Add native expression optimizer #23331
Draft
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
tdcmeehan
force-pushed
the
expr-spi
branch
4 times, most recently
from
5d6bd56 to
98d0623
Compare
tdcmeehan
force-pushed
the
expr-spi
branch
3 times, most recently
from
f9c1806 to
d79790f
Compare
tdcmeehan
force-pushed
the
expr-spi
branch
9 times, most recently
from
828a7a9 to
94f15ee
Compare
tdcmeehan
force-pushed
the
expr-spi
branch
4 times, most recently
from
ee395c9 to
2d0bf9b
Compare
facebook-github-bot
pushed a commit
to facebookincubator/velox
that referenced
this pull request
Nov 16, 2024
Summary: prestodb/presto#23331 adds a native expression optimizer to delegate expression evaluation to the native sidecar. This is used to constant fold expressions on the presto native sidecar, instead of on the presto java coordinator (which is the current behavior). prestodb/presto#22927 implements a proxygen endpoint to accept `RowExpression`s from `NativeSidecarExpressionInterpreter`, optimize them if possible (rewrite special form expressions), and compile the `RowExpression` to a velox expression with constant folding enabled. This velox expression is then converted back to a `RowExpression` and returned by the sidecar to the coordinator. When the constant folded velox expression is of type `velox::exec::ConstantExpr`, we need to return a `RowExpression` of type `ConstantExpression`. This requires us to serialize the constant value from `velox::exec::ConstantExpr` into `protocol::ConstantExpression::valueBlock`. This can be done by serializing the constant value vector to presto SerializedPage::column format, followed by base 64 encoding the result (reverse engineering the logic from `Base64Util.cpp::readBlock`). This PR adds a new function, `serializeSingleColumn`, to `PrestoVectorSerde`. This can be used to serialize input data from vectors containing a single element into a single PrestoPage column format (without the PrestoPage header). This function is not added to `PrestoBatchVectorSerializer` alongside the existing `serialize` function since that would require adding it as a virtual function in `BatchVectorSerializer` as well, and this is not desirable since the `PrestoPage` format is not relevant in this base class. There is an existing function `deserializeSingleColumn` in `PrestoVectorSerde` which is used to deserialize data from a single column, since `serializeSingleColumn` performs the inverse operation to this function, it is added alongside it in `PrestoVectorSerde`. Pull Request resolved: #10657 Reviewed By: amitkdutta Differential Revision: D66044754 Pulled By: pedroerp fbshipit-source-id: e509605067920f8207e5a3fa67552badc2ce0eba
athmaja-n
pushed a commit
to athmaja-n/velox
that referenced
this pull request
Jan 10, 2025
…bator#10657) Summary: prestodb/presto#23331 adds a native expression optimizer to delegate expression evaluation to the native sidecar. This is used to constant fold expressions on the presto native sidecar, instead of on the presto java coordinator (which is the current behavior). prestodb/presto#22927 implements a proxygen endpoint to accept `RowExpression`s from `NativeSidecarExpressionInterpreter`, optimize them if possible (rewrite special form expressions), and compile the `RowExpression` to a velox expression with constant folding enabled. This velox expression is then converted back to a `RowExpression` and returned by the sidecar to the coordinator. When the constant folded velox expression is of type `velox::exec::ConstantExpr`, we need to return a `RowExpression` of type `ConstantExpression`. This requires us to serialize the constant value from `velox::exec::ConstantExpr` into `protocol::ConstantExpression::valueBlock`. This can be done by serializing the constant value vector to presto SerializedPage::column format, followed by base 64 encoding the result (reverse engineering the logic from `Base64Util.cpp::readBlock`). This PR adds a new function, `serializeSingleColumn`, to `PrestoVectorSerde`. This can be used to serialize input data from vectors containing a single element into a single PrestoPage column format (without the PrestoPage header). This function is not added to `PrestoBatchVectorSerializer` alongside the existing `serialize` function since that would require adding it as a virtual function in `BatchVectorSerializer` as well, and this is not desirable since the `PrestoPage` format is not relevant in this base class. There is an existing function `deserializeSingleColumn` in `PrestoVectorSerde` which is used to deserialize data from a single column, since `serializeSingleColumn` performs the inverse operation to this function, it is added alongside it in `PrestoVectorSerde`. Pull Request resolved: facebookincubator#10657 Reviewed By: amitkdutta Differential Revision: D66044754 Pulled By: pedroerp fbshipit-source-id: e509605067920f8207e5a3fa67552badc2ce0eba
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
Adds an expression optimizer that delegates to the native sidecar for evaluation.
Motivation and Context
To resolve RFC-0005
Impact
With this change, we can now simplify expressions using only C++ functions and not Java analogs.
Test Plan
Tests are included. We are independently working on fuzzer testing to evaluate correctness, but that is out of scope for this PR.
Contributor checklist
Release Notes
Please follow release notes guidelines and fill in the release notes below.
If release note is NOT required, use: