enhance: support post filter execution #37363

chasingegg · 2024-11-01T07:21:57Z

sre-ci-robot · 2024-11-01T07:22:16Z

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by: chasingegg
To complete the pull request process, please assign congqixia after the PR has been reviewed.
You can assign the PR to them by writing /assign @congqixia in a comment when ready.

The full list of commands accepted by this bot can be found here.

Needs approval from an approver in each of these files:

internal/core/OWNERS

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

mergify · 2024-11-01T07:22:34Z

@chasingegg

Invalid PR Title Format Detected

Your PR submission does not adhere to our required standards. To ensure clarity and consistency, please meet the following criteria:

Title Format: The PR title must begin with one of these prefixes:

feat: for introducing a new feature.
fix: for bug fixes.
enhance: for improvements to existing functionality.
test: for add tests to existing functionality.
doc: for modifying documentation.
auto: for the pull request from bot.

Description Requirement: The PR must include a non-empty description, detailing the changes and their impact.

Required Title Structure:

[Type]: [Description of the PR]

Where Type is one of feat, fix, enhance, test or doc.

Example:

enhance: improve search performance significantly

Please review and update your PR to comply with these guidelines.

mergify · 2024-11-01T07:27:16Z

@chasingegg go-sdk check failed, comment rerun go-sdk can trigger the job again.

mergify · 2024-11-01T07:32:49Z

@chasingegg cpp-unit-test check failed, comment rerun cpp-unit-test can trigger the job again.

mergify · 2024-11-01T07:36:51Z

@chasingegg E2e jenkins job failed, comment /run-cpu-e2e can trigger the job again.

codecov · 2024-11-01T09:54:49Z

Codecov Report

Attention: Patch coverage is 84.39349% with 211 lines in your changes missing coverage. Please review.

Project coverage is 80.67%. Comparing base (5a23c80) to head (99d6e57).
Report is 4 commits behind head on master.

Files with missing lines	Patch %	Lines
internal/core/src/exec/expression/Expr.h	51.23%	59 Missing ⚠️
internal/core/src/exec/expression/CompareExpr.h	64.61%	23 Missing ⚠️
internal/core/src/common/Chunk.cpp	0.00%	15 Missing ⚠️
internal/core/src/exec/expression/ColumnExpr.cpp	65.78%	13 Missing ⚠️
internal/core/src/exec/operator/Utils.h	53.57%	13 Missing ⚠️
...rnal/core/src/exec/expression/JsonContainsExpr.cpp	96.03%	10 Missing ⚠️
...rnal/core/src/segcore/ChunkedSegmentSealedImpl.cpp	30.76%	9 Missing ⚠️
internal/core/src/segcore/SegmentInterface.h	33.33%	8 Missing ⚠️
internal/core/src/exec/operator/PostFilterNode.cpp	92.55%	7 Missing ⚠️
...ernal/core/src/exec/expression/BinaryRangeExpr.cpp	93.40%	6 Missing ⚠️
... and 13 more

Additional details and impacted files

@@            Coverage Diff             @@
##           master   #37363      +/-   ##
==========================================
- Coverage   80.67%   80.67%   -0.01%     
==========================================
  Files        1357     1360       +3     
  Lines      190679   191257     +578     
==========================================
+ Hits       153835   154294     +459     
- Misses      31416    31546     +130     
+ Partials     5428     5417      -11

Components	Coverage Δ
Client	`61.25% <ø> (ø)`
Core	`68.97% <84.39%> (+0.04%)`	⬆️
Go	`83.22% <ø> (+0.03%)`	⬆️

Files with missing lines	Coverage Δ
internal/core/src/common/Chunk.h	`60.24% <ø> (ø)`
internal/core/src/common/QueryInfo.h	`100.00% <ø> (ø)`
internal/core/src/exec/Driver.cpp	`81.39% <100.00%> (+0.44%)`	⬆️
internal/core/src/exec/QueryContext.h	`84.61% <100.00%> (+0.30%)`	⬆️
...ternal/core/src/exec/expression/AlwaysTrueExpr.cpp	`88.23% <100.00%> (+2.52%)`	⬆️
...e/src/exec/expression/BinaryArithOpEvalRangeExpr.h	`100.00% <100.00%> (ø)`
...nternal/core/src/exec/expression/BinaryRangeExpr.h	`94.00% <100.00%> (+1.31%)`	⬆️
internal/core/src/exec/expression/ColumnExpr.h	`44.73% <ø> (ø)`
internal/core/src/exec/expression/EvalCtx.h	`100.00% <100.00%> (ø)`
internal/core/src/exec/expression/ExistsExpr.cpp	`89.18% <100.00%> (+2.09%)`	⬆️
... and 40 more

... and 40 files with indirect coverage changes

PwzXxm · 2024-11-04T08:16:10Z

internal/core/src/exec/QueryContext.h

+    std::optional<int64_t>
+    get_iterator_batch_size() {
+        return milvus::index::GetValueFromConfig<int64_t>(
+            search_info_.search_params_, "batch_size");


Is it possible to rename it as conflicting to E2E Iterator parameters?

