { ArrayList<String> strs = new ArrayList<>(10); while (isQuote(skipWS())) { strs.add(string()); if (skipWS() == ',') eatChar(','); } return new AstStrList(strs); } /** * Parse a "num list". This could be either a plain list of numbers, or a range, or a list of ranges. For example * [2 3 4 5 6 7] can also be written as [2:6] or [2:2 4:4:1]. The format of each "range" is `start:count[:stride]`, * and it denotes the sequence {start, start + stride, ..., start + (count-1)*stride}. Here start and stride may * be real numbers, however count must be a non-negative integer. Negative strides are also not allowed. */ private AstNumList parseNumList() { ArrayList<Double> bases = new ArrayList<>(); ArrayList<Double> strides = new ArrayList<>(); ArrayList<Long> counts = new ArrayList<>(); while (skipWS() != ']') { double base = number(); double count = 1; double stride = 1; if (skipWS() == ':') { eatChar(':'); skipWS(); count = number(); if (count < 1 || ((long) count) != count) throw new IllegalASTException("Count must be a positive integer, got " + count); } if (skipWS() == ':') { eatChar(':'); skipWS(); stride = number(); if (stride < 0 || Double.isNaN(stride)) throw new IllegalASTException("Stride must be positive, got " + stride); } if (count == 1 && stride != 1) throw new IllegalASTException("If count is 1, then stride must be one (and ignored)"); bases.add(base); counts.add((long) count); strides.add(stride); // Optional comma separating span if (skipWS() == ',') eatChar(','); } return new AstNumList(bases, strides, counts); } private AstParameter parseNumberOrId() { String t = token(); if (NumberUtils.isNumber(t)) try { return new AstNum(parseDouble(t)); } catch (NumberFormatException e) { throw new IllegalASTException(e.toString()); } else return new AstId(t); } /** * Return the character at the current parse position (or `offset` chars in the future), without advancing it. * If there are no more characters to peek, return ' '. */ private char peek(int offset) { return _x + offset < _str.length() ? _str.charAt(_x + offset) : ' '; } /** * Consume the next character from the parse stream, throwing an exception if it is not `c`. */ private void eatChar(char c) { if (peek(0) != c) throw new IllegalASTException("Expected '" + c + "'. Got: '" + peek(0)); _x++; } /** * Advance parse pointer to the first non-whitespace character, and return that character. * If such non-whitespace character cannot be found, then return ' '. */ private char skipWS() { char c = ' '; while (_x < _str.length() && isWS(c = peek(0))) _x++; return c; } /** * Parse a "token" from the input stream. A token is terminated by the next whitespace, or any of the * following characters: )}],: * * NOTE: our notion of "token" is very permissive. We may want to restrict it in the future... */ private String token() { return token(invalidTokenCharacters); } private String token(Set<Character> invalidCharacters) { int start = _x; while (!invalidCharacters.contains(peek(0))) _x++; if (start == _x) throw new IllegalASTException("Missing token"); return _str.substring(start, _x); } /** * Parse a number from the token stream. */ private double number() { return parseDouble(token(invalidNumberCharacters)); } private double parseDouble(String s) { if (s.toLowerCase().equals("nan")) return Double.NaN; try { return Double.valueOf(s); } catch (NumberFormatException e) { throw new IllegalASTException(e.toString()); } } /** * Parse a string from the token stream. */ private String string() { char quote = peek(0); int start = ++_x; boolean has_escapes = false; while (_x < _str.length()) { char c = peek(0); if (c == '\\') { has_escapes = true; char cc = peek(1); if (simpleEscapeSequences.containsKey(cc)) { _x += 2; } else if (cc == 'x') { _x += 4; // e.g: \x5A } else if (cc == 'u') { _x += 6; // e.g: \u1234 } else if (cc == 'U') { _x += 10; // e.g: \U0010FFFF } else throw new IllegalASTException("Invalid escape sequence \\" + cc); } else if (c == quote) { _x++; if (has_escapes) { StringBuilder sb = new StringBuilder(); for (int i = start; i < _x - 1; i++) { char ch = _str.charAt(i); if (ch == '\\') { char cc = _str.charAt(++i); if (simpleEscapeSequences.containsKey(cc)) { sb.append(simpleEscapeSequences.get(cc)); } else { int n = (cc == 'x')? 2 : (cc == 'u')? 4 : (cc == 'U')? 8 : -1; int hex = -1; try { hex = StringUtils.unhex(_str.substring(i + 1, i + 1 + n)); } catch (NumberFormatException e) { throw new IllegalASTException(e.toString()); } if (hex > 0x10FFFF) throw new IllegalASTException("Illegal unicode codepoint " + hex); sb.append(Character.toChars(hex)); i += n; } } else { sb.append(ch); } } return sb.toString(); } else { return _str.substring(start, _x - 1); } } else { _x++; } } throw new IllegalASTException("Unterminated string at " + start); } /** * Return true if `c` is a whitespace character. */ private static boolean isWS(char c) { return c == ' ' || c == '\t' || c == '\n' || c == '\r'; } /** * Return true if `c` is a quote character. */ private static boolean isQuote(char c) { return c == '\'' || c == '\"'; } // Return unparsed text, useful in error messages and debugging // private String unparsed() { // return _str.substring(_x, _str.length()); // } // public AstRoot throwErr(String msg) { // int idx = _str.length() - 1; // int lo = _x, hi = idx; // // if (idx < lo) { // lo = idx; // hi = lo; // } // String s = msg + '\n' + _str + '\n'; // int i; // for (i = 0; i < lo; i++) s += ' '; // s += '^'; // i++; // for (; i < hi; i++) s += '-'; // if (i <= hi) s += '^'; // s += '\n'; // throw new IllegalASTException(s); // } public static class IllegalASTException extends IllegalArgumentException { public IllegalASTException(String s) { super(s); } } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/Session.java
package water.rapids; import water.*; import water.fvec.Frame; import water.fvec.Vec; import water.nbhm.*; import water.rapids.ast.AstFunction; import water.rapids.ast.AstRoot; import water.rapids.ast.prims.operators.AstPlus; import water.util.Log; import java.util.Collections; import java.util.Map; import java.util.Properties; /** * Session is a long-lasting environment supporting caching and Copy-On-Write optimization of Vecs. This session may * last over many different Rapids calls (provided they refer to the same session). When the session ends, all the * cached Vecs will be deleted (except those in user facing Frames). **/ public class Session { // How often to perform sanity checks. 0 means disable checks, 1000 is to always check. private final static int sanityChecksFrequency = 1000; private static int sanityChecksCounter = 0; private String id; // -------------------------------------------------------------------------- // Copy On Write optimization // -------------------------------------------------------------------------- // COW optimization: instead of copying Vecs, they are "virtually" copied by // simply pointer-sharing, and raising the ref-cnt here. Losing a copy can // lower the ref-cnt, and when it goes to zero the Vec can be removed. If // the Vec needs to be modified, and the ref-cnt is 1 - an update-in-place // can happen. Otherwise a true data copy is made, and the private copy is // modified. // Ref-counts per Vec. Always positive; zero is removed from the table; // negative is an error. At the end of any given Rapids expression the // counts should match all the Vecs in the FRAMES set. private NonBlockingHashMap<Key<Vec>, Integer> REFCNTS = new NonBlockingHashMap<>(); // Frames tracked by this Session and alive to the next Rapids call. When // the whole session ends, these frames can be removed from the DKV. These // Frames can share Vecs amongst themselves (tracked by the REFCNTS) and also // with other global frames. private NonBlockingHashMap<Key, Frame> FRAMES = new NonBlockingHashMap<>(); // Vec that came from global frames, and are considered immutable. Rapids // will always copy these Vecs before mutating or deleting. Total visible // refcnts are effectively the normal refcnts plus 1 for being in the GLOBALS // set. private NonBlockingHashSet<Key<Vec>> GLOBALS = new NonBlockingHashSet<>(); private final Properties properties = new Properties(); /** * Constructor */ public Session() { this(Key.make().toString()); } public Session(String id) { this.id = id; cluster_init(); } /** Return this session's id. */ public String id() { return id; } public void setProperty(String key, String value) { if (value != null) { properties.setProperty(key, value); } else { properties.remove(key); } } public String getProperty(String key, String defaultValue) { return properties.getProperty(key, defaultValue); } /** * Execute an AstRoot in the current Session with much assertion-checking * @param ast Rapids expression to execute * @param scope ? * @return the result from the Rapids expression */ public Val exec(AstRoot ast, AstFunction scope) { sanity_check_refs(null); // Execute Env env = new Env(this); env._scope = scope; Val val = ast.exec(env); // Execute assert env.sp() == 0; // Stack balanced at end sanity_check_refs(val); return val; // Can return a frame, which may point to session-shared Vecs } /** * Normal session exit. Returned Frames are fully deep-copied, and are responsibility of the caller to delete. * Returned Frames have their refcnts currently up by 1 (for the returned value itself). */ public Val end(Val returning) { sanity_check_refs(returning); // Remove all temp frames Futures fs = new Futures(); for (Frame fr : FRAMES.values()) { fs = downRefCnt(fr, fs); // Remove internal Vecs one by one DKV.remove(fr._key, fs); // Shallow remove, internal Vecs removed 1-by-1 } fs.blockForPending(); FRAMES.clear(); // No more temp frames // Copy (as needed) so the returning Frame is completely independent of the // (disappearing) session. if (returning != null && returning.isFrame()) { Frame fr = returning.getFrame(); Key<Vec>[] vecs = fr.keys(); for (int i = 0; i < vecs.length; i++) { _addRefCnt(vecs[i], -1); // Returning frame has refcnt +1, lower it now; should go to zero internal refcnts. if (GLOBALS.contains(vecs[i])) // Copy if shared with globals fr.replace(i, vecs[i].get().makeCopy()); } } GLOBALS.clear(); // No longer tracking globals sanity_check_refs(null); REFCNTS.clear(); return returning; } /** * The Rapids call threw an exception. Best-effort cleanup, no more exceptions */ public RuntimeException endQuietly(Throwable ex) { try { GLOBALS.clear(); Futures fs = new Futures(); for (Frame fr : FRAMES.values()) { for (Key<Vec> vec : fr.keys()) { Integer I = REFCNTS.get(vec); int i = (I == null ? 0 : I) - 1; if (i > 0) REFCNTS.put(vec, i); else { REFCNTS.remove(vec); Keyed.remove(vec, fs, true); } } DKV.remove(fr._key, fs); // Shallow remove, internal Vecs removed 1-by-1 } fs.blockForPending(); FRAMES.clear(); REFCNTS.clear(); } catch (Exception ex2) { Log.warn("Exception " + ex2 + " suppressed while cleaning up Rapids Session after already throwing " + ex); } return ex instanceof RuntimeException ? (RuntimeException) ex : new RuntimeException(ex); } /** * Internal ref cnts (not counting globals - which only ever keep things alive, and have a virtual +1 to refcnts * always). */ private int _getRefCnt(Key<Vec> vec) { Integer I = REFCNTS.get(vec); assert I == null || I > 0; // No zero or negative counts return I == null ? 0 : I; } private int _putRefCnt(Key<Vec> vec, int i) { assert i >= 0; // No negative counts if (i > 0) REFCNTS.put(vec, i); else REFCNTS.remove(vec); return i; } /** * Bump internal count, not counting globals */ private int _addRefCnt(Key<Vec> vec, int i) { return _putRefCnt(vec, _getRefCnt(vec) + i); } /** * External refcnt: internal refcnt plus 1 for being global */ private int getRefCnt(Key<Vec> vec) { return _getRefCnt(vec) + (GLOBALS.contains(vec) ? 1 : 0); } /** * RefCnt +i this Vec; Global Refs can be alive with zero internal counts */ private int addRefCnt(Key<Vec> vec, int i) { return _addRefCnt(vec, i) + (GLOBALS.contains(vec) ? 1 : 0); } /** * RefCnt +i all Vecs this Frame. */ Frame addRefCnt(Frame fr, int i) { if (fr != null) // Allow and ignore null Frame, easier calling convention for (Key<Vec> vec : fr.keys()) _addRefCnt(vec, i); return fr; // Flow coding } /** * Found in the DKV, if not a tracked TEMP make it a global */ Frame addGlobals(Frame fr) { if (!FRAMES.containsKey(fr._key)) Collections.addAll(GLOBALS, fr.keys()); return fr; // Flow coding } /** * Track a freshly minted tmp frame. This frame can be removed when the session ends (unlike global frames), or * anytime during the session when the client removes it. */ public Frame track_tmp(Frame fr) { assert fr._key != null; // Temps have names FRAMES.put(fr._key, fr); // Track for session addRefCnt(fr, 1); // Refcnt is also up: these Vecs stick around after single Rapids call for the next one DKV.put(fr); // Into DKV, so e.g. Flow can view for debugging return fr; // Flow coding } /** * Remove and delete a session-tracked frame. * Remove from all session tracking spaces. * Remove any newly-unshared Vecs, but keep the shared ones. */ public void remove(Frame fr) { if (fr == null) return; Futures fs = new Futures(); if (!FRAMES.containsKey(fr._key)) { // In globals and not temps? for (Key<Vec> vec : fr.keys()) { GLOBALS.remove(vec); // Not a global anymore if (REFCNTS.get(vec) == null) // If not shared with temps Keyed.remove(vec, fs, true); // Remove unshared dead global } } else { // Else a temp and not a global fs = downRefCnt(fr, fs); // Standard down-ref counting of all Vecs FRAMES.remove(fr._key); // And remove from temps } DKV.remove(fr._key, fs); // Shallow remove, internal were Vecs removed 1-by-1 fs.blockForPending(); } /** * Lower refcnt of all Vecs in frame, deleting Vecs that go to zero refs. * Passed in a Futures which is returned, and set to non-null if something gets deleted. */ Futures downRefCnt(Frame fr, Futures fs) { for (Key<Vec> vec : fr.keys()) // Refcnt -1 all Vecs if (addRefCnt(vec, -1) == 0) { if (fs == null) fs = new Futures(); Keyed.remove(vec, fs, true); } return fs; } /** * Update a global ID, maintaining sharing of Vecs */ public Frame assign(Key<Frame> id, Frame src) { if (FRAMES.containsKey(id)) throw new IllegalArgumentException("Cannot reassign temp " + id); Futures fs = new Futures(); // Vec lifetime invariant: Globals do not share with other globals (but can // share with temps). All the src Vecs are about to become globals. If // the ID already exists, and global Vecs within it are about to die, and thus // may be deleted. Frame fr = DKV.getGet(id); if (fr != null) { // Prior frame exists for (Key<Vec> vec : fr.keys()) { if (GLOBALS.remove(vec) && _getRefCnt(vec) == 0) Keyed.remove(vec, fs, true); // Remove unused global vec } } // Copy (defensive) the base vecs array. Then copy any vecs which are // already globals - this new global must be independent of any other // global Vecs - because global Vecs get side-effected by unrelated // operations. Vec[] svecs = src.vecs().clone(); for (int i = 0; i < svecs.length; i++) if (GLOBALS.contains(svecs[i]._key)) svecs[i] = svecs[i].makeCopy(); // Make and install new global Frame Frame fr2 = new Frame(id, src._names.clone(), svecs); DKV.put(fr2, fs); addGlobals(fr2); fs.blockForPending(); return fr2; } /** * Support C-O-W optimizations: the following list of columns are about to be updated. Copy them as-needed and * replace in the Frame. Return the updated Frame vecs for flow-coding. */ public Vec[] copyOnWrite(Frame fr, int[] cols) { Vec did_copy = null; // Did a copy? Vec[] vecs = fr.vecs(); for (int col : cols) { Vec vec = vecs[col]; int refcnt = getRefCnt(vec._key); assert refcnt > 0; if (refcnt > 1) // If refcnt is 1, we allow the update to take in-place fr.replace(col, (did_copy = vec.makeCopy())); } if (did_copy != null && fr._key != null) DKV.put(fr); // Then update frame in the DKV return vecs; } /** * Check that ref counts are in a consistent state. * This should only be called between calls to Rapids expressions (otherwise may blow false-positives). * @param returning If sanity check is done at the end of the session, there is a value being returned. This value * might be a Frame (which would not be in FRAMES). So we pass the "returning" value explicitly, * so that all its references can be properly accounted for. */ private void sanity_check_refs(Val returning) { if ((sanityChecksCounter++) % 1000 >= sanityChecksFrequency) return; // Compute refcnts from tracked frames only. Since we are between Rapids // calls the only tracked Vecs should be those from tracked frames. NonBlockingHashMap<Key<Vec>, Integer> refcnts = new NonBlockingHashMap<>(REFCNTS.size()); for (Frame fr : FRAMES.values()) for (Key<Vec> vec : fr.keys()) { Integer count = refcnts.get(vec); refcnts.put(vec, count == null ? 1 : count + 1); } // Now account for the returning frame (if it is a Frame). Note that it is entirely possible that this frame is // already in the FRAMES list, however we need to account for it anyways -- this is how Env works... if (returning != null && returning.isFrame()) for (Key<Vec> vec : returning.getFrame().keys()) { Integer count = refcnts.get(vec); refcnts.put(vec, count == null ? 1 : count + 1); } // Now compare computed refcnts to cached REFCNTS. // First check that every Vec in computed refcnt is also in REFCNTS, with equal counts. for (Map.Entry<Key<Vec>,Integer> pair : refcnts.entrySet()) { Key<Vec> vec = pair.getKey(); Integer count = pair.getValue(); Integer savedCount = REFCNTS.get(vec); if (savedCount == null) throw new IllegalStateException("REFCNTS missing vec " + vec); if (count.intValue() != savedCount.intValue()) throw new IllegalStateException( "Ref-count mismatch for vec " + vec + ": REFCNT = " + savedCount + ", should be " + count); } // Then check that every cached REFCNT is in the computed set as well. if (refcnts.size() != REFCNTS.size()) for (Map.Entry<Key<Vec>, Integer> pair : REFCNTS.entrySet()) { if (!refcnts.containsKey(pair.getKey())) throw new IllegalStateException( "REFCNTs contains an extra vec " + pair.getKey() + ", count = " + pair.getValue()); } } // To avoid a class-circularity hang, we need to force other members of the // cluster to load the Rapids & AstRoot classes BEFORE trying to execute code // remotely, because e.g. ddply runs functions on all nodes. private static volatile boolean _initialized; // One-shot init static void cluster_init() { if (_initialized) return; // Touch a common class to force loading new MRTask() { @Override public void setupLocal() { new AstPlus(); } }.doAllNodes(); _initialized = true; } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/SingleThreadRadixOrder.java
package water.rapids; // General principle here is that several parallel, tight, branch free loops, // faster than one heavy DKV pass per row // It is intended that several of these SingleThreadRadixOrder run on the same // node, to utilize the cores available. The initial MSB needs to split by num // nodes * cpus per node; e.g. 256 is pretty good for 10 nodes of 32 cores. // Later, use 9 bits, or a few more bits accordingly. // Its this 256 * 4kB = 1MB that needs to be < cache per core for cache write // efficiency in MoveByFirstByte(). 10 bits (1024 threads) would be 4MB which // still < L2 // Since o[] and x[] are arrays here (not Vecs) it's harder to see how to // parallelize inside this function. Therefore avoid that issue by using more // threads in calling split. import water.*; import water.fvec.Frame; import water.fvec.Vec; import water.util.ArrayUtils; import java.util.Arrays; class SingleThreadRadixOrder extends DTask<SingleThreadRadixOrder> { private final Frame _fr; private final int _MSBvalue; // only needed to be able to return the number of groups back to the caller RadixOrder private final int _keySize, _batchSize; private final boolean _isLeft; private transient long _o[/*batch*/][]; private transient byte _x[/*batch*/][]; private transient long _otmp[][]; private transient byte _xtmp[][]; // TEMPs private transient long counts[][]; private transient byte keytmp[]; //public long _groupSizes[][]; final long _mergeId; // outputs ... // o and x are changed in-place always // iff _groupsToo==true then the following are allocated and returned SingleThreadRadixOrder(Frame fr, boolean isLeft, int batchSize, int keySize, /*long nGroup[],*/ int MSBvalue, long mergeId) { _fr = fr; _isLeft = isLeft; _batchSize = batchSize; _keySize = keySize; _MSBvalue = MSBvalue; _mergeId = mergeId; } @Override public void compute2() { keytmp = MemoryManager.malloc1(_keySize); counts = new long[_keySize][256]; Key k; SplitByMSBLocal.MSBNodeHeader[] MSBnodeHeader = new SplitByMSBLocal.MSBNodeHeader[H2O.CLOUD.size()]; long numRows =0; for (int n=0; n<H2O.CLOUD.size(); n++) { // Log.info("Getting MSB " + MSBvalue + " Node Header from node " + n + "/" + H2O.CLOUD.size() + " for Frame " + _fr._key); // Log.info("Getting"); k = SplitByMSBLocal.getMSBNodeHeaderKey(_isLeft, _MSBvalue, n, _mergeId); MSBnodeHeader[n] = DKV.getGet(k); if (MSBnodeHeader[n]==null) continue; DKV.remove(k); numRows += ArrayUtils.sum(MSBnodeHeader[n]._MSBnodeChunkCounts); // This numRows is split into nbatch batches on that node. // This header has the counts of each chunk (the ordered chunk numbers on that node) } if (numRows == 0) { tryComplete(); return; } // Allocate final _o and _x for this MSB which is gathered together on this // node from the other nodes. // TO DO: as Arno suggested, wrap up into class for fixed width batching // (to save espc overhead) int nbatch = (int) ((numRows-1) / _batchSize +1); // at least one batch. // the size of the last batch (could be batchSize, too if happens to be // exact multiple of batchSize) int lastSize = (int) (numRows - (nbatch-1)*_batchSize); _o = new long[nbatch][]; _x = new byte[nbatch][]; int b; for (b = 0; b < nbatch-1; b++) { _o[b] = MemoryManager.malloc8(_batchSize); // TO DO?: use MemoryManager.malloc8() _x[b] = MemoryManager.malloc1(_batchSize * _keySize); } _o[b] = MemoryManager.malloc8(lastSize); _x[b] = MemoryManager.malloc1(lastSize * _keySize); SplitByMSBLocal.OXbatch ox[/*node*/] = new SplitByMSBLocal.OXbatch[H2O.CLOUD.size()]; int oxBatchNum[/*node*/] = new int[H2O.CLOUD.size()]; // which batch of OX are we on from that node? Initialized to 0. for (int node=0; node<H2O.CLOUD.size(); node++) { //TO DO: why is this serial? Relying on k = SplitByMSBLocal.getNodeOXbatchKey(_isLeft, _MSBvalue, node, /*batch=*/0, _mergeId); // assert k.home(); // TODO: PUBDEV-3074 ox[node] = DKV.getGet(k); // get the first batch for each node for this MSB DKV.remove(k); } int oxOffset[] = MemoryManager.malloc4(H2O.CLOUD.size()); int oxChunkIdx[] = MemoryManager.malloc4(H2O.CLOUD.size()); // that node has n chunks and which of those are we currently on? int targetBatch = 0, targetOffset = 0, targetBatchRemaining = _batchSize; final Vec vec = _fr.anyVec(); assert vec != null; for (int c=0; c<vec.nChunks(); c++) { int fromNode = vec.chunkKey(c).home_node().index(); // each chunk in the column may be on different nodes // See long comment at the top of SendSplitMSB. One line from there repeated here : // " When the helper node (i.e. this one, now) (i.e the node doing all // the A's) gets the A's from that node, it must stack all the nodes' A's // with the A's from the other nodes in chunk order in order to maintain // the original order of the A's within the global table. " // TODO: We could process these in node order and or/in parallel if we // cumulated the counts first to know the offsets - should be doable and // high value if (MSBnodeHeader[fromNode] == null) continue; // magically this works, given the outer for loop through global // chunk. Relies on LINE_ANCHOR_1 above. int numRowsToCopy = MSBnodeHeader[fromNode]._MSBnodeChunkCounts[oxChunkIdx[fromNode]++]; // _MSBnodeChunkCounts is a vector of the number of contributions from // each Vec chunk. Since each chunk is length int, this must less than // that, so int The set of data corresponding to the Vec chunk // contributions is stored packed in batched vectors _o and _x. // at most batchSize remaining. No need to actually put the number of rows left in here int sourceBatchRemaining = _batchSize - oxOffset[fromNode]; while (numRowsToCopy > 0) { // No need for class now, as this is a bit different to the other batch copier. Two isn't too bad. int thisCopy = Math.min(numRowsToCopy, Math.min(sourceBatchRemaining, targetBatchRemaining)); System.arraycopy(ox[fromNode]._o, oxOffset[fromNode], _o[targetBatch], targetOffset, thisCopy); System.arraycopy(ox[fromNode]._x, oxOffset[fromNode]*_keySize, _x[targetBatch], targetOffset*_keySize, thisCopy*_keySize); numRowsToCopy -= thisCopy; oxOffset[fromNode] += thisCopy; sourceBatchRemaining -= thisCopy; targetOffset += thisCopy; targetBatchRemaining -= thisCopy; if (sourceBatchRemaining == 0) { // fetch the next batch : k = SplitByMSBLocal.getNodeOXbatchKey(_isLeft, _MSBvalue, fromNode, ++oxBatchNum[fromNode], _mergeId); assert k.home(); ox[fromNode] = DKV.getGet(k); DKV.remove(k); if (ox[fromNode] == null) { // if the last chunksworth fills a batchsize exactly, the getGet above will have returned null. // TODO: Check will Cliff that a known fetch of a non-existent key is ok e.g. won't cause a delay/block? If ok, leave as good check. int numNonZero = 0; for (int tmp : MSBnodeHeader[fromNode]._MSBnodeChunkCounts) if (tmp>0) numNonZero++; assert oxBatchNum[fromNode]==numNonZero; assert ArrayUtils.sum(MSBnodeHeader[fromNode]._MSBnodeChunkCounts) % _batchSize == 0; } oxOffset[fromNode] = 0; sourceBatchRemaining = _batchSize; } if (targetBatchRemaining == 0) { targetBatch++; targetOffset = 0; targetBatchRemaining = _batchSize; } } } // We now have _o and _x collated from all the contributing nodes, in the correct original order. // TODO save this allocation and reuse per thread? Or will heap just take care of it. Time this allocation and copy as step 1 anyway. _xtmp = new byte[_x.length][]; _otmp = new long[_o.length][]; assert _x.length == _o.length; // i.e. aligned batch size between x and o (think 20 bytes keys and 8 bytes of long in o) // Seems like no deep clone available in Java. Maybe System.arraycopy but // maybe that needs target to be allocated first for (int i=0; i<_x.length; i++) { _xtmp[i] = Arrays.copyOf(_x[i], _x[i].length); _otmp[i] = Arrays.copyOf(_o[i], _o[i].length); } // TO DO: a way to share this working memory between threads. // Just create enough for the 4 threads active at any one time. Not 256 allocations and releases. // We need o[] and x[] in full for the result. But this way we don't need full size xtmp[] and otmp[] at any single time. // Currently Java will allocate and free these xtmp and otmp and maybe it does good enough job reusing heap that we don't need to explicitly optimize this reuse. // Perhaps iterating this task through the largest bins first will help java reuse heap. assert(_o != null); assert(numRows > 0); // The main work. Radix sort this batch ... run(0, numRows, _keySize-1); // if keySize is 6 bytes, first byte is byte 5 // don't need to clear these now using private transient // _counts = null; // keytmp = null; //_nGroup = null; // tell the world how many batches and rows for this MSB OXHeader msbh = new OXHeader(_o.length, numRows, _batchSize); Futures fs = new Futures(); DKV.put(getSortedOXHeaderKey(_isLeft, _MSBvalue, _mergeId), msbh, fs, true); assert _o.length == _x.length; for (b=0; b<_o.length; b++) { SplitByMSBLocal.OXbatch tmp = new SplitByMSBLocal.OXbatch(_o[b], _x[b]); Value v = new Value(SplitByMSBLocal.getSortedOXbatchKey(_isLeft, _MSBvalue, b, _mergeId), tmp); DKV.put(v._key, v, fs, true); // the OXbatchKey's on this node will be reused for the new keys v.freeMem(); } // TODO: check numRows is the total of the _x[b] lengths fs.blockForPending(); tryComplete(); } static Key getSortedOXHeaderKey(boolean isLeft, int MSBvalue, long mergeId) { // This guy has merges together data from all nodes and its data is not "from" // any particular node. Therefore node number should not be in the key. return Key.make("__radix_order__SortedOXHeader_MSB" + MSBvalue + "_" + mergeId + (isLeft ? "_LEFT" : "_RIGHT")); // If we don't say this it's random ... (byte) 1 /*replica factor*/, (byte) 31 /*hidden user-key*/, true, H2O.SELF); } static class OXHeader extends Iced<OXHeader> { OXHeader(int batches, long numRows, int batchSize) { _nBatch = batches; _numRows = numRows; _batchSize = batchSize; } final int _nBatch; final long _numRows; final int _batchSize; } private int keycmp(byte x[], int xi, byte y[], int yi) { // Same return value as strcmp in C. <0 => xi<yi xi *= _keySize; yi *= _keySize; int len = _keySize; while (len > 1 && x[xi] == y[yi]) { xi++; yi++; len--; } return ((x[xi] & 0xFF) - (y[yi] & 0xFF)); // 0xFF for getting back from -1 to 255 } // orders both x and o by reference in-place. Fast for small vectors, low // overhead. don't be tempted to binsearch backwards here because have to // shift anyway public void insert(long start, /*only for small len so len can be type int*/int len) { int batch0 = (int) (start / _batchSize); int batch1 = (int) ((start+len-1) / _batchSize); long origstart = start; // just for when straddle batch boundaries int len0 = 0; // same byte _xbatch[]; long _obatch[]; if (batch1 != batch0) { // small len straddles a batch boundary. Unlikely very often since len<=200 assert batch0 == batch1-1; len0 = _batchSize - (int)(start % _batchSize); // copy two halves to contiguous temp memory, do the below, then split it back to the two halves afterwards. // Straddles batches very rarely (at most once per batch) so no speed impact at all. _xbatch = new byte[len * _keySize]; System.arraycopy(_x[batch0], (int)((start % _batchSize)*_keySize),_xbatch, 0, len0*_keySize); System.arraycopy( _x[batch1], 0,_xbatch, len0*_keySize, (len-len0)*_keySize); _obatch = new long[len]; System.arraycopy(_o[batch0], (int)(start % _batchSize), _obatch, 0, len0); System.arraycopy(_o[batch1], 0, _obatch, len0, len-len0); start = 0; } else { _xbatch = _x[batch0]; // taking this outside the loop does indeed make quite a big different (hotspot isn't catching this, then) _obatch = _o[batch0]; } int offset = (int) (start % _batchSize); for (int i=1; i<len; i++) { // like bubble sort int cmp = keycmp(_xbatch, offset+i, _xbatch, offset+i-1); // TO DO: we don't need to compare the whole key here. Set cmpLen < keySize if (cmp < 0) { System.arraycopy(_xbatch, (offset+i)*_keySize, keytmp, 0, _keySize); int j = i-1; long otmp = _obatch[offset+i]; do { System.arraycopy(_xbatch, (offset+j)*_keySize, _xbatch, (offset+j+1)*_keySize, _keySize); _obatch[offset+j+1] = _obatch[offset+j]; j--; } while (j >= 0 && keycmp(keytmp, 0, _xbatch, offset+j)<0); System.arraycopy(keytmp, 0, _xbatch, (offset+j+1)*_keySize, _keySize); _obatch[offset + j + 1] = otmp; } } if (batch1 != batch0) { // Put the sorted data back into original two places straddling the boundary System.arraycopy(_xbatch, 0,_x[batch0], (int)(origstart % _batchSize) *_keySize, len0*_keySize); System.arraycopy(_xbatch, len0*_keySize,_x[batch1], 0, (len-len0)*_keySize); System.arraycopy( _obatch, 0,_o[batch0], (int)(origstart % _batchSize), len0); System.arraycopy(_obatch, len0,_o[batch1], 0, len-len0); } } public void run(final long start, final long len, final int Byte) { if (len < 200) { // N_SMALL=200 is guess based on limited testing. Needs calibrate(). // Was 50 based on sum(1:50)=1275 worst -vs- 256 cummulate + 256 memset + // allowance since reverse order is unlikely. insert(start, (int)len); // when nalast==0, iinsert will be called only from within iradix. // TO DO: inside insert it doesn't need to compare the bytes so far as // they're known equal, so pass Byte (NB: not Byte-1) through to insert() // TO DO: Maybe transposing keys to be a set of _keySize byte columns // might in fact be quicker - no harm trying. What about long and varying // length string keys? return; } final int batch0 = (int) (start / _batchSize); final int batch1 = (int) ((start+len-1) / _batchSize); // could well span more than one boundary when very large number of rows. final long thisHist[] = counts[Byte]; // thisHist reused and carefully set back to 0 below so we don't need to clear it now int idx = (int)(start%_batchSize)*_keySize + _keySize-Byte-1; int bin=-1; // the last bin incremented. Just to see if there is only one bin with a count. int thisLen = (int)Math.min(len, _batchSize - start%_batchSize); final int nbatch = batch1-batch0+1; // number of batches this span of len covers. Usually 1. Minimum 1. for (int b=0; b<nbatch; b++) { // taking this outside the loop below does indeed make quite a big different (hotspot isn't catching this, then) byte _xbatch[] = _x[batch0+b]; for (int i = 0; i < thisLen; i++) { bin = 0xff & _xbatch[idx]; thisHist[bin]++; idx += _keySize; // maybe TO DO: shorten key by 1 byte on each iteration, so we only // need to thisx && 0xFF. No, because we need for construction of // final table key columns. } idx = _keySize-Byte-1; thisLen = (b==nbatch-2/*next iteration will be last batch*/ ? (int)((start+len)%_batchSize) : _batchSize); // thisLen will be set to _batchSize for the middle batches when nbatch>=3 } if (thisHist[bin] == len) { // one bin has count len and the rest zero => next byte quick thisHist[bin] = 0; // important, clear for reuse if (Byte != 0) run(start, len, Byte-1); return; } long rollSum = 0; for (int c = 0; c < 256; c++) { if (rollSum == len) break; // done, all other bins are zero, no need to loop through them all final long tmp = thisHist[c]; // important to skip zeros for logic below to undo cumulate. Worth the // branch to save a deeply iterative memset back to zero if (tmp == 0) continue; thisHist[c] = rollSum; rollSum += tmp; } // Sigh. Now deal with batches here as well because Java doesn't have 64bit indexing. int oidx = (int)(start%_batchSize); int xidx = oidx*_keySize + _keySize-Byte-1; thisLen = (int)Math.min(len, _batchSize - start%_batchSize); for (int b=0; b<nbatch; b++) { // taking these outside the loop below does indeed make quite a big // different (hotspot isn't catching this, then) final long _obatch[] = _o[batch0+b]; final byte _xbatch[] = _x[batch0+b]; for (int i = 0; i < thisLen; i++) { long target = thisHist[0xff & _xbatch[xidx]]++; // now always write to the beginning of _otmp and _xtmp just to reuse the first hot pages _otmp[(int)(target/_batchSize)][(int)(target%_batchSize)] = _obatch[oidx+i]; // this must be kept in 8 bytes longs System.arraycopy(_xbatch, (oidx+i)*_keySize, _xtmp[(int)(target/_batchSize)], (int)(target%_batchSize)*_keySize, _keySize ); xidx += _keySize; // Maybe TO DO: this can be variable byte width and smaller widths as // descend through bytes (TO DO: reverse byte order so always doing &0xFF) } xidx = _keySize-Byte-1; oidx = 0; thisLen = (b==nbatch-2/*next iteration will be last batch*/ ? (int)((start+len)%_batchSize) : _batchSize); } // now copy _otmp and _xtmp back over _o and _x from the start position, allowing for boundaries // _o, _x, _otmp and _xtmp all have the same _batchsize runCopy(start,len,_keySize,_batchSize,_otmp,_xtmp,_o,_x); long itmp = 0; for (int i=0; i<256; i++) { if (thisHist[i]==0) continue; final long thisgrpn = thisHist[i] - itmp; if( !(thisgrpn == 1 || Byte == 0) ) run(start+itmp, thisgrpn, Byte-1); itmp = thisHist[i]; thisHist[i] = 0; // important, to save clearing counts on next iteration } } // Hot loop, pulled out from the main run code private static void runCopy(final long start, final long len, final int keySize, final int batchSize, final long otmp[][], final byte xtmp[][], final long o[][], final byte x[][]) { // now copy _otmp and _xtmp back over _o and _x from the start position, allowing for boundaries // _o, _x, _otmp and _xtmp all have the same _batchsize // Would be really nice if Java had 64bit indexing to save programmer time. long numRowsToCopy = len; int sourceBatch = 0, sourceOffset = 0; int targetBatch = (int)(start / batchSize), targetOffset = (int)(start % batchSize); int targetBatchRemaining = batchSize - targetOffset; // 'remaining' means of the the full batch, not of the numRowsToCopy int sourceBatchRemaining = batchSize - sourceOffset; // at most batchSize remaining. No need to actually put the number of rows left in here while (numRowsToCopy > 0) { // TO DO: put this into class as well, to ArrayCopy into batched final int thisCopy = (int)Math.min(numRowsToCopy, Math.min(sourceBatchRemaining, targetBatchRemaining)); System.arraycopy(otmp[sourceBatch], sourceOffset, o[targetBatch], targetOffset, thisCopy); System.arraycopy(xtmp[sourceBatch], sourceOffset*keySize, x[targetBatch], targetOffset*keySize, thisCopy*keySize); numRowsToCopy -= thisCopy; sourceOffset += thisCopy; sourceBatchRemaining -= thisCopy; targetOffset += thisCopy; targetBatchRemaining -= thisCopy; if (sourceBatchRemaining == 0) { sourceBatch++; sourceOffset = 0; sourceBatchRemaining = batchSize; } if (targetBatchRemaining == 0) { targetBatch++; targetOffset = 0; targetBatchRemaining = batchSize; } // 'source' and 'target' deliberately the same length variable names and long lines deliberately used so we // can easy match them up vertically to ensure they are the same } } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/SortCombine.java
