This is an automated email from the ASF dual-hosted git repository.
felixybw pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/incubator-gluten.git
The following commit(s) were added to refs/heads/main by this push:
new 3888f45e0 [GLUTEN-5745][VL] Add more comments for GenerateRel
conversion logic (#5746)
3888f45e0 is described below
commit 3888f45e060c4d8121c38190621282780b77a032
Author: James Xu <[email protected]>
AuthorDate: Wed May 15 02:20:43 2024 +0800
[GLUTEN-5745][VL] Add more comments for GenerateRel conversion logic (#5746)
Add more comments for GenerateRel conversion logic
---
cpp/velox/substrait/SubstraitToVeloxPlan.cc | 19 ++++++++++++++-----
1 file changed, 14 insertions(+), 5 deletions(-)
diff --git a/cpp/velox/substrait/SubstraitToVeloxPlan.cc
b/cpp/velox/substrait/SubstraitToVeloxPlan.cc
index 6ee7ad827..366ab5abd 100644
--- a/cpp/velox/substrait/SubstraitToVeloxPlan.cc
+++ b/cpp/velox/substrait/SubstraitToVeloxPlan.cc
@@ -733,8 +733,14 @@ core::PlanNodePtr
SubstraitToVeloxPlanConverter::toVeloxPlan(const ::substrait::
auto projNode = std::dynamic_pointer_cast<const
core::ProjectNode>(childNode);
if (projNode != nullptr && projNode->names().size() >
requiredChildOutput.size()) {
- // generator is a scalarfunction node -> explode(array(col, 'all'))
- // use the last one, this is ensure by scala code
+ // Generator function's input is not a field reference, e.g.
explode(array(1,2,3)), a sample
+ // input substrait plan is like the following(the plan structure is
ensured by scala code):
+ //
+ // Generate explode([1,2,3] AS _pre_0#129), false, [col#126]
+ // +- Project [fake_column#128, [1,2,3] AS _pre_0#129]
+ // +- RewrittenNodeWall Scan OneRowRelation[fake_column#128]
+ //
+ // The last projection column in GeneratorRel's child(Project) is the
column we need to unnest
auto innerName = projNode->names().back();
auto innerExpr = projNode->projections().back();
@@ -743,9 +749,12 @@ core::PlanNodePtr
SubstraitToVeloxPlanConverter::toVeloxPlan(const ::substrait::
VELOX_CHECK_NOT_NULL(unnestFieldExpr, " the key in unnest Operator only
support field");
unnest.emplace_back(unnestFieldExpr);
} else {
- // generator should be a array column -> explode(col)
- auto explodeFunc = generator.scalar_function();
- auto unnestExpr =
exprConverter_->toVeloxExpr(explodeFunc.arguments(0).value(), inputType);
+ // Generator function's input is a field reference, e.g. explode(col),
generator
+ // function's first argument is the field reference we need to unnest.
+ // This assumption holds for all the supported generator function:
+ // explode, posexplode, inline.
+ auto generatorFunc = generator.scalar_function();
+ auto unnestExpr =
exprConverter_->toVeloxExpr(generatorFunc.arguments(0).value(), inputType);
auto unnestFieldExpr = std::dynamic_pointer_cast<const
core::FieldAccessTypedExpr>(unnestExpr);
VELOX_CHECK_NOT_NULL(unnestFieldExpr, " the key in unnest Operator only
support field");
unnest.emplace_back(unnestFieldExpr);
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
