diff options
| author | maxkovalev <[email protected]> | 2025-10-20 14:21:50 +0300 |
|---|---|---|
| committer | maxkovalev <[email protected]> | 2025-10-20 15:18:43 +0300 |
| commit | c12bf4cfcead3a0172e4a140d0c89bbe18ee1791 (patch) | |
| tree | 101659067b46d1a4d43b48f76f8ba6d76f8e08af | |
| parent | c36cb32e8a206a8a5d591ec5eebff3bc25a6389a (diff) | |
YQL-19542: Save lineage in qstorage and check at replay
commit_hash:3693b0059b05bcf38d734e2d395dc3378902675c
| -rw-r--r-- | yql/essentials/core/facade/yql_facade.cpp | 8 | ||||
| -rw-r--r-- | yql/essentials/core/facade/yql_facade.h | 5 | ||||
| -rw-r--r-- | yql/essentials/core/services/yql_lineage.cpp | 23 | ||||
| -rw-r--r-- | yql/essentials/core/services/yql_lineage.h | 3 | ||||
| -rw-r--r-- | yql/essentials/core/services/yql_transform_pipeline.cpp | 88 | ||||
| -rw-r--r-- | yql/essentials/core/services/yql_transform_pipeline.h | 2 | ||||
| -rw-r--r-- | yql/essentials/tools/yql_facade_run/yql_facade_run.cpp | 3 | ||||
| -rw-r--r-- | yql/essentials/tools/yql_facade_run/yql_facade_run.h | 1 | ||||
| -rw-r--r-- | yql/tools/yqlrun/lib/yqlrun_lib.cpp | 1 |
9 files changed, 103 insertions, 31 deletions
diff --git a/yql/essentials/core/facade/yql_facade.cpp b/yql/essentials/core/facade/yql_facade.cpp index dfb44c1af3d..0b6f9eca6b6 100644 --- a/yql/essentials/core/facade/yql_facade.cpp +++ b/yql/essentials/core/facade/yql_facade.cpp @@ -1192,7 +1192,7 @@ TProgram::TFutureStatus TProgram::OptimizeAsync( .AddTypeAnnotation(TIssuesIds::CORE_TYPE_ANN, true) .AddPostTypeAnnotation() .Add(TExprOutputTransformer::Sync(ExprRoot_, traceOut), "ExprOutput") - .AddOptimization() + .AddOptimizationWithLineage(EnableLineage_) .Add(TExprOutputTransformer::Sync(ExprRoot_, exprOut, withTypes), "AstOutput") .Build(); @@ -1262,7 +1262,7 @@ TProgram::TFutureStatus TProgram::OptimizeAsyncWithConfig( pipeline.AddPostTypeAnnotation(); pipelineConf.AfterTypeAnnotation(&pipeline); - pipeline.AddOptimization(); + pipeline.AddOptimizationWithLineage(EnableLineage_); if (EnableRangeComputeFor_) { pipeline.Add(MakeExpandRangeComputeForTransformer(pipeline.GetTypeAnnotationContext()), "ExpandRangeComputeFor", TIssuesIds::CORE_EXEC); @@ -1402,7 +1402,7 @@ TProgram::TFutureStatus TProgram::RunAsync( pipeline.AddTypeAnnotation(TIssuesIds::CORE_TYPE_ANN, true); pipeline.AddPostTypeAnnotation(); pipeline.Add(TExprOutputTransformer::Sync(ExprRoot_, traceOut), "ExprOutput"); - pipeline.AddOptimizationWithLineage(); + pipeline.AddOptimizationWithLineage(EnableLineage_); if (EnableRangeComputeFor_) { pipeline.Add(MakeExpandRangeComputeForTransformer(pipeline.GetTypeAnnotationContext()), "ExpandRangeComputeFor", TIssuesIds::CORE_EXEC); @@ -1480,7 +1480,7 @@ TProgram::TFutureStatus TProgram::RunAsyncWithConfig( pipeline.AddPostTypeAnnotation(); pipelineConf.AfterTypeAnnotation(&pipeline); - pipeline.AddOptimizationWithLineage(); + pipeline.AddOptimizationWithLineage(EnableLineage_); if (EnableRangeComputeFor_) { pipeline.Add(MakeExpandRangeComputeForTransformer(pipeline.GetTypeAnnotationContext()), "ExpandRangeComputeFor", TIssuesIds::CORE_EXEC); diff --git a/yql/essentials/core/facade/yql_facade.h b/yql/essentials/core/facade/yql_facade.h index 018367a3b37..0f6767562ad 100644 --- a/yql/essentials/core/facade/yql_facade.h +++ b/yql/essentials/core/facade/yql_facade.h @@ -353,6 +353,10 @@ public: TString GetSourceCode() const; + void SetEnableLineage() { + EnableLineage_ = true; + } + private: TProgram( const NKikimr::NMiniKQL::IFunctionRegistry* functionRegistry, @@ -483,6 +487,7 @@ private: TMaybe<TString> GatewaysForMerge_; TIssues FinalIssues_; TMaybe<TIssue> ParametersIssue_; + bool EnableLineage_ = false; }; void UpdateSqlFlagsFromQContext(const TQContext& qContext, THashSet<TString>& flags, TMaybe<TString> gatewaysPatch = {}); diff --git a/yql/essentials/core/services/yql_lineage.cpp b/yql/essentials/core/services/yql_lineage.cpp index 57e3e02192f..35721515add 100644 --- a/yql/essentials/core/services/yql_lineage.cpp +++ b/yql/essentials/core/services/yql_lineage.cpp @@ -5,6 +5,7 @@ #include <yql/essentials/core/yql_opt_utils.h> #include <yql/essentials/core/yql_join.h> +#include <library/cpp/yson/node/node_io.h> #include <util/system/env.h> namespace NYql { @@ -966,4 +967,26 @@ TString CalculateLineage(const TExprNode& root, const TTypeAnnotationContext& ct return scanner.Process(); } +TString NormalizeLineage(const TString& lineageStr) { + THashMap<i64, TString> idToPath; + auto lineageNode = NYT::NodeFromYsonString(lineageStr); + auto& readsSection = lineageNode.AsMap()["Reads"]; + for (auto& readNode : readsSection.AsList()) { + auto& readMap = readNode.AsMap(); + idToPath[readMap["Id"].AsInt64()] = readMap["Name"].AsString(); + readMap["Id"] = readMap["Name"]; + } + auto& writesSection = lineageNode.AsMap()["Writes"]; + for (auto& writeNode : writesSection.AsList()) { + auto& writeMap = writeNode.AsMap(); + writeMap["Id"] = writeMap["Name"]; + for (auto& [fieldName, fieldLineage] : writeMap["Lineage"].AsMap()) { + for (auto& inputField : fieldLineage.AsList()) { + inputField["Input"] = idToPath[inputField["Input"].AsInt64()]; + } + } + } + return NYT::NodeToCanonicalYsonString(lineageNode); +} + } // namespace NYql diff --git a/yql/essentials/core/services/yql_lineage.h b/yql/essentials/core/services/yql_lineage.h index c1a71ae185b..2b66c064a4d 100644 --- a/yql/essentials/core/services/yql_lineage.h +++ b/yql/essentials/core/services/yql_lineage.h @@ -8,4 +8,7 @@ struct TExprContext; TString CalculateLineage(const TExprNode& root, const TTypeAnnotationContext& ctx, TExprContext& exprCtx, bool standalone); +// Replace input and output table's IDs with pathes for checking lineage equality +TString NormalizeLineage(const TString& lineageStr); + } // namespace NYql diff --git a/yql/essentials/core/services/yql_transform_pipeline.cpp b/yql/essentials/core/services/yql_transform_pipeline.cpp index 1b2df336500..45c0a5ae6a8 100644 --- a/yql/essentials/core/services/yql_transform_pipeline.cpp +++ b/yql/essentials/core/services/yql_transform_pipeline.cpp @@ -21,6 +21,9 @@ namespace NYql { +const TString LineageComponent = "Lineage"; +const TString LineageResultLabel = "LineageResult"; + TTransformationPipeline::TTransformationPipeline( TIntrusivePtr<TTypeAnnotationContext> ctx, TTypeAnnCallableFactory typeAnnCallableFactory) @@ -167,33 +170,66 @@ TTransformationPipeline& TTransformationPipeline::AddFinalCommonOptimization(EYq return *this; } -TTransformationPipeline& TTransformationPipeline::AddOptimizationWithLineage(bool checkWorld, bool withFinalOptimization, EYqlIssueCode issueCode) { +TTransformationPipeline& TTransformationPipeline::AddOptimizationWithLineage(bool enableLineage, bool checkWorld, bool withFinalOptimization, EYqlIssueCode issueCode) { AddCommonOptimization(false, issueCode); - Transformers_.push_back(TTransformStage( - CreateChoiceGraphTransformer( - [&typesCtx = std::as_const(*TypeAnnotationContext_)](const TExprNode::TPtr&, TExprContext&) { - return typesCtx.EnableLineage; - }, - TTransformStage( - CreateFunctorTransformer( - [typeCtx = TypeAnnotationContext_](const TExprNode::TPtr& input, TExprNode::TPtr& output, TExprContext& ctx) { - output = input; - try { - CalculateLineage(*input, *typeCtx, ctx, false); - } catch (const std::exception& e) { - YQL_LOG(ERROR) << "CalculateLineage error: " << e.what(); - typeCtx->CorrectLineage = false; - } - return IGraphTransformer::TStatus::Ok; - }), - "Lineage", - issueCode), - TTransformStage( - new TNullTransformer(), - "SkipLineage", - issueCode)), - "LineageCalculation", - issueCode)); + if (enableLineage) { + Transformers_.push_back(TTransformStage( + CreateChoiceGraphTransformer( + [&typesCtx = std::as_const(*TypeAnnotationContext_)](const TExprNode::TPtr&, TExprContext&) { + return typesCtx.EnableLineage; + }, + TTransformStage( + CreateSinglePassFunctorTransformer( + [typeCtx = TypeAnnotationContext_](const TExprNode::TPtr& input, TExprNode::TPtr& output, TExprContext& ctx) { + output = input; + TString calculatedLineage, loadedLineage; + if (typeCtx->QContext && typeCtx->QContext.CanRead()) { + auto loaded = typeCtx->QContext.GetReader()->Get({LineageComponent, LineageResultLabel}).GetValueSync(); + if (loaded.Defined()) { + loadedLineage = loaded->Value; + } else { + YQL_LOG(INFO) << "There is no lineage in QStorage, lineage calculation is skipped in replay mode"; + return IGraphTransformer::TStatus::Ok; + } + } + std::exception_ptr lineageError; + try { + calculatedLineage = CalculateLineage(*input, *typeCtx, ctx, false); + } catch (const std::exception& e) { + YQL_LOG(ERROR) << "Lineage calculation error: " << e.what(); + typeCtx->CorrectLineage = false; + lineageError = std::current_exception(); + } + if (!loadedLineage.empty()) { + // if lineage calculation is failed, but loaded lineage exists, rethrow exception for replay mode + if (lineageError) { + std::rethrow_exception(lineageError); + } + if (NormalizeLineage(calculatedLineage) != NormalizeLineage(loadedLineage)) { + YQL_LOG(INFO) << "Lineage in replay is different:" + << "\nCalculated lineage:\n" + << calculatedLineage + << "\nLoaded lineage:\n" + << loadedLineage; + throw yexception() << "Lineage in replay is different"; + } + YQL_LOG(INFO) << "Lineage replay is the same"; + } + if (typeCtx->QContext && typeCtx->QContext.CanWrite() && typeCtx->CorrectLineage) { + typeCtx->QContext.GetWriter()->Put({LineageComponent, LineageResultLabel}, calculatedLineage).GetValueSync(); + YQL_LOG(INFO) << "Lineage is saved to QStorage"; + } + return IGraphTransformer::TStatus::Ok; + }), + "Lineage", + issueCode), + TTransformStage( + new