Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

fixed bug limit push down of index scan. #2905

Merged
merged 4 commits into from
Sep 23, 2021
Merged

fixed bug limit push down of index scan. #2905

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
4560209
limit bug fix for indexScan
bright-starry-sky Sep 18, 2021
23bf1ab
add test case for limit = 1
bright-starry-sky Sep 23, 2021
2aca216
changed the init count to 0
bright-starry-sky Sep 23, 2021
a6e204a
Merge branch 'master' into limit_pushdown_fix
bright-starry-sky Sep 23, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 5 additions & 5 deletions src/storage/exec/IndexEdgeNode.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,7 +22,7 @@ class IndexEdgeNode final : public RelNode<T> {
IndexScanNode<T>* indexScanNode,
const std::vector<std::shared_ptr<const meta::NebulaSchemaProvider>>& schemas,
const std::string& schemaName,
int64_t limit)
int64_t limit = -1)
: context_(context),
indexScanNode_(indexScanNode),
schemas_(schemas),
Expand All @@ -42,11 +42,7 @@ class IndexEdgeNode final : public RelNode<T> {
data_.clear();
std::vector<storage::cpp2::EdgeKey> edges;
auto* iter = static_cast<EdgeIndexIterator*>(indexScanNode_->iterator());
int64_t count = 0;
while (iter && iter->valid()) {
if (limit_ > -1 && count++ == limit_) {
break;
}
if (context_->isPlanKilled()) {
return nebula::cpp2::ErrorCode::E_PLAN_IS_KILLED;
}
Expand All @@ -66,6 +62,7 @@ class IndexEdgeNode final : public RelNode<T> {
edges.emplace_back(std::move(edge));
iter->next();
}
int64_t count = 0;
for (const auto& edge : edges) {
auto key = NebulaKeyUtils::edgeKey(context_->vIdLen(),
partId,
Expand All @@ -82,6 +79,9 @@ class IndexEdgeNode final : public RelNode<T> {
} else {
return ret;
}
if (limit_ > 0 && ++count >= limit_) {
break;
}
}
return nebula::cpp2::ErrorCode::SUCCEEDED;
}
Expand Down
39 changes: 29 additions & 10 deletions src/storage/exec/IndexFilterNode.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,14 +26,16 @@ class IndexFilterNode final : public RelNode<T> {
// data anymore.
IndexFilterNode(RuntimeContext* context,
IndexScanNode<T>* indexScanNode,
StorageExpressionContext* exprCtx = nullptr,
Expression* exp = nullptr,
bool isEdge = false)
StorageExpressionContext* exprCtx,
Expression* exp,
bool isEdge,
int64_t limit = -1)
: context_(context),
indexScanNode_(indexScanNode),
exprCtx_(exprCtx),
filterExp_(exp),
isEdge_(isEdge) {
isEdge_(isEdge),
limit_(limit) {
evalExprByIndex_ = true;
RelNode<T>::name_ = "IndexFilterNode";
}
Expand All @@ -42,9 +44,14 @@ class IndexFilterNode final : public RelNode<T> {
// need to read data.
IndexFilterNode(RuntimeContext* context,
IndexEdgeNode<T>* indexEdgeNode,
StorageExpressionContext* exprCtx = nullptr,
Expression* exp = nullptr)
: context_(context), indexEdgeNode_(indexEdgeNode), exprCtx_(exprCtx), filterExp_(exp) {
StorageExpressionContext* exprCtx,
Expression* exp,
int64_t limit = -1)
: context_(context),
indexEdgeNode_(indexEdgeNode),
exprCtx_(exprCtx),
filterExp_(exp),
limit_(limit) {
evalExprByIndex_ = false;
isEdge_ = true;
}
Expand All @@ -53,9 +60,14 @@ class IndexFilterNode final : public RelNode<T> {
// need to read data.
IndexFilterNode(RuntimeContext* context,
IndexVertexNode<T>* indexVertexNode,
StorageExpressionContext* exprCtx = nullptr,
Expression* exp = nullptr)
: context_(context), indexVertexNode_(indexVertexNode), exprCtx_(exprCtx), filterExp_(exp) {
StorageExpressionContext* exprCtx,
Expression* exp,
int64_t limit = -1)
: context_(context),
indexVertexNode_(indexVertexNode),
exprCtx_(exprCtx),
filterExp_(exp),
limit_(limit) {
evalExprByIndex_ = false;
isEdge_ = false;
}
Expand All @@ -74,13 +86,15 @@ class IndexFilterNode final : public RelNode<T> {
} else {
data = indexVertexNode_->moveData();
}
int64_t count = 0;
for (const auto& k : data) {
if (context_->isPlanKilled()) {
return nebula::cpp2::ErrorCode::E_PLAN_IS_KILLED;
}
if (evalExprByIndex_) {
if (check(k.first)) {
data_.emplace_back(k.first, k.second);
count++;
}
} else {
const auto& schemas =
Expand All @@ -91,8 +105,12 @@ class IndexFilterNode final : public RelNode<T> {
}
if (check(reader.get(), k.first)) {
data_.emplace_back(k.first, k.second);
count++;
}
}
if (limit_ > 0 && count >= limit_) {
break;
}
}
return nebula::cpp2::ErrorCode::SUCCEEDED;
}
Expand Down Expand Up @@ -143,6 +161,7 @@ class IndexFilterNode final : public RelNode<T> {
Expression* filterExp_;
bool isEdge_;
bool evalExprByIndex_;
int64_t limit_;
std::vector<kvstore::KV> data_{};
};

