diff --git a/com.oracle.truffle.r.nodes.builtin/src/com/oracle/truffle/r/nodes/builtin/base/Split.java b/com.oracle.truffle.r.nodes.builtin/src/com/oracle/truffle/r/nodes/builtin/base/Split.java
index 8a5446c3d6af500002450e0b5686860bd2c4f872..f309975878a3af3f16c66c8e2a099601b24bb498 100644
--- a/com.oracle.truffle.r.nodes.builtin/src/com/oracle/truffle/r/nodes/builtin/base/Split.java
+++ b/com.oracle.truffle.r.nodes.builtin/src/com/oracle/truffle/r/nodes/builtin/base/Split.java
@@ -35,9 +35,11 @@ import com.oracle.truffle.r.runtime.Utils;
import com.oracle.truffle.r.runtime.builtins.RBuiltin;
import com.oracle.truffle.r.runtime.data.RDataFactory;
import com.oracle.truffle.r.runtime.data.RList;
+import com.oracle.truffle.r.runtime.data.RRawVector;
import com.oracle.truffle.r.runtime.data.RStringVector;
import com.oracle.truffle.r.runtime.data.model.RAbstractDoubleVector;
import com.oracle.truffle.r.runtime.data.model.RAbstractIntVector;
+import com.oracle.truffle.r.runtime.data.model.RAbstractListVector;
import com.oracle.truffle.r.runtime.data.model.RAbstractLogicalVector;
import com.oracle.truffle.r.runtime.data.model.RAbstractStringVector;
@@ -47,6 +49,9 @@ import com.oracle.truffle.r.runtime.data.model.RAbstractStringVector;
*
* TODO Can we find a way to efficiently write the specializations as generics? The code is
* identical except for the argument type.
+ *
+ * TODO: GNU R preserves the corresponding values of names attribute. There are (ignored) tests for
+ * this in TestBuiltin_split.
*/
@RBuiltin(name = "split", kind = INTERNAL, parameterNames = {"x", "f"}, behavior = PURE)
public abstract class Split extends RBuiltinNode.Arg2 {
@@ -62,9 +67,37 @@ public abstract class Split extends RBuiltinNode.Arg2 {
Casts.noCasts(Split.class);
}
- public static class SplitTemplate {
- @SuppressWarnings("unused") private int[] collectResultsSize;
- @SuppressWarnings("unused") private int nLevels;
+ @Specialization
+ protected RList split(RAbstractListVector x, RAbstractIntVector f) {
+ int[] factor = f.materialize().getDataWithoutCopying();
+ RStringVector names = getLevelNode.execute(f);
+ final int nLevels = getNLevels(names);
+
+ // initialise result arrays
+ Object[][] collectResults = new Object[nLevels][];
+ int[] collectResultSize = new int[nLevels];
+ for (int i = 0; i < collectResults.length; i++) {
+ collectResults[i] = new Object[INITIAL_SIZE];
+ }
+
+ // perform split
+ for (int i = 0, fi = 0; i < x.getLength(); ++i, fi = Utils.incMod(fi, factor.length)) {
+ int resultIndex = factor[fi] - 1; // a factor is a 1-based int vector
+ Object[] collect = collectResults[resultIndex];
+ if (collect.length == collectResultSize[resultIndex]) {
+ collectResults[resultIndex] = Arrays.copyOf(collect, collect.length * SCALE_FACTOR);
+ collect = collectResults[resultIndex];
+ }
+ collect[collectResultSize[resultIndex]++] = x.getDataAt(i);
+ }
+
+ // assemble result vectors and level names
+ Object[] results = new Object[nLevels];
+ for (int i = 0; i < nLevels; i++) {
+ results[i] = RDataFactory.createList(Arrays.copyOfRange(collectResults[i], 0, collectResultSize[i]));
+ }
+
+ return RDataFactory.createList(results, names);
}
@Specialization
@@ -199,6 +232,39 @@ public abstract class Split extends RBuiltinNode.Arg2 {
return RDataFactory.createList(results, names);
}
+ @Specialization
+ protected RList split(RRawVector x, RAbstractIntVector f) {
+ int[] factor = f.materialize().getDataWithoutCopying();
+ RStringVector names = getLevelNode.execute(f);
+ final int nLevels = getNLevels(names);
+
+ // initialise result arrays
+ byte[][] collectResults = new byte[nLevels][];
+ int[] collectResultSize = new int[nLevels];
+ for (int i = 0; i < collectResults.length; i++) {
+ collectResults[i] = new byte[INITIAL_SIZE];
+ }
+
+ // perform split
+ for (int i = 0, fi = 0; i < x.getLength(); ++i, fi = Utils.incMod(fi, factor.length)) {
+ int resultIndex = factor[fi] - 1; // a factor is a 1-based int vector
