Github user kiszk commented on a diff in the pull request:
https://github.com/apache/spark/pull/13909#discussion_r93814845
--- Diff:
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/complexTypeCreator.scala
---
@@ -58,77 +58,126 @@ case class CreateArray(children: Seq[Expression])
extends Expression {
}
override def doGenCode(ctx: CodegenContext, ev: ExprCode): ExprCode = {
- val array = ctx.freshName("array")
-
val et = dataType.elementType
val evals = children.map(e => e.genCode(ctx))
val isPrimitiveArray = ctx.isPrimitiveType(et)
- val (preprocess, arrayData) =
- GenArrayData.getCodeArrayData(ctx, et, children.size,
isPrimitiveArray, array)
-
- val assigns = if (isPrimitiveArray) {
- val primitiveTypeName = ctx.primitiveTypeName(et)
- evals.zipWithIndex.map { case (eval, i) =>
- eval.code + s"""
- if (${eval.isNull}) {
- $arrayData.setNullAt($i);
- } else {
- $arrayData.set$primitiveTypeName($i, ${eval.value});
- }
- """
- }
- } else {
- evals.zipWithIndex.map { case (eval, i) =>
- eval.code + s"""
- if (${eval.isNull}) {
- $array[$i] = null;
- } else {
- $array[$i] = ${eval.value};
- }
- """
- }
- }
- ev.copy(code =
- preprocess +
- ctx.splitExpressions(ctx.INPUT_ROW, assigns) +
- s"\nfinal ArrayData ${ev.value} = $arrayData;\n",
+ val (preprocess, postprocess, arrayData, array) =
+ GenArrayData.genCodeToCreateArrayData(ctx, et, children.size,
isPrimitiveArray)
+ val assigns = GenArrayData.genCodeToAssignArrayElements(
+ ctx, evals, et, isPrimitiveArray, arrayData, array, true)
+ /*
+ TODO: When we generate simpler code, we have to solve the following
exception
+ https://github.com/apache/spark/pull/13909/files#r93813725
+ ev.copy(
+ code = preprocess + ctx.splitExpressions(ctx.INPUT_ROW, assigns) +
postprocess
+ value = arrayData,
--- End diff --
Yes, it can avoid the exception. However, it may introduce a new
performance issue. At lines 82 and 86, two castings happens per loop iteration.
I would like to avoid these frequent castings.
```
/* 070 */ if (((ArrayData) arrayData1) instanceof UnsafeArrayData) {
/* 071 */
/* 072 */ final int sizeInBytes1 = ((UnsafeArrayData) ((ArrayData)
arrayData1)).getSizeInBytes();
/* 073 */ // grow the global buffer before writing data.
/* 074 */ holder.grow(sizeInBytes1);
/* 075 */ ((UnsafeArrayData) ((ArrayData)
arrayData1)).writeToMemory(holder.buffer, holder.cursor);
/* 076 */ holder.cursor += sizeInBytes1;
/* 077 */
/* 078 */ } else {
/* 079 */ final int numElements1 = ((ArrayData)
arrayData1).numElements();
/* 080 */ arrayWriter1.initialize(holder, numElements1, 8);
/* 081 */
/* 082 */ for (int index1 = 0; index1 < numElements1; index1++) {
/* 083 */ if (((ArrayData) arrayData1).isNullAt(index1)) {
/* 084 */ arrayWriter1.setNullLong(index1);
/* 085 */ } else {
/* 086 */ final long element1 = ((ArrayData)
arrayData1).getLong(index1);
/* 087 */ arrayWriter1.write(index1, element1);
/* 088 */ }
/* 089 */ }
/* 090 */ }
```
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
