This is an automated email from the ASF dual-hosted git repository.
lihaopeng pushed a commit to branch branch-3.0
in repository https://gitbox.apache.org/repos/asf/doris.git
The following commit(s) were added to refs/heads/branch-3.0 by this push:
new a6a008a5491 [opt](in expr) Optimize the IN expression by skipping
constant column… (#40917)
a6a008a5491 is described below
commit a6a008a5491d619b79c1147e850e11a865000594
Author: Mryange <[email protected]>
AuthorDate: Wed Sep 18 19:42:05 2024 +0800
[opt](in expr) Optimize the IN expression by skipping constant column…
(#40917)
…s. (#39912)
https://github.com/apache/doris/pull/39912
Optimize the IN expression by skipping constant columns
---
be/src/vec/exprs/vin_predicate.cpp | 19 ++++++++++++++++---
be/src/vec/exprs/vin_predicate.h | 3 +++
2 files changed, 19 insertions(+), 3 deletions(-)
diff --git a/be/src/vec/exprs/vin_predicate.cpp
b/be/src/vec/exprs/vin_predicate.cpp
index 1411254a2ca..e4a4969b00e 100644
--- a/be/src/vec/exprs/vin_predicate.cpp
+++ b/be/src/vec/exprs/vin_predicate.cpp
@@ -93,10 +93,24 @@ Status VInPredicate::open(RuntimeState* state,
VExprContext* context,
if (scope == FunctionContext::FRAGMENT_LOCAL) {
RETURN_IF_ERROR(VExpr::get_const_col(context, nullptr));
}
+
+ _is_args_all_constant = std::all_of(_children.begin() + 1, _children.end(),
+ [](const VExprSPtr& expr) { return
expr->is_constant(); });
_open_finished = true;
return Status::OK();
}
+size_t VInPredicate::skip_constant_args_size() const {
+ if (_is_args_all_constant && !_can_fast_execute) {
+ // This is an optimization. For expressions like colA IN (1, 2, 3, 4),
+ // where all values inside the IN clause are constants,
+ // a hash set is created during open, and it will not be accessed
again during execute
+ // Here, _children[0] is colA
+ return 1;
+ }
+ return _children.size();
+}
+
void VInPredicate::close(VExprContext* context,
FunctionContext::FunctionStateScope scope) {
VExpr::close_function_context(context, scope, _function);
VExpr::close(context, scope);
@@ -115,9 +129,8 @@ Status VInPredicate::execute(VExprContext* context, Block*
block, int* result_co
return Status::OK();
}
DCHECK(_open_finished || _getting_const_col);
- // TODO: not execute const expr again, but use the const column in
function context
- doris::vectorized::ColumnNumbers arguments(_children.size());
- for (int i = 0; i < _children.size(); ++i) {
+ doris::vectorized::ColumnNumbers arguments(skip_constant_args_size());
+ for (int i = 0; i < skip_constant_args_size(); ++i) {
int column_id = -1;
RETURN_IF_ERROR(_children[i]->execute(context, block, &column_id));
arguments[i] = column_id;
diff --git a/be/src/vec/exprs/vin_predicate.h b/be/src/vec/exprs/vin_predicate.h
index 4d227510b91..1b640056284 100644
--- a/be/src/vec/exprs/vin_predicate.h
+++ b/be/src/vec/exprs/vin_predicate.h
@@ -51,6 +51,8 @@ public:
std::string debug_string() const override;
+ size_t skip_constant_args_size() const;
+
const FunctionBasePtr function() { return _function; }
bool is_not_in() const { return _is_not_in; };
@@ -62,5 +64,6 @@ private:
const bool _is_not_in;
static const constexpr char* function_name = "in";
+ bool _is_args_all_constant = false;
};
} // namespace doris::vectorized
\ No newline at end of file
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