+ byte[] collect = collectResults[resultIndex];
+ if (collect.length == collectResultSize[resultIndex]) {
+ collectResults[resultIndex] = Arrays.copyOf(collect, collect.length * SCALE_FACTOR);
+ collect = collectResults[resultIndex];
+ }
+ collect[collectResultSize[resultIndex]++] = x.getDataAt(i).getValue();
+ }
+
+ // assemble result vectors and level names
+ Object[] results = new Object[nLevels];
+ for (int i = 0; i < nLevels; i++) {
+ results[i] = RDataFactory.createRawVector(Arrays.copyOfRange(collectResults[i], 0, collectResultSize[i]));
+ }
+
+ return RDataFactory.createList(results, names);
+ }
+
private static int getNLevels(RStringVector levels) {
return levels != null ? levels.getLength() : 0;
}
diff --git a/com.oracle.truffle.r.test/src/com/oracle/truffle/r/test/ExpectedTestOutput.test b/com.oracle.truffle.r.test/src/com/oracle/truffle/r/test/ExpectedTestOutput.test
index c03ce21ab0cb9070e7d00e468c82274b27f1c870..982b647f335990c04417558dc2d95376e77ca2bb 100644
--- a/com.oracle.truffle.r.test/src/com/oracle/truffle/r/test/ExpectedTestOutput.test
+++ b/com.oracle.truffle.r.test/src/com/oracle/truffle/r/test/ExpectedTestOutput.test
@@ -64273,6 +64273,14 @@ $`2`
[1] 2 4 6 8 10
+##com.oracle.truffle.r.test.builtins.TestBuiltin_split.testSplit#
+#{ split(as.raw(1:10), as.factor(c('a', 'b', 'a')); }
+Error: unexpected ';' in "{ split(as.raw(1:10), as.factor(c('a', 'b', 'a'));"
+
+##com.oracle.truffle.r.test.builtins.TestBuiltin_split.testSplit#
+#{ split(list(1, 2L, 'x', T), as.factor(c('a', 'b', 'a')); }
+Error: unexpected ';' in "{ split(list(1, 2L, 'x', T), as.factor(c('a', 'b', 'a'));"
+
##com.oracle.truffle.r.test.builtins.TestBuiltin_split.testSplit#
#{ x <- factor(c("a", "b", "a")); attr(x, "levels")<-c(7L, 42L) ; split(1:3, x) }
$`7`
@@ -64282,6 +64290,24 @@ $`42`
[1] 2
+##com.oracle.truffle.r.test.builtins.TestBuiltin_split.testSplitWithNames#Ignored.Unimplemented#
+#{ split(list(q=1, w=2L, e='x', r=T), as.factor(c('a', 'b', 'a')); }
+Error: unexpected ';' in "{ split(list(q=1, w=2L, e='x', r=T), as.factor(c('a', 'b', 'a'));"
+
+##com.oracle.truffle.r.test.builtins.TestBuiltin_split.testSplitWithNames#Ignored.Unimplemented#
+#{ tmp <- c(1,2,3); names(tmp) <- c('x','y','z'); split(tmp, as.factor(c('a','b'))); }
+$a
+x z
+1 3
+
+$b
+y
+2
+
+Warning message:
+In split.default(tmp, as.factor(c("a", "b"))) :
+ data length is not a multiple of split variable
+
##com.oracle.truffle.r.test.builtins.TestBuiltin_split.testsplit1#
#argv <- list(1:6, structure(1:2, .Label = c('1', '2'), class = 'factor')); .Internal(split(argv[[1]], argv[[2]]))
$`1`
diff --git a/com.oracle.truffle.r.test/src/com/oracle/truffle/r/test/builtins/TestBuiltin_split.java b/com.oracle.truffle.r.test/src/com/oracle/truffle/r/test/builtins/TestBuiltin_split.java
index 234bd45d0266926805884872852794e57ed378de..d551d2225e677fb93c52e60517b814c17b2d78ff 100644
--- a/com.oracle.truffle.r.test/src/com/oracle/truffle/r/test/builtins/TestBuiltin_split.java
+++ b/com.oracle.truffle.r.test/src/com/oracle/truffle/r/test/builtins/TestBuiltin_split.java
@@ -116,5 +116,13 @@ public class TestBuiltin_split extends TestBase {
assertEval("{ fu <- c(\"a\",\"b\") ; split(1:8,fu) }");
assertEval("{ g <- factor(round(c(0.4,1.3,0.6,1.8,2.5,4.1,2.2,1.0))) ; x <- c(0.1,3.2,1,0.6,1.9,3.3,1.6,1.7) + sqrt(as.numeric(g)) ; xg <- split(x, g) ; xg }");
assertEval("{ x <- factor(c(\"a\", \"b\", \"a\")); attr(x, \"levels\")<-c(7L, 42L) ; split(1:3, x) }");
+ assertEval("{ split(list(1, 2L, 'x', T), as.factor(c('a', 'b', 'a')); }");
+ assertEval("{ split(as.raw(1:10), as.factor(c('a', 'b', 'a')); }");
+ }
+
+ @Test
+ public void testSplitWithNames() {
+ assertEval(Ignored.Unimplemented, "{ split(list(q=1, w=2L, e='x', r=T), as.factor(c('a', 'b', 'a')); }");
+ assertEval(Ignored.Unimplemented, "{ tmp <- c(1,2,3); names(tmp) <- c('x','y','z'); split(tmp, as.factor(c('a','b'))); }");
}
}