Examples with Chunk water.fvec.Chunk used on opensource projects

Example 16 with Chunk

use of water.fvec.Chunk in project h2o-3 by h2oai.

the class AstBinOp method scalar_op_frame.

/**
   * Auto-widen the scalar to every element of the frame
   */
private ValFrame scalar_op_frame(final String str, Frame fr) {
    Frame res = new MRTask() {

        @Override
        public void map(Chunk[] chks, NewChunk[] cress) {
            BufferedString vstr = new BufferedString();
            for (int c = 0; c < chks.length; c++) {
                Chunk chk = chks[c];
                NewChunk cres = cress[c];
                Vec vec = chk.vec();
                // String Vectors: apply str_op as BufferedStrings to all elements
                if (vec.isString()) {
                    final BufferedString conStr = new BufferedString(str);
                    for (int i = 0; i < chk._len; i++) cres.addNum(str_op(conStr, chk.atStr(vstr, i)));
                } else if (vec.isCategorical()) {
                    // categorical Vectors: convert string to domain value; apply op (not
                    // str_op).  Not sure what the "right" behavior here is, can
                    // easily argue that should instead apply str_op to the categorical
                    // string domain value - except that this whole operation only
                    // makes sense for EQ/NE, and is much faster when just comparing
                    // doubles vs comparing strings.
                    final double d = (double) ArrayUtils.find(vec.domain(), str);
                    for (int i = 0; i < chk._len; i++) cres.addNum(op(d, chk.atd(i)));
                } else {
                    // mixing string and numeric
                    // false or true only
                    final double d = op(1, 2);
                    for (int i = 0; i < chk._len; i++) cres.addNum(d);
                }
            }
        }
    }.doAll(fr.numCols(), Vec.T_NUM, fr).outputFrame(fr._names, null);
    return new ValFrame(res);
}

Example 17 with Chunk

use of water.fvec.Chunk in project h2o-3 by h2oai.

the class AstRepLen method apply.

@Override
public ValFrame apply(Env env, Env.StackHelp stk, AstRoot[] asts) {
    Val v = asts[1].exec(env);
    long length = (long) asts[2].exec(env).getNum();
    Frame ff;
    if (v instanceof ValFrame)
        ff = stk.track(v).getFrame();
    else
        return new ValFrame(new Frame(Vec.makeCon(v.getNum(), length)));
    final Frame fr = ff;
    if (fr.numCols() == 1) {
        Vec vec = Vec.makeRepSeq(length, fr.numRows());
        new MRTask() {

            @Override
            public void map(Chunk c) {
                for (int i = 0; i < c._len; ++i) c.set(i, fr.anyVec().at((long) c.atd(i)));
            }
        }.doAll(vec);
        vec.setDomain(fr.anyVec().domain());
        return new ValFrame(new Frame(vec));
    } else {
        Frame f = new Frame();
        for (int i = 0; i < length; ++i) f.add(Frame.defaultColName(f.numCols()), fr.vec(i % fr.numCols()));
        return new ValFrame(f);
    }
}

Example 18 with Chunk

use of water.fvec.Chunk in project h2o-3 by h2oai.

the class AstAsDate method apply.

@Override
public ValFrame apply(Env env, Env.StackHelp stk, AstRoot[] asts) {
    Frame fr = stk.track(asts[1].exec(env)).getFrame();
    Vec vec = fr.vecs()[0];
    if (fr.vecs().length != 1 || !(vec.isCategorical() || vec.isString()))
        throw new IllegalArgumentException("as.Date requires a single column of factors or strings");
    final String format = asts[2].exec(env).getStr();
    if (format.isEmpty())
        throw new IllegalArgumentException("as.Date requires a non-empty format string");
    // check the format string more?
    final String[] dom = vec.domain();
    final boolean isStr = dom == null && vec.isString();
    assert isStr || dom != null : "as.Date error: domain is null, but vec is not String";
    Frame fr2 = new MRTask() {

        private transient DateTimeFormatter _fmt;

        @Override
        public void setupLocal() {
            _fmt = ParseTime.forStrptimePattern(format).withZone(ParseTime.getTimezone());
        }

        @Override
        public void map(Chunk c, NewChunk nc) {
            //done on each node in lieu of rewriting DateTimeFormatter as Iced
            String date;
            BufferedString tmpStr = new BufferedString();
            for (int i = 0; i < c._len; ++i) {
                if (!c.isNA(i)) {
                    if (isStr)
                        date = c.atStr(tmpStr, i).toString();
                    else
                        date = dom[(int) c.at8(i)];
                    nc.addNum(DateTime.parse(date, _fmt).getMillis(), 0);
                } else
                    nc.addNA();
            }
        }
    }.doAll(1, Vec.T_NUM, fr).outputFrame(fr._names, null);
    return new ValFrame(fr2);
}

Example 19 with Chunk

use of water.fvec.Chunk in project h2o-3 by h2oai.

the class AstBinOp method vec_op_frame.

private ValFrame vec_op_frame(Vec vec, Frame fr) {
    // Already checked for same rows, non-zero frame
    Frame rt = new Frame(fr);
    rt.add("", vec);
    Frame res = new MRTask() {

        @Override
        public void map(Chunk[] chks, NewChunk[] cress) {
            assert cress.length == chks.length - 1;
            Chunk clf = chks[cress.length];
            for (int c = 0; c < cress.length; c++) {
                Chunk crt = chks[c];
                NewChunk cres = cress[c];
                for (int i = 0; i < clf._len; i++) cres.addNum(op(clf.atd(i), crt.atd(i)));
            }
        }
    }.doAll(fr.numCols(), Vec.T_NUM, rt).outputFrame(fr._names, null);
    // Cleanup categorical misuse
    return cleanCategorical(fr, res);
}

Example 20 with Chunk

use of water.fvec.Chunk in project h2o-3 by h2oai.

the class AstBinOp method frame_op_row.

private ValFrame frame_op_row(Frame lf, Frame row) {
    final double[] rawRow = new double[row.numCols()];
    for (int i = 0; i < rawRow.length; ++i) // is numberlike, if not then NaN
    rawRow[i] = row.vec(i).isNumeric() || row.vec(i).isTime() ? row.vec(i).at(0) : Double.NaN;
    Frame res = new MRTask() {

        @Override
        public void map(Chunk[] chks, NewChunk[] cress) {
            for (int c = 0; c < cress.length; c++) {
                Chunk clf = chks[c];
                NewChunk cres = cress[c];
                for (int r = 0; r < clf._len; ++r) {
                    if (clf.vec().isString())
                        // TODO: improve
                        cres.addNum(Double.NaN);
                    else
                        cres.addNum(op(clf.atd(r), rawRow[c]));
                }
            }
        }
    }.doAll(lf.numCols(), Vec.T_NUM, lf).outputFrame(lf._names, null);
    return cleanCategorical(lf, res);
}