PwzXxm · 2024-11-04T08:29:48Z

internal/core/src/exec/expression/Expr.h

-            MoveCursorForIndex();
-            if (segment_->HasFieldData(field_id_)) {
+        // when we specify input, do not maintain states
+        if (has_offset_input_) {


Why it does not need to move cursor here when input is specified? Not the other way around as in internal/core/src/exec/expression/LogicalBinaryExpr.h per say?

PwzXxm · 2024-11-04T08:33:15Z

internal/core/src/exec/operator/FilterNode.cpp

+                      size_t hi,
+                      float dist) {
+    while (lo < hi) {
+        size_t mid = (lo + hi) >> 1;


Potential overflow, use size_t mid = lo + ((hi - lo) >> 1)

PwzXxm · 2024-11-04T08:37:47Z

internal/core/src/exec/operator/FilterNode.cpp

+    double scalar_cost =
+        std::chrono::duration<double, std::micro>(scalar_end - scalar_start)
+            .count();
+    monitor::internal_core_search_latency_postfilter.Observe(scalar_cost);


Great to see adding metrics as well!

PwzXxm · 2024-11-04T08:44:07Z

internal/core/src/exec/operator/FilterNode.cpp

+                auto col_vec_size = col_vec->size();
+                TargetBitmapView bitsetview(col_vec->GetRawData(),
+                                            col_vec_size);
+                Assert(bitsetview.size() <= batch_size);


Why this must be met? Is it possible that user gives a small batch_size?

PwzXxm · 2024-11-04T08:56:29Z

internal/core/src/exec/operator/FilterNode.cpp

+}
+
+RowVectorPtr
+PhyFilterNode::GetOutput() {


Does it support RangeSearch?

mergify · 2024-11-05T11:00:58Z

@chasingegg cpp-unit-test check failed, comment rerun cpp-unit-test can trigger the job again.

zhagnlu · 2024-11-13T03:28:29Z

internal/core/src/query/PlanProto.cpp

+            plan_node->search_info_.group_by_field_id_ == std::nullopt) {
+            plannode = std::make_shared<milvus::plan::MvccNode>(
+                milvus::plan::GetNextPlanNodeId());
+            sources = std::vector<milvus::plan::PlanNodePtr>{plannode};
+            plannode = std::make_shared<milvus::plan::VectorSearchNode>(
+                milvus::plan::GetNextPlanNodeId(), sources);
+            sources = std::vector<milvus::plan::PlanNodePtr>{plannode};
+
+            // add filter nodes after vector search node
+            auto expr = ParseExprs(anns_proto.predicates());
+            plannode = std::make_shared<plan::FilterNode>(
+                milvus::plan::GetNextPlanNodeId(), expr, sources);
+            sources = std::vector<milvus::plan::PlanNodePtr>{plannode};


wrapper as a function, same as below else

zhagnlu · 2024-11-13T03:34:25Z

internal/core/src/query/PlanProto.cpp

+            !search_info.search_params_.contains(RADIUS)) {
+            search_info.post_filter_execution =
+                search_info.search_params_[POST_FILTER];
+        }


Is that means user decide to whether using post filter or not ? Could decided by other method like stats info etc

zhagnlu · 2024-11-13T03:58:43Z

internal/core/src/exec/operator/FilterNode.h

+// FilterNode will accept offsets array and execute over these and generate result valid offsets
+namespace milvus {
+namespace exec {
+class PhyFilterNode : public Operator {


PhyPosterFilterBitsNode may more accurate. PhyFilterNode should not related with vector search node, not just return bits, it is more pure concept. When we support project function, add PhyFilterNode is better，same is FilterNode.h

zhagnlu · 2024-11-13T06:47:26Z

internal/core/src/exec/expression/EvalCtx.h

@@ -28,17 +28,26 @@ namespace milvus {
 namespace exec {

 class ExprSet;
+
+using OffsetVector = FixedVector<int64_t>;


int32_t is enough

mergify · 2024-11-14T04:03:42Z

@chasingegg cpp-unit-test check failed, comment rerun cpp-unit-test can trigger the job again.