Expand Down
8 changes: 4 additions & 4 deletions src/storage/exec/IndexScanNode.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,7 +21,7 @@ class IndexScanNode : public RelNode<T> {
IndexScanNode(RuntimeContext* context,
IndexID indexId,
std::vector<cpp2::IndexColumnHint> columnHints,
int64_t limit)
int64_t limit = -1)
: context_(context), indexId_(indexId), columnHints_(std::move(columnHints)), limit_(limit) {
/**
* columnHints's elements are {scanType = PREFIX|RANGE; beginStr; endStr},
Expand Down Expand Up @@ -74,9 +74,6 @@ class IndexScanNode : public RelNode<T> {
data_.clear();
int64_t count = 0;
while (!!iter_ && iter_->valid()) {
if (limit_ > -1 && count++ == limit_) {
break;
}
if (context_->isPlanKilled()) {
return {};
}
Expand All @@ -89,6 +86,9 @@ class IndexScanNode : public RelNode<T> {
}
}
data_.emplace_back(iter_->key(), "");
if (limit_ > 0 && ++count >= limit_) {
break;
}
iter_->next();
}
return std::move(data_);
Expand Down
11 changes: 6 additions & 5 deletions src/storage/exec/IndexVertexNode.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,7 +22,7 @@ class IndexVertexNode final : public RelNode<T> {
IndexScanNode<T>* indexScanNode,
const std::vector<std::shared_ptr<const meta::NebulaSchemaProvider>>& schemas,
const std::string& schemaName,
int64_t limit)
int64_t limit = -1)
: context_(context),
indexScanNode_(indexScanNode),
schemas_(schemas),
Expand All @@ -42,11 +42,8 @@ class IndexVertexNode final : public RelNode<T> {
data_.clear();
std::vector<VertexID> vids;
auto* iter = static_cast<VertexIndexIterator*>(indexScanNode_->iterator());
int64_t count = 0;

while (iter && iter->valid()) {
if (limit_ > -1 && count++ == limit_) {
break;
}
if (context_->isPlanKilled()) {
return nebula::cpp2::ErrorCode::E_PLAN_IS_KILLED;
}
Expand All @@ -61,6 +58,7 @@ class IndexVertexNode final : public RelNode<T> {
vids.emplace_back(iter->vId());
iter->next();
}
int64_t count = 0;
for (const auto& vId : vids) {
VLOG(1) << "partId " << partId << ", vId " << vId << ", tagId " << context_->tagId_;
auto key = NebulaKeyUtils::vertexKey(context_->vIdLen(), partId, vId, context_->tagId_);
Expand All @@ -73,6 +71,9 @@ class IndexVertexNode final : public RelNode<T> {
} else {
return ret;
}
if (limit_ > 0 && ++count >= limit_) {
break;
}
}
return nebula::cpp2::ErrorCode::SUCCEEDED;
}
Expand Down
28 changes: 15 additions & 13 deletions src/storage/index/LookupBaseProcessor-inl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -167,6 +167,8 @@ template <typename REQ, typename RESP>
StatusOr<StoragePlan<IndexID>> LookupBaseProcessor<REQ, RESP>::buildPlan(
IndexFilterItem* filterItem, nebula::DataSet* result) {
StoragePlan<IndexID> plan;
// TODO(sky) : Limit is not supported yet for de-dup node.
// Related to paging scan, the de-dup execution plan needs to be refactored
auto deDup = std::make_unique<DeDupNode<IndexID>>(result, deDupColPos_);
int32_t filterId = 0;
std::unique_ptr<IndexOutputNode<IndexID>> out;
Expand Down Expand Up @@ -319,8 +321,8 @@ std::unique_ptr<IndexOutputNode<IndexID>> LookupBaseProcessor<REQ, RESP>::buildP
auto indexId = ctx.get_index_id();
auto colHints = ctx.get_column_hints();

auto indexScan = std::make_unique<IndexScanNode<IndexID>>(
context_.get(), indexId, std::move(colHints), limit_);
auto indexScan =
std::make_unique<IndexScanNode<IndexID>>(context_.get(), indexId, std::move(colHints));
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why doesn't this plan need limit?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why doesn't this plan need limit?