package water.rapids; import water.*; import water.fvec.Chunk; import water.fvec.Frame; import water.fvec.NewChunk; import water.fvec.Vec; import water.parser.BufferedString; import water.util.Log; import java.util.Arrays; /** * Before this class is called, sorting has been completed on the sorted columns. The job of this class is to * gather the sorted rows of one MSB into chunks which will later be stitched together to form the whole frame. */ class SortCombine extends DTask<SortCombine> { long _numRowsInResult = 0; // returned to caller, so not transient int _chunkSizes[]; // TODO: only _chunkSizes.length is needed by caller, so return that length only double _timings[]; final FFSB _leftSB; private transient KeyOrder _leftKO; private transient long _leftFrom; private transient int _retBatchSize; private int[][] _numRowsPerCidx; private int _chunkNum; private boolean[] _stringCols; private boolean[] _intCols; final SingleThreadRadixOrder.OXHeader _leftSortedOXHeader; final long _mergeId; // around the cluster. static class FFSB extends Iced<FFSB> { private final Frame _frame; private final Vec _vec; private final int _chunkNode[]; // Chunk homenode index final int _msb; FFSB(Frame frame, int msb) { assert -1 <= msb && msb <= 255; // left ranges from 0 to 255, right from -1 to 255 _frame = frame; _msb = msb; // Create fast lookups to go from chunk index to node index of that chunk Vec vec = _vec = frame.anyVec(); _chunkNode = vec == null ? null : MemoryManager.malloc4(vec.nChunks()); if (vec == null) return; // Zero-columns for Sort for (int i = 0; i < _chunkNode.length; i++) _chunkNode[i] = vec.chunkKey(i).home_node().index(); } } SortCombine(FFSB leftSB, SingleThreadRadixOrder.OXHeader leftSortedOXHeader, long mergeId) { _leftSB = leftSB; _mergeId = mergeId; int columnsInResult = _leftSB._frame.numCols(); _stringCols = MemoryManager.mallocZ(columnsInResult); _intCols = MemoryManager.mallocZ(columnsInResult); if (_leftSB._frame!=null) { for (int col=0; col < _leftSB._frame.numCols(); col++) { if (_leftSB._frame.vec(col).isInt()) _intCols[col] = true; if (_leftSB._frame.vec(col).isString()) _stringCols[col] = true; } } _chunkNum = _leftSB._frame.anyVec().nChunks(); _leftSortedOXHeader = leftSortedOXHeader; } @Override public void compute2() { _timings = MemoryManager.malloc8d(20); long t0 = System.nanoTime(); _leftKO = new KeyOrder(_leftSortedOXHeader, _mergeId); _leftKO.initKeyOrder(_leftSB._msb,/*left=*/true); _retBatchSize = (int) _leftKO._batchSize; final long leftN = _leftSortedOXHeader._numRows; // store number of rows in left frame for the MSB assert leftN >= 1; _timings[0] += (System.nanoTime() - t0) / 1e9; _leftFrom = -1; long leftTo = leftN; // number of rows in left frame long retSize = leftTo - _leftFrom - 1; assert retSize >= 0; if (retSize == 0) { tryComplete(); return; } _numRowsInResult = retSize; setPerNodeNumsToFetch(); // find out the number of rows to fetch from H2O nodes, number of rows to fetch per chunk if (_numRowsInResult > 0) createChunksInDKV(_mergeId); tryComplete(); } /** * This method will find the number of rows per chunk to fetch per batch for a MSB */ private void setPerNodeNumsToFetch() { Vec anyVec = _leftSB._frame.anyVec(); int nbatch = _leftKO._order.length; _numRowsPerCidx = new int[nbatch][anyVec.nChunks()]; for (int batchNum =0; batchNum < nbatch; batchNum++) { int batchSize = _leftKO._order[batchNum].length; for (int index=0; index < batchSize; index++) { long globalRowNumber = _leftKO._order[batchNum][index]; int chkIdx = _leftSB._vec.elem2ChunkIdx(globalRowNumber); _leftKO._perNodeNumRowsToFetch[_leftSB._chunkNode[chkIdx]]++; _numRowsPerCidx[batchNum][chkIdx]++; // number of rows to fetch per Cidx for a certain MSB } } } // Holder for Key & Order info private static class KeyOrder { public final long _batchSize; private final transient byte _key[/*n2GB*/][/*i mod 2GB * _keySize*/]; private final transient long _order[/*n2GB*/][/*i mod 2GB * _keySize*/]; private final transient long _perNodeNumRowsToFetch[]; final long _mergeId; KeyOrder(SingleThreadRadixOrder.OXHeader sortedOXHeader, long mergeId) { _batchSize = sortedOXHeader._batchSize; final int nBatch = sortedOXHeader._nBatch; _key = new byte[nBatch][]; _order = new long[nBatch][]; _perNodeNumRowsToFetch = new long[H2O.CLOUD.size()]; _mergeId = mergeId; } void initKeyOrder(int msb, boolean isLeft) { for (int b = 0; b < _key.length; b++) { Value v = DKV.get(SplitByMSBLocal.getSortedOXbatchKey(isLeft, msb, b, _mergeId)); SplitByMSBLocal.OXbatch ox = v.get(); //mem version (obtained from remote) of the Values gets turned into POJO version v.freeMem(); //only keep the POJO version of the Value _key[b] = ox._x; _order[b] = ox._o; } } } private void chunksPopulatePerChunk(final long[][][] perNodeLeftRowsCidx, final long[][][] perNodeLeftIndices) { int numBatch = _leftKO._order.length; // note that _order is ordered as nbatch/batchIndex int[][] chkIndices = new int[numBatch][]; // store index for each batch per chunk for (int nbatch=0; nbatch < numBatch; nbatch++) { int sortedRowIndex = -1; chkIndices[nbatch] = new int[_chunkNum]; int batchSize = _leftKO._order[nbatch].length; for (int batchNum=0; batchNum < batchSize; batchNum++) { sortedRowIndex++; long row = _leftKO._order[nbatch][batchNum]; int chkIdx = _leftSB._vec.elem2ChunkIdx(row); //binary search in espc perNodeLeftRowsCidx[nbatch][chkIdx][chkIndices[nbatch][chkIdx]] = row; perNodeLeftIndices[nbatch][chkIdx][chkIndices[nbatch][chkIdx]] = sortedRowIndex; chkIndices[nbatch][chkIdx]++; } } } private void createChunksInDKV(long mergeId) { long t0 = System.nanoTime(), t1; // Create the chunks for the final frame from this MSB. final int batchSizeUUID = _retBatchSize; final int nbatch = (int) ((_numRowsInResult - 1) / batchSizeUUID + 1); final int cloudSize = H2O.CLOUD.size(); final long[][][] perMSBLeftRowsCidx = new long[nbatch][_chunkNum][]; // store the global row number per cidx final long[][][] perMSBLeftIndices = new long[nbatch][_chunkNum][]; // store the sorted index of that row // allocate memory to arrays for (int batchN=0; batchN < nbatch; batchN++) { for (int chkidx=0; chkidx < _chunkNum; chkidx++) { perMSBLeftRowsCidx[batchN][chkidx] = new long[_numRowsPerCidx[batchN][chkidx]]; perMSBLeftIndices[batchN][chkidx] = new long[_numRowsPerCidx[batchN][chkidx]]; } } _timings[2] += ((t1 = System.nanoTime()) - t0) / 1e9; t0 = t1; chunksPopulatePerChunk(perMSBLeftRowsCidx, perMSBLeftIndices); // populate perMSBLeftRowsCidx and perMSBLeftIndices _timings[3] += ((t1 = System.nanoTime()) - t0) / 1e9; t0 = t1; assert nbatch >= 1; final int lastSize = (int) (_numRowsInResult - (nbatch - 1) * batchSizeUUID); // if there is only 1 batch, this will be the size assert lastSize > 0; final int numColsInResult = _leftSB._frame.numCols(); final double[][][] frameLikeChunks = new double[numColsInResult][nbatch][]; //TODO: compression via int types final long[][][] frameLikeChunksLongs = new long[numColsInResult][nbatch][]; //TODO: compression via int types BufferedString[][][] frameLikeChunks4Strings = new BufferedString[numColsInResult][nbatch][]; // cannot allocate before hand _chunkSizes = new int[nbatch]; final GetRawRemoteRowsPerChunk grrrsLeftPerChunk[][] = new GetRawRemoteRowsPerChunk[cloudSize][]; for (int b = 0; b < nbatch; b++) { // divide rows of a MSB into batches to process to avoid overwhelming a machine allocateFrameLikeChunks(b, nbatch, lastSize, batchSizeUUID, frameLikeChunks, frameLikeChunks4Strings, frameLikeChunksLongs, numColsInResult); // allocate memory for frameLikeChunks... chunksPopulateRetFirstPerChunk(perMSBLeftRowsCidx, perMSBLeftIndices, b, grrrsLeftPerChunk, frameLikeChunks, frameLikeChunks4Strings, frameLikeChunksLongs); // fetch and populate rows of a MSB one batch at a time. _timings[10] += ((t1 = System.nanoTime()) - t0) / 1e9; t0 = t1; // compress all chunks and store them chunksCompressAndStore(b, numColsInResult, frameLikeChunks, frameLikeChunks4Strings, frameLikeChunksLongs, mergeId); if (nbatch > 1) { cleanUpMemory(grrrsLeftPerChunk, b); // clean up memory used by grrrsLeftperChunk } } _timings[11] += (System.nanoTime() - t0) / 1e9; } // collect all rows with the same MSB one batch at a time over all nodes in a sorted fashion private void chunksPopulateRetFirstPerChunk(final long[][][] perMSBLeftRowsCidx, final long[][][] perMSBLeftIndices, final int jb, final GetRawRemoteRowsPerChunk grrrsLeft[][], final double[][][] frameLikeChunks, BufferedString[][][] frameLikeChunks4String, final long[][][] frameLikeChunksLong) { RPC<GetRawRemoteRowsPerChunk> grrrsLeftRPC[][] = new RPC[H2O.CLOUD.size()][]; int batchSize = _leftKO._order[jb].length; for (H2ONode node : H2O.CLOUD._memary) { final int ni = node.index(); grrrsLeftRPC[ni] = new RPC[1]; grrrsLeft[ni] = new GetRawRemoteRowsPerChunk[1]; grrrsLeftRPC[ni][0] = new RPC<>(node, new GetRawRemoteRowsPerChunk(_leftSB._frame, batchSize, perMSBLeftRowsCidx[jb], perMSBLeftIndices[jb])).call(); } for (H2ONode node : H2O.CLOUD._memary) { int ni = node.index(); _timings[5] += (grrrsLeft[ni][0] = grrrsLeftRPC[ni][0].get()).timeTaken; } for (H2ONode node : H2O.CLOUD._memary) { final int ni = node.index(); // transfer entries from _chk to frameLikeChunks long[][] chksLong = grrrsLeft[ni][0]._chkLong; // indexed by col num per batchsize double[][] chks = grrrsLeft[ni][0]._chk; BufferedString[][] chksString = grrrsLeft[ni][0]._chkString; for (int cidx = 0; cidx < _chunkNum; cidx++) { if (_leftSB._chunkNode[cidx] == ni) { // copy over rows from this node int rowSize = perMSBLeftIndices[jb][cidx].length; for (int row = 0; row < rowSize; row++) { for (int col = 0; col < chks.length; col++) { int offset = (int) perMSBLeftIndices[jb][cidx][row]; if (this._stringCols[col]) { frameLikeChunks4String[col][jb][offset] = chksString[col][offset]; } else if (this._intCols[col]) { frameLikeChunksLong[col][jb][offset] = chksLong[col][offset]; } else { frameLikeChunks[col][jb][offset] = chks[col][offset]; } } } } } } } private void allocateFrameLikeChunks(final int b, final int nbatch, final int lastSize, final int batchSizeUUID, final double[][][] frameLikeChunks, final BufferedString[][][] frameLikeChunks4Strings, final long[][][] frameLikeChunksLongs, final int numColsInResult) { for (int col = 0; col < numColsInResult; col++) { // allocate memory for frameLikeChunks for this batch if (this._stringCols[col]) { frameLikeChunks4Strings[col][b] = new BufferedString[_chunkSizes[b] = (b == nbatch - 1 ? lastSize : batchSizeUUID)]; } else if (this._intCols[col]) { frameLikeChunksLongs[col][b] = MemoryManager.malloc8(_chunkSizes[b] = (b == nbatch - 1 ? lastSize : batchSizeUUID)); Arrays.fill(frameLikeChunksLongs[col][b], Long.MIN_VALUE); } else { frameLikeChunks[col][b] = MemoryManager.malloc8d(_chunkSizes[b] = (b == nbatch - 1 ? lastSize : batchSizeUUID)); Arrays.fill(frameLikeChunks[col][b], Double.NaN); // NA by default to save filling with NA for nomatches when allLeft } } } private void cleanUpMemory(GetRawRemoteRowsPerChunk[][] grrr, int batchIdx) { if (grrr != null) { int nodeNum = grrr.length; for (int nodeIdx = 0; nodeIdx < nodeNum; nodeIdx++) { int batchLimit = Math.min(batchIdx + 1, grrr[nodeIdx].length); if ((grrr[nodeIdx] != null) && (grrr[nodeIdx].length > 0)) { for (int bIdx = 0; bIdx < batchLimit; bIdx++) { // clean up memory int chkLen = grrr[nodeIdx][bIdx] == null ? 0 : (grrr[nodeIdx][bIdx]._chk == null ? 0 : grrr[nodeIdx][bIdx]._chk.length); for (int cindex = 0; cindex < chkLen; cindex++) { grrr[nodeIdx][bIdx]._chk[cindex] = null; grrr[nodeIdx][bIdx]._chkString[cindex] = null; grrr[nodeIdx][bIdx]._chkLong[cindex] = null; } if (chkLen > 0) { grrr[nodeIdx][bIdx]._chk = null; grrr[nodeIdx][bIdx]._chkString = null; grrr[nodeIdx][bIdx]._chkLong = null; } } } } } } // compress all chunks and store them private void chunksCompressAndStore(final int b, final int numColsInResult, final double[][][] frameLikeChunks, BufferedString[][][] frameLikeChunks4String, final long[][][] frameLikeChunksLong, long mergeId) { // compress all chunks and store them Futures fs = new Futures(); for (int col = 0; col < numColsInResult; col++) { if (this._stringCols[col]) { NewChunk nc = new NewChunk(null, 0); for (int index = 0; index < frameLikeChunks4String[col][b].length; index++) nc.addStr(frameLikeChunks4String[col][b][index]); Chunk ck = nc.compress(); DKV.put(BinaryMerge.getKeyForMSBComboPerCol(_leftSB._msb, -1, col, b, mergeId), ck, fs, true); frameLikeChunks4String[col][b] = null; //free mem as early as possible (it's now in the store) } else if (_intCols[col]) { NewChunk nc = new NewChunk(null, -1); for (long l : frameLikeChunksLong[col][b]) { if (l == Long.MIN_VALUE) nc.addNA(); else nc.addNum(l, 0); } Chunk ck = nc.compress(); DKV.put(BinaryMerge.getKeyForMSBComboPerCol(_leftSB._msb, -1, col, b, mergeId), ck, fs, true); frameLikeChunksLong[col][b] = null; //free mem as early as possible (it's now in the store) } else { Chunk ck = new NewChunk(frameLikeChunks[col][b]).compress(); DKV.put(BinaryMerge.getKeyForMSBComboPerCol(_leftSB._msb, -1, col, b, mergeId), ck, fs, true); frameLikeChunks[col][b] = null; //free mem as early as possible (it's now in the store) } } fs.blockForPending(); } static class GetRawRemoteRowsPerChunk extends DTask<GetRawRemoteRowsPerChunk> { Frame _fr; long[][] _perNodeLeftIndices; long[][] _perNodeLeftRowsCidx; double[/*col*/][] _chk; //null on the way to remote node, non-null on the way back BufferedString[][] _chkString; long[/*col*/][] _chkLong; int _batchSize; // deal with which batch we are working with int _nChunks; double timeTaken; GetRawRemoteRowsPerChunk(Frame fr, int batchSize, long[][] leftRowsCidx, long[][] leftRowsIndices) { _fr = fr; _batchSize = batchSize; // size of current batch we are dealing with _perNodeLeftIndices = leftRowsIndices; _perNodeLeftRowsCidx = leftRowsCidx; _nChunks = _perNodeLeftIndices.length; // number of chunks in fr. } private static long[][] malloc8A(int m, int n) { long[][] res = new long[m][]; for (int i = 0; i < m; ++i) res[i] = MemoryManager.malloc8(n); return res; } @Override public void compute2() { assert ((_perNodeLeftIndices != null) && (_perNodeLeftRowsCidx != null)); assert (_chk == null); long t0 = System.nanoTime(); _chk = MemoryManager.malloc8d(_fr.numCols(), _batchSize); _chkLong = malloc8A(_fr.numCols(), _batchSize); _chkString = new BufferedString[_fr.numCols()][_batchSize]; for (int cidx = 0; cidx < _nChunks; cidx++) { // go through each chunk and copy from frame to sorted arrays for (int col = 0; col < _fr.numCols(); col++) { Vec v = _fr.vec(col); if (!v.chunkKey(cidx).home()) break; // goto next cidex Chunk c = v.chunkForChunkIdx(cidx); int chunkSize = _perNodeLeftRowsCidx[cidx].length; if (v.isString()) { for (int row = 0; row < chunkSize; row++) { // copy string and numeric columns int offset = (int) (_perNodeLeftRowsCidx[cidx][row] - v.espc()[cidx]); // row number _chkString[col][(int) _perNodeLeftIndices[cidx][row]] = c.atStr(new BufferedString(), offset); // _chkString[col][row] store by reference here } } else if (v.isInt()) { for (int row = 0; row < chunkSize; row++) { // extract info from chunks to one place int offset = (int) (_perNodeLeftRowsCidx[cidx][row] - v.espc()[cidx]); // row number _chkLong[col][(int) _perNodeLeftIndices[cidx][row]] = (c.isNA(offset)) ? Long.MIN_VALUE : c.at8(offset); } } else { for (int row = 0; row < chunkSize; row++) { // extract info from chunks to one place int offset = (int) (_perNodeLeftRowsCidx[cidx][row] - v.espc()[cidx]); // row number _chk[col][(int) _perNodeLeftIndices[cidx][row]] = c.atd(offset); } } } } _perNodeLeftIndices = null; _perNodeLeftRowsCidx = null; _fr = null; assert (_chk != null && _chkLong != null && _chkString != null); timeTaken = (System.nanoTime() - t0) / 1e9; tryComplete(); } } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/SplitByMSBLocal.java
package water.rapids; import water.*; import water.fvec.Chunk; import water.util.ArrayUtils; import water.util.Log; import water.util.MathUtils; import water.util.PrettyPrint; import java.math.BigInteger; import java.util.Arrays; import java.util.Hashtable; import static java.math.BigInteger.ONE; import static java.math.BigInteger.ZERO; class SplitByMSBLocal extends MRTask<SplitByMSBLocal> { private final boolean _isLeft; private final int _shift, _batchSize, _bytesUsed[], _keySize; private final BigInteger _base[]; private final int _col[]; private final Key _linkTwoMRTask; private final int _id_maps[][]; private final int[] _ascending; private transient long _counts[][]; private transient long _o[][][]; // transient ok because there is no reduce here between nodes, and important to save shipping back to caller. private transient byte _x[][][]; private long _numRowsOnThisNode; final long _mergeId; static Hashtable<Key,SplitByMSBLocal> MOVESHASH = new Hashtable<>(); SplitByMSBLocal(boolean isLeft, BigInteger base[], int shift, int keySize, int batchSize, int bytesUsed[], int[] col, Key linkTwoMRTask, int[][] id_maps, int[] ascending, long mergeId) { _isLeft = isLeft; // we only currently use the shift (in bits) for the first column for the // MSB (which we don't know from bytesUsed[0]). Otherwise we use the // bytesUsed to write the key's bytes. _shift = shift; _batchSize=batchSize; _bytesUsed=bytesUsed; _col=col; _base=base; _keySize = keySize; _linkTwoMRTask = linkTwoMRTask; _id_maps = id_maps; _ascending = ascending; _mergeId = mergeId; } @Override protected void setupLocal() { Key k = RadixCount.getKey(_isLeft, _col[0], _mergeId, H2O.SELF); _counts = ((RadixCount.Long2DArray) DKV.getGet(k))._val; // get the sparse spine for this node, created and DKV-put above DKV.remove(k); // First cumulate MSB count histograms across the chunks in this node long MSBhist[] = MemoryManager.malloc8(256); int nc = _fr.anyVec().nChunks(); assert nc == _counts.length; for (int c = 0; c < nc; c++) { if (_counts[c]!=null) { for (int h = 0; h < 256; h++) { MSBhist[h] += _counts[c][h]; } } } _numRowsOnThisNode = ArrayUtils.sum(MSBhist); // we just use this count for the DKV data transfer rate message, number of rows having values with current MSB if (ArrayUtils.maxValue(MSBhist) > Math.max(1000, _fr.numRows() / 20 / H2O.CLOUD.size())) { // TO DO: better test of a good even split Log.warn("RadixOrder(): load balancing on this node not optimal (max value should be <= " + (Math.max(1000, _fr.numRows() / 20 / H2O.CLOUD.size())) + " " + Arrays.toString(MSBhist) + ")"); } // shared between threads on the same node, all mappers write into distinct // locations (no conflicts, no need to atomic updates, etc.) Log.info("Allocating _o and _x buckets on this node with known size up front ... "); long t0 = System.nanoTime(); _o = new long[256][][]; _x = new byte[256][][]; // for each bucket, there might be > 2^31 bytes, so an extra dimension for that for (int msb = 0; msb < 256; msb++) { if (MSBhist[msb] == 0) continue; int nbatch = (int) ((MSBhist[msb]-1)/_batchSize +1); // at least one batch int lastSize = (int) (MSBhist[msb] - (nbatch-1) * _batchSize); // the size of the last batch (could be batchSize) assert nbatch > 0; assert lastSize > 0; _o[msb] = new long[nbatch][]; _x[msb] = new byte[nbatch][]; int b; for (b = 0; b < nbatch-1; b++) { _o[msb][b] = MemoryManager.malloc8(_batchSize); // TODO?: use MemoryManager.malloc8() _x[msb][b] = MemoryManager.malloc1(_batchSize * _keySize); } _o[msb][b] = MemoryManager.malloc8(lastSize); _x[msb][b] = MemoryManager.malloc1(lastSize * _keySize); } Log.debug("done in " + (System.nanoTime() - t0) / 1e9+" seconds."); // TO DO: otherwise, expand width. Once too wide (and interestingly large // width may not be a problem since small buckets won't impact cache), // start rolling up bins (maybe into pairs or even quads) for (int msb = 0; msb < 256; msb++) { // each of the 256 columns starts at 0 for the 0th chunk. This 0 offsets // into x[MSBvalue][batch div][mod] and o[MSBvalue][batch div][mod] long rollSum = 0; for (int c = 0; c < nc; c++) { if (_counts[c] == null) continue; long tmp = _counts[c][msb]; // Warning: modify the POJO DKV cache, but that's fine since this node // won't ask for the original DKV.get() version again _counts[c][msb] = rollSum; rollSum += tmp; } } MOVESHASH.put(_linkTwoMRTask, this); // NB: no radix skipping in this version (unlike data.table we'll use // biggestBit and assume further bits are used). } @Override public void map(Chunk chk[]) { long myCounts[] = _counts[chk[0].cidx()]; //cumulative offsets into o and x if (myCounts == null) { Log.debug("myCounts empty for chunk " + chk[0].cidx()); return; } // Loop through this chunk and write the byte key and the source row number // into the local MSB buckets // TODO: make this branch free and write the already-compressed _mem // directly. Just need to normalize compression across all chunks. This // has to loop through rows because we need the MSBValue from the first // column to use on the others, by row. Nothing to do cache efficiency, // although, it will be most cache efficient (holding one page of each // column's _mem, plus a page of this_x, all contiguous. At the cost of // more instructions. boolean[] isIntCols = MemoryManager.mallocZ(chk.length); for (int c=0; c < chk.length; c++){ isIntCols[c] = chk[c].vec().isCategorical() || chk[c].vec().isInt(); } for (int r=0; r<chk[0]._len; r++) { // tight, branch free and cache efficient (surprisingly) int MSBvalue = 0; // default for NA BigInteger thisx = ZERO; if (!chk[0].isNA(r)) { // TODO: restore branch-free again, go by column and retain original // compression with no .at8() if (_id_maps[0]!=null) { // dealing with enum column thisx = BigInteger.valueOf(_isLeft?_id_maps[0][(int)chk[0].at8(r)] + 1:(int)chk[0].at8(r) + 1); MSBvalue = thisx.shiftRight(_shift).intValue(); // may not be worth that as has to be global minimum so will rarely be // able to use as raw, but when we can maybe can do in bulk } else { // dealing with numeric columns (int or double), translate row value into MSB Bucket value thisx = isIntCols[0] ? BigInteger.valueOf(_ascending[0]*chk[0].at8(r)).subtract(_base[0]).add(ONE): MathUtils.convertDouble2BigInteger(_ascending[0]*chk[0].atd(r)).subtract(_base[0]).add(ONE); MSBvalue = thisx.shiftRight(_shift).intValue(); } } long target = myCounts[MSBvalue]++; int batch = (int) (target / _batchSize); int offset = (int) (target % _batchSize); assert _o[MSBvalue] != null; _o[MSBvalue][batch][offset] = (long) r + chk[0].start(); // move i and the index. byte this_x[] = _x[MSBvalue][batch]; offset *= _keySize; // can't overflow because batchsize was chosen above to be maxByteSize/max(keysize,8) byte keyArray[] = thisx.toByteArray(); // switched already here. int offIndex = keyArray.length > 8 ? -1 : _bytesUsed[0] - keyArray.length; int endLen = _bytesUsed[0] - (keyArray.length > 8 ? 8 : keyArray.length); for (int i = _bytesUsed[0] - 1; (i >= endLen && i >= 0); i--) { this_x[offset + i] = keyArray[i - offIndex]; } // add on the key values with values from other columns for (int c=1; c<chk.length; c++) { // TO DO: left align subsequent offset += _bytesUsed[c-1]; // advance offset by the previous field width if (chk[c].isNA(r)) continue; // NA is a zero field so skip over as java initializes memory to 0 for us always if (_id_maps[c] != null) { thisx = BigInteger.valueOf(_isLeft?_id_maps[c][(int)chk[c].at8(r)] + 1:(int)chk[c].at8(r) + 1); } else { // for numerical/integer columns thisx = isIntCols[c]? BigInteger.valueOf(_ascending[c]*chk[c].at8(r)).subtract(_base[c]).add(ONE): MathUtils.convertDouble2BigInteger(_ascending[c]*chk[c].atd(r)).subtract(_base[c]).add(ONE); } keyArray = thisx.toByteArray(); // switched already here. offIndex = keyArray.length > 8 ? -1 : _bytesUsed[c] - keyArray.length; endLen = _bytesUsed[c] - (keyArray.length > 8 ? 8 : keyArray.length); for (int i = _bytesUsed[c] - 1; (i >= endLen && i >= 0); i--) { this_x[offset + i] = keyArray[i - offIndex]; } } } } static H2ONode ownerOfMSB(int MSBvalue) { // TO DO: this isn't properly working for efficiency. This value should pick the value of where it is, somehow. // Why not getSortedOXHeader(MSBvalue).home_node() ? //int blocksize = (int) Math.ceil(256. / H2O.CLOUD.size()); //H2ONode node = H2O.CLOUD._memary[MSBvalue / blocksize]; return H2O.CLOUD._memary[MSBvalue % H2O.CLOUD.size()]; // spread it around more. } static Key getNodeOXbatchKey(boolean isLeft, int MSBvalue, int node, int batch, long mergeId) { return Key.make("__radix_order__NodeOXbatch_MSB" + MSBvalue + "_node" + node + "_batch" + batch + "_" + mergeId + (isLeft ? "_LEFT" : "_RIGHT"), Key.HIDDEN_USER_KEY, false, SplitByMSBLocal.ownerOfMSB(MSBvalue)); } static Key getSortedOXbatchKey(boolean isLeft, int MSBvalue, int batch, long mergeId) { return Key.make("__radix_order__SortedOXbatch_MSB" + MSBvalue + "_batch" + batch + "_" + mergeId + (isLeft ? "_LEFT" : "_RIGHT"), Key.HIDDEN_USER_KEY, false, SplitByMSBLocal.ownerOfMSB(MSBvalue)); } static class OXbatch extends Iced { OXbatch(long[] o, byte[] x) { _o = o; _x = x; } final long[/*batchSize or lastSize*/] _o; final byte[/*batchSize or lastSize*/] _x; } static Key getMSBNodeHeaderKey(boolean isLeft, int MSBvalue, int node, long mergeId) { return Key.make("__radix_order__OXNodeHeader_MSB" + MSBvalue + "_node" + node + "_" + mergeId + (isLeft ? "_LEFT" : "_RIGHT"), Key.HIDDEN_USER_KEY, false, SplitByMSBLocal.ownerOfMSB(MSBvalue)); } static class MSBNodeHeader extends Iced { MSBNodeHeader(int MSBnodeChunkCounts[/*chunks*/]) { _MSBnodeChunkCounts = MSBnodeChunkCounts;} int _MSBnodeChunkCounts[]; // a vector of the number of contributions from each chunk. Since each chunk is length int, this must less than that, so int } // Push o/x in chunks to owning nodes void sendSplitMSB() { // The map() above ran above for each chunk on this node. Although this // data was written to _o and _x in the order of chunk number (because we // calculated those offsets in order in the prior step), the chunk numbers // will likely have gaps because chunks are distributed across nodes not // using a modulo approach but something like chunk1 on node1, chunk2 on // node2, etc then modulo after that. Also, as tables undergo changes as a // result of user action, their distribution of chunks to nodes could // change or be changed (e.g. 'Tomas' rebalance()') for various reasons. // When the helper node (i.e the node doing all the A's) gets the A's from // this node, it must stack all this nodes' A's with the A's from the other // nodes in chunk order in order to maintain the original order of the A's // within the global table. To to do that, this node must tell the helper // node where the boundaries are in _o and _x. That's what the first for // loop below does. The helper node doesn't need to be sent the // corresponding chunk numbers. He already knows them from the Vec header // which he already has locally. // TODO: perhaps write to o_ and x_ in batches in the first place, and just // send more and smaller objects via the DKV. This makes the stitching // much easier on the helper node too, as it doesn't need to worry about // batch boundaries in the source data. Then it might be easier to // parallelize that helper part. The thinking was that if each chunk // generated 256 objects, that would flood the DKV with keys? // TODO: send nChunks * 256. Currently we do nNodes * 256. Or avoid DKV // altogether if possible. Log.info("Starting SendSplitMSB on this node (keySize is " + _keySize + " as ["); for( int bs : _bytesUsed ) Log.debug(" "+bs); Log.debug(" ]) ..."); long t0 = System.nanoTime(); Futures myfs = new Futures(); // Private Futures instead of _fs, so can block early and get timing results for (int msb =0; msb <_o.length /*256*/; ++msb) { // TODO this can be done in parallel, surely // "I found my A's (msb=0) and now I'll send them to the node doing all the A's" // "I'll send you a long vector of _o and _x (batched if very long) along with where the boundaries are." // "You don't need to know the chunk numbers of these boundaries, because you know the node of each chunk from your local Vec header" if(_o[msb] == null) continue; myfs.add(H2O.submitTask(new SendOne(msb,myfs))); } myfs.blockForPending(); double timeTaken = (System.nanoTime() - t0) / 1e9; long bytes = _numRowsOnThisNode*( 8/*_o*/ + _keySize) + 64; Log.debug("took : " + timeTaken+" seconds."); Log.debug(" DKV.put " + PrettyPrint.bytes(bytes) + " @ " + String.format("%.3f", bytes / timeTaken / (1024*1024*1024)) + " GByte/sec [10Gbit = 1.25GByte/sec]"); } class SendOne extends H2O.H2OCountedCompleter<SendOne> { // Nothing on remote node here, just a local parallel loop private final int _msb; private final Futures _myfs; SendOne(int msb, Futures myfs) { _msb = msb; _myfs = myfs; } @Override public void compute2() { int numChunks = 0; // how many of the chunks are on this node for( long[] cnts : _counts ) if (cnts != null) // the map() allocated the 256 vector in the spine slots for this node's chunks // even if cnts[_msb]==0 (no _msb for this chunk) we'll store // that because needed by line marked LINE_ANCHOR_1 below. numChunks++; // make dense. And by construction (i.e. cumulative counts) these chunks // contributed in order int msbNodeChunkCounts[] = MemoryManager.malloc4(numChunks); int j=0; long lastCount = 0; // _counts are cumulative at this stage so need to diff for( long[] cnts : _counts ) { if (cnts != null) { if (cnts[_msb] == 0) { // robust in case we skipped zeros when accumulating msbNodeChunkCounts[j] = 0; } else { // _counts is long so it can be accumulated in-place iirc. // TODO: check msbNodeChunkCounts[j] = (int)(cnts[_msb] - lastCount); lastCount = cnts[_msb]; } j++; } } MSBNodeHeader msbh = new MSBNodeHeader(msbNodeChunkCounts); // Need dontCache==true, so data does not remain both locally and on remote. // Use private Futures so can block independent of MRTask Futures. DKV.put(getMSBNodeHeaderKey(_isLeft, _msb, H2O.SELF.index(), _mergeId), msbh, _myfs, true); for (int b=0;b<_o[_msb].length; b++) { OXbatch ox = new OXbatch(_o[_msb][b], _x[_msb][b]); // this does not copy in Java, just references DKV.put(getNodeOXbatchKey(_isLeft, _msb, H2O.SELF.index(), b, _mergeId), ox, _myfs, true); } tryComplete(); } } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/Val.java
package water.rapids; import hex.Model; import water.Keyed; import water.fvec.Frame; import water.Iced; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import java.util.Map; /** * Generic execution values for the untyped stack. */ abstract public class Val extends Iced { // Things on the execution stack final public static int NUM = 1; // double final public static int NUMS = 2; // array of doubles final public static int STR = 3; // string final public static int STRS = 4; // array of strings final public static int FRM = 5; // Frame, not a Vec. Can be a Frame of 1 Vec final public static int ROW = 6; // Row of data; limited to a single array of doubles final public static int FUN = 7; // Function final public static int MOD = 8; // Model final public static int MFRM = 9; // Map of (String, Frame) final public static int KEYED = 10; // Keyed abstract public int type(); // One of these methods is overridden in each subclass public boolean isEmpty() { return isNums() && getNums().length == 0; } public boolean isNum() { return false; } public boolean isNums() { return false; } public boolean isStr() { return false; } public boolean isStrs() { return false; } public boolean isFrame() { return false; } public boolean isMapFrame() { return false; } public boolean isRow() { return false; } public boolean isFun() { return false; } public boolean isModel() { return false; } public boolean isKeyed() { return false; } // One of these methods is overridden in each subclass public double getNum() { throw badValue("number"); } public boolean getBool() { return getNum() == 1; } public double[] getNums() { throw badValue("number array"); } public String getStr() { throw badValue("String"); } public String[] getStrs() { throw badValue("String array"); } public Frame getFrame() { throw badValue("Frame"); } public Map<String, Frame> getMapFrame() { throw badValue("MapFrame"); } public double[] getRow() { throw badValue("Row"); } public AstPrimitive getFun() { throw badValue("function"); } public Model getModel() { throw badValue("Model"); } public Keyed getKeyed() { throw badValue("Keyed"); } private IllegalArgumentException badValue(String expectedType) { return new IllegalArgumentException("Expected a " + expectedType + " but found a " + getClass()); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/AstBuiltin.java
package water.rapids.ast; import water.H2O; import water.rapids.Env; import water.rapids.Val; /** * (Replacement for AstPrimitive). */ public abstract class AstBuiltin<T extends AstBuiltin<T>> extends AstPrimitive<T> { /** * <p>Primary method to invoke this function, passing all the parameters * as the `asts` list.</p> * * <p>The default implementation of this method executes all Asts within * the provided environment, and then calls {@link #exec(Val[])} passing it * the arguments as the list of {@link Val}s. A derived class will then only * need to override the second `exec()` function which is much simpler.</p> * * <p>However for certain functions (such as short-circuit boolean operators) * executing all arguments is not desirable -- these functions would have to * override this more general method.</p> * * @param env Current execution environment. Variables are looked up here. * @param stk TODO need clarification * @param asts List of AstRoot expressions that are arguments to the * function. First element in this list is the function itself. * @return value resulting from calling the function with the provided list * of arguments. */ public Val apply(Env env, Env.StackHelp stk, AstRoot[] asts) { Val[] args = new Val[asts.length]; args[0] = null; for (int i = 1; i < asts.length; i++) { args[i] = stk.track(asts[i].exec(env)); } return exec(args); } /** * Most Ast* functions will want to override this method. The semantics is * "call this function with the provided list of arguments". */ @SuppressWarnings("UnusedParameters") protected Val exec(Val[] args) { throw H2O.unimpl(); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/AstExec.java
package water.rapids.ast; import water.rapids.Env; import water.rapids.Val; import water.rapids.vals.ValFun; import water.util.SB; import java.util.ArrayList; /** * Apply A Function. Basic function execution. */ public class AstExec extends AstRoot { public final AstRoot[] _asts; public AstExec() { this((AstRoot[])null); } public AstExec(AstRoot[] asts) { _asts = asts; } public AstExec(ArrayList<AstRoot> asts) { _asts = asts.toArray(new AstRoot[asts.size()]); } @Override public String str() { SB sb = new SB().p('('); for (AstRoot ast : _asts) sb.p(ast.toString()).p(' '); return sb.p(')').toString(); } @Override public String example() { return "(func ...args)"; } @Override public String description() { return "List of whitespace-separated tokens within parenthesis is interpreted as a function application. The " + "first argument must be a function name (or an expression returning a function), all other tokens are passed " + "to the function as arguments. For example: `(sqrt 16)`, `(+ 2 3)`, `(getTimeZone)`, etc."; } // Function application. Execute the first AstRoot and verify that it is a // function. Then call that function's apply method. Do not evaluate other // arguments; e.g. short-circuit logicals' apply calls may choose to not ever // evalute some arguments. @Override public Val exec(Env env) { Val fun = _asts[0].exec(env); if (!fun.isFun()) throw new IllegalArgumentException("Expected a function but found " + fun.getClass()); AstPrimitive ast = fun.getFun(); int nargs = ast.nargs(); if (nargs != -1 && nargs != _asts.length) throw new IllegalArgumentException( "Incorrect number of arguments; '" + ast + "' expects " + (nargs - 1) + " but was passed " + (_asts.length - 1)); try (Env.StackHelp stk = env.stk()) { return env.returning(ast.apply(env, stk, _asts)); } } public String[] getArgs() { return ((ValFun) _asts[0].exec(new Env(null))).getArgs(); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/AstFrame.java
package water.rapids.ast; import water.fvec.Frame; import water.rapids.Env; import water.rapids.Val; import water.rapids.vals.ValFrame; /** * A Frame. Execution is just to return the constant. */ public class AstFrame extends AstRoot { final ValFrame _fr; public AstFrame() { _fr = null; } public AstFrame(Frame fr) { _fr = new ValFrame(fr); } @Override public String str() { return _fr == null ? null : _fr.toString(); } @Override public String example() { return null; } @Override public String description() { return null; } @Override public Val exec(Env env) { return env.returning(_fr); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/AstFunction.java
package water.rapids.ast; import water.rapids.Env; import water.rapids.Rapids; import water.rapids.Val; import water.rapids.vals.ValFun; import water.util.SB; import java.util.ArrayList; /** * Define a function * Syntax: { ids... . expr } * IDs are bound within expr */ public class AstFunction extends AstPrimitive { final String[] _ids; // Identifier names final AstRoot _body; // The function body // If this function is being evaluated, record the arguments and parent lexical scope final Val[] _args; // Evaluated arguments to a function final AstFunction _parent; // Parent lexical scope public AstFunction() { _ids = null; _body = null; _args = null; _parent = null; } public AstFunction(ArrayList<String> ids, AstRoot body) { _ids = ids.toArray(new String[ids.size()]); _body = body; _args = null; // This is a template of an uncalled function _parent = null; } // A function applied to arguments public AstFunction(AstFunction fun, Val[] args, AstFunction parent) { _ids = fun._ids; _body = fun._body; _parent = parent; _args = args; } @Override public String str() { SB sb = new SB().p('{'); penv(sb); for (String id : _ids) sb.p(id).p(' '); sb.p(". ").p(_body.toString()).p('}'); return sb.toString(); } @Override public String example() { return "{ ...args . expr }"; } @Override public String description() { return "Function definition: a list of tokens in curly braces. All initial tokens (which must be valid " + "identifiers) become function arguments, then a single dot '.' must follow, and finally an expression which " + "is the body of the function. Functions with variable number of arguments are not supported. Example: " + "squaring function `{x . (^ x 2)}`"; } // Print environment private void penv(SB sb) { if (_parent != null) _parent.penv(sb); if (_args != null) for (int i = 1; i < _ids.length; i++) sb.p(_ids[i]).p('=').p(_args[i].toString()).p(' '); } // Function execution. Just throw self on stack like a constant. However, // capture the existing global scope. @Override public ValFun exec(Env env) { return new ValFun(new AstFunction(this, null, env._scope)); } // Expected argument count, plus self @Override public int nargs() { return _ids.length; } @Override public String[] args() { return _ids; } // Do a ID lookup, returning the matching argument if found public Val lookup(String id) { for (int i = 1; i < _ids.length; i++) if (id.equals(_ids[i])) return _args[i]; // Hit, return found argument return _parent == null ? null : _parent.lookup(id); } // Apply this function: evaluate all arguments, push a lexical scope mapping // the IDs to the ARGs, then evaluate the body. After execution pop the // lexical scope and return the results. @Override public Val apply(Env env, Env.StackHelp stk, AstRoot asts[]) { // Evaluation all arguments Val[] args = new Val[asts.length]; for (int i = 1; i < asts.length; i++) args[i] = stk.track(asts[i].exec(env)); AstFunction old = env._scope; env._scope = new AstFunction(this, args, _parent); // Push a new lexical scope, extended from the old Val res = stk.untrack(_body.exec(env)); env._scope = old; // Pop the lexical scope off (by restoring the old unextended scope) return res; } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/AstParameter.java
package water.rapids.ast; public abstract class AstParameter extends AstRoot { @Override public String example() { return null; } @Override public String description() { return null; } public String toJavaString() { return str(); } // Select columns by number or String. public int[] columns(String[] names) { throw new IllegalArgumentException("Requires a number-list, but found an " + getClass().getSimpleName()); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/AstPrimitive.java
package water.rapids.ast; import water.rapids.Env; import water.rapids.Val; import water.rapids.vals.ValFun; import water.util.StringUtils; /** * A primitive operation. Execution just returns the function. *Application* (not execution) applies the function * to the arguments. */ public abstract class AstPrimitive<T extends AstPrimitive<T>> extends AstRoot<T> { private transient ValFun _v; /** * Number of function's arguments + 1. Thus, a binary operator like '+' * should be declared with 3 nargs: ["+", lhs, rhs]. * For variable-argument functions this method should return -1. */ public abstract int nargs(); /** * List of argument names. The length of the returned array should be equal * to `nargs() - 1` (unless `nargs()` returns -1, in which case this function * may return {"..."} or something similar). */ public abstract String[] args(); /** * <p>Primary method to invoke this function, passing all the parameters * as the `asts` list.</p> * * @param env Current execution environment. Variables are looked up here. * @param stk TODO need clarification * @param asts List of AstRoot expressions that are arguments to the * function. First element in this list is the function itself. * @return value resulting from calling the function with the provided list * of arguments. */ public abstract Val apply(Env env, Env.StackHelp stk, AstRoot[] asts); @Override public ValFun exec(Env env) { if (_v == null) _v = new ValFun(this); return _v; } @Override public String example() { int nargs = nargs(); return nargs == 1? "(" + str() + ")" : nargs >= 2? "(" + str() + " " + StringUtils.join(" ", args()) + ")" : nargs == -1? "(" + str() + " ...)" : null; // shouldn't be possible, but who knows? } @Override public String description() { return ""; } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/AstRoot.java
package water.rapids.ast; import water.Iced; import water.rapids.*; import water.rapids.ast.prims.reducers.AstMean; import water.rapids.ast.params.*; import water.rapids.vals.*; /** * Base class for all nodes in Rapids language Abstract Syntax Tree. */ public abstract class AstRoot<T extends AstRoot<T>> extends Iced<T> { /** * <p>"Execute" this AST expression, and return the result. For different ASTs * this may have different interpretation. For example, consider this Rapids * expression:</p> * <pre> (mean frame True False)</pre> * * <p>It will be parsed into the following structure:</p> * <pre> * AstExec() instance with * _asts = [AstMean() singleton instance, * new AstId(frame), * AstConst.TRUE, * AstConst.FALSE] * </pre> * * <p>Execution of {@link AstExec} will execute its first argument, _asts[0], * verify that it produces a function ({@link ValFun}), then call * {@link AstPrimitive#apply(Env, Env.StackHelp, AstRoot[])} on that function * passing down the list of _asts arguments.</p> * * <p>The {@link AstMean} class will in turn execute all its arguments, * where execution of {@link AstId} fetches the referred symbol from the * environment, and execution of {@link AstConst} returns the value of that * constant.</p> * * <p>Certain other functions may choose not to evaluate all their arguments * (for example boolean expressions providing short-circuit evaluation).</p> */ public abstract Val exec(Env env); /** * String representation of this Ast object in the Rapids language. For * {@link AstPrimitive}s this is the name of the function; for * {@link AstParameter}s this is either the name of the variable, or the * value of the numeric constant that the parameter represents. For more * complicated constructs such as {@link AstExec} or {@link AstFunction} * this method should return those objects as a Rapids string. */ public abstract String str(); // Note: the following 2 methods example() and description() really // ought to be static. Unfortunately, Java doesn't support overriding // static methods in subclasses, and "abstract static ..." is even a // syntax error. /** * <p>Return an example of how this Ast construct ought to be used. This * method is used to build documentation for the Rapids language. It is * different from {@link #str()}, in particular it must provide a valid * example even in a static context. For example, an {@link AstStr} may * return <code>"Hello, world!"</code> as an example. At the same time, * for different {@link AstPrimitive}s this method should generally provide * a typical example of how that function is to be used.</p> * * <p>Return <code>null</code> to indicate that the object should not be * included in the documentation.</p> */ public abstract String example(); /** * <p>Return the detailed description (help) for what this language construct * does or how it is supposed to be used. This method is used in conjunction * with {@link #example()} to build the documentation of the Rapids * language.</p> * * <p>If you need to include any formatting, then please use Markup language. * Although it is up to the client to support it, Markup is one of the * simplest and easiest alternatives.</p> * * <p>Return <code>null</code> to indicate that the object should not be * included in the documentation.</p> */ public abstract String description(); @Override public String toString() { return str(); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/AstRow.java
package water.rapids.ast; import water.rapids.Env; import water.rapids.vals.ValRow; /** * A Row. Execution is just to return the constant. */ public class AstRow extends AstRoot { final ValRow _row; public AstRow(double[] ds, String[] names) { _row = new ValRow(ds, names); } @Override public String str() { return _row.toString(); } @Override public String example() { return null; } @Override public String description() { return null; } @Override public ValRow exec(Env env) { return _row; } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/params/AstConst.java
package water.rapids.ast.params; import water.rapids.Env; import water.rapids.Rapids; import water.rapids.ast.AstParameter; import water.rapids.vals.ValNum; /** * Class for constants */ public class AstConst extends AstParameter { private final ValNum _v; private final String name; final public static AstConst FALSE = new AstConst("False", 0); final public static AstConst TRUE = new AstConst("True", 1); final public static AstConst NAN = new AstConst("NaN", Double.NaN); final public static AstConst PI = new AstConst("Pi", Math.PI); final public static AstConst E = new AstConst("E", Math.E); public AstConst() { name = null; _v = null; } public AstConst(String name, double d) { this.name = name; this._v = new ValNum(d); } @Override public String str() { return name; } @Override public ValNum exec(Env env) { return _v; } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/params/AstId.java
package water.rapids.ast.params; import water.fvec.Frame; import water.rapids.Env; import water.rapids.Val; import water.rapids.ast.AstParameter; /** * An ID. Execution does lookup in the current scope. */ public class AstId extends AstParameter { private final String _id; public AstId() { _id = null; } public AstId(String id) { _id = id; } public AstId(Frame f) { _id = f._key.toString(); } @Override public String str() { return _id; } @Override public Val exec(Env env) { return env.returning(env.lookup(_id)); } @Override public String toJavaString() { return "\"" + str() + "\""; } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/params/AstNum.java
package water.rapids.ast.params; import water.rapids.Env; import water.rapids.vals.ValNum; import water.rapids.ast.AstParameter; /** * A number literal. Execution simply returns its value. */ public class AstNum extends AstParameter { private final ValNum _v; public AstNum() { this(0); } public AstNum(double d) { _v = new ValNum(d); } @Override public String str() { return _v.toString(); } @Override public int[] columns(String[] names) { return new int[]{(int) _v.getNum()}; } public void setNum(double d) { _v.setNum(d); } public double getNum() { return _v.getNum(); } @Override public ValNum exec(Env env) { return _v; } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/params/AstNumList.java
package water.rapids.ast.params; import water.H2O; import water.rapids.Env; import water.rapids.Val; import water.rapids.ast.AstParameter; import water.rapids.vals.ValNums; import water.util.ArrayUtils; import water.util.SB; import java.util.ArrayList; import java.util.Arrays; /** * A collection of base/stride/cnts. * Syntax: { {num | num:cnt | num:cnt:stride},* } * <p/> * The bases can be unordered with dups (often used for column selection where * repeated columns are allowed, and order matters). The _isList flag tracks * that all cnts are 1 (and hence all strides are ignored and 1); these lists * may or may not be sorted. Note that some column selection is dense * (typical all-columns is: {0:MAX_INT}), and this has cnt>1. * <p/> * When cnts are > 1, bases must be sorted, with base+stride*cnt always less * than the next base. Typical use-case might be a list of probabilities for * computing quantiles, or grid-search parameters. * <p/> * Asking for a sorted integer expansion will sort the bases internally, and * also demand no overlap between bases. The has(), min() and max() calls * require a sorted list. */ public class AstNumList extends AstParameter { public final double[] _bases; final double _strides[]; final long _cnts[]; public final boolean _isList; // True if an unordered list of numbers (cnts are 1, stride is ignored) public boolean _isSort; // True if bases are sorted. May get updated later. public AstNumList(ArrayList<Double> bases, ArrayList<Double> strides, ArrayList<Long> counts) { int n = bases.size(); // Convert to fixed-sized arrays _bases = new double[n]; _strides = new double[n]; _cnts = new long[n]; boolean isList = true; for (int i = 0; i < n; i++) { _bases[i] = bases.get(i); _cnts[i] = counts.get(i); _strides[i] = strides.get(i); if (_cnts[i] != 1) isList = false; } _isList = isList; // Complain about unordered bases, unless it's a simple number list boolean isSorted = true; for (int i = 1; i < n; i++) if (_bases[i-1] + (_cnts[i-1] - 1) * _strides[i-1] >= _bases[i]) { if (_isList) isSorted = false; else throw new IllegalArgumentException("Overlapping numeric ranges"); } _isSort = isSorted; } // A simple AstNumList of 1 number public AstNumList(double d) { _bases = new double[]{d}; _strides = new double[]{1}; _cnts = new long[]{1}; _isList = _isSort = true; } // A simple dense range AstNumList public AstNumList(long lo, long hi_exclusive) { _bases = new double[]{lo}; _strides = new double[]{1}; _cnts = new long[]{hi_exclusive - lo}; _isList = false; _isSort = true; } // An empty number list public AstNumList() { _bases = new double[0]; _strides = new double[0]; _cnts = new long[0]; _isList = _isSort = true; } public AstNumList(double[] list) { _bases = list; _strides = new double[list.length]; _cnts = new long[list.length]; _isList = true; Arrays.fill(_strides, 1); Arrays.fill(_cnts, 1); } public AstNumList(int[] list) { this(ArrayUtils.copyFromIntArray(list)); } @Override public Val exec(Env env) { return new ValNums(expand()); } @Override public String str() { SB sb = new SB().p('['); for (int i = 0; i < _bases.length; i++) { sb.p(_bases[i]); if (_cnts[i] != 1) { sb.p(':').p(_bases[i] + _cnts[i] * _strides[i]); if (_strides[i] != 1 || ((long) _bases[i]) != _bases[i]) sb.p(':').p(_strides[i]); } if (i < _bases.length - 1) sb.p(','); } return sb.p(']').toString(); } @Override public String toJavaString() { double[] ary = expand(); if (ary == null || ary.length == 0) return "\"null\""; SB sb = new SB().p('{'); for (int i = 0; i < ary.length - 1; ++i) sb.p(ary[i]).p(','); return sb.p('}').toString(); } // Expand the compressed form into an array of doubles. public double[] expand() { // Count total values int nrows = (int) cnt(), r = 0; // Fill in values double[] vals = new double[nrows]; for (int i = 0; i < _bases.length; i++) { if (Double.isNaN(_bases[i])) { vals[r++] = Double.NaN; } else { for (double d = _bases[i]; d < _bases[i] + _cnts[i] * _strides[i]; d += _strides[i]) vals[r++] = d; } } return vals; } // Update-in-place sort of bases public AstNumList sort() { if (_isSort) return this; // Flow coding fast-path cutout int[] idxs = ArrayUtils.seq(0, _bases.length); ArrayUtils.sort(idxs, _bases); double[] bases = _bases.clone(); double[] strides = _strides.clone(); long[] cnts = _cnts.clone(); for (int i = 0; i < idxs.length; i++) { _bases[i] = bases[idxs[i]]; _strides[i] = strides[idxs[i]]; _cnts[i] = cnts[idxs[i]]; } _isSort = true; return this; } // Expand the compressed form into an array of ints; // often used for unordered column lists public int[] expand4() { // Count total values int nrows = (int) cnt(), r = 0; // Fill in values int[] vals = new int[nrows]; for (int i = 0; i < _bases.length; i++) for (double d = _bases[i]; d < _bases[i] + _cnts[i] * _strides[i]; d += _strides[i]) vals[r++] = (int) d; return vals; } // Expand the compressed form into an array of ints; // often used for sorted column lists int[] expand4Sort() { return sort().expand4(); } // Expand the compressed form into an array of longs; // often used for unordered row lists public long[] expand8() { // Count total values int nrows = (int) cnt(), r = 0; // Fill in values long[] vals = new long[nrows]; for (int i = 0; i < _bases.length; i++) for (double d = _bases[i]; d < _bases[i] + _cnts[i] * _strides[i]; d += _strides[i]) vals[r++] = (long) d; return vals; } // Expand the compressed form into an array of longs; // often used for sorted row lists public long[] expand8Sort() { return sort().expand8(); } public double max() { assert _isSort; return _bases[_bases.length - 1] + _cnts[_cnts.length - 1] * _strides[_strides.length - 1]; } // largest exclusive value (weird rite?!) public double min() { assert _isSort; return _bases[0]; } public long cnt() { return water.util.ArrayUtils.sum(_cnts); } public boolean isDense() { return _cnts.length == 1 && _bases[0] == 0 && _strides[0] == 1; } public boolean isEmpty() { return _bases.length == 0; } // check if n is in this list of numbers // NB: all contiguous ranges have already been checked to have stride 1 public boolean has(long v) { int idx = findBase(v); if (idx >= 0) return true; idx = -idx - 2; // See Arrays.binarySearch; returns (-idx-1), we want +idx-1 ... if idx == -1 => then this transformation has no effect if (idx < 0) return false; assert _bases[idx] < v; // Sanity check binary search, AND idx >= 0 return v < _bases[idx] + _cnts[idx] * _strides[idx] && (v - _bases[idx]) % _strides[idx] == 0; } /** * Finds index of a given value in this number sequence, indexing start at 0. * @param v value * @return value index (>= 0) or -1 if value is not a member of this sequence */ public long index(long v) { int bIdx = findBase(v); if (bIdx >= 0) return water.util.ArrayUtils.sum(_cnts, 0, bIdx); bIdx = -bIdx - 2; if (bIdx < 0) return -1L; assert _bases[bIdx] < v; long offset = v - (long) _bases[bIdx]; long stride = (long) _strides[bIdx]; if ((offset >= _cnts[bIdx] * stride) || (offset % stride != 0)) return -1L; return water.util.ArrayUtils.sum(_cnts, 0, bIdx) + (offset / stride); } private int findBase(long v) { assert _isSort; // Only called when already sorted // do something special for negative indexing... that does not involve // allocating arrays, once per list element! if (v < 0) throw H2O.unimpl(); return Arrays.binarySearch(_bases, v); } // Select columns by number. Numbers are capped to the number of columns +1 // - this allows R to see a single out-of-range value and throw a range check // - this allows Python to see a single out-of-range value and ignore it // - this allows Python to pass [0:MAXINT] without blowing out the max number of columns. // Note that the Python front-end does not want to cap the max column size, because // this will force eager evaluation on a standard column slice operation. // Note that the list is often unsorted (_isSort is false). // Note that the list is often dense with cnts>1 (_isList is false). @Override public int[] columns(String[] names) { // Count total values, capped by max len+1 int nrows = 0, r = 0; for (int i = 0; i < _bases.length; i++) nrows += Math.min(_bases[i] + _cnts[i], names.length + 1) - Math.min(_bases[i], names.length + 1); // Fill in values int[] vals = new int[nrows]; for (int i = 0; i < _bases.length; i++) { int lim = Math.min((int) (_bases[i] + _cnts[i]), names.length + 1); for (int d = Math.min((int) _bases[i], names.length + 1); d < lim; d++) vals[r++] = d; } return vals; } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/params/AstStr.java
package water.rapids.ast.params; import water.rapids.Env; import water.rapids.ast.AstParameter; import water.rapids.vals.ValStr; /** * A String. Execution is just to return the constant. */ public class AstStr extends AstParameter { private final ValStr _v; public AstStr() { this(null); } public AstStr(String str) { _v = new ValStr(str); } @Override public String str() { return _v.toString().replaceAll("^\"|^\'|\"$|\'$", ""); } @Override public ValStr exec(Env env) { return _v; } @Override public String toJavaString() { return "\"" + str() + "\""; } @Override public int[] columns(String[] names) { int i = water.util.ArrayUtils.find(names, _v.getStr()); if (i == -1) throw new IllegalArgumentException("Column " + _v.getStr() + " not found"); return new int[]{i}; } public String getStr() { return _v.getStr(); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/params/AstStrList.java
package water.rapids.ast.params; import water.rapids.Env; import water.rapids.Val; import water.rapids.ast.AstParameter; import water.rapids.vals.ValStrs; import java.util.ArrayList; import java.util.Arrays; /** * A collection of Strings only. This is a syntatic form only, and never executes and never gets on the execution * stack. */ public class AstStrList extends AstParameter { public String[] _strs; public AstStrList() { _strs = null; } public AstStrList(ArrayList<String> strs) { _strs = strs.toArray(new String[strs.size()]); } @Override public Val exec(Env env) { return new ValStrs(_strs); } @Override public String str() { return Arrays.toString(_strs); } // Select columns by number or String. @Override public int[] columns(String[] names) { int[] idxs = new int[_strs.length]; for (int i = 0; i < _strs.length; i++) { int idx = idxs[i] = water.util.ArrayUtils.find(names, _strs[i]); if (idx == -1) throw new IllegalArgumentException("Column " + _strs[i] + " not found"); } return idxs; } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/advmath/AstCorrelation.java
package water.rapids.ast.prims.advmath; import water.Key; import water.MRTask; import water.fvec.*; import water.rapids.Env; import water.rapids.Val; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import water.rapids.vals.ValFrame; import water.rapids.vals.ValNum; import water.util.ArrayUtils; import water.util.EnumUtils; import java.util.Arrays; /** * Calculate Pearson's Correlation Coefficient between columns of a frame * <p/> * Formula: * Pearson's Correlation Coefficient = Cov(X,Y)/sigma(X) * sigma(Y) */ public class AstCorrelation extends AstPrimitive { @Override public String[] args() { return new String[]{"ary", "x", "y", "use", "method"}; } protected enum Mode {Everything, AllObs, CompleteObs} private enum Method {Pearson, Spearman} @Override public int nargs() { return 1 + 4; /* (cor X Y use method) */ } @Override public String str() { return "cor"; } @Override public Val apply(Env env, Env.StackHelp stk, AstRoot asts[]) { Frame frx = stk.track(asts[1].exec(env)).getFrame(); Frame fry = stk.track(asts[2].exec(env)).getFrame(); if (frx.numRows() != fry.numRows()) throw new IllegalArgumentException("Frames must have the same number of rows, found " + frx.numRows() + " and " + fry.numRows()); String use = stk.track(asts[3].exec(env)).getStr(); Mode mode; switch (use) { case "everything": mode = Mode.Everything; break; case "all.obs": mode = Mode.AllObs; break; case "complete.obs": mode = Mode.CompleteObs; break; default: throw new IllegalArgumentException("unknown use mode: " + use); } final Method method = getMethodFromUserInput(stk.track(asts[4].exec(env)).getStr()); switch (method) { case Pearson: return fry.numRows() == 1 ? scalar(frx, fry, mode) : array(frx, fry, mode); case Spearman: return spearman(frx, fry, mode); default: throw new IllegalStateException(String.format("Given method input'%s' is not supported. Available options are: %s", method, Arrays.toString(Method.values()))); } } private static Method getMethodFromUserInput(final String methodUserInput) { return EnumUtils.valueOfIgnoreCase(Method.class, methodUserInput) .orElseThrow(() -> new IllegalArgumentException(String.format("Unknown correlation method '%s'. Available options are: %s", methodUserInput, Arrays.toString(Method.values())))); } private Val spearman(final Frame frameX, final Frame frameY, final Mode mode) { final Frame spearmanMatrix = SpearmanCorrelation.calculate(frameX, frameY, mode); if (frameY.numCols() == 1) { // If there are only two columns compared, return a single number with the correlation coefficient return new ValNum(spearmanMatrix.vec(0).at(0)); } else { // Otherwise just return the correlation matrix return new ValFrame(spearmanMatrix); } } // Pearson Correlation for one row, which will return a scalar value. private ValNum scalar(Frame frx, Frame fry, Mode mode) { if (frx.numCols() != fry.numCols()) throw new IllegalArgumentException("Single rows must have the same number of columns, found " + frx.numCols() + " and " + fry.numCols()); Vec vecxs[] = frx.vecs(); Vec vecys[] = fry.vecs(); double xmean = 0; double ymean = 0; double xvar = 0; double yvar = 0; double xsd; double ysd; double ncols = fry.numCols(); double NACount = 0; double xval; double yval; double ss = 0; for (int r = 0; r < ncols; r++) { xval = vecxs[r].at(0); yval = vecys[r].at(0); if (Double.isNaN(xval) || Double.isNaN(yval)) NACount++; else { xmean += xval; ymean += yval; } } xmean /= (ncols - NACount); ymean /= (ncols - NACount); for (int r = 0; r < ncols; r++) { xval = vecxs[r].at(0); yval = vecys[r].at(0); if (!(Double.isNaN(xval) || Double.isNaN(yval))) { //Compute variance of x and y vars xvar += Math.pow((vecxs[r].at(0) - xmean), 2); yvar += Math.pow((vecys[r].at(0) - ymean), 2); //Compute sum of squares of x and y ss += (vecxs[r].at(0) - xmean) * (vecys[r].at(0) - ymean); } } xsd = Math.sqrt(xvar / (ncols - 1 - NACount)); //Sample Standard Deviation ysd = Math.sqrt(yvar / (ncols - 1 - NACount)); //Sample Standard Deviation double denom = xsd * ysd; //sd(x) * sd(y) if (NACount != 0) { if (mode.equals(Mode.AllObs)) throw new IllegalArgumentException("Mode is 'all.obs' but NAs are present"); if (mode.equals(Mode.Everything)) return new ValNum(Double.NaN); } return new ValNum((ss / (ncols - NACount - 1)) / denom); //Pearson's Correlation Coefficient } // Matrix correlation. Compute correlation between all columns from each Frame // against each other. Return a matrix of correlations which is frx.numCols // wide and fry.numCols tall. private Val array(Frame frx, Frame fry, Mode mode) { Vec[] vecxs = frx.vecs(); int ncolx = vecxs.length; Vec[] vecys = fry.vecs(); int ncoly = vecys.length; if (mode.equals(Mode.Everything) || mode.equals(Mode.AllObs)) { if (mode.equals(Mode.AllObs)) { if (mode.equals(Mode.AllObs)) { for (Vec v : vecxs) if (v.naCnt() != 0) throw new IllegalArgumentException("Mode is 'all.obs' but NAs are present"); } } //Set up CoVarTask CoVarTask[] cvs = new CoVarTask[ncoly]; //Get mean of x vecs double[] xmeans = new double[ncolx]; for (int x = 0; x < ncolx; x++) { xmeans[x] = vecxs[x].mean(); } //Set up double arrays to capture sd(x), sd(y) and sd(x) * sd(y) double[] sigmay = new double[ncoly]; double[] sigmax = new double[ncolx]; double[][] denom = new double[ncoly][ncolx]; // Launch tasks; each does all Xs vs one Y for (int y = 0; y < ncoly; y++) { //Get covariance between x and y cvs[y] = new CoVarTask(vecys[y].mean(), xmeans).dfork(new Frame(vecys[y]).add(frx)); //Get sigma of y vecs sigmay[y] = vecys[y].sigma(); } //Get sigma of x vecs for (int x = 0; x < ncolx; x++) { sigmax[x] = vecxs[x].sigma(); } //Denominator for correlation calculation is sigma_y * sigma_x (All x sigmas vs one Y) for (int y = 0; y < ncoly; y++) { for (int x = 0; x < ncolx; x++) { denom[y][x] = sigmay[y] * sigmax[x]; } } // 1-col returns scalar if (ncolx == 1 && ncoly == 1) { return new ValNum((cvs[0].getResult()._covs[0] / (fry.numRows() - 1)) / denom[0][0]); } //Gather final result, which is the correlation coefficient per column Vec[] res = new Vec[ncoly]; Key<Vec>[] keys = Vec.VectorGroup.VG_LEN1.addVecs(ncoly); for (int y = 0; y < ncoly; y++) { res[y] = Vec.makeVec(ArrayUtils.div(ArrayUtils.div(cvs[y].getResult()._covs, (fry.numRows() - 1)), denom[y]), keys[y]); } return new ValFrame(new Frame(fry._names, res)); } else { //if (mode.equals(Mode.CompleteObs)) //Omit NA rows between X and Y. //This will help with cov, sigma & mean calculations later as we only want to calculate cov, sigma, & mean //for rows with no NAs Frame frxy_naomit = new MRTask() { private void copyRow(int row, Chunk[] cs, NewChunk[] ncs) { for (int i = 0; i < cs.length; ++i) { if (cs[i] instanceof CStrChunk) ncs[i].addStr(cs[i], row); else if (cs[i] instanceof C16Chunk) ncs[i].addUUID(cs[i], row); else if (cs[i].hasFloat()) ncs[i].addNum(cs[i].atd(row)); else ncs[i].addNum(cs[i].at8(row), 0); } } @Override public void map(Chunk[] cs, NewChunk[] ncs) { int col; for (int row = 0; row < cs[0]._len; ++row) { for (col = 0; col < cs.length; ++col) if (cs[col].isNA(row)) break; if (col == cs.length) copyRow(row, cs, ncs); } } }.doAll(new Frame(frx).add(fry).types(), new Frame(frx).add(fry)).outputFrame(new Frame(frx).add(fry).names(), new Frame(frx).add(fry).domains()); //Collect new vecs that do not contain NA rows Vec[] vecxs_naomit = frxy_naomit.subframe(0, ncolx).vecs(); int ncolx_naomit = vecxs_naomit.length; Vec[] vecys_naomit = frxy_naomit.subframe(ncolx, frxy_naomit.vecs().length).vecs(); int ncoly_naomit = vecys_naomit.length; //Set up CoVarTask CoVarTask[] cvs = new CoVarTask[ncoly_naomit]; //Get mean of X vecs double[] xmeans = new double[ncolx_naomit]; for (int x = 0; x < ncolx_naomit; x++) { xmeans[x] = vecxs_naomit[x].mean(); } //Set up double arrays to capture sd(x), sd(y) and sd(x) * sd(y) double[] sigmay = new double[ncoly_naomit]; double[] sigmax = new double[ncolx_naomit]; double[][] denom = new double[ncoly_naomit][ncolx_naomit]; // Launch tasks; each does all Xs vs one Y for (int y = 0; y < ncoly_naomit; y++) { //Get covariance between x and y cvs[y] = new CoVarTask(vecys_naomit[y].mean(), xmeans).dfork(new Frame(vecys_naomit[y]).add(frxy_naomit.subframe(0, ncolx))); //Get sigma of y vecs sigmay[y] = vecys_naomit[y].sigma(); } //Get sigma of x vecs for (int x = 0; x < ncolx_naomit; x++) { sigmax[x] = vecxs_naomit[x].sigma(); } //Denominator for correlation calculation is sigma_y * sigma_x (All x sigmas vs one Y) for (int y = 0; y < ncoly_naomit; y++) { for (int x = 0; x < ncolx_naomit; x++) { denom[y][x] = sigmay[y] * sigmax[x]; } } // 1-col returns scalar if (ncolx_naomit == 1 && ncoly_naomit == 1) { return new ValNum((cvs[0].getResult()._covs[0] / (frxy_naomit.numRows() - 1)) / denom[0][0]); } //Gather final result, which is the correlation coefficient per column Vec[] res = new Vec[ncoly_naomit]; Key<Vec>[] keys = Vec.VectorGroup.VG_LEN1.addVecs(ncoly_naomit); for (int y = 0; y < ncoly_naomit; y++) { res[y] = Vec.makeVec(ArrayUtils.div(ArrayUtils.div(cvs[y].getResult()._covs, (frxy_naomit.numRows() - 1)), denom[y]), keys[y]); } return new ValFrame(new Frame(frxy_naomit.subframe(ncolx, frxy_naomit.vecs().length)._names, res)); } } private static class CoVarTask extends MRTask<CoVarTask> { double[] _covs; final double _xmeans[], _ymean; CoVarTask(double ymean, double[] xmeans) { _ymean = ymean; _xmeans = xmeans; } @Override public void map(Chunk cs[]) { final int ncolsx = cs.length - 1; final Chunk cy = cs[0]; final int len = cy._len; _covs = new double[ncolsx]; double sum; for (int x = 0; x < ncolsx; x++) { sum = 0; final Chunk cx = cs[x + 1]; final double xmean = _xmeans[x]; for (int row = 0; row < len; row++) sum += (cx.atd(row) - xmean) * (cy.atd(row) - _ymean); _covs[x] = sum; } } @Override public void reduce(CoVarTask cvt) { ArrayUtils.add(_covs, cvt._covs); } } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/advmath/AstDistance.java
package water.rapids.ast.prims.advmath; import water.H2O; import water.MRTask; import water.fvec.*; import water.rapids.Env; import water.rapids.Val; import water.rapids.ast.AstBuiltin; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import water.rapids.vals.ValFrame; import water.util.ArrayUtils; import water.util.Log; import java.util.Arrays; /** * Calculate Distance Metric between pairs of rows */ public class AstDistance extends AstBuiltin<AstDistance> { @Override public String[] args() { return new String[]{"ary", "x", "y", "measure"}; } @Override public int nargs() { return 1 + 3; /* (distance X Y measure) */ } @Override public String str() { return "distance"; } @Override public String description() { return "Compute a pairwise distance measure between all rows of two numeric H2OFrames.\n" + "For a given (usually larger) reference frame (N rows x p cols),\n" + "and a (usually smaller) query frame (M rows x p cols), we return a numeric Frame of size (N rows x M cols),\n" + "where the ij-th element is the distance measure between the i-th reference row and the j-th query row.\n" + "Note1: The output frame is symmetric.\n" + "Note2: Since N x M can be very large, it may be more efficient (memory-wise) to make multiple calls with smaller query Frames."; } @Override public Val apply(Env env, Env.StackHelp stk, AstRoot asts[]) { Frame frx = stk.track(asts[1].exec(env)).getFrame(); Frame fry = stk.track(asts[2].exec(env)).getFrame(); String measure = stk.track(asts[3].exec(env)).getStr(); return computeCosineDistances(frx, fry, measure); } public Val computeCosineDistances(Frame references, Frame queries, String distanceMetric) { Log.info("Number of references: " + references.numRows()); Log.info("Number of queries : " + queries.numRows()); String[] options = new String[]{"cosine","cosine_sq","l1","l2"}; if (!ArrayUtils.contains(options, distanceMetric.toLowerCase())) throw new IllegalArgumentException("Invalid distance measure provided: " + distanceMetric + ". Mustbe one of " + Arrays.toString(options)); if (references.numRows() * queries.numRows() * 8 > H2O.CLOUD.free_mem() ) throw new IllegalArgumentException("Not enough free memory to allocate the distance matrix (" + references.numRows() + " rows and " + queries.numRows() + " cols. Try specifying a smaller query frame."); if (references.numCols() != queries.numCols()) throw new IllegalArgumentException("Frames must have the same number of cols, found " + references.numCols() + " and " + queries.numCols()); if (queries.numRows() > Integer.MAX_VALUE) throw new IllegalArgumentException("Queries can't be larger than 2 billion rows."); if (queries.numCols() != references.numCols()) throw new IllegalArgumentException("Queries and References must have the same dimensionality"); for (int i=0;i<queries.numCols();++i) { if (!references.vec(i).isNumeric()) throw new IllegalArgumentException("References column " + references.name(i) + " is not numeric."); if (!queries.vec(i).isNumeric()) throw new IllegalArgumentException("Queries column " + references.name(i) + " is not numeric."); if (references.vec(i).naCnt()>0) throw new IllegalArgumentException("References column " + references.name(i) + " contains missing values."); if (queries.vec(i).naCnt()>0) throw new IllegalArgumentException("Queries column " + references.name(i) + " contains missing values."); } return new ValFrame(new DistanceComputer(queries, distanceMetric).doAll((int)queries.numRows(), Vec.T_NUM, references).outputFrame()); } static public class DistanceComputer extends MRTask<DistanceComputer> { Frame _queries; String _measure; DistanceComputer(Frame queries, String measure) { _queries = queries; _measure = measure; } @Override public void map(Chunk[] cs, NewChunk[] ncs) { int p = cs.length; //dimensionality int Q = (int) _queries.numRows(); int R = cs[0]._len; Vec.Reader[] Qs = new Vec.Reader[p]; for (int i = 0; i < p; ++i) { Qs[i] = _queries.vec(i).new Reader(); } double[] denomR = null; double[] denomQ = null; final boolean cosine = _measure.toLowerCase().equals("cosine"); final boolean cosine_sq = _measure.toLowerCase().equals("cosine_sq"); final boolean l1 = _measure.toLowerCase().equals("l1"); final boolean l2 = _measure.toLowerCase().equals("l2"); if (cosine || cosine_sq) { denomR = new double[R]; denomQ = new double[Q]; for (int r = 0; r < R; ++r) { // Reference row (chunk-local) for (int c = 0; c < p; ++c) { //cols denomR[r] += Math.pow(cs[c].atd(r), 2); } } for (int q = 0; q < Q; ++q) { // Query row (global) for (int c = 0; c < p; ++c) { //cols denomQ[q] += Math.pow(Qs[c].at(q), 2); } } } for (int r = 0; r < cs[0]._len; ++r) { // Reference row (chunk-local) for (int q = 0; q < Q; ++q) { // Query row (global) double distRQ = 0; if (l1) { for (int c = 0; c < p; ++c) { //cols distRQ += Math.abs(cs[c].atd(r) - Qs[c].at(q)); } } else if (l2) { for (int c = 0; c < p; ++c) { //cols distRQ += Math.pow(cs[c].atd(r) - Qs[c].at(q), 2); } distRQ = Math.sqrt(distRQ); } else if (cosine || cosine_sq) { for (int c = 0; c < p; ++c) { //cols distRQ += cs[c].atd(r) * Qs[c].at(q); } if (cosine_sq) { distRQ *= distRQ; distRQ /= denomR[r] * denomQ[q]; } else { distRQ /= Math.sqrt(denomR[r] * denomQ[q]); } } ncs[q].addNum(distRQ); // one Q distance per Reference } } } } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/advmath/AstHist.java
package water.rapids.ast.prims.advmath; import sun.misc.Unsafe; import water.MRTask; import water.fvec.Chunk; import water.fvec.Frame; import water.fvec.NewChunk; import water.fvec.Vec; import water.nbhm.UtilUnsafe; import water.rapids.ast.prims.reducers.AstMad; import water.rapids.Env; import water.rapids.Val; import water.rapids.vals.ValFrame; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import water.rapids.ast.params.AstNum; import water.rapids.ast.params.AstNumList; import water.rapids.ast.params.AstStr; import water.util.ArrayUtils; public class AstHist extends AstPrimitive { @Override public String[] args() { return new String[]{"ary", "breaks"}; } @Override public int nargs() { return 1 + 2; } // (hist x breaks) @Override public String str() { return "hist"; } @Override public ValFrame apply(Env env, Env.StackHelp stk, AstRoot asts[]) { // stack is [ ..., ary, breaks] // handle the breaks Frame fr2; Frame f = stk.track(asts[1].exec(env)).getFrame(); if (f.numCols() != 1) throw new IllegalArgumentException("Hist only applies to single numeric columns."); Vec vec = f.anyVec(); if (!vec.isNumeric()) throw new IllegalArgumentException("Hist only applies to single numeric columns."); //TODO Add case when vec is a constant numeric if(vec.isConst()) throw new IllegalArgumentException("Hist does not apply to constant numeric columns."); AstRoot a = asts[2]; String algo = null; int numBreaks = -1; double[] breaks = null; if (a instanceof AstStr) algo = a.str().toLowerCase(); else if (a instanceof AstNumList) breaks = ((AstNumList) a).expand(); else if (a instanceof AstNum) numBreaks = (int) a.exec(env).getNum(); AstHist.HistTask t; double h; double x1 = vec.max(); double x0 = vec.min(); if (breaks != null) t = new AstHist.HistTask(breaks, -1, -1/*ignored if _h==-1*/).doAll(vec); else if (algo != null) { switch (algo) { case "sturges": numBreaks = sturges(vec); h = (x1 - x0) / numBreaks; break; case "rice": numBreaks = rice(vec); h = (x1 - x0) / numBreaks; break; case "sqrt": numBreaks = sqrt(vec); h = (x1 - x0) / numBreaks; break; case "doane": numBreaks = doane(vec); h = (x1 - x0) / numBreaks; break; case "scott": h = scotts_h(vec); numBreaks = scott(vec, h); break; // special bin width computation case "fd": h = fds_h(vec); numBreaks = fd(vec, h); break; // special bin width computation default: numBreaks = sturges(vec); h = (x1 - x0) / numBreaks; // just do sturges even if junk passed in } t = new AstHist.HistTask(computeCuts(vec, numBreaks), h, x0).doAll(vec); } else { h = (x1 - x0) / numBreaks; t = new AstHist.HistTask(computeCuts(vec, numBreaks), h, x0).doAll(vec); } // wanna make a new frame here [breaks,counts,mids] final double[] brks = t._breaks; final long[] cnts = t._counts; final double[] mids_true = t._mids; final double[] mids = new double[t._breaks.length - 1]; for (int i = 1; i < brks.length; ++i) mids[i - 1] = .5 * (t._breaks[i - 1] + t._breaks[i]); Vec layoutVec = Vec.makeZero(brks.length); fr2 = new MRTask() { @Override public void map(Chunk[] c, NewChunk[] nc) { int start = (int) c[0].start(); for (int i = 0; i < c[0]._len; ++i) { nc[0].addNum(brks[i + start]); if (i == 0) { nc[1].addNA(); nc[2].addNA(); nc[3].addNA(); } else { nc[1].addNum(cnts[(i - 1) + start]); nc[2].addNum(mids_true[(i - 1) + start]); nc[3].addNum(mids[(i - 1) + start]); } } } }.doAll(4, Vec.T_NUM, new Frame(layoutVec)).outputFrame(null, new String[]{"breaks", "counts", "mids_true", "mids"}, null); layoutVec.remove(); return new ValFrame(fr2); } public static int sturges(Vec v) { return (int) Math.ceil(1 + log2(v.length())); } public static int rice(Vec v) { return (int) Math.ceil(2 * Math.pow(v.length(), 1. / 3.)); } public static int sqrt(Vec v) { return (int) Math.sqrt(v.length()); } public static int doane(Vec v) { return (int) (1 + log2(v.length()) + log2(1 + (Math.abs(third_moment(v)) / sigma_g1(v)))); } public static int scott(Vec v, double h) { return (int) Math.ceil((v.max() - v.min()) / h); } public static int fd(Vec v, double h) { return (int) Math.ceil((v.max() - v.min()) / h); } // Freedman-Diaconis slightly modified to use MAD instead of IQR public static double fds_h(Vec v) { return 2 * AstMad.mad(new Frame(v), null, 1.4826) * Math.pow(v.length(), -1. / 3.); } public static double scotts_h(Vec v) { return 3.5 * Math.sqrt(AstVariance.getVar(v)) / (Math.pow(v.length(), 1. / 3.)); } public static double log2(double numerator) { return (Math.log(numerator)) / Math.log(2) + 1e-10; } public static double sigma_g1(Vec v) { return Math.sqrt((6 * (v.length() - 2)) / ((v.length() + 1) * (v.length() + 3))); } public static double third_moment(Vec v) { final double mean = v.mean(); AstHist.ThirdMomTask t = new AstHist.ThirdMomTask(mean).doAll(v); double m2 = t._ss / v.length(); double m3 = t._sc / v.length(); return m3 / Math.pow(m2, 1.5); } public static class ThirdMomTask extends MRTask<AstHist.ThirdMomTask> { double _ss; double _sc; final double _mean; ThirdMomTask(double mean) { _mean = mean; } @Override public void setupLocal() { _ss = 0; _sc = 0; } @Override public void map(Chunk c) { for (int i = 0; i < c._len; ++i) { if (!c.isNA(i)) { double d = c.atd(i) - _mean; double d2 = d * d; _ss += d2; _sc += d2 * d; } } } @Override public void reduce(AstHist.ThirdMomTask t) { _ss += t._ss; _sc += t._sc; } } public static double fourth_moment(Vec v) { final double mean = v.mean(); AstHist.FourthMomTask t = new AstHist.FourthMomTask(mean).doAll(v); double m2 = t._ss / v.length(); double m4 = t._sc / v.length(); return m4 / Math.pow(m2, 2.0); } public static class FourthMomTask extends MRTask<AstHist.FourthMomTask> { double _ss; double _sc; final double _mean; FourthMomTask(double mean) { _mean = mean; } @Override public void setupLocal() { _ss = 0; _sc = 0; } @Override public void map(Chunk c) { for (int i = 0; i < c._len; ++i) { if (!c.isNA(i)) { double d = c.atd(i) - _mean; double d2 = d * d; _ss += d2; _sc += d2 * d * d; } } } @Override public void reduce(AstHist.FourthMomTask t) { _ss += t._ss; _sc += t._sc; } } public double[] computeCuts(Vec v, int numBreaks) { if (numBreaks <= 0) throw new IllegalArgumentException("breaks must be a positive number"); // just make numBreaks cuts equidistant from each other spanning range of [v.min, v.max] double min; double w = (v.max() - (min = v.min())) / numBreaks; double[] res = new double[numBreaks]; for (int i = 0; i < numBreaks; ++i) res[i] = min + w * (i + 1); return res; } public static class HistTask extends MRTask<AstHist.HistTask> { final private double _h; // bin width final private double _x0; // far left bin edge final private double[] _min; // min for each bin, updated atomically final private double[] _max; // max for each bin, updated atomically // unsafe crap for mins/maxs of bins private static final Unsafe U = UtilUnsafe.getUnsafe(); // double[] offset and scale private static final int _dB = U.arrayBaseOffset(double[].class); private static final int _dS = U.arrayIndexScale(double[].class); private static long doubleRawIdx(int i) { return _dB + _dS * i; } // long[] offset and scale private static final int _8B = U.arrayBaseOffset(long[].class); private static final int _8S = U.arrayIndexScale(long[].class); private static long longRawIdx(int i) { return _8B + _8S * i; } // out private final double[] _breaks; private final long[] _counts; private final double[] _mids; HistTask(double[] cuts, double h, double x0) { _breaks = cuts; _min = new double[_breaks.length - 1]; _max = new double[_breaks.length - 1]; _counts = new long[_breaks.length - 1]; _mids = new double[_breaks.length - 1]; _h = h; _x0 = x0; } @Override public void map(Chunk c) { // if _h==-1, then don't have fixed bin widths... must loop over bins to obtain the correct bin # for (int i = 0; i < c._len; ++i) { int x = 1; if (c.isNA(i)) continue; double r = c.atd(i); if (_h == -1) { for (; x < _counts.length; x++) if (r <= _breaks[x]) break; x--; // back into the bin where count should go } else x = Math.min(_counts.length - 1, (int) Math.floor((r - _x0) / _h)); // Pick the bin floor( (x - x0) / h ) or ceil( (x-x0)/h - 1 ), choose the first since fewer ops bumpCount(x); setMinMax(Double.doubleToRawLongBits(r), x); } } @Override public void reduce(AstHist.HistTask t) { if (_counts != t._counts) ArrayUtils.add(_counts, t._counts); for (int i = 0; i < _mids.length; ++i) { _min[i] = t._min[i] < _min[i] ? t._min[i] : _min[i]; _max[i] = t._max[i] > _max[i] ? t._max[i] : _max[i]; } } @Override public void postGlobal() { for (int i = 0; i < _mids.length; ++i) _mids[i] = 0.5 * (_max[i] + _min[i]); } private void bumpCount(int x) { long o = _counts[x]; while (!U.compareAndSwapLong(_counts, longRawIdx(x), o, o + 1)) o = _counts[x]; } private void setMinMax(long v, int x) { double o = _min[x]; double vv = Double.longBitsToDouble(v); while (vv < o && U.compareAndSwapLong(_min, doubleRawIdx(x), Double.doubleToRawLongBits(o), v)) o = _min[x]; setMax(v, x); } private void setMax(long v, int x) { double o = _max[x]; double vv = Double.longBitsToDouble(v); while (vv > o && U.compareAndSwapLong(_min, doubleRawIdx(x), Double.doubleToRawLongBits(o), v)) o = _max[x]; } } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/advmath/AstImpute.java
package water.rapids.ast.prims.advmath; import hex.quantile.QuantileModel; import water.Freezable; import water.H2O; import water.MRTask; import water.fvec.Chunk; import water.fvec.Frame; import water.fvec.Vec; import water.rapids.*; import water.rapids.ast.AstFrame; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import water.rapids.ast.params.AstNum; import water.rapids.ast.params.AstNumList; import water.rapids.ast.params.AstStr; import water.rapids.ast.params.AstStrList; import water.rapids.ast.prims.mungers.AstGroup; import water.rapids.ast.prims.reducers.AstMean; import water.rapids.ast.prims.reducers.AstMedian; import water.rapids.vals.ValFrame; import water.rapids.vals.ValNums; import water.util.ArrayUtils; import water.util.IcedDouble; import water.util.IcedHashMap; import java.util.Arrays; import java.util.HashSet; import java.util.Set; /** * Impute columns of a data frame in place. * <p/> * This impute can impute whole Frames or a specific Vec within the Frame. Imputation * will be by the default mean (for numeric columns) or mode (for categorical columns). * String, date, and UUID columns are never imputed. * <p/> * When a Vec is specified to be imputed, it can alternatively be imputed by grouping on * some other columns in the Frame. If groupByCols is specified, but the user does not * supply a column to be imputed then an IllegalArgumentException will be raised. Further, * if the user specifies the column to impute within the groupByCols, exceptions will be * raised. * <p/> * The methods that a user may impute by are as follows: * - mean: Vec.T_NUM * - median: Vec.T_NUM * - mode: Vec.T_CAT * - bfill: Any valid Vec type * - ffill: Any valid Vec type * <p/> * All methods of imputation are done in place! The first three methods (mean, median, * mode) are self-explanatory. The bfill and ffill methods will attempt to fill NAs using * adjacent cell value (either before or forward): * <p/> * Vec = [ bfill_value, NA, ffill_value] * | ^^ | * -> || <- * impute * <p/> * If the impute method is median then the combineMethod can be one of the Enum variants * of QuantileModel.CombineMethod = { INTERPOLATE, AVERAGE, LOW, HIGH }. The Enum * specifies how to combine quantiles on even sample sizes. This parameter is ignored in * all other cases. * <p/> * Finally, the groupByFrame can be used to impute a column with a pre-computed groupby * result. * <p/> * Other notes: * <p/> * If col is -1, then the entire Frame will be imputed using mean/mode where appropriate. */ public class AstImpute extends AstPrimitive { @Override public String[] args() { return new String[]{"ary", "col", "method", "combineMethod", "groupByCols", "groupByFrame", "values"}; } @Override public String str() { return "h2o.impute"; } @Override public int nargs() { return 1 + 7; } // (h2o.impute data col method combine_method groupby groupByFrame values) @Override public Val apply(Env env, Env.StackHelp stk, AstRoot asts[]) { // Argument parsing and sanity checking // Whole frame being imputed Frame fr = stk.track(asts[1].exec(env)).getFrame(); // Column within frame being imputed final int col = (int) asts[2].exec(env).getNum(); if (col >= fr.numCols()) throw new IllegalArgumentException("Column not -1 or in range 0 to " + fr.numCols()); final boolean doAllVecs = col == -1; final Vec vec = doAllVecs ? null : fr.vec(col); // Technique used for imputation AstRoot method = null; boolean ffill0 = false, bfill0 = false; switch (asts[3].exec(env).getStr().toUpperCase()) { case "MEAN": method = new AstMean(); break; case "MEDIAN": method = new AstMedian(); break; case "MODE": method = new AstMode(); break; case "FFILL": ffill0 = true; break; case "BFILL": bfill0 = true; break; default: throw new IllegalArgumentException("Method must be one of mean, median or mode"); } // Only for median, how is the median computed on even sample sizes? QuantileModel.CombineMethod combine = QuantileModel.CombineMethod.valueOf(asts[4].exec(env).getStr().toUpperCase()); // Group-by columns. Empty is allowed, and perfectly normal. AstRoot ast = asts[5]; AstNumList by2; if (ast instanceof AstNumList) by2 = (AstNumList) ast; else if (ast instanceof AstNum) by2 = new AstNumList(((AstNum) ast).getNum()); else if (ast instanceof AstStrList) { String[] names = ((AstStrList) ast)._strs; double[] list = new double[names.length]; int i = 0; for (String name : ((AstStrList) ast)._strs) list[i++] = fr.find(name); Arrays.sort(list); by2 = new AstNumList(list); } else throw new IllegalArgumentException("Requires a number-list, but found a " + ast.getClass()); Frame groupByFrame = asts[6].str().equals("_") ? null : stk.track(asts[6].exec(env)).getFrame(); AstRoot vals = asts[7]; AstNumList values; if (vals instanceof AstNumList) values = (AstNumList) vals; else if (vals instanceof AstNum) values = new AstNumList(((AstNum) vals).getNum()); else values = null; boolean doGrpBy = !by2.isEmpty() || groupByFrame != null; // Compute the imputed value per-group. Empty groups are allowed and OK. IcedHashMap<AstGroup.G, Freezable[]> group_impute_map; if (!doGrpBy) { // Skip the grouping work if (ffill0 || bfill0) { // do a forward/backward fill on the NA // TODO: requires chk.previousNonNA and chk.nextNonNA style methods (which may go across chk boundaries)s final boolean ffill = ffill0; final boolean bfill = bfill0; throw H2O.unimpl("No ffill or bfill imputation supported"); // new MRTask() { // @Override public void map(Chunk[] cs) { // int len=cs[0]._len; // end of this chk // long start=cs[0].start(); // absolute beginning of chk s.t. start-1 bleeds into previous chk // long absEnd = start+len; // absolute end of the chk s.t. absEnd+1 bleeds into next chk // for(int c=0;c<cs.length;++c ) // for(int r=0;r<cs[0]._len;++r ) { // if( cs[c].isNA(r) ) { // if( r > 0 && r < len-1 ) { // cs[c].set(r,ffill?) // } // } // } // } // }.doAll(doAllVecs?fr:new Frame(vec)); // return new ValNum(Double.NaN); } else { final double[] res = values == null ? new double[fr.numCols()] : values.expand(); if (values == null) { // fill up res if no values supplied user, common case if (doAllVecs) { for (int i = 0; i < res.length; ++i) if (fr.vec(i).isNumeric() || fr.vec(i).isCategorical()) res[i] = fr.vec(i).isNumeric() ? fr.vec(i).mean() : ArrayUtils.maxIndex(fr.vec(i).bins()); } else { Arrays.fill(res, Double.NaN); if (method instanceof AstMean) res[col] = vec.mean(); if (method instanceof AstMedian) res[col] = AstMedian.median(new Frame(vec), combine); if (method instanceof AstMode) res[col] = AstMode.mode(vec); } } new MRTask() { @Override public void map(Chunk[] cs) { int len = cs[0]._len; // run down each chk for (int c = 0; c < cs.length; ++c) if (!Double.isNaN(res[c])) for (int row = 0; row < len; ++row) if (cs[c].isNA(row)) cs[c].set(row, res[c]); } }.doAll(fr); return new ValNums(res); } } else { if (col >= fr.numCols()) throw new IllegalArgumentException("Column not -1 or in range 0 to " + fr.numCols()); Frame imputes = groupByFrame; if (imputes == null) { // Build and run a GroupBy command AstGroup ast_grp = new AstGroup(); // simple case where user specified a column... col == -1 means do all columns if (doAllVecs) { AstRoot[] aggs = new AstRoot[(int) (3 + 3 * (fr.numCols() - by2.cnt()))]; aggs[0] = ast_grp; aggs[1] = new AstFrame(fr); aggs[2] = by2; int c = 3; for (int i = 0; i < fr.numCols(); ++i) { if (!by2.has(i) && (fr.vec(i).isCategorical() || fr.vec(i).isNumeric())) { aggs[c] = fr.vec(i).isNumeric() ? new AstMean() : new AstMode(); aggs[c + 1] = new AstNumList(i, i + 1); aggs[c + 2] = new AstStr("rm"); c += 3; } } imputes = ast_grp.apply(env, stk, aggs).getFrame(); } else imputes = ast_grp.apply(env, stk, new AstRoot[]{ast_grp, new AstFrame(fr), by2, /**/method, new AstNumList(col, col + 1), new AstStr("rm") /**/}).getFrame(); } if (by2.isEmpty() && imputes.numCols() > 2) // >2 makes it ambiguous which columns are groupby cols and which are aggs, throw IAE throw new IllegalArgumentException("Ambiguous group-by frame. Supply the `by` columns to proceed."); final int[] bycols0 = ArrayUtils.seq(0, Math.max((int) by2.cnt(), 1 /* imputes.numCols()-1 */)); group_impute_map = new Gather(by2.expand4(), bycols0, fr.numCols(), col).doAll(imputes)._group_impute_map; // Now walk over the data, replace NAs with the imputed results final IcedHashMap<AstGroup.G, Freezable[]> final_group_impute_map = group_impute_map; if (by2.isEmpty()) { int[] byCols = new int[imputes.numCols() - 1]; for (int i = 0; i < byCols.length; ++i) byCols[i] = fr.find(imputes.name(i)); by2 = new AstNumList(byCols); } final int[] bycols = by2.expand4(); new MRTask() { @Override public void map(Chunk cs[]) { Set<Integer> _bycolz = new HashSet<>(); for (int b : bycols) _bycolz.add(b); AstGroup.G g = new AstGroup.G(bycols.length, null); for (int row = 0; row < cs[0]._len; row++) for (int c = 0; c < cs.length; ++c) if (!_bycolz.contains(c)) if (cs[c].isNA(row)) cs[c].set(row, ((IcedDouble) final_group_impute_map.get(g.fill(row, cs, bycols))[c])._val); } }.doAll(fr); return new ValFrame(imputes); } } // flatten the GroupBy result Frame back into a IcedHashMap private static class Gather extends MRTask<Gather> { private final int _imputedCol; private final int _ncol; private final int[] _byCols0; // actual group-by indexes private final int[] _byCols; // index into the grouped-by frame result private IcedHashMap<AstGroup.G, Freezable[]> _group_impute_map; private transient Set<Integer> _localbyColzSet; Gather(int[] byCols0, int[] byCols, int ncol, int imputeCol) { _byCols = byCols; _byCols0 = byCols0; _ncol = ncol; _imputedCol = imputeCol; } @Override public void setupLocal() { _localbyColzSet = new HashSet<>(); for (int by : _byCols0) _localbyColzSet.add(by); } @Override public void map(Chunk cs[]) { _group_impute_map = new IcedHashMap<>(); for (int row = 0; row < cs[0]._len; ++row) { IcedDouble[] imputes = new IcedDouble[_ncol]; for (int c = 0, z = _byCols.length; c < imputes.length; ++c, ++z) { // z used to skip over the gby cols into the columns containing the aggregated columns if (_imputedCol != -1) imputes[c] = c == _imputedCol ? new IcedDouble(cs[cs.length - 1].atd(row)) : new IcedDouble(Double.NaN); else imputes[c] = _localbyColzSet.contains(c) ? new IcedDouble(Double.NaN) : new IcedDouble(cs[z].atd(row)); } _group_impute_map.put(new AstGroup.G(_byCols.length, null).fill(row, cs, _byCols), imputes); } } @Override public void reduce(Gather mrt) { _group_impute_map.putAll(mrt._group_impute_map); } } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/advmath/AstKFold.java
package water.rapids.ast.prims.advmath; import water.MRTask; import water.fvec.Chunk; import water.fvec.Frame; import water.fvec.Vec; import water.rapids.Env; import water.rapids.vals.ValFrame; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import water.util.VecUtils; import java.util.Random; import static water.util.RandomUtils.getRNG; public class AstKFold extends AstPrimitive { @Override public String[] args() { return new String[]{"ary", "nfolds", "seed"}; } @Override public int nargs() { return 1 + 3; } // (kfold_column x nfolds seed) @Override public String str() { return "kfold_column"; } public static Vec kfoldColumn(Vec v, final int nfolds, final long seed) { new MRTask() { @Override public void map(Chunk c) { long start = c.start(); for (int i = 0; i < c._len; ++i) { int fold = Math.abs(getRNG(start + seed + i).nextInt()) % nfolds; c.set(i, fold); } } }.doAll(v); return v; } public static Vec moduloKfoldColumn(Vec v, final int nfolds) { new MRTask() { @Override public void map(Chunk c) { long start = c.start(); for (int i = 0; i < c._len; ++i) c.set(i, (int) ((start + i) % nfolds)); } }.doAll(v); return v; } public static Vec stratifiedKFoldColumn(Vec y, final int nfolds, final long seed) { // for each class, generate a fold column (never materialized) // therefore, have a seed per class to be used by the map call if (!(y.isCategorical() || (y.isNumeric() && y.isInt()))) throw new IllegalArgumentException("stratification only applies to integer and categorical columns. Got: " + y.get_type_str()); final long[] classes = new VecUtils.CollectIntegerDomain().doAll(y).domain(); final int nClass = y.isNumeric() ? classes.length : y.domain().length; final long[] seeds = new long[nClass]; // seed for each regular fold column (one per class) for (int i = 0; i < nClass; ++i) seeds[i] = getRNG(seed + i).nextLong(); return new MRTask() { private int getFoldId(long absoluteRow, long seed) { return Math.abs(getRNG(absoluteRow + seed).nextInt()) % nfolds; } // dress up the foldColumn (y[1]) as follows: // 1. For each testFold and each classLabel loop over the response column (y[0]) // 2. If the classLabel is the current response and the testFold is the foldId // for the current row and classLabel, then set the foldColumn to testFold // // How this balances labels per fold: // Imagine that a KFold column was generated for each class. Observe that this // makes the outer loop a way of selecting only the test rows from each fold // (i.e., the holdout rows). Each fold is balanced sequentially in this way // since y[1] is only updated if the current row happens to be a holdout row // for the given classLabel. // // Next observe that looping over each classLabel filters down each KFold // so that it contains labels for just THAT class. This is how the balancing // can be made so that it is independent of the chunk distribution and the // per chunk class distribution. // // Downside is this performs nfolds*nClass passes over each Chunk. For // "reasonable" classification problems, this could be 100 passes per Chunk. @Override public void map(Chunk[] y) { long start = y[0].start(); for (int testFold = 0; testFold < nfolds; ++testFold) { for (int classLabel = 0; classLabel < nClass; ++classLabel) { for (int row = 0; row < y[0]._len; ++row) { // missing response gets spread around if (y[0].isNA(row)) { if ((start + row) % nfolds == testFold) y[1].set(row, testFold); } else { if (y[0].at8(row) == (classes == null ? classLabel : classes[classLabel])) { if (testFold == getFoldId(start + row, seeds[classLabel])) y[1].set(row, testFold); } } } } } } }.doAll(new Frame(y, y.makeZero()))._fr.vec(1); } @Override public ValFrame apply(Env env, Env.StackHelp stk, AstRoot asts[]) { Vec foldVec = stk.track(asts[1].exec(env)).getFrame().anyVec().makeZero(); int nfolds = (int) asts[2].exec(env).getNum(); long seed = (long) asts[3].exec(env).getNum(); return new ValFrame(new Frame(kfoldColumn(foldVec, nfolds, seed == -1 ? new Random().nextLong() : seed))); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/advmath/AstKurtosis.java
package water.rapids.ast.prims.advmath; import water.fvec.Frame; import water.fvec.Vec; import water.rapids.Env; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import water.rapids.vals.ValNums; public class AstKurtosis extends AstPrimitive { @Override public String[] args() { return new String[]{"ary", "na_rm"}; } @Override public String str() {return "kurtosis";} @Override public int nargs() {return 1 + 2;} // (kurtosis ary na.rm) @Override public ValNums apply(Env env, Env.StackHelp stk, AstRoot asts[]) { Frame fr = stk.track(asts[1].exec(env)).getFrame(); boolean narm = asts[2].exec(env).getNum() == 1; double[] ds = new double[fr.numCols()]; Vec[] vecs = fr.vecs(); for (int i = 0; i < fr.numCols(); i++) ds[i] = kurtosis(vecs[i], narm); return new ValNums(ds); } public static double kurtosis(Vec v, boolean narm) { return !v.isNumeric() || v.length() == 0 || (!narm && v.naCnt() > 0) ? Double.NaN : AstHist.fourth_moment(v); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/advmath/AstMode.java
package water.rapids.ast.prims.advmath; import water.fvec.Frame; import water.fvec.Vec; import water.rapids.Env; import water.rapids.vals.ValNum; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import water.util.ArrayUtils; import water.util.MRUtils; /** * Find the mode: the most popular element. */ public class AstMode extends AstPrimitive { @Override public String[] args() { return new String[]{"ary"}; } @Override public String str() { return "mode"; } @Override public int nargs() { return 1 + 1; } // (mode ary) @Override public ValNum apply(Env env, Env.StackHelp stk, AstRoot asts[]) { Frame fr = stk.track(asts[1].exec(env)).getFrame(); if (fr.numCols() != 1 || !fr.anyVec().isCategorical()) throw new IllegalArgumentException("mode only works on a single categorical column"); return new ValNum(mode(fr.anyVec())); } public static int mode(Vec v) { if (v.isNumeric()) { MRUtils.Dist t = new MRUtils.Dist().doAll(v); int mode = ArrayUtils.maxIndex(t.dist()); return (int) t.keys()[mode]; } double[] dist = new MRUtils.ClassDist(v).doAll(v).dist(); return ArrayUtils.maxIndex(dist); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/advmath/AstModuloKFold.java
package water.rapids.ast.prims.advmath; import water.fvec.Frame; import water.fvec.Vec; import water.rapids.Env; import water.rapids.vals.ValFrame; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; public class AstModuloKFold extends AstPrimitive { @Override public String[] args() { return new String[]{"ary", "nfolds"}; } @Override public int nargs() { return 1 + 2; } // (modulo_kfold_column x nfolds) @Override public String str() { return "modulo_kfold_column"; } @Override public ValFrame apply(Env env, Env.StackHelp stk, AstRoot asts[]) { Vec foldVec = stk.track(asts[1].exec(env)).getFrame().anyVec().makeZero(); int nfolds = (int) asts[2].exec(env).getNum(); return new ValFrame(new Frame(AstKFold.moduloKfoldColumn(foldVec, nfolds))); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/advmath/AstQtile.java