TNullTransformer(), + "SkipLineage", + issueCode)), + "LineageCalculation", + issueCode)); + } AddProviderOptimization(issueCode); if (withFinalOptimization) { AddFinalCommonOptimization(issueCode); diff --git a/yql/essentials/core/services/yql_transform_pipeline.h b/yql/essentials/core/services/yql_transform_pipeline.h index 4cd60c76079..ca1dbfae83d 100644 --- a/yql/essentials/core/services/yql_transform_pipeline.h +++ b/yql/essentials/core/services/yql_transform_pipeline.h @@ -36,7 +36,7 @@ public: TTransformationPipeline& AddFinalCommonOptimization(EYqlIssueCode issueCode = TIssuesIds::CORE_OPTIMIZATION); TTransformationPipeline& AddOptimization(bool checkWorld = true, bool withFinalOptimization = true, EYqlIssueCode issueCode = TIssuesIds::CORE_OPTIMIZATION); TTransformationPipeline& AddProviderOptimization(EYqlIssueCode issueCode = TIssuesIds::CORE_OPTIMIZATION); - TTransformationPipeline& AddOptimizationWithLineage(bool checkWorld = true, bool withFinalOptimization = true, EYqlIssueCode issueCode = TIssuesIds::CORE_OPTIMIZATION); + TTransformationPipeline& AddOptimizationWithLineage(bool enableLineage, bool checkWorld = true, bool withFinalOptimization = true, EYqlIssueCode issueCode = TIssuesIds::CORE_OPTIMIZATION); TTransformationPipeline& AddLineageOptimization(TMaybe<TString>& lineageOut, EYqlIssueCode issueCode = TIssuesIds::CORE_OPTIMIZATION); TTransformationPipeline& AddCheckExecution(bool checkWorld = true, EYqlIssueCode issueCode = TIssuesIds::CORE_OPTIMIZATION); TTransformationPipeline& AddRun(TOperationProgressWriter writer, EYqlIssueCode issueCode = TIssuesIds::CORE_EXEC); diff --git a/yql/essentials/tools/yql_facade_run/yql_facade_run.cpp b/yql/essentials/tools/yql_facade_run/yql_facade_run.cpp index c754381e910..5062d712f23 100644 --- a/yql/essentials/tools/yql_facade_run/yql_facade_run.cpp +++ b/yql/essentials/tools/yql_facade_run/yql_facade_run.cpp @@ -777,6 +777,9 @@ int TFacadeRunner::DoRun(TProgramFactory& factory) { } program->SetUseTableMetaFromGraph(RunOptions_.UseMetaFromGrpah); program->SetValidateOptions(RunOptions_.ValidateMode); + if (RunOptions_.EnableLineage) { + program->SetEnableLineage(); + } bool fail = false; if (RunOptions_.ProgramType != EProgramType::SExpr) { diff --git a/yql/essentials/tools/yql_facade_run/yql_facade_run.h b/yql/essentials/tools/yql_facade_run/yql_facade_run.h index b7fdcbbb0ae..b0a08d370aa 100644 --- a/yql/essentials/tools/yql_facade_run/yql_facade_run.h +++ b/yql/essentials/tools/yql_facade_run/yql_facade_run.h @@ -141,6 +141,7 @@ public: bool EnableQPlayer = false; bool OptimizeLibs = true; bool CustomTests = false; + bool EnableLineage = false; void Parse(int argc, const char* argv[]); diff --git a/yql/tools/yqlrun/lib/yqlrun_lib.cpp b/yql/tools/yqlrun/lib/yqlrun_lib.cpp index aad7657ea4a..c0f23c972d3 100644 --- a/yql/tools/yqlrun/lib/yqlrun_lib.cpp +++ b/yql/tools/yqlrun/lib/yqlrun_lib.cpp @@ -49,6 +49,7 @@ TYqlRunTool::TYqlRunTool() GetRunOptions().UseRepeatableRandomAndTimeProviders = true; GetRunOptions().ResultsFormat = NYson::EYsonFormat::Pretty; GetRunOptions().CustomTests = true; + GetRunOptions().EnableLineage = true; GetRunOptions().AddOptExtension([this](NLastGetopt::TOpts& opts) { opts.AddLongOption('t', "table", "Table mapping").RequiredArgument("table@file") |