if FilterNode exists. will check the limit in FilterNode. skip the IndexScanNode.

if (context_->isEdge()) {
auto edge = std::make_unique<IndexEdgeNode<IndexID>>(
context_.get(), indexScan.get(), schemas_, context_->edgeName_, limit_);
Expand Down Expand Up @@ -370,11 +372,11 @@ std::unique_ptr<IndexOutputNode<IndexID>> LookupBaseProcessor<REQ, RESP>::buildP
auto indexId = ctx.get_index_id();
auto colHints = ctx.get_column_hints();

auto indexScan = std::make_unique<IndexScanNode<IndexID>>(
context_.get(), indexId, std::move(colHints), limit_);
auto indexScan =
std::make_unique<IndexScanNode<IndexID>>(context_.get(), indexId, std::move(colHints));

auto filter = std::make_unique<IndexFilterNode<IndexID>>(
context_.get(), indexScan.get(), exprCtx, exp, context_->isEdge());
context_.get(), indexScan.get(), exprCtx, exp, context_->isEdge(), limit_);
filter->addDependency(indexScan.get());
auto output =
std::make_unique<IndexOutputNode<IndexID>>(result, context_.get(), filter.get(), true);
Expand Down Expand Up @@ -421,14 +423,14 @@ LookupBaseProcessor<REQ, RESP>::buildPlanWithDataAndFilter(nebula::DataSet* resu
auto indexId = ctx.get_index_id();
auto colHints = ctx.get_column_hints();

auto indexScan = std::make_unique<IndexScanNode<IndexID>>(
context_.get(), indexId, std::move(colHints), limit_);
auto indexScan =
std::make_unique<IndexScanNode<IndexID>>(context_.get(), indexId, std::move(colHints));
if (context_->isEdge()) {
auto edge = std::make_unique<IndexEdgeNode<IndexID>>(
context_.get(), indexScan.get(), schemas_, context_->edgeName_, limit_);
context_.get(), indexScan.get(), schemas_, context_->edgeName_);
edge->addDependency(indexScan.get());
auto filter =
std::make_unique<IndexFilterNode<IndexID>>(context_.get(), edge.get(), exprCtx, exp);
auto filter = std::make_unique<IndexFilterNode<IndexID>>(
context_.get(), edge.get(), exprCtx, exp, limit_);
filter->addDependency(edge.get());

auto output = std::make_unique<IndexOutputNode<IndexID>>(result, context_.get(), filter.get());
Expand All @@ -439,10 +441,10 @@ LookupBaseProcessor<REQ, RESP>::buildPlanWithDataAndFilter(nebula::DataSet* resu
return output;
} else {
auto vertex = std::make_unique<IndexVertexNode<IndexID>>(
context_.get(), indexScan.get(), schemas_, context_->tagName_, limit_);
context_.get(), indexScan.get(), schemas_, context_->tagName_);
vertex->addDependency(indexScan.get());
auto filter =
std::make_unique<IndexFilterNode<IndexID>>(context_.get(), vertex.get(), exprCtx, exp);
auto filter = std::make_unique<IndexFilterNode<IndexID>>(
context_.get(), vertex.get(), exprCtx, exp, limit_);
filter->addDependency(vertex.get());

auto output = std::make_unique<IndexOutputNode<IndexID>>(result, context_.get(), filter.get());
Expand Down
5 changes: 5 additions & 0 deletions src/storage/index/LookupProcessor.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,6 +23,11 @@ void LookupProcessor::process(const cpp2::LookupIndexRequest& req) {

void LookupProcessor::doProcess(const cpp2::LookupIndexRequest& req) {
auto retCode = requestCheck(req);
if (limit_ == 0) {
onProcessFinished();
onFinished();
return;
}
if (retCode != nebula::cpp2::ErrorCode::SUCCEEDED) {
for (auto& p : req.get_parts()) {
pushResultCode(retCode, p);
Expand Down
Loading