package water.rapids.ast.prims.advmath; import hex.quantile.Quantile; import hex.quantile.QuantileModel; import water.DKV; import water.Job; import water.fvec.Frame; import water.fvec.Vec; import water.rapids.Env; import water.rapids.Val; import water.rapids.vals.ValFrame; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import water.rapids.ast.params.AstNumList; /** * Quantiles: * (quantile %frame [numnber_list_probs] "string_interpolation_type") */ public class AstQtile extends AstPrimitive { @Override public String[] args() { return new String[]{"ary", "probs", "interpolationMethod", "weights_column"}; } @Override public int nargs() { return 1 + 4; } @Override public String str() { return "quantile"; } @Override public ValFrame apply(Env env, Env.StackHelp stk, AstRoot asts[]) { QuantileModel.QuantileParameters parms = new QuantileModel.QuantileParameters(); Frame fr = stk.track(asts[1].exec(env)).getFrame(); Frame fr_wkey = new Frame(fr); // Force a bogus Key for Quantiles ModelBuilder DKV.put(fr_wkey); parms._train = fr_wkey._key; parms._probs = ((AstNumList) asts[2]).expand(); for (double d : parms._probs) if (d < 0 || d > 1) throw new IllegalArgumentException("Probability must be between 0 and 1: " + d); String inter = asts[3].exec(env).getStr(); parms._combine_method = QuantileModel.CombineMethod.valueOf(inter.toUpperCase()); parms._weights_column = asts[4].str().equals("_") ? null : asts[4].str(); // Compute Quantiles Job j = new Quantile(parms).trainModel(); QuantileModel q = (QuantileModel) j.get(); DKV.remove(j._key); // Remove bogus Key DKV.remove(fr_wkey._key); // Reshape all outputs as a Frame, with probs in col 0 and the // quantiles in cols 1 thru fr.numCols() - except the optional weights vec int ncols = fr.numCols(); if (parms._weights_column != null) ncols--; Vec[] vecs = new Vec[1 /*1 more for the probs themselves*/ + ncols]; String[] names = new String[vecs.length]; vecs[0] = Vec.makeCon(null, parms._probs); names[0] = "Probs"; int w = 0; for (int i = 0; i < vecs.length - 1; ++i) { if (fr._names[i].equals(parms._weights_column)) w = 1; vecs[i + 1] = Vec.makeCon(null, q._output._quantiles[i]); names[i + 1] = fr._names[w + i] + "Quantiles"; } q.delete(); return new ValFrame(new Frame(names, vecs)); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/advmath/AstRunif.java
package water.rapids.ast.prims.advmath; import water.fvec.Frame; import water.fvec.Vec; import water.rapids.Env; import water.rapids.Val; import water.rapids.vals.ValFrame; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import java.util.Random; public class AstRunif extends AstPrimitive { @Override public String[] args() { return new String[]{"ary", "seed"}; } @Override public int nargs() { return 1 + 2; } // (h2o.runif frame seed) @Override public String str() { return "h2o.runif"; } @Override public ValFrame apply(Env env, Env.StackHelp stk, AstRoot asts[]) { Frame fr = stk.track(asts[1].exec(env)).getFrame(); long seed = (long) asts[2].exec(env).getNum(); if (seed == -1) seed = new Random().nextLong(); return new ValFrame(new Frame(new String[]{"rnd"}, new Vec[]{fr.anyVec().makeRand(seed)})); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/advmath/AstSkewness.java
package water.rapids.ast.prims.advmath; import water.fvec.Frame; import water.fvec.Vec; import water.rapids.Env; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import water.rapids.vals.ValNums; public class AstSkewness extends AstPrimitive { @Override public String[] args() { return new String[]{"ary", "na_rm"}; } @Override public String str() { return "skewness"; } @Override public int nargs() { return 1 + 2; } // (skewness ary na.rm) @Override public ValNums apply(Env env, Env.StackHelp stk, AstRoot asts[]) { Frame fr = stk.track(asts[1].exec(env)).getFrame(); boolean narm = asts[2].exec(env).getNum() == 1; double[] ds = new double[fr.numCols()]; Vec[] vecs = fr.vecs(); for (int i = 0; i < fr.numCols(); i++) ds[i] = skewness(vecs[i], narm); return new ValNums(ds); } public static double skewness(Vec v, boolean narm) { return !v.isNumeric() || v.length() == 0 || (!narm && v.naCnt() > 0) ? Double.NaN : AstHist.third_moment(v); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/advmath/AstStratifiedKFold.java
package water.rapids.ast.prims.advmath; import water.fvec.Frame; import water.fvec.Vec; import water.rapids.Env; import water.rapids.vals.ValFrame; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import java.util.Random; public class AstStratifiedKFold extends AstPrimitive { @Override public String[] args() { return new String[]{"ary", "nfolds", "seed"}; } @Override public int nargs() { return 1 + 3; } // (stratified_kfold_column x nfolds seed) @Override public String str() { return "stratified_kfold_column"; } @Override public ValFrame apply(Env env, Env.StackHelp stk, AstRoot asts[]) { Vec foldVec = stk.track(asts[1].exec(env)).getFrame().anyVec().makeZero(); int nfolds = (int) asts[2].exec(env).getNum(); long seed = (long) asts[3].exec(env).getNum(); return new ValFrame(new Frame(AstKFold.stratifiedKFoldColumn(foldVec, nfolds, seed == -1 ? new Random().nextLong() : seed))); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/advmath/AstStratifiedSplit.java
package water.rapids.ast.prims.advmath; import water.*; import water.fvec.Chunk; import water.fvec.Frame; import water.fvec.Vec; import water.rapids.Env; import water.rapids.vals.ValFrame; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import water.util.VecUtils; import java.util.*; import static water.util.RandomUtils.getRNG; public class AstStratifiedSplit extends AstPrimitive { public static final String OUTPUT_COLUMN_NAME = "test_train_split"; public static final String[] OUTPUT_COLUMN_DOMAIN = new String[]{"train", "test"}; @Override public String[] args() { return new String[]{"ary", "test_frac", "seed"}; } @Override public int nargs() { return 1 + 3; } // (h2o.random_stratified_split y test_frac seed) @Override public String str() { return "h2o.random_stratified_split"; } @Override public ValFrame apply(Env env, Env.StackHelp stk, AstRoot asts[]) { Frame frame = stk.track(asts[1].exec(env)).getFrame(); final double testFrac = asts[2].exec(env).getNum(); long seed = (long) asts[3].exec(env).getNum(); // It is just a single column if (frame.numCols() != 1) throw new IllegalArgumentException("Must give a single column to stratify against. Got: " + frame.numCols() + " columns."); Vec stratifyingColumn = frame.anyVec(); Frame result = new Frame(Key.<Frame>make(), new String[] {OUTPUT_COLUMN_NAME}, new Vec[] { split(stratifyingColumn, testFrac, seed, OUTPUT_COLUMN_DOMAIN)} ); return new ValFrame(result); } public static Vec split(Vec stratifyingColumn, double splittingFraction, long randomizationSeed, String[] splittingDom) { checkIfCanStratifyBy(stratifyingColumn); randomizationSeed = randomizationSeed == -1 ? new Random().nextLong() : randomizationSeed; // Collect input vector domain final long[] classes = new VecUtils.CollectIntegerDomain().doAll(stratifyingColumn).domain(); // Number of output classes final int numClasses = classes.length; // Make a new column based on input column - this needs to follow layout of input vector! // Save vector into DKV Vec outputVec = stratifyingColumn.makeCon(0.0, Vec.T_CAT); outputVec.setDomain(splittingDom); DKV.put(outputVec); // Collect index frame // FIXME: This is in fact collecting inverse index class -> {row indices} ClassIdxTask finTask = new ClassIdxTask(numClasses,classes).doAll(stratifyingColumn); // Loop through each class in the input column HashSet<Long> usedIdxs = new HashSet<>(); for (int classLabel = 0; classLabel < numClasses; classLabel++) { // extract frame with index locations of the minority class // calculate target number of this class to go to test final LongAry indexAry = finTask._indexes[classLabel]; long tnum = Math.max(Math.round(indexAry.size() * splittingFraction), 1); HashSet<Long> tmpIdxs = new HashSet<>(); // randomly select the target number of indexes int generated = 0; int count = 0; while (generated < tnum) { int i = (int) (getRNG(count+ randomizationSeed).nextDouble() * indexAry.size()); if (tmpIdxs.contains(indexAry.get(i))) { count+=1;continue; } tmpIdxs.add(indexAry.get(i)); generated += 1; count += 1; } usedIdxs.addAll(tmpIdxs); } // Update class assignments new ClassAssignMRTask(usedIdxs).doAll(outputVec); return outputVec; } static void checkIfCanStratifyBy(Vec vec) { if (!(vec.isCategorical() || (vec.isNumeric() && vec.isInt()))) throw new IllegalArgumentException("Stratification only applies to integer and categorical columns. Got: " + vec.get_type_str()); if (vec.length() > Integer.MAX_VALUE) { throw new IllegalArgumentException("Cannot stratified the frame because it is too long: nrows=" + vec.length()); } } public static class ClassAssignMRTask extends MRTask<AstStratifiedSplit.ClassAssignMRTask> { HashSet<Long> _idx; ClassAssignMRTask(HashSet<Long> idx) { _idx = idx; } @Override public void map(Chunk ck) { for (int i = 0; i<ck.len(); i++) { if (_idx.contains(ck.start() + i)) { ck.set(i,1.0); } } _idx = null; // Do not send it back } } public static class ClassIdxTask extends MRTask<AstStratifiedSplit.ClassIdxTask> { LongAry[] _indexes; private final int _nclasses; private long[] _classes; private transient HashMap<Long, Integer> _classMap; public ClassIdxTask(int nclasses, long[] classes) { _nclasses = nclasses; _classes = classes; } @Override protected void setupLocal() { _classMap = new HashMap<>(2*_classes.length); for (int i = 0; i < _classes.length; i++) { _classMap.put(_classes[i], i); } } @Override public void map(Chunk[] ck) { _indexes = new LongAry[_nclasses]; for (int i = 0; i < _nclasses; i++) { _indexes[i] = new LongAry(); } for (int i = 0; i < ck[0].len(); i++) { long clas = ck[0].at8(i); Integer clas_idx = _classMap.get(clas); if (clas_idx != null) _indexes[clas_idx].add(ck[0].start() + i); } _classes = null; } @Override public void reduce(AstStratifiedSplit.ClassIdxTask c) { for (int i = 0; i < c._indexes.length; i++) { for (int j = 0; j < c._indexes[i].size(); j++) { _indexes[i].add(c._indexes[i].get(j)); } } } } public static class LongAry extends Iced<AstStratifiedSplit.LongAry> { public LongAry(long ...vals){_ary = vals; _sz = vals.length;} long [] _ary = new long[4]; int _sz; public void add(long i){ if (_sz == _ary.length) _ary = Arrays.copyOf(_ary, Math.max(4, _ary.length * 2)); _ary[_sz++] = i; } public long get(int i){ if(i >= _sz) throw new ArrayIndexOutOfBoundsException(i); return _ary[i]; } public int size(){return _sz;} public long[] toArray(){return Arrays.copyOf(_ary,_sz);} public void clear() {_sz = 0;} } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/advmath/AstTable.java
package water.rapids.ast.prims.advmath; import water.AutoBuffer; import water.MRTask; import water.fvec.Chunk; import water.fvec.Frame; import water.fvec.NewChunk; import water.fvec.Vec; import water.nbhm.NonBlockingHashMapLong; import water.rapids.Env; import water.rapids.Val; import water.rapids.vals.ValFrame; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import water.util.ArrayUtils; import java.util.Arrays; import java.util.concurrent.atomic.AtomicLong; /** * Variance between columns of a frame * TODO: Define "table" in terms of "groupby" * TODO: keep dense format for two-column comparison (like in previous version of Rapids) * (table X Y) ==> * (groupby (cbind X Y) [X Y] nrow TRUE) */ public class AstTable extends AstPrimitive { @Override public String[] args() { return new String[]{"X", "Y", "dense"}; } @Override public int nargs() { return -1; } // (table X dense) or (table X Y dense) @Override public String str() { return "table"; } @Override public ValFrame apply(Env env, Env.StackHelp stk, AstRoot asts[]) { Frame fr1 = stk.track(asts[1].exec(env)).getFrame(); final boolean dense = asts[asts.length - 1].exec(env).getNum() == 1; Frame fr2 = asts.length == 4 ? stk.track(asts[2].exec(env)).getFrame() : null; int ncols = fr1.numCols() + (fr2 == null ? 0 : fr2.numCols()); Vec vec1 = fr1.vec(0); ValFrame res = fast_table(vec1, ncols, fr1._names[0]); if (res != null) return res; if (!(asts.length == 3 || asts.length == 4) || ncols > 2) throw new IllegalArgumentException("table expects one or two columns"); Vec vec2 = fr1.numCols() == 2 ? fr1.vec(1) : fr2 != null ? fr2.vec(0) : null; int sz = fr1._names.length + (fr2 != null ? fr2._names.length : 0); String[] colnames = new String[sz]; int i = 0; for (String name : fr1._names) colnames[i++] = name; if (fr2 != null) for (String name : fr2._names) colnames[i++] = name; return slow_table(vec1, vec2, colnames, dense); } // ------------------------------------------------------------------------- // Fast-path for 1 integer column private ValFrame fast_table(Vec v1, int ncols, String colname) { if (ncols != 1 || !v1.isInt()) return null; long spanl = (long) v1.max() - (long) v1.min() + 1; if (spanl > 1000000) return null; // Cap at decent array size, for performance // First fast-pass counting AstTable.FastCnt fastCnt = new AstTable.FastCnt((long) v1.min(), (int) spanl).doAll(v1); final long cnts[] = fastCnt._cnts; final long minVal = fastCnt._min; // Second pass to build the result frame, skipping zeros Vec dataLayoutVec = Vec.makeCon(0, cnts.length); Frame fr = new MRTask() { @Override public void map(Chunk cs[], NewChunk nc0, NewChunk nc1) { final Chunk c = cs[0]; for (int i = 0; i < c._len; ++i) { int idx = (int) (i + c.start()); if (cnts[idx] > 0) { nc0.addNum(idx + minVal); nc1.addNum(cnts[idx]); } } } }.doAll(new byte[]{Vec.T_NUM, Vec.T_NUM}, dataLayoutVec).outputFrame(new String[]{colname, "Count"}, new String[][]{v1.domain(), null}); dataLayoutVec.remove(); return new ValFrame(fr); } // Fast-pass for counting unique integers in a span private static class FastCnt extends MRTask<AstTable.FastCnt> { final long _min; final int _span; long _cnts[]; FastCnt(long min, int span) { _min = min; _span = span; } @Override public void map(Chunk c) { _cnts = new long[_span]; for (int i = 0; i < c._len; i++) if (!c.isNA(i)) _cnts[(int) (c.at8(i) - _min)]++; } @Override public void reduce(AstTable.FastCnt fc) { ArrayUtils.add(_cnts, fc._cnts); } } // ------------------------------------------------------------------------- // Count unique combos in 1 or 2 columns, where the values are not integers, // or cover a very large span. private ValFrame slow_table(Vec v1, Vec v2, String[] colnames, boolean dense) { // For simplicity, repeat v1 if v2 is missing; this will end up filling in // only the diagonal of a 2-D array (in what is otherwise a 1-D array). // This should be nearly the same cost as a 1-D array, since everything is // sparsely filled in. // If this is the 1-column case (all counts on the diagonals), just build a // 1-d result. if (v2 == null) { // Slow-pass group counting, very sparse hashtables. Note that Vec v2 is // used as the left-most arg, or OUTER dimension - which will be columns in // the final result. AstTable.SlowCnt sc = new AstTable.SlowCnt().doAll(v1, v1); // Get the column headers as sorted doubles double dcols[] = collectDomain(sc._col0s); Frame res = new Frame(); Vec rowlabel = Vec.makeVec(dcols, Vec.VectorGroup.VG_LEN1.addVec()); rowlabel.setDomain(v1.domain()); res.add(colnames[0], rowlabel); long cnts[] = new long[dcols.length]; for (int col = 0; col < dcols.length; col++) { long lkey = Double.doubleToRawLongBits(dcols[col]); NonBlockingHashMapLong<AtomicLong> colx = sc._col0s.get(lkey); AtomicLong al = colx.get(lkey); cnts[col] = al.get(); } Vec vec = Vec.makeVec(cnts, null, Vec.VectorGroup.VG_LEN1.addVec()); res.add("Counts", vec); return new ValFrame(res); } // 2-d table result. Frame res = new Frame(); if (!dense) { // Slow-pass group counting, very sparse hashtables. Note that Vec v2 is // used as the left-most arg, or OUTER dimension - which will be columns in // the final result. AstTable.SlowCnt sc = new AstTable.SlowCnt().doAll(v2, v1); // Get the column headers as sorted doubles double dcols[] = collectDomain(sc._col0s); // Need the row headers as sorted doubles also, but these are scattered // throughout the nested tables. Fold 'em into 1 table. NonBlockingHashMapLong<AtomicLong> rows = new NonBlockingHashMapLong<>(); for (NonBlockingHashMapLong.IteratorLong i = iter(sc._col0s); i.hasNext(); ) rows.putAll(sc._col0s.get(i.nextLong())); double drows[] = collectDomain(rows); // Now walk the columns one by one, building a Vec per column, building a // Frame result. Rowlabel for first column. Vec rowlabel = Vec.makeVec(drows, Vec.VectorGroup.VG_LEN1.addVec()); rowlabel.setDomain(v1.domain()); res.add(colnames[0], rowlabel); long cnts[] = new long[drows.length]; for (int col = 0; col < dcols.length; col++) { NonBlockingHashMapLong<AtomicLong> colx = sc._col0s.get(Double.doubleToRawLongBits(dcols[col])); for (int row = 0; row < drows.length; row++) { AtomicLong al = colx.get(Double.doubleToRawLongBits(drows[row])); cnts[row] = al == null ? 0 : al.get(); } Vec vec = Vec.makeVec(cnts, null, Vec.VectorGroup.VG_LEN1.addVec()); res.add(v2.isCategorical() ? v2.domain()[col] : Double.toString(dcols[col]), vec); } } else { AstTable.SlowCnt sc = new AstTable.SlowCnt().doAll(v1, v2); double dcols[] = collectDomain(sc._col0s); NonBlockingHashMapLong<AtomicLong> rows = new NonBlockingHashMapLong<>(); for (NonBlockingHashMapLong.IteratorLong i = iter(sc._col0s); i.hasNext(); ) rows.putAll(sc._col0s.get(i.nextLong())); double drows[] = collectDomain(rows); int x = 0; int sz = 0; for (NonBlockingHashMapLong.IteratorLong i = iter(sc._col0s); i.hasNext(); ) { sz += sc._col0s.get(i.nextLong()).size(); } long cnts[] = new long[sz]; double[] left_categ = new double[sz]; double[] right_categ = new double[sz]; for (double dcol : dcols) { NonBlockingHashMapLong<AtomicLong> colx = sc._col0s.get(Double.doubleToRawLongBits(dcol)); for (double drow : drows) { AtomicLong al = colx.get(Double.doubleToRawLongBits(drow)); if (al != null) { left_categ[x] = dcol; right_categ[x] = drow; cnts[x] = al.get(); x++; } } } Vec vec = Vec.makeVec(left_categ, Vec.VectorGroup.VG_LEN1.addVec()); if (v1.isCategorical()) vec.setDomain(v1.domain()); res.add(colnames[0], vec); vec = Vec.makeVec(right_categ, Vec.VectorGroup.VG_LEN1.addVec()); if (v2.isCategorical()) vec.setDomain(v2.domain()); res.add(colnames[1], vec); vec = Vec.makeVec(cnts, null, Vec.VectorGroup.VG_LEN1.addVec()); res.add("Counts", vec); } return new ValFrame(res); } // Collect the unique longs from this NBHML, convert to doubles and return // them as a sorted double[]. private static double[] collectDomain(NonBlockingHashMapLong ls) { int sz = ls.size(); // Uniques double ds[] = new double[sz]; int x = 0; for (NonBlockingHashMapLong.IteratorLong i = iter(ls); i.hasNext(); ) ds[x++] = Double.longBitsToDouble(i.nextLong()); Arrays.sort(ds); return ds; } private static NonBlockingHashMapLong.IteratorLong iter(NonBlockingHashMapLong nbhml) { return (NonBlockingHashMapLong.IteratorLong) nbhml.keySet().iterator(); } // Implementation is a double-dimension NBHML. Each dimension key is the raw // long bits of the double column. Bottoms out in an AtomicLong. private static class SlowCnt extends MRTask<AstTable.SlowCnt> { transient NonBlockingHashMapLong<NonBlockingHashMapLong<AtomicLong>> _col0s; @Override public void setupLocal() { _col0s = new NonBlockingHashMapLong<>(); } @Override public void map(Chunk c0, Chunk c1) { for (int i = 0; i < c0._len; i++) { double d0 = c0.atd(i); if (Double.isNaN(d0)) continue; long l0 = Double.doubleToRawLongBits(d0); double d1 = c1.atd(i); if (Double.isNaN(d1)) continue; long l1 = Double.doubleToRawLongBits(d1); // Atomically fetch/create nested NBHM NonBlockingHashMapLong<AtomicLong> col1s = _col0s.get(l0); if (col1s == null) { // Speed filter pre-filled entries col1s = new NonBlockingHashMapLong<>(); NonBlockingHashMapLong<AtomicLong> old = _col0s.putIfAbsent(l0, col1s); if (old != null) col1s = old; // Lost race, use old value } // Atomically fetch/create nested AtomicLong AtomicLong cnt = col1s.get(l1); if (cnt == null) { // Speed filter pre-filled entries cnt = new AtomicLong(); AtomicLong old = col1s.putIfAbsent(l1, cnt); if (old != null) cnt = old; // Lost race, use old value } // Atomically bump counter cnt.incrementAndGet(); } } @Override public void reduce(AstTable.SlowCnt sc) { if (_col0s == sc._col0s) return; throw water.H2O.unimpl(); } public final AutoBuffer write_impl(AutoBuffer ab) { if (_col0s == null) return ab.put8(0); ab.put8(_col0s.size()); for (long col0 : _col0s.keySetLong()) { ab.put8(col0); NonBlockingHashMapLong<AtomicLong> col1s = _col0s.get(col0); ab.put8(col1s.size()); for (long col1 : col1s.keySetLong()) { ab.put8(col1); ab.put8(col1s.get(col1).get()); } } return ab; } public final AstTable.SlowCnt read_impl(AutoBuffer ab) { long len0 = ab.get8(); if (len0 == 0) return this; _col0s = new NonBlockingHashMapLong<>(); for (long i = 0; i < len0; i++) { NonBlockingHashMapLong<AtomicLong> col1s = new NonBlockingHashMapLong<>(); _col0s.put(ab.get8(), col1s); long len1 = ab.get8(); for (long j = 0; j < len1; j++) col1s.put(ab.get8(), new AtomicLong(ab.get8())); } return this; } @Override public String toString() { StringBuilder sb = new StringBuilder(); for (NonBlockingHashMapLong.IteratorLong i = iter(_col0s); i.hasNext(); ) { long l = i.nextLong(); double d = Double.longBitsToDouble(l); sb.append(d).append(": {"); NonBlockingHashMapLong<AtomicLong> col1s = _col0s.get(l); for (NonBlockingHashMapLong.IteratorLong j = iter(col1s); j.hasNext(); ) { long l2 = j.nextLong(); double d2 = Double.longBitsToDouble(l2); AtomicLong al = col1s.get(l2); sb.append(d2).append(": ").append(al.get()).append(", "); } sb.append("}\n"); } return sb.toString(); } } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/advmath/AstTfIdf.java
package water.rapids.ast.prims.advmath; import hex.tfidf.DocumentFrequencyTask; import hex.tfidf.InverseDocumentFrequencyTask; import hex.tfidf.TermFrequencyTask; import hex.tfidf.TfIdfPreprocessorTask; import org.apache.log4j.Logger; import water.Key; import water.MRTask; import water.Scope; import water.fvec.*; import water.rapids.Env; import water.rapids.Merge; import water.rapids.Rapids; import water.rapids.Val; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import water.rapids.ast.prims.string.AstToLower; import water.rapids.vals.ValFrame; import water.util.ArrayUtils; /** * Primitive AST operation to compute TF-IDF values for given document corpus.<br> * * <br> * <b>Parameters:</b> * <p><ul> * <li><code>frame</code> - Input frame containing data for whose TF-IDF values should be computed * <li><code>doc_id_idx</code> - Index of a column containing document IDs * <li><code>text_idx</code> - Index of a column containing words/documents (depending on the <code>preprocess</code> parameter) * <li><code>preprocess</code> - Flag indicating whether input should be pre-processed or not * <li><code>case_sensitive</code> - Flag indicating whether input should be treated as a case sensitive data * </ul><p> * * <br> * <b>Content of a column with index <code>content_idx</code>:</b> * <p><ul> * <li>See {@link TfIdfPreprocessorTask} - (default) when pre-processing is enabled * <li><code>word</code> - when pre-processing is disabled * </ul><p> */ public class AstTfIdf extends AstPrimitive<AstTfIdf> { /** * Name to be used for a column containing Inverse Document Frequency values in the output frame of this operation. */ private static final String IDF_COL_NAME = "IDF"; /** * Name to be used for a column containing TF-IDF values in the output frame of this operation. */ private static final String TF_IDF_COL_NAME = "TF-IDF"; /** * Column names to be used for preprocessed frame. */ private static final String[] PREPROCESSED_FRAME_COL_NAMES = new String[] { "DocID", "Words" }; /** * Class logger. */ private static Logger log = Logger.getLogger(AstTfIdf.class); @Override public int nargs() { return 1 + 5; // (tf-idf input_frame_name doc_id_col_idx text_col_idx preprocess case_sensitive) } @Override public String[] args() { return new String[]{ "frame", "doc_id_idx", "text_idx", "preprocess", "case_sensitive"}; } @Override public Val apply(Env env, Env.StackHelp stk, AstRoot[] asts) { Frame inputFrame = stk.track(asts[1].exec(env).getFrame()); final int docIdIdx = (int) asts[2].exec(env).getNum(); final int contentIdx = (int) asts[3].exec(env).getNum(); final boolean preprocess = asts[4].exec(env).getBool(); final boolean caseSensitive = asts[5].exec(env).getBool(); if (inputFrame.anyVec().length() <= 0) throw new IllegalArgumentException("Empty input frame provided."); Scope.enter(); Frame tfIdfFrame = null; try { // Input checks int inputFrameColsCnt = inputFrame.numCols(); if (docIdIdx >= inputFrameColsCnt || contentIdx >= inputFrameColsCnt) throw new IllegalArgumentException("Provided column index is out of bounds. Number of columns in the input frame: " + inputFrameColsCnt); Vec docIdVec = inputFrame.vec(docIdIdx); Vec contentVec = inputFrame.vec(contentIdx); if (!docIdVec.isNumeric() || !contentVec.isString()) throw new IllegalArgumentException("Incorrect format of input frame." + "Following row format is expected: (numeric) documentID, (string) " + (preprocess ? "documentContent." : "words. " + "Got "+docIdVec.get_type_str() + " and " + contentVec.get_type_str() +" instead.")); // Case sensitivity if (!caseSensitive) { Scope.track(inputFrame.replace(contentIdx, AstToLower.toLowerStringCol(inputFrame.vec(contentIdx)))); } // Pre-processing Frame wordFrame; long documentsCnt; if (preprocess) { byte[] outputTypes = new byte[]{ Vec.T_NUM, Vec.T_STR }; wordFrame = new TfIdfPreprocessorTask(docIdIdx, contentIdx).doAll(outputTypes, inputFrame) .outputFrame(PREPROCESSED_FRAME_COL_NAMES, null); documentsCnt = inputFrame.numRows(); } else { String[] columnsNames = ArrayUtils.select(inputFrame.names(), new int[]{ docIdIdx, contentIdx }); wordFrame = inputFrame.subframe(columnsNames); String countDocumentsRapid = "(unique (cols " + asts[1].toString() + " [" + docIdIdx + "]) false)"; documentsCnt = Rapids.exec(countDocumentsRapid).getFrame().anyVec().length(); } Scope.track(wordFrame); // TF Frame tfOutFrame = TermFrequencyTask.compute(wordFrame); Scope.track(tfOutFrame); // DF Frame dfOutFrame = DocumentFrequencyTask.compute(tfOutFrame); Scope.track(dfOutFrame); // IDF InverseDocumentFrequencyTask idf = new InverseDocumentFrequencyTask(documentsCnt); Vec idfValues = idf.doAll(new byte[]{ Vec.T_NUM }, dfOutFrame.lastVec()).outputFrame().anyVec(); Scope.track(idfValues); // Replace DF column with IDF column Vec removedCol = dfOutFrame.remove(dfOutFrame.numCols() - 1); Scope.track(removedCol); dfOutFrame.add(IDF_COL_NAME, idfValues); // Intermediate frame containing both TF and IDF values Scope.track(tfOutFrame.replace(1, tfOutFrame.vecs()[1].toCategoricalVec())); Scope.track(dfOutFrame.replace(0, dfOutFrame.vecs()[0].toCategoricalVec())); int[][] levelMaps = { CategoricalWrappedVec.computeMap(tfOutFrame.vec(1).domain(), dfOutFrame.vec(0).domain()) }; Frame tfIdfIntermediate = Merge.merge(tfOutFrame, dfOutFrame, new int[]{1}, new int[]{0}, false, levelMaps); Scope.track(tfIdfIntermediate.replace(1, tfIdfIntermediate.vecs()[1].toStringVec())); // TF-IDF int tfOutFrameColCnt = tfIdfIntermediate.numCols(); TfIdfTask tfIdfTask = new TfIdfTask(tfOutFrameColCnt - 2, tfOutFrameColCnt - 1); Vec tfIdfValues = tfIdfTask.doAll(new byte[]{Vec.T_NUM}, tfIdfIntermediate).outputFrame().anyVec(); Scope.track(tfIdfValues); // Construct final frame containing TF, IDF and TF-IDF values tfIdfIntermediate.add(TF_IDF_COL_NAME, tfIdfValues); tfIdfIntermediate._key = Key.make(); if (log.isDebugEnabled()) log.debug(tfIdfIntermediate.toTwoDimTable().toString()); tfIdfFrame = tfIdfIntermediate; } finally { Key[] keysToKeep = tfIdfFrame != null ? tfIdfFrame.keys() : new Key[]{}; Scope.exit(keysToKeep); } return new ValFrame(tfIdfFrame); } @Override public String str() { return "tf-idf"; } /** * Final TF-IDF Map-Reduce task used to combine TF and IDF values together. */ private static class TfIdfTask extends MRTask<TfIdfTask> { // IN /** * Index of a column containing Term Frequency values in the input frame of this task. */ private final int _tfColIndex; /** * Index of a column containing Inverse Document Frequency values in the input frame of this task. */ private final int _idfColIndex; private TfIdfTask(int tfColIndex, int idfColIndex) { _tfColIndex = tfColIndex; _idfColIndex = idfColIndex; } @Override public void map(Chunk[] cs, NewChunk nc) { Chunk tfValues = cs[_tfColIndex]; Chunk idfValues = cs[_idfColIndex]; for (int row = 0; row < tfValues._len; row++) { nc.addNum(tfValues.at8(row) * idfValues.atd(row)); } } } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/advmath/AstUnique.java
package water.rapids.ast.prims.advmath; import water.DKV; import water.H2O; import water.fvec.Frame; import water.fvec.Vec; import water.fvec.task.UniqOldTask; import water.fvec.task.UniqTask; import water.rapids.Env; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import water.rapids.vals.ValFrame; import water.util.Log; import water.util.VecUtils; public class AstUnique extends AstPrimitive { @Override public String[] args() { return new String[]{"ary"}; } @Override public int nargs() { return 2 + 1; } // (unique col) @Override public String str() { return "unique"; } @Override public ValFrame apply(Env env, Env.StackHelp stk, AstRoot asts[]) { final Frame fr = stk.track(asts[1].exec(env)).getFrame(); final boolean includeNAs = asts[2].exec(env).getBool(); return new ValFrame(uniqueValuesBy(fr,0, includeNAs)); } /** return a frame with unique values from the specified column */ public static Frame uniqueValuesBy(final Frame fr, final int columnIndex, final boolean includeNAs) { final Vec vec0 = fr.vec(columnIndex); final Vec v; if (vec0.isCategorical()) { // Vector domain might contain levels not actually present in the vector - collection of actual values is required. final String[] actualVecDomain = VecUtils.collectDomainFast(vec0); final boolean contributeNAs = vec0.naCnt() > 0 && includeNAs; final long uniqueVecLength = contributeNAs ? actualVecDomain.length + 1 : actualVecDomain.length; v = Vec.makeSeq(0, uniqueVecLength, true); if(contributeNAs) { v.setNA(uniqueVecLength - 1); } v.setDomain(actualVecDomain); DKV.put(v); } else { long start = System.currentTimeMillis(); String uniqImpl = H2O.getSysProperty("rapids.unique.impl", "IcedDouble"); switch (uniqImpl) { case "IcedDouble": v = new UniqTask().doAll(vec0).toVec(); break; case "GroupBy": v = new UniqOldTask().doAll(vec0).toVec(); break; default: throw new UnsupportedOperationException("Unknown unique implementation: " + uniqImpl); } Log.info("Unique on a numerical Vec (len=" + vec0.length() + ") took " + (System.currentTimeMillis() - start) + "ms and returned " + v.length() + " unique values (impl: " + uniqImpl + ")."); } return new Frame(v); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/advmath/AstVariance.java
