[FLINK-15430][table-planner-blink] Fix Java 64K method compiling limi… #10737

libenchao · 2020-01-01T15:55:48Z

…tation for blink planner

What is the purpose of the change

Fix blink planner code generation exceeds 64k java method limitation.

Brief change log

Add isSplit field in CodeGeneratorContext
Split expressions when exceeds threshold in ExprCodeGenerator
Use reuseableLocalVariable in GeneratedExpression.deepCopy

Verifying this change

This change added tests and can be verified as follows:

CalcITCase. testLongProjectionList

Does this pull request potentially affect one of the following parts:

Dependencies (does it add or upgrade a dependency): (yes / no)
The public API, i.e., is any changed class annotated with @Public(Evolving): (yes / no)
The serializers: (yes / no / don't know)
The runtime per-record code paths (performance sensitive): (yes / no / don't know)
Anything that affects deployment or recovery: JobManager (and its components), Checkpointing, Yarn/Mesos, ZooKeeper: (yes / no / don't know)
The S3 file system connector: (yes / no / don't know)

Documentation

Does this pull request introduce a new feature? (yes / no)
If yes, how is the feature documented? (not applicable / docs / JavaDocs / not documented)

…tation for blink planner

flinkbot · 2020-01-01T15:59:34Z

Thanks a lot for your contribution to the Apache Flink project. I'm the @flinkbot. I help the community
to review your pull request. We will use this comment to track the progress of the review.

Automated Checks

Last check on commit fdd8856 (Wed Jan 01 15:59:34 UTC 2020)

Warnings:

No documentation files were touched! Remember to keep the Flink docs up to date!

_{Mention the bot in a comment to re-run the automated checks.}

Review Progress

❓ 1. The [description] looks good.
❓ 2. There is [consensus] that the contribution should go into to Flink.
❓ 3. Needs [attention] from.
❓ 4. The change fits into the overall [architecture].
❓ 5. Overall code [quality] is good.

Please see the Pull Request Review Guide for a full explanation of the review process.

Details

The Bot is tracking the review progress through labels. Labels are applied according to the order of the review items. For consensus, approval by a Flink committer of PMC member is required

Bot commands

The @flinkbot bot supports the following commands:

@flinkbot approve description to approve one or more aspects (aspects: description, consensus, architecture and quality)
@flinkbot approve all to approve all aspects
@flinkbot approve-until architecture to approve everything until architecture
@flinkbot attention @username1 [@username2 ..] to require somebody's attention
@flinkbot disapprove architecture to remove an approval you gave earlier

flinkbot · 2020-01-01T16:48:49Z

CI report:

fdd8856 Travis: FAILURE Azure: FAILURE
263bb0c Travis: SUCCESS Azure: SUCCESS
fe83218 Travis: FAILURE Azure: FAILURE
c5bcead Travis: SUCCESS Azure: SUCCESS
780431a Travis: SUCCESS Azure: SUCCESS

Bot commands

The @flinkbot bot supports the following commands:

@flinkbot run travis re-run the last Travis build
@flinkbot run azure re-run the last Azure build

libenchao · 2020-01-02T03:06:59Z

@flinkbot run travis

wuchong · 2020-01-02T08:31:06Z

cc @JingsongLi

JingsongLi

Thanks @libenchao , left two thought:

Can we just copy codes from legacy planner? It seems that there are some corner cases in CodeGenerator.makeReusableInSplits.
If we just have some codes like boolean $myNullTerm = ..., will lead to compile exception after code splitting?

libenchao · 2020-01-02T14:47:56Z

@JingsongLi Thanks for the review.

I'll try to align the behavior with old planner, and give you feedback after that.
Yes. I also realized that, and fixed all places I found for now. However, this can only happen when the Expression is longer than TableConfig.maxGeneratedCodeLength.

This reverts commit 263bb0c.

libenchao · 2020-01-05T07:09:34Z

@JingsongLi @wuchong I go over the implementation again, and find that it indeed has many corner cases.

I changed TableConfig.maxGeneratedCodeLength to 1, and run all the tests in blink planner, 762/6117 tests failed.
For comparison, I did the same thing to old planner, and found 21/2321 tests failed.

So, it's indeed not easy to do code split currently. And I propose to narrow down the scope of this simple fix, which just targets long projection list. Then we can just focus on CalcCodeGenerator and add a param allowSplit to ExprCodeGenerator.generateResultExpression.
What's your opinion?

JingsongLi · 2020-01-06T02:39:11Z

@libenchao Thanks for you updating.

I changed TableConfig.maxGeneratedCodeLength to 1, and run all the tests in blink planner

Good job!

narrow down the scope of this simple fix, which just targets long projection list

+1

JingsongLi · 2020-01-06T03:43:21Z

...-planner-blink/src/main/scala/org/apache/flink/table/planner/codegen/ExprCodeGenerator.scala

+        val methodName = newName("split")
+        val method =
+          s"""
+            |private void $methodName() throw Exception {


Thanks for the review, fixed.

JingsongLi · 2020-01-06T06:37:51Z

This temporary fixing looks good to me. Now we limit this fixing to the Calc only. It's not a perfect solution. But it does solve the problem.

@wuchong @KurtYoung What do you think?

wuchong

Thanks @libenchao @JingsongLi . I'm fine with current temporary solution.

I only left a minor comment about the method naming.

wuchong · 2020-01-06T07:12:47Z

...anner-blink/src/main/scala/org/apache/flink/table/planner/codegen/CodeGeneratorContext.scala

+  /**
+   * Flag indicating whether split has occurred.
+   */
+  private var isSplit = false


Rename the variable to isCodeSplit to make it clear that the code is splitted?
The comment can also be improved Flag that indicates whether the generated code is split into several methods.

wuchong · 2020-01-06T07:20:42Z

...anner-blink/src/main/scala/org/apache/flink/table/planner/codegen/CodeGeneratorContext.scala

    reusableLocalVariableStatements(methodName) = mutable.LinkedHashSet[String]()
  }

+  def setSplit(): Unit = {


enableCodeSplit() or setCodeSplit()? Please also add Javadoc above this method.

libenchao · 2020-01-06T07:43:18Z

@wuchong Thanks for the review. I've updated the code.

wuchong

LGTM.

…o prevent compiler exceptions (#10737)

[FLINK-15430][table-planner-blink] Fix Java 64K method compiling limi…

fdd8856

…tation for blink planner

rmetzger added the review=description? label Jan 1, 2020

rmetzger added the component=TableSQL/Planner label Jan 1, 2020

libenchao changed the title ~~[FLINK-15430][table-planner-blink] Fix Java 64K method compiling limi…~~ [WIP][FLINK-15430][table-planner-blink] Fix Java 64K method compiling limi… Jan 2, 2020

Change MatchCodeGenerator to declare reusable local variables

263bb0c

libenchao changed the title ~~[WIP][FLINK-15430][table-planner-blink] Fix Java 64K method compiling limi…~~ [FLINK-15430][table-planner-blink] Fix Java 64K method compiling limi… Jan 2, 2020

JingsongLi reviewed Jan 2, 2020

View reviewed changes

Revert "Change MatchCodeGenerator to declare reusable local variables"

5852f35

This reverts commit 263bb0c.

Add allowSplit param, to narrow down code split scope

fe83218

JingsongLi reviewed Jan 6, 2020

View reviewed changes

Fix typo throw to throws

c5bcead

wuchong reviewed Jan 6, 2020

View reviewed changes

Refine naming and comment

780431a

wuchong approved these changes Jan 6, 2020

View reviewed changes

wuchong merged commit 744ba5d into apache:master Jan 6, 2020

wuchong pushed a commit that referenced this pull request Jan 6, 2020

[FLINK-15430][table-planner-blink] Split generated methods for Calc t…

03eeb11

…o prevent compiler exceptions (#10737)

[FLINK-15430][table-planner-blink] Fix Java 64K method compiling limi… #10737

[FLINK-15430][table-planner-blink] Fix Java 64K method compiling limi… #10737

Uh oh!

Conversation

libenchao commented Jan 1, 2020

What is the purpose of the change

Brief change log

Verifying this change

Does this pull request potentially affect one of the following parts:

Documentation

Uh oh!

flinkbot commented Jan 1, 2020

Automated Checks

Review Progress

Uh oh!

flinkbot commented Jan 1, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

CI report:

Uh oh!

libenchao commented Jan 2, 2020

Uh oh!

wuchong commented Jan 2, 2020

Uh oh!

JingsongLi left a comment

Choose a reason for hiding this comment

Uh oh!

libenchao commented Jan 2, 2020

Uh oh!

libenchao commented Jan 5, 2020

Uh oh!

JingsongLi commented Jan 6, 2020

Uh oh!

JingsongLi Jan 6, 2020

Choose a reason for hiding this comment

Uh oh!

libenchao Jan 6, 2020

Choose a reason for hiding this comment

Uh oh!

JingsongLi commented Jan 6, 2020

Uh oh!

wuchong left a comment

Choose a reason for hiding this comment

Uh oh!

wuchong Jan 6, 2020

Choose a reason for hiding this comment

Uh oh!

wuchong Jan 6, 2020

Choose a reason for hiding this comment

Uh oh!

libenchao commented Jan 6, 2020

Uh oh!

wuchong left a comment

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

flinkbot commented Jan 1, 2020 •

edited

Loading