This is an automated email from the ASF dual-hosted git repository.
yiguolei pushed a commit to branch branch-4.0
in repository https://gitbox.apache.org/repos/asf/doris.git
The following commit(s) were added to refs/heads/branch-4.0 by this push:
new 98fbe2bf2d4 branch-4.0: [fix](predicate) Avoid recomputing predicates
#60484 (#60514)
98fbe2bf2d4 is described below
commit 98fbe2bf2d445628699f948738166af927d167e7
Author: github-actions[bot]
<41898282+github-actions[bot]@users.noreply.github.com>
AuthorDate: Thu Feb 5 14:53:10 2026 +0800
branch-4.0: [fix](predicate) Avoid recomputing predicates #60484 (#60514)
Cherry-picked from #60484
Co-authored-by: Gabriel <[email protected]>
---
be/src/olap/column_predicate.h | 2 ++
be/src/olap/comparison_predicate.h | 6 ++++++
be/src/olap/in_list_predicate.h | 7 +++++++
be/src/olap/null_predicate.h | 1 +
be/src/pipeline/exec/olap_scan_operator.cpp | 23 ++++++++++++++++++++++-
5 files changed, 38 insertions(+), 1 deletion(-)
diff --git a/be/src/olap/column_predicate.h b/be/src/olap/column_predicate.h
index 9e6cf291a26..c4a5736beb0 100644
--- a/be/src/olap/column_predicate.h
+++ b/be/src/olap/column_predicate.h
@@ -225,6 +225,8 @@ public:
}
virtual double get_ignore_threshold() const { return 0; }
+ // If this predicate acts on the key column, this predicate should be
erased.
+ virtual bool could_be_erased() const { return false; }
// Return the size of value set for IN/NOT IN predicates and 0 for others.
virtual std::string debug_string() const {
fmt::memory_buffer debug_string_buffer;
diff --git a/be/src/olap/comparison_predicate.h
b/be/src/olap/comparison_predicate.h
index 7b86e8522f7..53e8ae6898e 100644
--- a/be/src/olap/comparison_predicate.h
+++ b/be/src/olap/comparison_predicate.h
@@ -52,6 +52,12 @@ public:
ColumnPredicate::debug_string());
return fmt::to_string(debug_string_buffer);
}
+ bool could_be_erased() const override {
+ if ((PT == PredicateType::NE && !_opposite) || (PT ==
PredicateType::EQ && _opposite)) {
+ return false;
+ }
+ return true;
+ }
PredicateType type() const override { return PT; }
diff --git a/be/src/olap/in_list_predicate.h b/be/src/olap/in_list_predicate.h
index 2c34f5cfa38..20003ab7845 100644
--- a/be/src/olap/in_list_predicate.h
+++ b/be/src/olap/in_list_predicate.h
@@ -145,6 +145,13 @@ public:
PredicateType type() const override { return PT; }
+ bool could_be_erased() const override {
+ if ((PT == PredicateType::NOT_IN_LIST && !_opposite) ||
+ (PT == PredicateType::IN_LIST && _opposite)) {
+ return false;
+ }
+ return true;
+ }
Status evaluate(const vectorized::IndexFieldNameAndTypePair&
name_with_type,
IndexIterator* iterator, uint32_t num_rows,
roaring::Roaring* result) const override {
diff --git a/be/src/olap/null_predicate.h b/be/src/olap/null_predicate.h
index 7345e4b2703..4db04add084 100644
--- a/be/src/olap/null_predicate.h
+++ b/be/src/olap/null_predicate.h
@@ -60,6 +60,7 @@ public:
ColumnPredicate::debug_string(), _is_null);
return fmt::to_string(debug_string_buffer);
}
+ bool could_be_erased() const override { return true; }
PredicateType type() const override;
diff --git a/be/src/pipeline/exec/olap_scan_operator.cpp
b/be/src/pipeline/exec/olap_scan_operator.cpp
index 05a6ddbfd2c..d8f698c4081 100644
--- a/be/src/pipeline/exec/olap_scan_operator.cpp
+++ b/be/src/pipeline/exec/olap_scan_operator.cpp
@@ -21,6 +21,7 @@
#include <memory>
#include <numeric>
+#include <optional>
#include "cloud/cloud_meta_mgr.h"
#include "cloud/cloud_storage_engine.h"
@@ -878,6 +879,8 @@ Status OlapScanLocalState::_build_key_ranges_and_filters() {
DCHECK(_slot_id_to_predicates.count(iter->first) > 0);
const auto& value_range = iter->second;
+ std::optional<int> key_to_erase;
+
RETURN_IF_ERROR(std::visit(
[&](auto&& range) {
// make a copy or range and pass to extend_scan_key,
keep the range unchanged
@@ -889,7 +892,7 @@ Status OlapScanLocalState::_build_key_ranges_and_filters() {
_scan_keys.extend_scan_key(temp_range,
p._max_scan_key_num,
&exact_range,
&eos, &should_break));
if (exact_range) {
- _slot_id_to_value_range.erase(iter->first);
+ key_to_erase = iter->first;
}
} else {
// if exceed max_pushdown_conditions_per_column,
use whole_value_rang instead
@@ -902,6 +905,24 @@ Status OlapScanLocalState::_build_key_ranges_and_filters()
{
return Status::OK();
},
value_range));
+
+ // Perform the erase operation after the visit is complete, still
under lock
+ if (key_to_erase.has_value()) {
+ _slot_id_to_value_range.erase(*key_to_erase);
+
+ std::vector<std::shared_ptr<ColumnPredicate>> new_predicates;
+ for (const auto& it : _slot_id_to_predicates[*key_to_erase]) {
+ if (!it->could_be_erased()) {
+ new_predicates.push_back(it);
+ }
+ }
+ if (new_predicates.empty()) {
+ _slot_id_to_predicates.erase(*key_to_erase);
+ } else {
+ _slot_id_to_predicates[*key_to_erase] = new_predicates;
+ }
+ }
+ // lock is released here when it goes out of scope
}
if (eos) {
_eos = true;
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