package water.rapids.ast.prims.advmath; import water.Key; import water.MRTask; import water.fvec.Chunk; import water.fvec.Frame; import water.fvec.Vec; import water.rapids.Env; import water.rapids.Val; import water.rapids.vals.ValFrame; import water.rapids.vals.ValNum; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import water.util.ArrayUtils; import java.util.Arrays; /** * Variance between columns of a frame */ public class AstVariance extends AstPrimitive { @Override public String[] args() { return new String[]{"ary", "x", "y", "use", "symmetric"}; } private enum Mode {Everything, AllObs, CompleteObs} @Override public int nargs() { return 1 + 4; /* (var X Y use symmetric) */ } @Override public String str() { return "var"; } @Override public Val apply(Env env, Env.StackHelp stk, AstRoot asts[]) { Frame frx = stk.track(asts[1].exec(env)).getFrame(); Frame fry = stk.track(asts[2].exec(env)).getFrame(); if (frx.numRows() != fry.numRows()) throw new IllegalArgumentException("Frames must have the same number of rows, found " + frx.numRows() + " and " + fry.numRows()); String use = stk.track(asts[3].exec(env)).getStr(); boolean symmetric = asts[4].exec(env).getNum() == 1; Mode mode; switch (use) { case "everything": mode = Mode.Everything; break; case "all.obs": mode = Mode.AllObs; break; case "complete.obs": mode = Mode.CompleteObs; break; default: throw new IllegalArgumentException("unknown use mode: " + use); } return fry.numRows() == 1 ? scalar(frx, fry, mode) : array(frx, fry, mode, symmetric); } // Scalar covariance for 1 row private ValNum scalar(Frame frx, Frame fry, Mode mode) { if (frx.numCols() != fry.numCols()) throw new IllegalArgumentException("Single rows must have the same number of columns, found " + frx.numCols() + " and " + fry.numCols()); Vec vecxs[] = frx.vecs(); Vec vecys[] = fry.vecs(); double xmean = 0, ymean = 0, ncols = frx.numCols(), NACount = 0, xval, yval, ss = 0; for (int r = 0; r < ncols; r++) { xval = vecxs[r].at(0); yval = vecys[r].at(0); if (Double.isNaN(xval) || Double.isNaN(yval)) NACount++; else { xmean += xval; ymean += yval; } } xmean /= (ncols - NACount); ymean /= (ncols - NACount); if (NACount != 0) { if (mode.equals(Mode.AllObs)) throw new IllegalArgumentException("Mode is 'all.obs' but NAs are present"); if (mode.equals(Mode.Everything)) return new ValNum(Double.NaN); } for (int r = 0; r < ncols; r++) { xval = vecxs[r].at(0); yval = vecys[r].at(0); if (!(Double.isNaN(xval) || Double.isNaN(yval))) ss += (vecxs[r].at(0) - xmean) * (vecys[r].at(0) - ymean); } return new ValNum(ss / (ncols - NACount - 1)); } // Matrix covariance. Compute covariance between all columns from each Frame // against each other. Return a matrix of covariances which is frx.numCols // wide and fry.numCols tall. private Val array(Frame frx, Frame fry, Mode mode, boolean symmetric) { Vec[] vecxs = frx.vecs(); int ncolx = vecxs.length; Vec[] vecys = fry.vecs(); int ncoly = vecys.length; if (mode.equals(Mode.Everything) || mode.equals(Mode.AllObs)) { if (mode.equals(Mode.AllObs)) { for (Vec v : vecxs) if (v.naCnt() != 0) throw new IllegalArgumentException("Mode is 'all.obs' but NAs are present"); if (!symmetric) for (Vec v : vecys) if (v.naCnt() != 0) throw new IllegalArgumentException("Mode is 'all.obs' but NAs are present"); } CoVarTaskEverything[] cvs = new CoVarTaskEverything[ncoly]; double[] xmeans = new double[ncolx]; for (int x = 0; x < ncoly; x++) xmeans[x] = vecxs[x].mean(); if (symmetric) { //1-col returns scalar if (ncoly == 1) return new ValNum(vecys[0].naCnt() == 0 ? vecys[0].sigma() * vecys[0].sigma() : Double.NaN); int[] idx = new int[ncoly]; for (int y = 1; y < ncoly; y++) idx[y] = y; int[] first_index = new int[]{0}; //compute covariances between column_i and column_i+1, column_i+2, ... Frame reduced_fr; for (int y = 0; y < ncoly - 1; y++) { idx = ArrayUtils.removeIds(idx, first_index); reduced_fr = new Frame(frx.vecs(idx)); cvs[y] = new CoVarTaskEverything(vecys[y].mean(), xmeans).dfork(new Frame(vecys[y]).add(reduced_fr)); } double[][] res_array = new double[ncoly][ncoly]; //fill in the diagonals (variances) using sigma from rollupstats for (int y = 0; y < ncoly; y++) res_array[y][y] = vecys[y].naCnt() == 0 ? vecys[y].sigma() * vecys[y].sigma() : Double.NaN; //arrange the results into the bottom left of res_array. each successive cvs is 1 smaller in length for (int y = 0; y < ncoly - 1; y++) System.arraycopy(ArrayUtils.div(cvs[y].getResult()._covs, (fry.numRows() - 1)), 0, res_array[y], y + 1, ncoly - y - 1); //copy over the bottom left of res_array to its top right for (int y = 0; y < ncoly - 1; y++) { for (int x = y + 1; x < ncoly; x++) { res_array[x][y] = res_array[y][x]; } } //set Frame Vec[] res = new Vec[ncoly]; Key<Vec>[] keys = Vec.VectorGroup.VG_LEN1.addVecs(ncoly); for (int y = 0; y < ncoly; y++) { res[y] = Vec.makeVec(res_array[y], keys[y]); } return new ValFrame(new Frame(fry._names, res)); } // Launch tasks; each does all Xs vs one Y for (int y = 0; y < ncoly; y++) cvs[y] = new CoVarTaskEverything(vecys[y].mean(), xmeans).dfork(new Frame(vecys[y]).add(frx)); // 1-col returns scalar if (ncolx == 1 && ncoly == 1) { return new ValNum(cvs[0].getResult()._covs[0] / (fry.numRows() - 1)); } // Gather all the Xs-vs-Y covariance arrays; divide by rows Vec[] res = new Vec[ncoly]; Key<Vec>[] keys = Vec.VectorGroup.VG_LEN1.addVecs(ncoly); for (int y = 0; y < ncoly; y++) res[y] = Vec.makeVec(ArrayUtils.div(cvs[y].getResult()._covs, (fry.numRows() - 1)), keys[y]); return new ValFrame(new Frame(fry._names, res)); } else { //if (mode.equals(Mode.CompleteObs)) { //two-pass algorithm for computation of variance for numerical stability if (symmetric) { if (ncoly == 1) return new ValNum(vecys[0].sigma() * vecys[0].sigma()); CoVarTaskCompleteObsMeanSym taskCompleteObsMeanSym = new CoVarTaskCompleteObsMeanSym().doAll(fry); long NACount = taskCompleteObsMeanSym._NACount; double[] ymeans = ArrayUtils.div(taskCompleteObsMeanSym._ysum, fry.numRows() - NACount); // 1 task with all Ys CoVarTaskCompleteObsSym cvs = new CoVarTaskCompleteObsSym(ymeans).doAll(new Frame(fry)); double[][] res_array = new double[ncoly][ncoly]; for (int y = 0; y < ncoly; y++) { System.arraycopy(ArrayUtils.div(cvs._covs[y], (fry.numRows() - 1 - NACount)), y, res_array[y], y, ncoly - y); } //copy over the bottom left of res_array to its top right for (int y = 0; y < ncoly - 1; y++) { for (int x = y + 1; x < ncoly; x++) { res_array[x][y] = res_array[y][x]; } } //set Frame Vec[] res = new Vec[ncoly]; Key<Vec>[] keys = Vec.VectorGroup.VG_LEN1.addVecs(ncoly); for (int y = 0; y < ncoly; y++) { res[y] = Vec.makeVec(res_array[y], keys[y]); } return new ValFrame(new Frame(fry._names, res)); } CoVarTaskCompleteObsMean taskCompleteObsMean = new CoVarTaskCompleteObsMean(ncoly, ncolx).doAll(new Frame(fry).add(frx)); long NACount = taskCompleteObsMean._NACount; double[] ymeans = ArrayUtils.div(taskCompleteObsMean._ysum, fry.numRows() - NACount); double[] xmeans = ArrayUtils.div(taskCompleteObsMean._xsum, fry.numRows() - NACount); // 1 task with all Xs and Ys CoVarTaskCompleteObs cvs = new CoVarTaskCompleteObs(ymeans, xmeans).doAll(new Frame(fry).add(frx)); // 1-col returns scalar if (ncolx == 1 && ncoly == 1) { return new ValNum(cvs._covs[0][0] / (fry.numRows() - 1 - NACount)); } // Gather all the Xs-vs-Y covariance arrays; divide by rows Vec[] res = new Vec[ncoly]; Key<Vec>[] keys = Vec.VectorGroup.VG_LEN1.addVecs(ncoly); for (int y = 0; y < ncoly; y++) res[y] = Vec.makeVec(ArrayUtils.div(cvs._covs[y], (fry.numRows() - 1 - NACount)), keys[y]); return new ValFrame(new Frame(fry._names, res)); } } private static class CoVarTaskEverything extends MRTask<CoVarTaskEverything> { double[] _covs; final double _xmeans[], _ymean; CoVarTaskEverything(double ymean, double[] xmeans) { _ymean = ymean; _xmeans = xmeans; } @Override public void map(Chunk cs[]) { final int ncolsx = cs.length - 1; final Chunk cy = cs[0]; final int len = cy._len; _covs = new double[ncolsx]; double sum; for (int x = 0; x < ncolsx; x++) { sum = 0; final Chunk cx = cs[x + 1]; final double xmean = _xmeans[x]; for (int row = 0; row < len; row++) sum += (cx.atd(row) - xmean) * (cy.atd(row) - _ymean); _covs[x] = sum; } } @Override public void reduce(CoVarTaskEverything cvt) { ArrayUtils.add(_covs, cvt._covs); } } private static class CoVarTaskCompleteObsMean extends MRTask<CoVarTaskCompleteObsMean> { double[] _xsum, _ysum; long _NACount; int _ncolx, _ncoly; CoVarTaskCompleteObsMean(int ncoly, int ncolx) { _ncolx = ncolx; _ncoly = ncoly; } @Override public void map(Chunk cs[]) { _xsum = new double[_ncolx]; _ysum = new double[_ncoly]; double[] xvals = new double[_ncolx]; double[] yvals = new double[_ncoly]; double xval, yval; boolean add; int len = cs[0]._len; for (int row = 0; row < len; row++) { add = true; //reset existing arrays to 0 rather than initializing new ones to save on garbage collection Arrays.fill(xvals, 0); Arrays.fill(yvals, 0); for (int y = 0; y < _ncoly; y++) { final Chunk cy = cs[y]; yval = cy.atd(row); //if any yval along a row is NA, discard the entire row if (Double.isNaN(yval)) { _NACount++; add = false; break; } yvals[y] = yval; } if (add) { for (int x = 0; x < _ncolx; x++) { final Chunk cx = cs[x + _ncoly]; xval = cx.atd(row); //if any xval along a row is NA, discard the entire row if (Double.isNaN(xval)) { _NACount++; add = false; break; } xvals[x] = xval; } } //add is true iff row has been traversed and found no NAs among yvals and xvals if (add) { ArrayUtils.add(_xsum, xvals); ArrayUtils.add(_ysum, yvals); } } } @Override public void reduce(CoVarTaskCompleteObsMean cvt) { ArrayUtils.add(_xsum, cvt._xsum); ArrayUtils.add(_ysum, cvt._ysum); _NACount += cvt._NACount; } } private static class CoVarTaskCompleteObs extends MRTask<CoVarTaskCompleteObs> { double[][] _covs; final double _xmeans[], _ymeans[]; CoVarTaskCompleteObs(double[] ymeans, double[] xmeans) { _ymeans = ymeans; _xmeans = xmeans; } @Override public void map(Chunk cs[]) { int ncolx = _xmeans.length; int ncoly = _ymeans.length; double[] xvals = new double[ncolx]; double[] yvals = new double[ncoly]; _covs = new double[ncoly][ncolx]; double[] _covs_y; double xval, yval, ymean; boolean add; int len = cs[0]._len; for (int row = 0; row < len; row++) { add = true; //reset existing arrays to 0 rather than initializing new ones to save on garbage collection Arrays.fill(xvals, 0); Arrays.fill(yvals, 0); for (int y = 0; y < ncoly; y++) { final Chunk cy = cs[y]; yval = cy.atd(row); //if any yval along a row is NA, discard the entire row if (Double.isNaN(yval)) { add = false; break; } yvals[y] = yval; } if (add) { for (int x = 0; x < ncolx; x++) { final Chunk cx = cs[x + ncoly]; xval = cx.atd(row); //if any xval along a row is NA, discard the entire row if (Double.isNaN(xval)) { add = false; break; } xvals[x] = xval; } } //add is true iff row has been traversed and found no NAs among yvals and xvals if (add) { for (int y = 0; y < ncoly; y++) { _covs_y = _covs[y]; yval = yvals[y]; ymean = _ymeans[y]; for (int x = 0; x < ncolx; x++) _covs_y[x] += (xvals[x] - _xmeans[x]) * (yval - ymean); } } } } @Override public void reduce(CoVarTaskCompleteObs cvt) { ArrayUtils.add(_covs, cvt._covs); } } private static class CoVarTaskCompleteObsMeanSym extends MRTask<CoVarTaskCompleteObsMeanSym> { double[] _ysum; long _NACount; @Override public void map(Chunk cs[]) { int ncoly = cs.length; _ysum = new double[ncoly]; double[] yvals = new double[ncoly]; double yval; boolean add; int len = cs[0]._len; for (int row = 0; row < len; row++) { add = true; Arrays.fill(yvals, 0); for (int y = 0; y < ncoly; y++) { final Chunk cy = cs[y]; yval = cy.atd(row); //if any yval along a row is NA, discard the entire row if (Double.isNaN(yval)) { _NACount++; add = false; break; } yvals[y] = yval; } if (add) { ArrayUtils.add(_ysum, yvals); } } } @Override public void reduce(CoVarTaskCompleteObsMeanSym cvt) { ArrayUtils.add(_ysum, cvt._ysum); _NACount += cvt._NACount; } } private static class CoVarTaskCompleteObsSym extends MRTask<CoVarTaskCompleteObsSym> { double[][] _covs; final double _ymeans[]; CoVarTaskCompleteObsSym(double[] ymeans) { _ymeans = ymeans; } @Override public void map(Chunk cs[]) { int ncoly = _ymeans.length; double[] yvals = new double[ncoly]; _covs = new double[ncoly][ncoly]; double[] _covs_y; double yval, ymean; boolean add; int len = cs[0]._len; for (int row = 0; row < len; row++) { add = true; //reset existing arrays to 0 rather than initializing new ones to save on garbage collection Arrays.fill(yvals, 0); for (int y = 0; y < ncoly; y++) { final Chunk cy = cs[y]; yval = cy.atd(row); //if any yval along a row is NA, discard the entire row if (Double.isNaN(yval)) { add = false; break; } yvals[y] = yval; } //add is true iff row has been traversed and found no NAs among yvals if (add) { for (int y = 0; y < ncoly; y++) { _covs_y = _covs[y]; yval = yvals[y]; ymean = _ymeans[y]; for (int x = y; x < ncoly; x++) _covs_y[x] += (yvals[x] - _ymeans[x]) * (yval - ymean); } } } } @Override public void reduce(CoVarTaskCompleteObsSym cvt) { ArrayUtils.add(_covs, cvt._covs); } } public static double getVar(Vec v) { return v.naCnt() == 0 ? v.sigma() * v.sigma() : Double.NaN; } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/advmath/SpearmanCorrelation.java
package water.rapids.ast.prims.advmath; import water.Key; import water.MRTask; import water.Scope; import water.fvec.Chunk; import water.fvec.Frame; import water.fvec.Vec; import water.rapids.Merge; import water.util.FrameUtils; import java.math.BigDecimal; import java.math.MathContext; import java.util.Objects; public class SpearmanCorrelation { public static Frame calculate(final Frame frameX, Frame frameY, final AstCorrelation.Mode mode) { Objects.requireNonNull(frameX); Objects.requireNonNull(frameY); checkCorrelationDoable(frameX, frameY, mode); final Frame correlationMatrix = createCorrelationMatrix(frameX, frameY); // If the two frame contain the same vectors (key-wise), then the diagonal of the correlation matrix can be automatically filled with 1s. // Unless there mode is "Everything", which enforces NaN correlation values if there is a NaN observation. final boolean framesAreEqual = !AstCorrelation.Mode.Everything.equals(mode) && framesContainSameVecs(frameX, frameY); for (int vecIdX = 0; vecIdX < frameX.numCols(); vecIdX++) { for (int vecIdY = 0; vecIdY < frameY.numCols(); vecIdY++) { Scope.enter(); try { if (framesAreEqual && vecIdX == vecIdY) { // If the correlation is calculated within frame frame, comparing the same vecs always resultings in 1.0 // correlation coefficient. Therefore, there is no need to calculate it. correlationMatrix.vec(vecIdX) .set(vecIdY, 1d); } else if (isNaNCorrelation(frameX.vec(vecIdX), frameY.vec(vecIdY), mode)) { correlationMatrix.vec(vecIdX) .set(vecIdY, Double.NaN); } else { // Actual SCC calculation final SpearmanRankedVectors rankedVectors = rankedVectors(frameX, frameY, vecIdX, vecIdY, mode); // Means must be calculated separately - those are not calculated for categorical columns in rollup stats. final double[] means = calculateMeans(rankedVectors._x, rankedVectors._y); final SpearmanCorrelationCoefficientTask spearman = new SpearmanCorrelationCoefficientTask(means[0], means[1]) .doAll(rankedVectors._x, rankedVectors._y); correlationMatrix.vec(vecIdX) .set(vecIdY, spearman.getSpearmanCorrelationCoefficient()); } } finally { Scope.exit(); } } } return correlationMatrix; } /** * Compares two frames for their vecs being the same key-wise. * * @param frameX An instance of {@link Frame} to compare * @param frameY A second instance of {@link Frame} to compare * @return True if and only if the frames both contain the same vectors in the same order and quantity. Otherwise false. */ private static boolean framesContainSameVecs(final Frame frameX, final Frame frameY) { final Vec[] vecsX = frameX.vecs(); final Vec[] vecsY = frameY.vecs(); if (vecsX.length != vecsY.length) return false; for (int i = 0; i < vecsX.length; i++) { if (!vecsX[i]._key.equals(vecsY[i]._key)) return false; } return true; } /** * @param frameX Frame X candiate for SCC calculation * @param frameY Frame Y candidate for SCC calculation * @param mode An instance of {@link AstCorrelation.Mode} * @return False if AstCorrelation.Mode is set to AllObs and any of the vectors in compared frames contains NaNs, otherwise True. * @throws IllegalArgumentException When the {@link AstCorrelation.Mode} is set to AllObs and any of the vectors contains NaN * or when any of the frames is empty */ private static void checkCorrelationDoable(final Frame frameX, final Frame frameY, final AstCorrelation.Mode mode) throws IllegalArgumentException { if (!AstCorrelation.Mode.AllObs.equals(mode)) return; // Those frames are not guaranteed to have a key - Frames generated by rapids are keyless if (frameX.numCols() == 0) throw new IllegalArgumentException("First given frame for Spearman calculation has no columnns."); if (frameY.numCols() == 0) throw new IllegalArgumentException("Second given frame for Spearman calculation has no columnns."); final Vec[] vecsX = frameX.vecs(); final Vec[] vecsY = frameY.vecs(); for (int i = 0; i < vecsX.length; i++) { if (vecsX[i].naCnt() != 0 || vecsY[i].naCnt() != 0) { throw new IllegalArgumentException("Mode is 'AllObs' but NAs are present"); } } } /** * @param vecX Vec X candiate for SCC calculation * @param vecY Vec Y candidate for SCC calculation * @param mode An instance of {@link AstCorrelation.Mode} * @return True if AstCorrelation.Mode is set to EVERYTHING and any of the vectors compared contains NaNs, otherwise False. */ private static boolean isNaNCorrelation(final Vec vecX, final Vec vecY, final AstCorrelation.Mode mode) { return AstCorrelation.Mode.Everything.equals(mode) && (vecX.naCnt() > 0 || vecY.naCnt() > 0); } private static Frame createCorrelationMatrix(final Frame frameX, final Frame frameY) { final Vec[] correlationVecs = new Vec[frameX.numCols()]; final int height = frameY.numCols(); for (int width = 0; width < frameX.numCols(); width++) { correlationVecs[width] = Vec.makeCon(Double.NaN, height); } return new Frame(Key.make(), correlationVecs); } /** * Sorts and ranks the vectors of which SCC is calculated. Original Frame is not modified. * * @param frameX Original frame containing the vectors compared. * @param vecIdX First compared vector * @param vecIdY Second compared vector * @return An instance of {@link SpearmanRankedVectors}, holding two new vectors with row rank. */ private static SpearmanRankedVectors rankedVectors(final Frame frameX, final Frame frameY, final int vecIdX, final int vecIdY, final AstCorrelation.Mode mode) { Frame comparedVecsWithNas = new Frame(frameX.vec(vecIdX).makeCopy(), frameY.vec(vecIdY).makeCopy()); Frame unsortedVecs; if (AstCorrelation.Mode.CompleteObs.equals(mode)) { unsortedVecs = comparedVecsWithNas; } else { unsortedVecs = new Merge.RemoveNAsTask(0, 1) .doAll(comparedVecsWithNas.types(), comparedVecsWithNas) .outputFrame(comparedVecsWithNas.names(), comparedVecsWithNas.domains()); } Frame sortedX = new Frame(unsortedVecs.vec(0).makeCopy()); Scope.track(sortedX); Frame sortedY = new Frame(unsortedVecs.vec(1).makeCopy()); Scope.track(sortedY); final boolean xIsOrdered = needsOrdering(sortedX.vec(0)); final boolean yIsOrdered = needsOrdering(sortedY.vec(0)); if (xIsOrdered) { FrameUtils.labelRows(sortedX, "label"); sortedX = sortedX.sort(new int[]{0}); Scope.track(sortedX); } if (yIsOrdered) { FrameUtils.labelRows(sortedY, "label"); sortedY = sortedY.sort(new int[]{0}); Scope.track(sortedY); } assert sortedX.numRows() == sortedY.numRows(); final Vec orderX = needsOrdering(sortedX.vec(0)) ? Vec.makeZero(sortedX.numRows()) : frameX.vec(vecIdX); final Vec orderY = needsOrdering(sortedY.vec(0)) ? Vec.makeZero(sortedY.numRows()) : frameY.vec(vecIdY); final Vec xLabel = sortedX.vec("label") == null ? sortedX.vec(0) : sortedX.vec("label"); final Vec xValue = sortedX.vec(0); final Vec yLabel = sortedY.vec("label") == null ? sortedY.vec(0) : sortedY.vec("label"); final Vec yValue = sortedY.vec(0); Scope.track(xLabel); Scope.track(yLabel); final Vec.Writer orderXWriter = orderX.open(); final Vec.Writer orderYWriter = orderY.open(); final Vec.Reader xValueReader = xValue.new Reader(); final Vec.Reader yValueReader = yValue.new Reader(); final Vec.Reader xLabelReader = xLabel.new Reader(); final Vec.Reader yLabelReader = yLabel.new Reader(); // Put the actual rank into the vectors with ranks. Ensure equal values share the same rank. double lastX = Double.NaN; double lastY = Double.NaN; long skippedX = 0; long skippedY = 0; for (int i = 0; i < orderX.length(); i++) { if (xIsOrdered) { if (lastX == xValueReader.at(i)) { skippedX++; } else { skippedX = 0; } lastX = xValueReader.at(i); orderXWriter.set(xLabelReader.at8(i) - 1, i - skippedX); } if (yIsOrdered) { if (lastY == yValueReader.at(i)) { skippedY++; } else { skippedY = 0; } lastY = yValueReader.at(i); orderYWriter.set(yLabelReader.at8(i) - 1, i - skippedY); } } orderXWriter.close(); orderYWriter.close(); // Ensure chunk layout is the same by adding the vectors into a new Frame. final Frame sameChunkLayoutFrame = new Frame(new String[]{"X"}, new Vec[]{orderX}); sameChunkLayoutFrame.add("Y", orderY); // Return the vectors with ranks - the ones with same chunk layout from inside the frame return new SpearmanRankedVectors(sameChunkLayoutFrame.vec("X"), sameChunkLayoutFrame.vec("Y")); } /** * Ranked vectors prepared to calculate Spearman's correlation coefficient */ private static class SpearmanRankedVectors { private final Vec _x; private final Vec _y; public SpearmanRankedVectors(Vec x, Vec y) { this._x = x; this._y = y; } } private static boolean needsOrdering(final Vec vec) { return !vec.isCategorical(); } /** * A task to do calculate Spearman's correlation coefficient. Not using the "approximation equation", but the * fully-fledged equation resistant against noise from repeated values. * The intermediate calculations required for standard deviation of both columns could be calculated by existing code, * however the point is to perform the calculations by going through the data only once. * * @see {@link water.rapids.ast.prims.advmath.AstVariance} */ private static class SpearmanCorrelationCoefficientTask extends MRTask<SpearmanCorrelationCoefficientTask> { // Arguments obtained externally private final double _xMean; private final double _yMean; private double spearmanCorrelationCoefficient; // Required to later finish calculation of standard deviation private double _xDiffSquared = 0; private double _yDiffSquared = 0; private double _xyMul = 0; // If at least one of the vectors contains NaN, such line is skipped private long _linesVisited; /** * @param xMean Mean value of the first 'x' vector, with NaNs skipped * @param yMean Mean value of the second 'y' vector, with NaNs skipped */ private SpearmanCorrelationCoefficientTask(final double xMean, final double yMean) { this._xMean = xMean; this._yMean = yMean; } @Override public void map(Chunk[] chunks) { assert chunks.length == 2; // Amount of linear correlation only calculated between two vectors at once final Chunk xChunk = chunks[0]; final Chunk yChunk = chunks[1]; for (int row = 0; row < chunks[0].len(); row++) { final double x = xChunk.atd(row); final double y = yChunk.atd(row); _linesVisited++; _xyMul += x * y; final double xDiffFromMean = x - _xMean; final double yDiffFromMean = y - _yMean; _xDiffSquared += Math.pow(xDiffFromMean, 2); _yDiffSquared += Math.pow(yDiffFromMean, 2); } } @Override public void reduce(final SpearmanCorrelationCoefficientTask mrt) { // The intermediate results are addable. The final calculations are done afterwards. this._xDiffSquared += mrt._xDiffSquared; this._yDiffSquared += mrt._yDiffSquared; this._linesVisited += mrt._linesVisited; this._xyMul += mrt._xyMul; } @Override protected void postGlobal() { final double xStdDev = Math.sqrt(_xDiffSquared / _linesVisited); final double yStdDev = Math.sqrt(_yDiffSquared / _linesVisited); spearmanCorrelationCoefficient = (_xyMul - (_linesVisited * _xMean * _yMean)) / (_linesVisited * xStdDev * yStdDev); } public double getSpearmanCorrelationCoefficient() { return spearmanCorrelationCoefficient; } } /** * Calculates means of given numerical Vectors. Provided there is a NaN on a row in any of the give vectors, * the row is skipped and involved in the mean calculation. * * @param vecs An array of {@link Vec}, must not be empty, nor null. All vectors must be of same length. * @return An array of doubles with means for given vectors in the order they were given as arguments. * @throws IllegalArgumentException Zero vectors provided, */ private static double[] calculateMeans(final Vec... vecs) throws IllegalArgumentException { if (vecs.length < 1) { throw new IllegalArgumentException("There are no vectors to calculate means from."); } final long referenceVectorLength = vecs[0].length(); for (int i = 0; i < vecs.length; i++) { if (!vecs[i].isCategorical() && !vecs[i].isNumeric()) { throw new IllegalArgumentException(String.format("Given vector '%s' is not numerical or categorical.", vecs[i]._key.toString())); } if (referenceVectorLength != vecs[i].length()) { throw new IllegalArgumentException("Vectors to calculate means from do not have the same length." + String.format(" Vector '%s' is of length '%d'", vecs[i]._key.toString(), vecs[i].length())); } } return new MeanTask() .doAll(vecs)._means; } /** * Calculates means of given numerical Vectors. Provided there is a NaN on a row in any of the give vectors, * the row is skipped and involved in the mean calculation. */ private static class MeanTask extends MRTask<MeanTask> { private double[] _means; private long _linesVisited = 0; @Override public void map(Chunk[] cs) { // Sums might get big, BigDecimal ensures local accuracy and no overflow final BigDecimal[] averages = new BigDecimal[cs.length]; for (int i = 0; i < averages.length; i++) { averages[i] = new BigDecimal(0, MathContext.DECIMAL128); } row: for (int row = 0; row < cs[0].len(); row++) { final double[] values = new double[cs.length]; for (int col = 0; col < cs.length; col++) { values[col] = cs[col].atd(row); if (Double.isNaN(values[col])) break row; // If a NaN is detected in any of the columns, just skip the row } _linesVisited++; for (int i = 0; i < values.length; i++) { averages[i] = averages[i].add(new BigDecimal(values[i], MathContext.DECIMAL128), MathContext.DECIMAL128); } } this._means = new double[cs.length]; for (int i = 0; i < averages.length; i++) { this._means[i] = averages[i].divide(new BigDecimal(_linesVisited), MathContext.DECIMAL64).doubleValue(); } } @Override public void reduce(MeanTask mrt) { final int numChunks = _means.length; for (int i = 0; i < numChunks; i++) { _means[i] = (_means[i] * _linesVisited + mrt._means[i] * mrt._linesVisited) / (_linesVisited + mrt._linesVisited); } _linesVisited += mrt._linesVisited; } } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/assign/AstAppend.java
package water.rapids.ast.prims.assign; import water.fvec.Frame; import water.fvec.Vec; import water.rapids.Env; import water.rapids.Val; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import water.rapids.vals.ValFrame; /** * Attach a named column(s) to a destination frame. * * Syntax: destinationFrame (sourceFrame columnName)+ */ public class AstAppend extends AstPrimitive { @Override public String[] args() { return new String[]{"dst (src colName)+"}; } @Override public int nargs() { return -1; } // (append dst src "colName") @Override public String str() { return "append"; } @Override public ValFrame apply(Env env, Env.StackHelp stk, AstRoot asts[]) { assert asts.length >= 1 /* append */ + 3 /* args */: "Append needs at least 3 parameters"; assert (asts.length & 1) == 0 : "Wrong number of parameters"; Frame dst = stk.track(asts[1].exec(env)).getFrame(); dst = new Frame(dst._names.clone(), dst.vecs().clone()); for (int i = 2; i < asts.length; i+=2) { Val vsrc = stk.track(asts[i].exec(env)); String newColName = asts[i+1].exec(env).getStr(); Vec vec = dst.anyVec(); switch (vsrc.type()) { case Val.NUM: vec = vec.makeCon(vsrc.getNum()); break; case Val.STR: vec = vec.makeCon(vsrc.getStr()); break; case Val.FRM: if (vsrc.getFrame().numCols() != 1) throw new IllegalArgumentException("Can only append one column"); vec = vsrc.getFrame().anyVec(); break; default: throw new IllegalArgumentException( "Source must be a Frame or Number, but found a " + vsrc.getClass()); } dst.add(newColName, vec); } return new ValFrame(dst); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/assign/AstAssign.java
package water.rapids.ast.prims.assign; import water.*; import water.fvec.Frame; import water.rapids.Env; import water.rapids.vals.ValFrame; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; /** * Assign a whole frame over a global. Copy-On-Write optimizations make this cheap. */ public class AstAssign extends AstPrimitive { @Override public String[] args() { return new String[]{"id", "frame"}; } @Override public int nargs() { return 1 + 2; } // (assign id frame) @Override public String str() { return "assign"; } @Override public ValFrame apply(Env env, Env.StackHelp stk, AstRoot asts[]) { Key<Frame> id = Key.make(asts[1].str()); Frame src = stk.track(asts[2].exec(env)).getFrame(); return new ValFrame(env._ses.assign(id, src)); // New global Frame over shared Vecs } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/assign/AstRecAsgnHelper.java
package water.rapids.ast.prims.assign; import water.Iced; import water.fvec.Chunk; import water.fvec.Vec; import java.util.UUID; public class AstRecAsgnHelper { /** * Generic abstraction over Chunk setter methods. */ public static abstract class ValueSetter extends Iced<ValueSetter> { /** * Sets a value (possibly a constant) to a position of the Chunk. * @param idx Chunk-local index */ public abstract void setValue(Chunk chk, int idx); /** * Sets a value (possibly a constant) to a given index of a Vec. * @param vec Vec * @param idx absolute index */ public abstract void setValue(Vec vec, long idx); } /** * Create an instance of ValueSetter for a given scalar value. * It creates setter of the appropriate type based on the type of the underlying Vec. * @param v Vec * @param value scalar value * @return instance of ValueSetter */ public static ValueSetter createValueSetter(Vec v, Object value) { if (value == null) { return new NAValueSetter(); } switch (v.get_type()) { case Vec.T_CAT: return new CatValueSetter(v.domain(), value); case Vec.T_NUM: case Vec.T_TIME: return new NumValueSetter(value); case Vec.T_STR: return new StrValueSetter(value); case Vec.T_UUID: return new UUIDValueSetter(value); default: throw new IllegalArgumentException("Cannot create ValueSetter for a Vec of type = " + v.get_type_str()); } } private static class NAValueSetter extends ValueSetter { public NAValueSetter() {} // for Externalizable @Override public void setValue(Chunk chk, int idx) { chk.setNA(idx); } @Override public void setValue(Vec vec, long idx) { vec.setNA(idx); } } private static class CatValueSetter extends ValueSetter { private int _val; public CatValueSetter() {} // for Externalizable private CatValueSetter(String[] domain, Object val) { if (! (val instanceof String)) { throw new IllegalArgumentException("Value needs to be categorical, value = " + val); } int factorIdx = -1; for (int i = 0; i < domain.length; i++) if (val.equals(domain[i])) { factorIdx = i; break; } if (factorIdx == -1) throw new IllegalArgumentException("Value is not in the domain of the Vec, value = " + val); _val = factorIdx; } @Override public void setValue(Chunk chk, int idx) { chk.set(idx, _val); } @Override public void setValue(Vec vec, long idx) { vec.set(idx, (double) _val); } } private static class NumValueSetter extends ValueSetter { private double _val; public NumValueSetter() {} // for Externalizable private NumValueSetter(Object val) { if (! (val instanceof Number)) { throw new IllegalArgumentException("Value needs to be numeric, value = " + val); } _val = ((Number) val).doubleValue(); } @Override public void setValue(Chunk chk, int idx) { chk.set(idx, _val); } @Override public void setValue(Vec vec, long idx) { vec.set(idx, _val); } } private static class StrValueSetter extends ValueSetter { private String _val; public StrValueSetter() {} // for Externalizable private StrValueSetter(Object val) { if (! (val instanceof String)) { throw new IllegalArgumentException("Value needs to be string, value = " + val); } _val = (String) val; } @Override public void setValue(Chunk chk, int idx) { chk.set(idx, _val); } @Override public void setValue(Vec vec, long idx) { vec.set(idx, _val); } } private static class UUIDValueSetter extends ValueSetter { private UUID _val; public UUIDValueSetter() {} // for Externalizable private UUIDValueSetter(Object val) { if (val instanceof String) { val = UUID.fromString((String) val); } else if (! (val instanceof UUID)) { throw new IllegalArgumentException("Value needs to be an UUID, value = " + val); } _val = (UUID) val; } @Override public void setValue(Chunk chk, int idx) { chk.set(idx, _val); } @Override public void setValue(Vec vec, long idx) { vec.set(idx, _val); } } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/assign/AstRectangleAssign.java
package water.rapids.ast.prims.assign; import water.DKV; import water.H2O; import water.MRTask; import water.fvec.Chunk; import water.fvec.Frame; import water.fvec.Vec; import water.parser.BufferedString; import water.rapids.*; import water.rapids.ast.AstParameter; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import water.rapids.ast.params.AstNum; import water.rapids.ast.params.AstNumList; import water.rapids.ast.prims.mungers.AstColSlice; import water.rapids.vals.ValFrame; import water.util.ArrayUtils; import java.util.Arrays; import static water.rapids.ast.prims.assign.AstRecAsgnHelper.