chasingegg · 2024-11-14T04:18:39Z

rerun cpp-unit-test

Signed-off-by: chasingegg <[email protected]>

mergify · 2024-11-15T06:31:26Z

@chasingegg go-sdk check failed, comment rerun go-sdk can trigger the job again.

mergify · 2024-11-15T06:32:37Z

@chasingegg cpp-unit-test check failed, comment rerun cpp-unit-test can trigger the job again.

chasingegg · 2024-11-15T08:26:44Z

rerun go-sdk

chasingegg · 2024-11-15T08:26:54Z

rerun cpp-unit-test

sre-ci-robot added area/compilation size/XXL Denotes a PR that changes 1000+ lines. labels Nov 1, 2024

sre-ci-robot requested review from godchen0212 and xiaocai2333 November 1, 2024 07:22

mergify bot added dco-passed DCO check passed. do-not-merge/invalid-pr-format labels Nov 1, 2024

chasingegg changed the title ~~Support post filter~~ enhance: support post filter execution Nov 1, 2024

mergify bot added kind/enhancement Issues or changes related to enhancement and removed do-not-merge/invalid-pr-format labels Nov 1, 2024

PwzXxm reviewed Nov 4, 2024

View reviewed changes

chasingegg force-pushed the support-post-filter branch from 6a7ea43 to 83d7a5e Compare November 5, 2024 10:23

chasingegg force-pushed the support-post-filter branch 3 times, most recently from 909b13b to 82aa236 Compare November 8, 2024 10:14

mergify bot added the ci-passed label Nov 8, 2024

zhagnlu reviewed Nov 13, 2024

View reviewed changes

chasingegg force-pushed the support-post-filter branch from 82aa236 to 8dcc2db Compare November 14, 2024 03:50

mergify bot removed the ci-passed label Nov 14, 2024

mergify bot added the ci-passed label Nov 14, 2024

chasingegg force-pushed the support-post-filter branch from 8dcc2db to 3a9224e Compare November 14, 2024 09:07

mergify bot removed the ci-passed label Nov 14, 2024

mergify bot added the ci-passed label Nov 14, 2024

alexanderguzhva mentioned this pull request Nov 14, 2024

[Enhancement]: post filter execution SIMD #37700

Open

1 task

chasingegg force-pushed the support-post-filter branch from 3a9224e to 8e67a30 Compare November 15, 2024 06:11

Support post filter

99d6e57

Signed-off-by: chasingegg <[email protected]>

chasingegg force-pushed the support-post-filter branch from 8e67a30 to 99d6e57 Compare November 15, 2024 06:15

mergify bot removed the ci-passed label Nov 15, 2024

mergify bot added the ci-passed label Nov 15, 2024

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

enhance: support post filter execution #37363

enhance: support post filter execution #37363

chasingegg commented Nov 1, 2024

sre-ci-robot commented Nov 1, 2024

mergify bot commented Nov 1, 2024

mergify bot commented Nov 1, 2024

mergify bot commented Nov 1, 2024

mergify bot commented Nov 1, 2024

codecov bot commented Nov 1, 2024 •

edited

Loading

PwzXxm Nov 4, 2024

PwzXxm Nov 4, 2024

PwzXxm Nov 4, 2024

PwzXxm Nov 4, 2024

PwzXxm Nov 4, 2024

PwzXxm Nov 4, 2024

mergify bot commented Nov 5, 2024

zhagnlu Nov 13, 2024

zhagnlu Nov 13, 2024

zhagnlu Nov 13, 2024

zhagnlu Nov 13, 2024

mergify bot commented Nov 14, 2024

chasingegg commented Nov 14, 2024

mergify bot commented Nov 15, 2024

mergify bot commented Nov 15, 2024

chasingegg commented Nov 15, 2024

chasingegg commented Nov 15, 2024

enhance: support post filter execution #37363

Are you sure you want to change the base?

enhance: support post filter execution #37363

Conversation

chasingegg commented Nov 1, 2024

sre-ci-robot commented Nov 1, 2024

mergify bot commented Nov 1, 2024

mergify bot commented Nov 1, 2024

mergify bot commented Nov 1, 2024

mergify bot commented Nov 1, 2024

codecov bot commented Nov 1, 2024 • edited Loading

Codecov Report

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mergify bot commented Nov 5, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mergify bot commented Nov 14, 2024

chasingegg commented Nov 14, 2024

mergify bot commented Nov 15, 2024

mergify bot commented Nov 15, 2024

chasingegg commented Nov 15, 2024

chasingegg commented Nov 15, 2024

codecov bot commented Nov 1, 2024 •

edited

Loading