*; /** * Rectangular assign into a row and column slice. The destination must * already exist. The output is conceptually a new copy of the data, with a * fresh Frame. Copy-On-Write optimizations lower the cost to be proportional * to the over-written sections. */ public class AstRectangleAssign extends AstPrimitive { @Override public String[] args() { return new String[]{"dst", "src", "col_expr", "row_expr"}; } @Override public int nargs() { return 5; } // (:= dst src col_expr row_expr) @Override public String str() { return ":="; } @Override public ValFrame apply(Env env, Env.StackHelp stk, AstRoot[] asts) { Frame dst = stk.track(asts[1].exec(env)).getFrame(); Val vsrc = stk.track(asts[2].exec(env)); AstParameter col_list = (AstParameter) asts[3]; // Column selection AstNumList cols_numlist = new AstNumList(col_list.columns(dst.names())); // Special for AstAssign: "empty" really means "all" if (cols_numlist.isEmpty()) cols_numlist = new AstNumList(0, dst.numCols()); // Allow R-like number list expansion: negative column numbers mean exclusion int[] cols = AstColSlice.col_select(dst.names(), cols_numlist); // Any COW optimized path changes Vecs in dst._vecs, and so needs a // defensive copy. Any update-in-place path updates Chunks instead of // dst._vecs, and does not need a defensive copy. To make life easier, // just make the copy now. dst = new Frame(dst._names, dst.vecs().clone()); // Assign over the column slice if (asts[4] instanceof AstNum || asts[4] instanceof AstNumList) { // Explictly named row assignment AstNumList rows = (asts[4] instanceof AstNum) ? new AstNumList(((AstNum) asts[4]).getNum()) : ((AstNumList) asts[4]); if (rows.isEmpty()) rows = new AstNumList(0, dst.numRows()); // Empty rows is really: all rows switch (vsrc.type()) { case Val.NUM: assign_frame_scalar(dst, cols, rows, nanToNull(vsrc.getNum()), env._ses); break; case Val.STR: assign_frame_scalar(dst, cols, rows, vsrc.getStr(), env._ses); break; case Val.FRM: assign_frame_frame(dst, cols, rows, vsrc.getFrame(), env._ses); break; default: throw new IllegalArgumentException("Source must be a Frame or Number, but found a " + vsrc.getClass()); } } else { // Boolean assignment selection? Frame rows = stk.track(asts[4].exec(env)).getFrame(); switch (vsrc.type()) { case Val.NUM: assign_frame_scalar(dst, cols, rows, nanToNull(vsrc.getNum()), env._ses); break; case Val.STR: assign_frame_scalar(dst, cols, rows, vsrc.getStr(), env._ses); break; case Val.FRM: throw H2O.unimpl(); default: throw new IllegalArgumentException("Source must be a Frame or Number, but found a " + vsrc.getClass()); } } return new ValFrame(dst); } // Rectangular array copy from src into dst private void assign_frame_frame(Frame dst, int[] cols, AstNumList rows, Frame src, Session ses) { // Sanity check if (cols.length != src.numCols()) throw new IllegalArgumentException("Source and destination frames must have the same count of columns"); long nrows = rows.cnt(); if (src.numRows() != nrows) throw new IllegalArgumentException("Requires same count of rows in the number-list (" + nrows + ") as in the source (" + src.numRows() + ")"); // Whole-column assignment? Directly reuse columns: Copy-On-Write // optimization happens here on the apply() exit. if (dst.numRows() == nrows && rows.isDense()) { for (int i = 0; i < cols.length; i++) dst.replace(cols[i], src.vecs()[i]); if (dst._key != null) DKV.put(dst); return; } // Partial update; needs to preserve type, and may need to copy to support // copy-on-write Vec[] dvecs = dst.vecs(); final Vec[] svecs = src.vecs(); for (int col = 0; col < cols.length; col++) { int dtype = dvecs[cols[col]].get_type(); if (dtype != svecs[col].get_type()) throw new IllegalArgumentException("Columns must be the same type; " + "column " + col + ", \'" + dst._names[cols[col]] + "\', is of type " + dvecs[cols[col]].get_type_str() + " and the source is " + svecs[col].get_type_str()); if ((dtype == Vec.T_CAT) && (! Arrays.equals(dvecs[cols[col]].domain(), svecs[col].domain()))) throw new IllegalArgumentException("Cannot assign to a categorical column with a different domain; " + "source column " + src._names[col] + ", target column " + dst._names[cols[col]]); } // Frame fill // Handle fast small case if (nrows <= 1 || (cols.length * nrows) <= 1000) { // Go parallel for more than 1000 random updates // Copy dst columns as-needed to allow update-in-place dvecs = ses.copyOnWrite(dst, cols); // Update dst columns long[] rownums = rows.expand8(); // Just these rows for (int col = 0; col < svecs.length; col++) if (svecs[col].get_type() == Vec.T_STR) { BufferedString bStr = new BufferedString(); for (int ridx = 0; ridx < rownums.length; ridx++) { BufferedString s = svecs[col].atStr(bStr, ridx); dvecs[cols[col]].set(rownums[ridx], s != null ? s.toString() : null); } } else { for (int ridx = 0; ridx < rownums.length; ridx++) dvecs[cols[col]].set(rownums[ridx], svecs[col].at(ridx)); } return; } // Handle large case Vec[] vecs = ses.copyOnWrite(dst, cols); Vec[] vecs2 = new Vec[cols.length]; // Just the selected columns get updated for (int i = 0; i < cols.length; i++) vecs2[i] = vecs[cols[i]]; rows.sort(); // Side-effect internal sort; needed for fast row lookup new AssignFrameFrameTask(rows, svecs).doAll(vecs2); } private static class AssignFrameFrameTask extends RowSliceTask { private Vec[] _svecs; private AssignFrameFrameTask(AstNumList rows, Vec[] svecs) { super(rows); _svecs = svecs; } @Override void mapChunkSlice(Chunk[] cs, int chkOffset) { long start = cs[0].start(); Chunk[] scs = null; for (int i = chkOffset; i < cs[0]._len; ++i) { long idx = _rows.index(start + i); if (idx < 0) continue; if ((scs == null) || (scs[0].start() < idx) || (idx >= scs[0].start() + scs[0].len())) { int sChkIdx = _svecs[0].elem2ChunkIdx(idx); scs = new Chunk[_svecs.length]; for (int j = 0; j < _svecs.length; j++) { scs[j] = _svecs[j].chunkForChunkIdx(sChkIdx); } } BufferedString bStr = new BufferedString(); int si = (int) (idx - scs[0].start()); for (int j = 0; j < cs.length; j++) { Chunk chk = cs[j]; Chunk schk = scs[j]; if (_svecs[j].get_type() == Vec.T_STR) { BufferedString s = schk.atStr(bStr, si); chk.set(i, s != null ? s.toString() : null); BufferedString bss = chk.atStr(new BufferedString(), i); if (s == null && bss != null) { chk.set(i, s != null ? s.toString() : null); } } else { chk.set(i, schk.atd(si)); } } } } } // Assign a SCALAR over some dst rows; optimize for all rows private void assign_frame_scalar(Frame dst, int[] cols, AstNumList rows, Object src, Session ses) { long nrows = rows.cnt(); // Bulk assign a numeric constant (probably zero) over a frame. Directly set // columns: Copy-On-Write optimization happens here on the apply() exit. // Note: this skips "scalar to Vec" compatibility check because the whole Vec is overwritten if (dst.numRows() == nrows && rows.isDense() && (src instanceof Number)) { Vec anyVec = dst.anyVec(); assert anyVec != null; // if anyVec was null, then dst.numRows() would have been 0 Vec vsrc = anyVec.makeCon((double) src); for (int col : cols) dst.replace(col, vsrc); if (dst._key != null) DKV.put(dst); return; } // Make sure the scalar value is compatible with the target vector for (int col: cols) { if (! isScalarCompatible(src, dst.vec(col))) { throw new IllegalArgumentException("Cannot assign value " + src + " into a vector of type " + dst.vec(col).get_type_str() + "."); } } // Handle fast small case if (nrows == 1) { Vec[] vecs = ses.copyOnWrite(dst, cols); long drow = (long) rows._bases[0]; for (int col : cols) createValueSetter(vecs[col], src).setValue(vecs[col], drow); return; } // Handle large case Vec[] vecs = ses.copyOnWrite(dst, cols); Vec[] vecs2 = new Vec[cols.length]; // Just the selected columns get updated for (int i = 0; i < cols.length; i++) vecs2[i] = vecs[cols[i]]; rows.sort(); // Side-effect internal sort; needed for fast row lookup AssignFrameScalarTask.doAssign(rows, vecs2, src); } private static class AssignFrameScalarTask extends RowSliceTask { final ValueSetter[] _setters; AssignFrameScalarTask(AstNumList rows, Vec[] vecs, Object value) { super(rows); _setters = new ValueSetter[vecs.length]; for (int i = 0; i < _setters.length; i++) _setters[i] = createValueSetter(vecs[i], value); } @Override void mapChunkSlice(Chunk[] cs, int chkOffset) { long start = cs[0].start(); for (int i = chkOffset; i < cs[0]._len; ++i) if (_rows.has(start + i)) for (int col = 0; col < cs.length; col++) _setters[col].setValue(cs[col], i); } /** * Assigns a given value to a specified rows of given Vecs. * @param rows row specification * @param dst target Vecs * @param src source Value */ static void doAssign(AstNumList rows, Vec[] dst, Object src) { new AssignFrameScalarTask(rows, dst, src).doAll(dst); } } private boolean isScalarCompatible(Object scalar, Vec v) { if (scalar == null) return true; else if (scalar instanceof Number) return v.get_type() == Vec.T_NUM || v.get_type() == Vec.T_TIME; else if (scalar instanceof String) { if (v.get_type() == Vec.T_CAT) { return ArrayUtils.contains(v.domain(), (String) scalar); } else return v.get_type() == Vec.T_STR || (v.get_type() == Vec.T_UUID); } else return false; } private static Double nanToNull(double value) { return Double.isNaN(value) ? null : value; } // Boolean assignment with a scalar private void assign_frame_scalar(Frame dst, int[] cols, Frame rows, Object src, Session ses) { Vec bool = rows.vec(0); if (dst.numRows() != rows.numRows()) { throw new IllegalArgumentException("Frame " + dst._key + " has different number of rows than frame " + rows._key + " (" + dst.numRows() + " vs " + rows.numRows() + ")."); } // Bulk assign a numeric constant over a frame. Directly set columns without checking target type // assuming the user just wants to overwrite everything: Copy-On-Write optimization happens here on the apply() exit. // Note: this skips "scalar to Vec" compatibility check because the whole Vec is overwritten if (bool.isConst() && ((int) bool.min() == 1) && (src instanceof Number)) { Vec anyVec = dst.anyVec(); assert anyVec != null; Vec vsrc = anyVec.makeCon((double) src); for (int col : cols) dst.replace(col, vsrc); if (dst._key != null) DKV.put(dst); return; } // Make sure the scalar value is compatible with the target vector for (int col: cols) { if (! isScalarCompatible(src, dst.vec(col))) { throw new IllegalArgumentException("Cannot assign value " + src + " into a vector of type " + dst.vec(col).get_type_str() + "."); } } Vec[] vecs = ses.copyOnWrite(dst, cols); Vec[] vecs2 = new Vec[cols.length]; // Just the selected columns get updated for (int i = 0; i < cols.length; i++) vecs2[i] = vecs[cols[i]]; ConditionalAssignTask.doAssign(vecs2, src, rows.vec(0)); } private static class ConditionalAssignTask extends MRTask<ConditionalAssignTask> { final ValueSetter[] _setters; ConditionalAssignTask(Vec[] vecs, Object value) { _setters = new ValueSetter[vecs.length]; for (int i = 0; i < _setters.length; i++) _setters[i] = AstRecAsgnHelper.createValueSetter(vecs[i], value); } @Override public void map(Chunk[] cs) { Chunk bool = cs[cs.length - 1]; for (int row = 0; row < cs[0]._len; row++) { if (bool.at8(row) == 1) for (int col = 0; col < cs.length - 1; col++) _setters[col].setValue(cs[col], row); } } /** * Sets a given value to all cells where given predicateVec is true. * @param dst target Vecs * @param src source Value * @param predicateVec predicate Vec */ static void doAssign(Vec[] dst, Object src, Vec predicateVec) { Vec[] vecs = new Vec[dst.length + 1]; System.arraycopy(dst, 0, vecs, 0, dst.length); vecs[vecs.length - 1] = predicateVec; new ConditionalAssignTask(dst, src).doAll(vecs); } } private static abstract class RowSliceTask extends MRTask<RowSliceTask> { final AstNumList _rows; RowSliceTask(AstNumList rows) { _rows = rows; } @Override public void map(Chunk[] cs) { long start = cs[0].start(); long end = start + cs[0]._len; long min = (long) _rows.min(), max = (long) _rows.max() - 1; // exclusive max to inclusive max when stride == 1 // [ start, ..., end ] the chunk //1 [] rows out left: rows.max() < start //2 [] rows out rite: rows.min() > end //3 [ rows ] rows run left: rows.min() < start && rows.max() <= end //4 [ rows ] rows run in : start <= rows.min() && rows.max() <= end //5 [ rows ] rows run rite: start <= rows.min() && end < rows.max() if (!(max < start || min > end)) { // not situation 1 or 2 above long startOffset = min > start ? min : start; // situation 4 and 5 => min > start; int chkOffset = (int) (startOffset - start); mapChunkSlice(cs, chkOffset); } } abstract void mapChunkSlice(Chunk[] cs, int chkOffset); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/assign/AstRm.java
package water.rapids.ast.prims.assign; import water.DKV; import water.Key; import water.Keyed; import water.Value; import water.fvec.Frame; import water.rapids.Env; import water.rapids.vals.ValNum; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; /** * Remove by ID. Removing a Frame updates refcnts. Returns 1 for removing, 0 if id does not exist. */ public class AstRm extends AstPrimitive { @Override public String[] args() { return new String[]{"id"}; } @Override public int nargs() { return 1 + 1; } // (rm id) @Override public String str() { return "rm"; } @Override public ValNum apply(Env env, Env.StackHelp stk, AstRoot[] asts) { Key id = Key.make(env.expand(asts[1].str())); Value val = DKV.get(id); if (val == null) return new ValNum(0); if (val.isFrame()) env._ses.remove(val.<Frame>get()); // Remove unshared Vecs else Keyed.remove(id); // Normal (e.g. Model) remove return new ValNum(1); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/assign/AstTmpAssign.java
package water.rapids.ast.prims.assign; import water.DKV; import water.Key; import water.Value; import water.fvec.Frame; import water.rapids.Env; import water.rapids.Val; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import water.rapids.vals.ValFrame; /** * Assign a temp. All such assignments are final (cannot change), but the temp can be deleted. Temp is returned for * immediate use, and also set in the DKV. Must be globally unique in the DKV. */ public class AstTmpAssign extends AstPrimitive { @Override public String[] args() { return new String[]{"id", "frame"}; } @Override public int nargs() { return 1 + 2; } // (tmp= id frame) @Override public String str() { return "tmp="; } @Override public ValFrame apply(Env env, Env.StackHelp stk, AstRoot[] asts) { // Note: non-standard evaluation of the first argument! Instead of being // executed, it is stringified. This, for example, allows us to write an // expression as // (tmp= newid (* frame 3)) // instead of // (tmp= "newid" (* frame 3)) // On the other hand, this makes us unable to create dynamic identifiers // in Rapids, for example this is invalid: // (tmp= (+ "id" 3) (* frame 3)) // Right now there is no need for dynamically generated identifiers, since // we don't even have proper variables or loops or control structures yet. // Key<Frame> id = Key.make(env.expand(asts[1].str())); Val srcVal = stk.track(asts[2].exec(env)); Frame srcFrame = srcVal.getFrame(); Value v = DKV.get(id); if (v != null) { if (v.get().equals(srcFrame)) return (ValFrame) srcVal; else throw new IllegalArgumentException("Temp ID " + id + " already exists"); } Frame dst = new Frame(id, srcFrame._names, srcFrame.vecs()); return new ValFrame(env._ses.track_tmp(dst)); // Track new session-wide ID } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/filters
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/filters/dropduplicates/AstDropDuplicates.java
package water.rapids.ast.prims.filters.dropduplicates; import water.fvec.Frame; import water.rapids.Env; import water.rapids.Val; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import water.rapids.vals.ValFrame; import water.util.EnumUtils; import java.util.Arrays; /** * Removes duplicated rows, leaving only the first or last observed duplicate in place. */ public class AstDropDuplicates extends AstPrimitive<AstDropDuplicates> { @Override public int nargs() { return 1 + 3; } @Override public String[] args() { return new String[]{"ary", "frame", "cols", "droporder"}; } @Override public Val apply(Env env, Env.StackHelp stk, AstRoot[] asts) { final Frame deduplicatedFrame = stk.track(asts[1].exec(env)).getFrame(); final int[] comparedColumnsIndices = ColumnIndicesParser.parseAndCheckComparedColumnIndices(deduplicatedFrame, stk.track(asts[2].exec(env))); final String dropOrderString = asts[3].str(); final KeepOrder keepOrder = EnumUtils.valueOfIgnoreCase(KeepOrder.class, dropOrderString) .orElseThrow(() -> new IllegalArgumentException(String.format("Unknown drop order: '%s'. Known types: %s", dropOrderString, Arrays.toString(KeepOrder.values())))); final DropDuplicateRows dropDuplicateRows = new DropDuplicateRows(deduplicatedFrame, comparedColumnsIndices, keepOrder); final Frame outputFrame = dropDuplicateRows.dropDuplicates(); return new ValFrame(outputFrame); } @Override public String str() { return "dropdup"; } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/filters
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/filters/dropduplicates/CollectChunkBorderValuesTask.java
package water.rapids.ast.prims.filters.dropduplicates; import water.MRTask; import water.fvec.Chunk; import water.fvec.NewChunk; /** * Collects border values from a chunk - last element of each chunk. */ public class CollectChunkBorderValuesTask extends MRTask<CollectChunkBorderValuesTask> { @Override public void map(final Chunk[] oldChunks, NewChunk[] newChunks) { for (int columnIndex = 0; columnIndex < oldChunks.length; columnIndex++) { oldChunks[columnIndex].extractRows(newChunks[columnIndex], oldChunks[columnIndex].len() - 1); } } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/filters
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/filters/dropduplicates/ColumnIndicesParser.java
package water.rapids.ast.prims.filters.dropduplicates; import water.fvec.Frame; import water.fvec.Vec; import water.rapids.Val; public class ColumnIndicesParser { /** * @param deduplicatedFrame Deduplicated frame to look for vectors in * @param comparedColumns A {@link Val} instance of columns to be compared during row de-duplication process. * Accepts {@link water.rapids.ast.params.AstStr}, {@link water.rapids.ast.params.AstStrList}, * and {@link water.rapids.ast.params.AstNumList}. * @return An array if primitive integers with indices of vectorss * @throws IllegalArgumentException If comparedColumns field is not of given type. If any given column is not * recognized in deduplicatedFrame or if any column compared has unsupported type. */ public static int[] parseAndCheckComparedColumnIndices(final Frame deduplicatedFrame, final Val comparedColumns) throws IllegalArgumentException { final int[] columnIndices; if (comparedColumns.isStr()) { final String columnName = comparedColumns.getStr(); final int columnIndex = deduplicatedFrame.find(columnName); if (columnIndex == -1) { throw new IllegalArgumentException(String.format("Unknown column name: '%s'", columnName)); } columnIndices = new int[]{columnIndex}; } else if (comparedColumns.isStrs()) { final String[] columnNames = comparedColumns.getStrs(); columnIndices = new int[columnNames.length]; for (int i = 0; i < columnNames.length; i++) { final String columnName = columnNames[i]; final int columnIndex = deduplicatedFrame.find(columnName); if (columnIndex == -1) { throw new IllegalArgumentException(String.format("Unknown column name: '%s'", columnName)); } else if (isUnsupportedVecType(deduplicatedFrame.types()[columnIndex])) { throw new IllegalArgumentException(String.format("Column '%s' is of unsupported type %s for row de-duplication.", columnName, Vec.TYPE_STR[deduplicatedFrame.types()[columnIndex]])); } columnIndices[i] = columnIndex; } } else if (comparedColumns.isNums()) { final double[] columnRangeDouble = comparedColumns.getNums(); columnIndices = new int[columnRangeDouble.length]; for (int i = 0; i < columnRangeDouble.length; i++) { columnIndices[i] = (int) columnRangeDouble[i]; if (columnIndices[i] < 0 || columnIndices[i] > deduplicatedFrame.numCols() - 1) { throw new IllegalArgumentException(String.format("No such column index: '%d', frame has %d columns," + "maximum index is %d. ", columnIndices[i], deduplicatedFrame.numCols(), deduplicatedFrame.numCols() - 1)); } else if (isUnsupportedVecType(deduplicatedFrame.types()[columnIndices[i]])) { throw new IllegalArgumentException(String.format("Column '%s' is of unsupported type %s for row de-duplication.", deduplicatedFrame.name(columnIndices[i]), Vec.TYPE_STR[deduplicatedFrame.types()[columnIndices[i]]])); } } } else { throw new IllegalArgumentException(String.format("Column range for deduplication must either be a set of columns, or a " + "column range. Given type: %s", comparedColumns.type())); } return columnIndices; } private static boolean isUnsupportedVecType(final byte vecType) { return vecType == Vec.T_STR || vecType == Vec.T_BAD || vecType == Vec.T_UUID; } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/filters
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/filters/dropduplicates/DropDuplicateRows.java
package water.rapids.ast.prims.filters.dropduplicates; import water.Scope; import water.fvec.Frame; import water.fvec.Vec; import water.rapids.Merge; import water.util.ArrayUtils; import java.util.Arrays; /** * Drops duplicated rows of a Frame */ public class DropDuplicateRows { private static final String LABEL_COLUMN_NAME = "label"; final Frame sourceFrame; final int[] comparedColumnIndices; final KeepOrder keepOrder; /** * @param sourceFrame Frame to perform the row de-duplication on * @param comparedColumnIndices Indices of columns to consider during the comparison * @param keepOrder Which rows to keep. */ public DropDuplicateRows(final Frame sourceFrame, final int[] comparedColumnIndices, final KeepOrder keepOrder) { this.sourceFrame = sourceFrame; this.comparedColumnIndices = comparedColumnIndices; this.keepOrder = keepOrder; } public Frame dropDuplicates() { Frame outputFrame = null; try { Scope.enter(); final Vec labelVec = Scope.track(Vec.makeSeq(1, sourceFrame.numRows())); final Frame fr = new Frame(sourceFrame); fr.add(LABEL_COLUMN_NAME, labelVec); final Frame sortedFrame = Scope.track(sortByComparedColumns(fr)); final Frame chunkBoundaries = Scope.track(new CollectChunkBorderValuesTask() .doAll(sortedFrame.types(), sortedFrame) .outputFrame(null, sortedFrame.names(), sortedFrame.domains())); final Frame deDuplicatedFrame = Scope.track(new DropDuplicateRowsTask(chunkBoundaries, comparedColumnIndices) .doAll(sortedFrame.types(), sortedFrame) .outputFrame(null, sortedFrame.names(), sortedFrame.domains())); // Removing duplicates, domains remain the same // Before the final sorted duplicated is created, remove the unused datasets to free some space early chunkBoundaries.remove(); sortedFrame.remove(); outputFrame = Scope.track(Merge.sort(deDuplicatedFrame, deDuplicatedFrame.numCols() - 1)); outputFrame.remove(outputFrame.numCols() - 1).remove(); return outputFrame; } finally { if (outputFrame != null) { Scope.exit(outputFrame.keys()); } else { Scope.exit(); // Clean up in case of any exception/error. } } } /** * Creates a copy of the original dataset, sorted by all compared columns. * The sort is done with respect to {@link KeepOrder} value. * * @return A new Frame sorted by all compared columns. */ private Frame sortByComparedColumns(Frame fr) { final int labelColumnIndex = fr.find(LABEL_COLUMN_NAME); final int[] sortByColumns = ArrayUtils.append(comparedColumnIndices, labelColumnIndex); final boolean ascendingSort = KeepOrder.First == keepOrder; final int[] sortOrder = new int[sortByColumns.length]; // Compared columns are always sorted in the same order Arrays.fill(sortOrder, 0, sortOrder.length - 1, Merge.ASCENDING); // Label column is sorted differently based on DropOrder sortOrder[sortOrder.length - 1] = ascendingSort ? Merge.ASCENDING : Merge.DESCENDING; return Merge.sort(fr, sortByColumns, sortOrder); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/filters
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/filters/dropduplicates/DropDuplicateRowsTask.java
package water.rapids.ast.prims.filters.dropduplicates; import water.MRTask; import water.fvec.Chunk; import water.fvec.Frame; import water.fvec.NewChunk; import water.fvec.Vec; /** * Performs the row de-duplication itself. */ public class DropDuplicateRowsTask extends MRTask<DropDuplicateRowsTask> { final Frame chunkBoundaries; private final int[] comparedColumnIndices; /** * @param chunkBoundaries Frame with border values of each chunk * @param comparedColumnIndices Columns indices to include in row deduplication comparison */ public DropDuplicateRowsTask(final Frame chunkBoundaries, final int[] comparedColumnIndices) { this.chunkBoundaries = chunkBoundaries; this.comparedColumnIndices = comparedColumnIndices; } @Override public void map(Chunk[] chunks, NewChunk[] newChunks) { final int chunkLength = chunks[0].len(); final int chunkId = chunks[0].cidx(); for (int row = 0; row < chunkLength; row++) { final boolean equal; if (chunkId == 0 && row == 0) { for (int columnIndex = 0; columnIndex < chunks.length; columnIndex++) { chunks[columnIndex].extractRows(newChunks[columnIndex], row); } continue; } else if (chunkId != 0 && row == 0) { equal = compareFirstRowWithPreviousChunk(chunks, row, chunkId); } else { equal = compareRows(chunks, row, chunks, row - 1); } if (!equal) { for (int columnIndex = 0; columnIndex < chunks.length; columnIndex++) { chunks[columnIndex].extractRows(newChunks[columnIndex], row); } } } } private boolean compareFirstRowWithPreviousChunk(final Chunk[] chunks, final int row, final int chunkId) { final Chunk[] previousRowChunks = new Chunk[chunkBoundaries.numCols()]; for (int column = 0; column < chunkBoundaries.numCols(); column++) { previousRowChunks[column] = chunkBoundaries.vec(column).chunkForChunkIdx(chunkId - 1); } return compareRows(chunks, row, previousRowChunks, 0); } private boolean compareRows(final Chunk[] chunksA, final int rowA, final Chunk[] chunksB, final int rowB) { for (final int column : comparedColumnIndices) { final boolean isPreviousNA = chunksA[column].isNA(rowA); final boolean isCurrentNA = chunksB[column].isNA(rowB); if (isPreviousNA || isCurrentNA) { return isPreviousNA && isCurrentNA; } switch (chunksA[column].vec().get_type()) { case Vec.T_NUM: final double previousDoubleValue = chunksA[column].atd(rowA); final double currentDoubleValue = chunksB[column].atd(rowB); if (previousDoubleValue != currentDoubleValue) return false; break; case Vec.T_CAT: case Vec.T_TIME: final long previousTimeValue = chunksA[column].at8(rowA); final long currentTimeValue = chunksB[column].at8(rowB); if (previousTimeValue != currentTimeValue) return false; break; default: throw new IllegalStateException("Unexpected value: " + chunksA[column].vec().get_type()); } } return true; } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/filters
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/filters/dropduplicates/KeepOrder.java
package water.rapids.ast.prims.filters.dropduplicates; /** * Determines which duplicated row is kept during row de-duplication process. */ public enum KeepOrder { First, // Retain first, drop rest Last // Retain last, drop rest }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/internal/AstRunTool.java
package water.rapids.ast.prims.internal; import water.rapids.Env; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import water.rapids.vals.ValStr; import java.lang.reflect.Method; public class AstRunTool extends AstPrimitive<AstRunTool> { private static final String TOOLS_PACKAGE = "water.tools."; @Override public String[] args() { return new String[]{"tool_class", "tool_parameters"}; } @Override public int nargs() { return 1 + 2; } // (run_tool tool_class tool_parameters) @Override public String str() { return "run_tool"; } @Override public ValStr apply(Env env, Env.StackHelp stk, AstRoot[] asts) { String toolClassName = stk.track(asts[1].exec(env)).getStr(); String[] args = stk.track(asts[2].exec(env)).getStrs(); try { // only allow to run approved tools (from our package), not just anything on classpath Class<?> clazz = Class.forName(TOOLS_PACKAGE + toolClassName); Method mainMethod = clazz.getDeclaredMethod("mainInternal", String[].class); mainMethod.invoke(null, new Object[]{args}); } catch (Exception e) { RuntimeException shorterException = new RuntimeException(e.getCause().getMessage()); shorterException.setStackTrace(new StackTraceElement[0]); throw shorterException; } return new ValStr("OK"); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstAbs.java
package water.rapids.ast.prims.math; /** */ public class AstAbs extends AstUniOp { @Override public String str() { return "abs"; } @Override public double op(double d) { return Math.abs(d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstAcos.java
package water.rapids.ast.prims.math; /** */ public class AstAcos extends AstUniOp { @Override public String str() { return "acos"; } @Override public double op(double d) { return Math.acos(d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstAcosh.java
package water.rapids.ast.prims.math; import org.apache.commons.math3.util.FastMath; /** */ public class AstAcosh extends AstUniOp { @Override public String str() { return "acosh"; } @Override public double op(double d) { return FastMath.acosh(d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstAsin.java
package water.rapids.ast.prims.math; /** */ public class AstAsin extends AstUniOp { @Override public String str() { return "asin"; } @Override public double op(double d) { return Math.asin(d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstAsinh.java
package water.rapids.ast.prims.math; import org.apache.commons.math3.util.FastMath; /** */ public class AstAsinh extends AstUniOp { @Override public String str() { return "asinh"; } @Override public double op(double d) { return FastMath.asinh(d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstAtan.java
package water.rapids.ast.prims.math; /** */ public class AstAtan extends AstUniOp { @Override public String str() { return "atan"; } @Override public double op(double d) { return Math.atan(d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstAtanh.java
package water.rapids.ast.prims.math; import org.apache.commons.math3.util.FastMath; /** */ public class AstAtanh extends AstUniOp { @Override public String str() { return "atanh"; } @Override public double op(double d) { return FastMath.atanh(d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstCeiling.java
package water.rapids.ast.prims.math; /** */ public class AstCeiling extends AstUniOp { @Override public String str() { return "ceiling"; } @Override public double op(double d) { return Math.ceil(d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstCos.java
package water.rapids.ast.prims.math; /** */ public class AstCos extends AstUniOp { @Override public String str() { return "cos"; } @Override public double op(double d) { return Math.cos(d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstCosPi.java
package water.rapids.ast.prims.math; /** */ public class AstCosPi extends AstUniOp { @Override public String str() { return "cospi"; } @Override public double op(double d) { return Math.cos(Math.PI * d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstCosh.java
package water.rapids.ast.prims.math; /** */ public class AstCosh extends AstUniOp { @Override public String str() { return "cosh"; } @Override public double op(double d) { return Math.cosh(d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstDiGamma.java
package water.rapids.ast.prims.math; import org.apache.commons.math3.special.Gamma; /** */ public class AstDiGamma extends AstUniOp { @Override public String str() { return "digamma"; } @Override public double op(double d) { return Double.isNaN(d) ? Double.NaN : Gamma.digamma(d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstExp.java
package water.rapids.ast.prims.math; /** */ public class AstExp extends AstUniOp { @Override public String str() { return "exp"; } @Override public double op(double d) { return Math.exp(d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstExpm1.java
package water.rapids.ast.prims.math; /** */ public class AstExpm1 extends AstUniOp { @Override public String str() { return "expm1"; } @Override public double op(double d) { return Math.expm1(d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstFloor.java
package water.rapids.ast.prims.math; /** */ public class AstFloor extends AstUniOp { @Override public String str() { return "floor"; } @Override public double op(double d) { return Math.floor(d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstGamma.java
package water.rapids.ast.prims.math; import org.apache.commons.math3.special.Gamma; /** */ public class AstGamma extends AstUniOp { @Override public String str() { return "gamma"; } @Override public double op(double d) { return Gamma.gamma(d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstLGamma.java
package water.rapids.ast.prims.math; import org.apache.commons.math3.special.Gamma; /** */ public class AstLGamma extends AstUniOp { @Override public String str() { return "lgamma"; } @Override public double op(double d) { return Gamma.logGamma(d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstLog.java
package water.rapids.ast.prims.math; /** */ public class AstLog extends AstUniOp { @Override public String str() { return "log"; } @Override public double op(double d) { return Math.log(d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstLog10.java
package water.rapids.ast.prims.math; /** */ public class AstLog10 extends AstUniOp { @Override public String str() { return "log10"; } @Override public double op(double d) { return Math.log10(d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstLog1P.java
package water.rapids.ast.prims.math; /** */ public class AstLog1P extends AstUniOp { @Override public String str() { return "log1p"; } @Override public double op(double d) { return Math.log1p(d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstLog2.java
package water.rapids.ast.prims.math; /** */ public class AstLog2 extends AstUniOp { @Override public String str() { return "log2"; } @Override public double op(double d) { return Math.log(d) / Math.log(2); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstNoOp.java
package water.rapids.ast.prims.math; /** */ public class AstNoOp extends AstUniOp { @Override public String str() { return "none"; } @Override public double op(double d) { return d; } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstNot.java
package water.rapids.ast.prims.math; /** */ public class AstNot extends AstUniOp { public String str() { return "not"; } public double op(double d) { return Double.isNaN(d) ? Double.NaN : d == 0 ? 1 : 0; } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstRound.java
package water.rapids.ast.prims.math; import water.rapids.ast.prims.operators.AstBinOp; /** */ public class AstRound extends AstBinOp { public String str() { return "round"; } public double op(double x, double digits) { // e.g.: floor(2.676*100 + 0.5) / 100 => 2.68 if (Double.isNaN(x)) return x; double sgn = x < 0 ? -1 : 1; x = Math.abs(x); if ((int) digits != digits) digits = Math.round(digits); double power_of_10 = (int) Math.pow(10, (int) digits); return sgn * (digits == 0 // go to the even digit ? (x % 1 > 0.5 || (x % 1 == 0.5 && !(Math.floor(x) % 2 == 0))) ? Math.ceil(x) : Math.floor(x) : Math.floor(x * power_of_10 + 0.5) / power_of_10); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstSgn.java
package water.rapids.ast.prims.math; /** */ public class AstSgn extends AstUniOp { @Override public String str() { return "sign"; } @Override public double op(double d) { return Math.signum(d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstSignif.java
package water.rapids.ast.prims.math; import water.rapids.ast.prims.operators.AstBinOp; /** */ public class AstSignif extends AstBinOp { public String str() { return "signif"; } public double op(double x, double digits) { if (Double.isNaN(x)) return x; if (digits < 1) digits = 1; //mimic R's base::signif if ((int) digits != digits) digits = Math.round(digits); java.math.BigDecimal bd = new java.math.BigDecimal(x); bd = bd.round(new java.math.MathContext((int) digits, java.math.RoundingMode.HALF_EVEN)); return bd.doubleValue(); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstSin.java
package water.rapids.ast.prims.math; /** */ public class AstSin extends AstUniOp { @Override public String str() { return "sin"; } @Override public double op(double d) { return Math.sin(d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstSinPi.java
package water.rapids.ast.prims.math; /** */ public class AstSinPi extends AstUniOp { @Override public String str() { return "sinpi"; } @Override public double op(double d) { return Math.sin(Math.PI * d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstSinh.java
package water.rapids.ast.prims.math; /** */ public class AstSinh extends AstUniOp { @Override public String str() { return "sinh"; } @Override public double op(double d) { return Math.sinh(d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstSqrt.java
package water.rapids.ast.prims.math; /** */ public class AstSqrt extends AstUniOp { @Override public String str() { return "sqrt"; } @Override public double op(double d) { return Math.sqrt(d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstTan.java
package water.rapids.ast.prims.math; /** */ public class AstTan extends AstUniOp { @Override public String str() { return "tan"; } @Override public double op(double d) { return Math.tan(d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstTanPi.java
package water.rapids.ast.prims.math; /** */ public class AstTanPi extends AstUniOp { @Override public String str() { return "tanpi"; } @Override public double op(double d) { return Math.tan(Math.PI * d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstTanh.java
package water.rapids.ast.prims.math; /** */ public class AstTanh extends AstUniOp { @Override public String str() { return "tanh"; } @Override public double op(double d) { return Math.tanh(d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstTriGamma.java
package water.rapids.ast.prims.math; import org.apache.commons.math3.special.Gamma; /** */ public class AstTriGamma extends AstUniOp { @Override public String str() { return "trigamma"; } @Override public double op(double d) { return Double.isNaN(d) ? Double.NaN : Gamma.trigamma(d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstTrunc.java
package water.rapids.ast.prims.math; /** */ public class AstTrunc extends AstUniOp { @Override public String str() { return "trunc"; } @Override public double op(double d) { return d >= 0 ? Math.floor(d) : Math.ceil(d); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/math/AstUniOp.java
package water.rapids.ast.prims.math; import water.H2O; import water.MRTask; import water.fvec.Chunk; import water.fvec.Frame; import water.fvec.NewChunk; import water.fvec.Vec; import water.rapids.Val; import water.rapids.ast.AstBuiltin; import water.rapids.vals.ValFrame; import water.rapids.vals.ValNum; import water.rapids.vals.ValRow; /** * Subclasses auto-widen between scalars and Frames, and have exactly one argument */ public abstract class AstUniOp<T extends AstUniOp<T>> extends AstBuiltin<T> { @Override public String[] args() { return new String[]{"ary"}; } @Override public int nargs() { return 1 + 1; } @Override public Val exec(Val... args) { Val val = args[1]; switch (val.type()) { case Val.NUM: return new ValNum(op(val.getNum())); case Val.FRM: Frame fr = val.getFrame(); for (int i = 0; i < fr.numCols(); i++) if (!fr.vec(i).isNumeric()) throw new IllegalArgumentException( "Operator " + str() + "() cannot be applied to non-numeric column " + fr.name(i)); // Get length of columns in fr and append `op(colName)`. For example, a column named "income" that had // a log transformation would now be changed to `log(income)`. String[] newNames = new String[fr.numCols()]; for (int i = 0; i < newNames.length; i++) { newNames[i] = str() + "(" + fr.name(i) + ")"; } return new ValFrame(new MRTask() { @Override public void map(Chunk cs[], NewChunk ncs[]) { for (int col = 0; col < cs.length; col++) { Chunk c = cs[col]; NewChunk nc = ncs[col]; for (int i = 0; i < c._len; i++) nc.addNum(op(c.atd(i))); } } }.doAll(fr.numCols(), Vec.T_NUM, fr).outputFrame(newNames, null)); case Val.ROW: double[] ds = new double[val.getRow().length]; for (int i = 0; i < ds.length; ++i) ds[i] = op(val.getRow()[i]); String[] names = ((ValRow) val).getNames().clone(); return new ValRow(ds, names); default: throw H2O.unimpl("unop unimpl: " + val.getClass()); } } public abstract double op(double d); }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/matrix/AstMMult.java
package water.rapids.ast.prims.matrix; import hex.DMatrix; import water.fvec.Frame; import water.rapids.Env; import water.rapids.vals.ValFrame; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; /** * Matrix multiplication */ public class AstMMult extends AstPrimitive { @Override public String[] args() { return new String[]{"ary", "ary2"}; } @Override public int nargs() { return 1 + 2; } // (x X1 X2) @Override public String str() { return "x"; } @Override public ValFrame apply(Env env, Env.StackHelp stk, AstRoot asts[]) { Frame X1 = stk.track(asts[1].exec(env)).getFrame(); Frame X2 = stk.track(asts[2].exec(env)).getFrame(); return new ValFrame(DMatrix.mmul(X1, X2)); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/matrix/AstTranspose.java
package water.rapids.ast.prims.matrix; import hex.DMatrix; import water.fvec.Frame; import water.rapids.Env; import water.rapids.vals.ValFrame; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; /** * Matrix transposition */ public class AstTranspose extends AstPrimitive { @Override public String[] args() { return new String[]{"ary"}; } @Override public int nargs() { return 1 + 1; } // (t X) @Override public String str() { return "t"; } @Override public ValFrame apply(Env env, Env.StackHelp stk, AstRoot asts[]) { Frame f = stk.track(asts[1].exec(env)).getFrame(); return new ValFrame(DMatrix.transpose(f)); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/misc/AstComma.java
package water.rapids.ast.prims.misc; import water.rapids.Env; import water.rapids.Val; import water.rapids.vals.ValNum; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; /** * Evaluate any number of expressions, returning the last one */ public class AstComma extends AstPrimitive { @Override public String[] args() { return new String[]{"..."}; } @Override public int nargs() { return -1; } // variable args @Override public String str() { return ","; } @Override public Val apply(Env env, Env.StackHelp stk, AstRoot asts[]) { Val val = new ValNum(0); for (int i = 1; i < asts.length; i++) val = stk.track(asts[i].exec(env)); // Evaluate all expressions for side-effects return val; // Return the last one } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/misc/AstLs.java
package water.rapids.ast.prims.misc; import water.Futures; import water.Key; import water.KeySnapshot; import water.fvec.AppendableVec; import water.fvec.Frame; import water.fvec.NewChunk; import water.fvec.Vec; import water.rapids.Env; import water.rapids.Val; import water.rapids.vals.ValFrame; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import java.util.ArrayList; /** * R 'ls' command. * <p/> * This method is purely for the console right now. Print stuff into the string buffer. * JSON response is not configured at all. */ public class AstLs extends AstPrimitive { @Override public String[] args() { return null; } @Override public int nargs() { return 1; } @Override public String str() { return "ls"; } @Override public ValFrame apply(Env env, Env.StackHelp stk, AstRoot asts[]) { ArrayList<String> domain = new ArrayList<>(); Futures fs = new Futures(); AppendableVec av = new AppendableVec(Vec.VectorGroup.VG_LEN1.addVec(), Vec.T_CAT); NewChunk keys = new NewChunk(av, 0); int r = 0; for (Key key : KeySnapshot.globalSnapshot().keys()) { keys.addCategorical(r++); domain.add(key.toString()); } String[] key_domain = domain.toArray(new String[domain.size()]); av.setDomain(key_domain); keys.close(fs); Vec c0 = av.layout_and_close(fs); // c0 is the row index vec fs.blockForPending(); return new ValFrame(new Frame(Key.<Frame>make("h2o_ls"), new String[]{"key"}, new Vec[]{c0})); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/misc/AstSetProperty.java
package water.rapids.ast.prims.misc; import water.MRTask; import water.rapids.Val; import water.rapids.ast.AstBuiltin; import water.rapids.vals.ValStr; import water.util.ArrayUtils; import water.util.Log; import water.util.StringUtils; /** * Internal operator that lets user set a given system property on all nodes of H2O cluster. * It is meant for debugging of running clusters and it is not meant to be directly exposed to users. */ public class AstSetProperty extends AstBuiltin<AstSetProperty> { @Override public String[] args() { return new String[]{"property", "value"}; } @Override public int nargs() { return 1 + 2; } // (setproperty property value) @Override public String str() { return "setproperty"; } @Override protected ValStr exec(Val[] args) { String property = args[1].getStr(); String value = args[2].getStr(); String debugMessage = setClusterProperty(property, value); return new ValStr(debugMessage); } private static String setClusterProperty(String property, String value) { String[] oldValues = new SetClusterPropertyTask(property, value).doAllNodes()._oldValues; return "Old values of " + property + " (per node): " + StringUtils.join(",", oldValues); } private static class SetClusterPropertyTask extends MRTask<SetClusterPropertyTask> { private String _property; private String _value; private String[] _oldValues; private SetClusterPropertyTask(String property, String value) { _property = property; _value = value; _oldValues = new String[0]; } @Override protected void setupLocal() { _oldValues = ArrayUtils.append(_oldValues, String.valueOf(System.getProperty(_property))); Log.info("Setting property: " + _property + "=" + _value); System.setProperty(_property, _value); } @Override public void reduce(SetClusterPropertyTask mrt) { _oldValues = ArrayUtils.append(_oldValues, mrt._oldValues); } } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/models/AstFairnessMetrics.java
package water.rapids.ast.prims.models; import hex.AUC2; import hex.Model; import org.apache.commons.math3.distribution.HypergeometricDistribution; import org.apache.commons.math3.stat.inference.GTest; import water.DKV; import water.Key; import water.MRTask; import water.exceptions.H2OIllegalArgumentException; import water.fvec.Chunk; import water.fvec.Frame; import water.fvec.Vec; import water.rapids.Env; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import water.rapids.vals.ValMapFrame; import water.util.ArrayUtils; import water.util.TwoDimTable; import java.lang.reflect.Field; import java.util.Arrays; import java.util.HashMap; import java.util.Map; import java.util.stream.IntStream; public class AstFairnessMetrics extends AstPrimitive { static public class FairnessMetrics { double tp; double fp; double tn; double fn; double total; double relativeSize; double accuracy; double precision; double f1; double tpr; double tnr; double fpr; double fnr; double auc; double aucpr; double gini; double selected; double selectedRatio; double logloss; public FairnessMetrics(double tp, double tn, double fp, double fn, double logLossSum, AUC2.AUCBuilder aucBuilder, double nrows) { this.tp = tp; this.tn = tn; this.fp = fp; this.fn = fn; total = tp + fp + tn + fn; logloss = logLossSum / total; relativeSize = total / nrows; accuracy = (tp + tn) / total; precision = tp / (fp + tp); f1 = (2 * tp) / (2 * tp + fp + fn); tpr = tp / (tp + fn); tnr = tn / (tn + fp); fpr = fp / (fp + tn); fnr = fn / (fn + tp); if (aucBuilder != null) { AUC2 auc2 = new AUC2(aucBuilder); auc = auc2._auc; aucpr = auc2._pr_auc; gini = auc2._gini; } else { auc = Double.NaN; aucpr = Double.NaN; gini = Double.NaN; } selected = tp + fp; selectedRatio = (tp + fp) / total; } } public static class FairnessMRTask extends MRTask { // Threshold to switch from Fisher's exact test to G-test. // Fisher's exact test gets slower with increasing size of the population // fortunately G-test approximates it very well for bigger population. // Recommendation is to use Fisher's test if the population size is lower than 1000 // (http://www.biostathandbook.com/small.html) but nowadays even 10000 is computed in the blink of an eye public static final int GTEST_THRESHOLD = 10000; // The magic constant REL1+1e-7 is taken from the R implementation of fisher.test in order to make // the results the same => easier to test public static final double FISHER_TEST_REL_ERROR = (1 + 1e-7); int[] protectedColsIdx; int[] cardinalities; int responseIdx; int predictionIdx; final int tpIdx = 0; final int tnIdx = 1; final int fpIdx = 2; final int fnIdx = 3; final int llsIdx = 4; // Log Loss Sum final int essentialMetrics = 5; final int maxIndex; final int favourableClass; int[] _results; AUC2.AUCBuilder[] _aucs; public FairnessMRTask(int[] protectedColsIdx, int[] cardinalities, int responseIdx, int predictionIdx, int favourableClass) { super(); this.protectedColsIdx = protectedColsIdx; this.cardinalities = cardinalities; this.responseIdx = responseIdx; this.predictionIdx = predictionIdx; this.favourableClass = favourableClass; double maxIndexDbl = Arrays.stream(cardinalities).asDoubleStream().reduce((a, b) -> a * b).getAsDouble(); if (maxIndexDbl > Integer.MAX_VALUE) throw new RuntimeException("Too many combinations of categories! Maximum number of category combinations is " + Integer.MAX_VALUE + "!"); this.maxIndex = (int) maxIndexDbl; } private int pColsToKey(Chunk[] cs, int row) { int[] indices = new int[protectedColsIdx.length]; for (int i = 0; i < protectedColsIdx.length; i++) { if (cs[protectedColsIdx[i]].isNA(row)) indices[i] = (cardinalities[i] - 1); else indices[i] += cs[protectedColsIdx[i]].at8(row); } return pColsToKey(indices); } public int pColsToKey(int[] indices) { int result = 0; int base = 1; for (int i = 0; i < protectedColsIdx.length; i++) { result += indices[i] * base; base *= cardinalities[i]; } return result; } private double[] keyToPCols(int value) { double[] result = new double[cardinalities.length]; for (int i = 0; i < cardinalities.length; i++) { final int tmp = value % cardinalities[i]; value /= cardinalities[i]; if (tmp == cardinalities[i] - 1) result[i] = Double.NaN; else result[i] = tmp; } return result; } protected String keyToString(int value, Frame fr) { double[] pcolIdx = keyToPCols(value); StringBuilder result = new StringBuilder(); for (int i = 0; i < protectedColsIdx.length; i++) { if (i > 0) result.append(","); if (Double.isFinite(pcolIdx[i])) { result.append(fr.vec(protectedColsIdx[i]).domain()[(int) pcolIdx[i]]); } else { result.append("NaN"); } } return result.toString().replaceAll("[^A-Za-z0-9,]", "_"); } @Override public void map(Chunk[] cs) { assert _results == null; _results = new int[maxIndex * essentialMetrics]; _aucs = new AUC2.AUCBuilder[maxIndex]; for (int i = 0; i < cs[0]._len; i++) { final int key = pColsToKey(cs, i); final long response = favourableClass == 1 ? cs[responseIdx].at8(i) : 1 - cs[responseIdx].at8(i); final long prediction = favourableClass == 1 ? cs[predictionIdx].at8(i) : 1 - cs[predictionIdx].at8(i); final double predictionProb = favourableClass == 1 ? cs[predictionIdx + 2].atd(i) : cs[predictionIdx + 1].atd(i); if (response == prediction) { if (response == 1) _results[essentialMetrics * key + tpIdx]++; else _results[essentialMetrics * key + tnIdx]++; } else { if (prediction == 1) _results[essentialMetrics * key + fpIdx]++; else _results[essentialMetrics * key + fnIdx]++; } _results[essentialMetrics * key + llsIdx] += -(response * Math.log(predictionProb) + (1 - response) * Math.log(1 - predictionProb)); if (_aucs[key] == null) _aucs[key] = new AUC2.AUCBuilder(400); _aucs[key].perRow(predictionProb, (int) response, 1); } } @Override public void reduce(MRTask mrt) { FairnessMRTask other = (FairnessMRTask) mrt; if (this._results == other._results) return; for (int i = 0; i < _results.length; i++) { _results[i] += other._results[i]; } for (int i = 0; i < maxIndex; i++) { if (_aucs[i] == null) _aucs[i] = other._aucs[i]; else if (other._aucs[i] != null) _aucs[i].reduce(other._aucs[i]); } } public Frame getMetrics(String[] protectedCols, Frame fr, Model model, String[] reference, final String frName) { // Calculate additional metrics FairnessMetrics[] results = new FairnessMetrics[maxIndex]; final long nrows = fr.numRows(); for (int i = 0; i < maxIndex; i++) { results[i] = new FairnessMetrics( _results[i * essentialMetrics + tpIdx], _results[i * essentialMetrics + tnIdx], _results[i * essentialMetrics + fpIdx], _results[i * essentialMetrics + fnIdx], _results[i * essentialMetrics + llsIdx], _aucs[i], nrows ); } // Get reference - If not specified, use the biggest group as a reference. int referenceIdx = 0; if (reference != null) { int[] indices = new int[protectedCols.length]; for (int i = 0; i < protectedCols.length; i++) { indices[i] = ArrayUtils.find(fr.vec(protectedCols[i]).domain(), reference[i]); } referenceIdx = pColsToKey(indices); } else { double max = 0; for (int key = 0; key < maxIndex; key++) { if (results[key].total > max) { max = results[key].total; referenceIdx = key; } } } int emptyResults = 0; for (FairnessMetrics fm : results) emptyResults += fm.total == 0 ? 1 : 0; // Fill in a frame final String[] skipAIR = new String[]{"total", "relativeSize"}; Field[] metrics = FairnessMetrics.class.getDeclaredFields(); final int protectedColsCnt = protectedCols.length; final int metricsCount = metrics.length + (metrics.length - skipAIR.length) + 1/*p-value*/; double[][] resultCols = new double[protectedColsCnt + metricsCount][results.length - emptyResults]; FairnessMetrics ref = results[referenceIdx]; int nonEmptyKey = 0; for (int key = 0; key < maxIndex; key++) { if (results[key].total == 0) continue; int counter = 0; double[] decodedKey = keyToPCols(key); for (int i = 0; i < protectedCols.length; i++) { resultCols[i][nonEmptyKey] = decodedKey[i]; } for (int i = 0; i < metrics.length; i++) { try { resultCols[protectedColsCnt + i][nonEmptyKey] = metrics[i].getDouble(results[key]); if (!ArrayUtils.contains(skipAIR, metrics[i].getName())) { final double air = metrics[i].getDouble(results[key]) / metrics[i].getDouble(ref); resultCols[protectedColsCnt + metrics.length + i - counter][nonEmptyKey] = air; } else counter++; } catch (IllegalAccessException e) { throw new RuntimeException(e); } } try { resultCols[resultCols.length - 1][nonEmptyKey] = getPValue(ref, results[key]); } catch (Exception e) { resultCols[resultCols.length - 1][nonEmptyKey] = Double.NaN; } nonEmptyKey++; } String[] colNames = new String[protectedColsCnt + metricsCount]; System.arraycopy(protectedCols, 0, colNames, 0, protectedCols.length); int counter = 0; for (int i = 0; i < metrics.length; i++) { colNames[protectedColsCnt + i] = metrics[i].getName(); if (!ArrayUtils.contains(skipAIR, metrics[i].getName())) { colNames[protectedColsCnt + metrics.length + i - counter] = "AIR_" + metrics[i].getName(); } else counter++; } colNames[colNames.length - 1] = "p.value"; Vec[] vecs = new Vec[protectedColsCnt + metricsCount]; for (int i = 0; i < protectedColsCnt; i++) { vecs[i] = Vec.makeVec(resultCols[i], fr.domains()[protectedColsIdx[i]], Vec.newKey()); } for (int i = 0; i < metricsCount; i++) { vecs[protectedColsCnt + i] = Vec.makeVec(resultCols[protectedColsCnt + i], Vec.newKey()); } return new Frame(Key.make("fairness_metrics_" + frName + "_for_model_" + model._key), colNames, vecs); } public Map<String, Frame> getROCInfo(Model model, Frame fr, final String frName) { Map<String, Frame> result = new HashMap<>(); for (int id = 0; id < maxIndex; id++) { if (_aucs[id] == null) continue; AUC2 auc = new AUC2(_aucs[id]); // Fill TwoDimTable String[] thresholds = new String[auc._nBins]; for (int i = 0; i < auc._nBins; i++) thresholds[i] = Double.toString(auc._ths[i]); AUC2.ThresholdCriterion crits[] = AUC2.ThresholdCriterion.VALUES; String[] colHeaders = new String[crits.length + 2]; String[] types = new String[crits.length + 2]; String[] formats = new String[crits.length + 2]; colHeaders[0] = "Threshold"; types[0] = "double"; formats[0] = "%f"; int i; for (i = 0; i < crits.length; i++) { colHeaders[i + 1] = crits[i].toString(); types[i + 1] = crits[i]._isInt ? "long" : "double"; formats[i + 1] = crits[i]._isInt ? "%d" : "%f"; } colHeaders[i + 1] = "idx"; types[i + 1] = "int"; formats[i + 1] = "%d"; TwoDimTable thresholdsByMetrics = new TwoDimTable("Metrics for Thresholds", "Binomial metrics as a function of classification thresholds", new String[auc._nBins], colHeaders, types, formats, null); for (i = 0; i < auc._nBins; i++) { int j = 0; thresholdsByMetrics.set(i, j, Double.valueOf(thresholds[i])); for (j = 0; j < crits.length; j++) { double d = crits[j].exec(auc, i); // Note: casts to Object are NOT redundant thresholdsByMetrics.set(i, 1 + j, crits[j]._isInt ? (Object) ((long) d) : d); } thresholdsByMetrics.set(i, 1 + j, i); } String groupName = keyToString(id, fr); Frame f = thresholdsByMetrics.asFrame(Key.make("thresholds_and_metrics_" + groupName + "_for_model_" + model._key + "_for_frame_" + frName)); DKV.put(f); result.put("thresholds_and_metrics_" + groupName, f); } return result; } /** * Calculate p-value using Fisher's exact test * <p> * | | Protected Group | Reference | * |--------------+-----------------+-----------| * | Selected | a | b | * | Not Selected | c | d | * * @param a * @param b * @param c * @param d * @return */ private static double fishersTest(long a, long b, long c, long d) { long popSize = a + b + c + d; if (popSize > Integer.MAX_VALUE) return Double.NaN; // Make sure we don't get stuck on p-value computation HypergeometricDistribution hgd = new HypergeometricDistribution((int) popSize, (int) (a + b), (int) (a + c)); double p = hgd.probability((int) a); double pValue = 0; // sum up pValues in all more extreme cases - like in R, sum all less probable cases in to the p-value for (int i = (int) Math.max(a - d, 0); i <= Math.min(a + b, a + c); i++) { final double proposal = hgd.probability(i); if (proposal <= p * FISHER_TEST_REL_ERROR) pValue += proposal; } return pValue; } /** * Calculate p-value. If there is a high number of instances use G-test as approximation of Fisher's exact test, * otherwise use Fisher's exact test. * * @param ref Metrics for the reference group * @param results Metrics for a protected group * @return p-value */ private static double getPValue(FairnessMetrics ref, FairnessMetrics results) { long a = (long) results.selected; long b = (long) ref.selected; long c = (long) (results.total - results.selected); long d = (long) (ref.total - ref.selected); if ((ref.total < GTEST_THRESHOLD && results.total < GTEST_THRESHOLD) || a == 0 || b == 0 || c == 0 || d == 0) { // fisher's exact test return fishersTest(a, b, c, d); } else { return new GTest().gTestDataSetsComparison( new long[]{a, c}, new long[]{b, d} ); } } } @Override public String[] args() { return new String[]{"model", "test_frame", "protected_columns", "reference", "favourable_class"}; } @Override public int nargs() { return 1 + 5; } @Override public String str() { return "fairnessMetrics"; } @Override public ValMapFrame apply(Env env, Env.StackHelp stk, AstRoot asts[]) { Model model = stk.track(asts[1].exec(env)).getModel(); Frame fr = stk.track((asts[2].exec(env)).getFrame()); String[] protectedCols = stk.track(asts[3].exec(env)).getStrs(); String[] reference = stk.track(asts[4].exec(env)).getStrs(); String favourableClass = stk.track(asts[5].exec(env)).getStr(); final String frameName = asts[2].str(); // Used only for naming the derived metrics final int responseIdx = fr.find(model._parms._response_column); if (!model._output.isBinomialClassifier()) { throw new H2OIllegalArgumentException("Model has to be a binomial model!"); } for (String pc : protectedCols) { if (fr.find(pc) == -1) throw new RuntimeException(pc + " was not found in the frame!"); if (!fr.vec(pc).isCategorical()) throw new H2OIllegalArgumentException(pc + " has to be a categorical column!"); } if (reference.length != protectedCols.length) reference = null; else for (int i = 0; i < protectedCols.length; i++) { if (!ArrayUtils.contains(fr.vec(protectedCols[i]).domain(), reference[i])) { throw new RuntimeException("Reference group is not present in the protected column"); } } if (!ArrayUtils.contains(fr.vec(responseIdx).domain(), favourableClass)) throw new RuntimeException("Favourable class is not present in the response!"); final int favorableClassId = ArrayUtils.find(fr.vec(responseIdx).domain(), favourableClass); final int[] protectedColsIdx = fr.find(protectedCols); final int[] cardinalities = IntStream.of(protectedColsIdx).map(colId -> fr.vec(colId).cardinality() + 1).toArray(); // +1 for missing value // Sanity check - the number of subgroups grows very quickly and higher values are practically unexplainable // but I don't want to limit the user too much if (Arrays.stream(cardinalities).asDoubleStream().reduce((a, b) -> a * b).orElse(Double.MAX_VALUE) > 1e6) throw new RuntimeException("Too many combinations of categories! Maximum number of category combinations is 1e6."); Frame predictions = new Frame(fr).add(model.score(fr)); DKV.put(predictions); try { FairnessMRTask fairnessMRTask = (FairnessMRTask) new FairnessMRTask( protectedColsIdx, cardinalities, responseIdx, fr.numCols(), favorableClassId ).doAll(predictions); Frame metrics = fairnessMRTask.getMetrics(protectedCols, fr, model, reference, frameName); Map<String, Frame> results = fairnessMRTask.getROCInfo(model, fr, frameName); DKV.put(metrics); results.put("overview", metrics); return new ValMapFrame(results); } finally { DKV.remove(predictions.getKey()); } } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/models/AstMakeLeaderboard.java
package water.rapids.ast.prims.models; import hex.Model; import hex.ModelContainer; import hex.leaderboard.*; import water.DKV; import water.Key; import water.fvec.Frame; import water.logging.Logger; import water.logging.LoggerFactory; import water.rapids.Env; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import water.rapids.vals.ValFrame; import water.util.Log; import java.util.Arrays; import java.util.stream.Stream; /** * Compute Leaderboard */ public class AstMakeLeaderboard extends AstPrimitive { @Override public String[] args() { return new String[]{"models", "leaderboardFrame", "sortMetric", "extensions", "scoringData"}; } @Override public int nargs() { return 1 + 5; } // (makeLeaderboard models leaderboardFrame sortMetric extensions projectName) @Override public String str() { return "makeLeaderboard"; } private static LeaderboardExtensionsProvider createLeaderboardExtensionProvider(Frame leaderboardFrame) { return new LeaderboardExtensionsProvider() { @Override public LeaderboardCell[] createExtensions(Model model) { return new LeaderboardCell[]{ new TrainingTime(model), new ScoringTimePerRow(model, leaderboardFrame), new AlgoName(model), }; } }; } @Override public ValFrame apply(Env env, Env.StackHelp stk, AstRoot[] asts) { Key[] models = Arrays.stream(stk.track(asts[1].exec(env)).getStrs()) .flatMap(model_id -> { Object obj = DKV.getGet(model_id); if (obj instanceof Model) { return Stream.of(Key.make(model_id)); } else if (obj instanceof ModelContainer) { ModelContainer mc = (ModelContainer) obj; return Stream.of(mc.getModelKeys()); } else { throw new RuntimeException("Unsupported model/grid id: " + model_id + "!"); } }).toArray(Key[]::new); // Get Frame clones Frame and the clone is key-less and since we are doing just read only ops here we can use the original frame String leaderboardFrameKey = stk.track(asts[2].exec(env)).getStr(); Frame leaderboardFrame = null; if (!leaderboardFrameKey.isEmpty()) leaderboardFrame = DKV.getGet(leaderboardFrameKey); else leaderboardFrameKey = null; String sortMetric = stk.track(asts[3].exec(env)).getStr(); String[] extensions = stk.track(asts[4].exec(env)).getStrs(); String scoringData = stk.track(asts[5].exec(env)).getStr().toLowerCase(); String projectName = leaderboardFrameKey + "_" + Arrays.deepHashCode(models) + "_" + scoringData; Arrays.stream(models).forEach(DKV::prefetch); if (sortMetric.equalsIgnoreCase("auto")) sortMetric = null; final boolean oneTrainingFrame = Arrays.stream(models).map(m -> ((Model) DKV.getGet(m))._parms._train).distinct().count() == 1; final boolean oneValidationFrame = Arrays.stream(models).map(m -> ((Model) DKV.getGet(m))._parms._valid).distinct().count() == 1; final boolean oneNFoldsSetting = Arrays.stream(models) .map(m -> ((Model) DKV.getGet(m))._parms) .filter(parms -> !parms.algoName().equalsIgnoreCase("stackedensemble")) .map(parameters -> parameters._nfolds) .distinct() .count() == 1; final boolean allCV = Arrays.stream(models).allMatch(m -> ((Model) DKV.getGet(m))._output._cross_validation_metrics != null); final boolean allHasValid = Arrays.stream(models).allMatch(m -> ((Model) DKV.getGet(m))._output._validation_metrics != null); boolean warnAboutTrain = false; boolean warnAboutValid = false; boolean warnAboutNFolds = false; boolean warnAboutLeaderboard = false; if (scoringData.equals("auto") && leaderboardFrame == null) { warnAboutTrain = true; warnAboutNFolds = true; scoringData = "xval"; } if (scoringData.equals("xval")) { warnAboutTrain = true; warnAboutNFolds = true; warnAboutLeaderboard = true; if (!allCV) scoringData = "valid"; } if (scoringData.equals("valid")) { warnAboutTrain = false; warnAboutValid = true; warnAboutLeaderboard = true; if (!allHasValid) scoringData = "train"; } if (scoringData.equals("train")) { warnAboutTrain = true; warnAboutValid = false; warnAboutLeaderboard = true; } // One training frame can be false positive if models from two different automls are used. if (warnAboutTrain && !oneTrainingFrame) Log.warn("More than one training frame was used amongst the models provided to the leaderboard."); if (warnAboutValid && !oneValidationFrame) Log.warn("More than one validation frame was used amongst the models provided to the leaderboard."); if (warnAboutNFolds && !oneNFoldsSetting) Log.warn("More than one n-folds settings are present."); //had to exclude SEs if (warnAboutLeaderboard && leaderboardFrame != null) Log.warn("Leaderboard frame present but scoring data are set to " + scoringData + ". Using scores from " + scoringData + "."); final Logger logger = LoggerFactory.getLogger(Leaderboard.class); Leaderboard ldb = Leaderboard.getOrMake(projectName, logger, leaderboardFrame, sortMetric, Leaderboard.ScoreData.valueOf(scoringData)); ldb.setExtensionsProvider(createLeaderboardExtensionProvider(leaderboardFrame)); ldb.addModels(models); ldb.ensureSorted(); Frame leaderboard = ldb.toTwoDimTable(extensions).asFrame(Key.make()); return new ValFrame(leaderboard); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/models/AstModelResetThreshold.java
package water.rapids.ast.prims.models; import hex.Model; import water.rapids.Env; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import water.rapids.vals.ValFrame; /** * Reset a model threshold and return the old one. */ public class AstModelResetThreshold extends AstPrimitive { @Override public String[] args() { return new String[]{"model", "threshold"}; } @Override public int nargs() { return 1 + 2; } // (+ Model + threshold) @Override public String str() { return "model.reset.threshold"; } @Override public ValFrame apply(Env env, Env.StackHelp stk, AstRoot asts[]) { Model model = stk.track(asts[1].exec(env)).getModel(); double oldThreshold = model._output.defaultThreshold(); double newThreshold = stk.track(asts[2].exec(env)).getNum(); model.resetThreshold(newThreshold); return ValFrame.fromRow(oldThreshold); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/models/AstPerfectAUC.java
package water.rapids.ast.prims.models; import hex.AUC2; import water.fvec.Frame; import water.fvec.Vec; import water.rapids.Env; import water.rapids.Val; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import water.rapids.vals.ValFrame; /** * Calculates a "perfect" (= not approximated) AUC */ public class AstPerfectAUC extends AstPrimitive { @Override public String[] args() { return new String[]{"probs", "acts"}; } @Override public int nargs() { return 1 + 2; } // (perfectAUC probs acts) @Override public String str() { return "perfectAUC"; } @Override public ValFrame apply(Env env, Env.StackHelp stk, AstRoot asts[]) { Vec probs = getSingleVec(stk.track(asts[1].exec(env)), "probabilities"); Vec acts = getSingleVec(stk.track(asts[2].exec(env)), "actuals"); double auc = AUC2.perfectAUC(probs, acts); return ValFrame.fromRow(auc); } private static Vec getSingleVec(Val v, String what) { Frame f = v.getFrame(); if (f == null || f.numCols() != 1) { throw new IllegalArgumentException("Expected a frame containing a single vector of " + what + ". Instead got " + String.valueOf(f)); } return f.vec(0); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/models/AstPermutationVarImp.java
package water.rapids.ast.prims.models; import hex.Model; import water.Key; import water.Scope; import water.fvec.Frame; import water.fvec.Vec; import water.rapids.Env; import water.rapids.PermutationVarImp; import water.rapids.Val; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import water.rapids.vals.ValFrame; import water.util.ArrayUtils; import water.util.TwoDimTable; import java.util.Arrays; import java.util.Locale; /** * Ast class for passing the model, frame and metric to calculate Permutation Variable Importance */ public class AstPermutationVarImp extends AstPrimitive { @Override public int nargs() { return 1 + 7; } @Override public String[] args() { return new String[]{"model", "frame", "metric", "n_samples", "n_repeats", "features", "seed"}; } @Override public String str() { return "PermutationVarImp"; } @Override public ValFrame apply(Env env, Env.StackHelp stk, AstRoot asts[]) { Model model = stk.track(asts[1].exec(env)).getModel(); Frame fr = stk.track(asts[2].exec(env)).getFrame(); String metric = stk.track(asts[3].exec(env)).getStr().toLowerCase(); long n_samples = (long) stk.track(asts[4].exec(env)).getNum(); int n_repeats = (int) stk.track(asts[5].exec(env)).getNum(); String[] features = null; Val featuresVal = stk.track(asts[6].exec(env)); if (!featuresVal.isEmpty()) // empty string list is interpreted as nums features = featuresVal.getStrs(); long seed = (long) stk.track(asts[7].exec(env)).getNum(); if (n_samples < -1 || n_samples == 0 || n_samples == 1 || n_samples > fr.numRows()) { throw new IllegalArgumentException("Argument n_samples has to be either -1 to use the whole frame " + "or greater than 2 and lower than or equal to the number of rows of the provided frame!"); } if (n_repeats < 1) { throw new IllegalArgumentException("Argument n_repeats must be greater than 0!"); } if (features != null) { String[] notInFrame = Arrays.stream(features).filter((f) -> !ArrayUtils.contains(fr.names(), f)).toArray(String[]::new); if (notInFrame.length > 0) { throw new IllegalArgumentException("Features " + String.join(", ", notInFrame) + " are not present in the provided frame!"); } String[] notUsedInModel = Arrays.stream(features).filter((f) -> !ArrayUtils.contains(model._output._origNames == null ? model._output._names : model._output._origNames, f)).toArray(String[]::new); if (notUsedInModel.length > 0) { throw new IllegalArgumentException("Features " + String.join(", ", notInFrame) + " weren't used for training!"); } } Scope.enter(); Frame pviFr = null; try { // Calculate Permutation Variable Importance PermutationVarImp pvi = new PermutationVarImp(model, fr); TwoDimTable varImpTable = null; if (n_repeats > 1) { varImpTable = pvi.getRepeatedPermutationVarImp(metric, n_samples, n_repeats, features, seed); } else { varImpTable = pvi.getPermutationVarImp(metric, n_samples, features, seed); } // Create Frame from TwoDimTable pviFr = varimpToFrame(varImpTable, Key.make(model._key + "permutationVarImp")); Scope.track(pviFr); } finally { Key[] keysToKeep = pviFr != null ? pviFr.keys() : new Key[]{}; Scope.exit(keysToKeep); } return new ValFrame(pviFr); } private static Frame varimpToFrame(TwoDimTable twoDimTable, Key frameKey) { String[] colNames = new String[twoDimTable.getColDim() + 1]; colNames[0] = "Variable"; System.arraycopy(twoDimTable.getColHeaders(),0, colNames, 1, twoDimTable.getColDim()); Vec[] vecs = new Vec[colNames.length]; vecs[0] = Vec.makeVec(twoDimTable.getRowHeaders(), Vec.newKey()); double[] tmpRow = new double[twoDimTable.getRowDim()]; for (int j = 0; j < twoDimTable.getColDim(); j++) { for (int i = 0; i < twoDimTable.getRowDim(); i++) { tmpRow[i] = (double) twoDimTable.get(i, j); } vecs[j + 1] = Vec.makeVec(tmpRow, Vec.newKey()); } Frame fr = new Frame(frameKey, colNames, vecs); return fr; } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/models/AstResultFrame.java
package water.rapids.ast.prims.models; import hex.Model; import water.rapids.Env; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import water.rapids.vals.ValFrame; public class AstResultFrame extends AstPrimitive { @Override public String[] args() { return new String[]{"model key"}; } @Override public int nargs() { return 1 + 1; } // (segment_models_as_frame segment_models_id) @Override public String str() { return "result"; } @Override public ValFrame apply(Env env, Env.StackHelp stk, AstRoot asts[]) { Model model = stk.track(asts[1].exec(env)).getModel(); return new ValFrame(model.result()); } }
0
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims
java-sources/ai/h2o/h2o-core/3.46.0.7/water/rapids/ast/prims/models/AstSegmentModelsAsFrame.java
package water.rapids.ast.prims.models; import hex.segments.SegmentModels; import water.rapids.Env; import water.rapids.ast.AstPrimitive; import water.rapids.ast.AstRoot; import water.rapids.vals.ValFrame; public class AstSegmentModelsAsFrame extends AstPrimitive { @Override public String[] args() { return new String[]{"segment_models"}; } @Override public int nargs() { return 1 + 1; } // (segment_models_as_frame segment_models_id) @Override public String str() { return "segment_models_as_frame"; } @Override public ValFrame apply(Env env, Env.StackHelp stk, AstRoot asts[]) { SegmentModels models = (SegmentModels) stk.track(asts[1].exec(env)).getKeyed(); return new ValFrame(models.toFrame()); } }