3.x: Change how the cause of CompositeException is generated #6748

akarnokd · 2019-12-06T10:22:37Z

This PR changes how CompositeException.getCause creates a cause exception on demand. In 1.x and 2.x, the code tried to link up the various inner exceptions via their initCause, which was in on itself fishy as well as could lead to excessive memory usage.

Instead, the new code will present the inner exceptions as part of a formatted message, which in theory, should be still picked up by IDE exception listings and allow navigation:

Multiple exceptions (2)
|-- io.reactivex.rxjava3.exceptions.TestException: ex3
    at io.reactivex.rxjava3.exceptions.CompositeExceptionTest.nestedMultilineMessage(CompositeExceptionTest.java:341)
|-- io.reactivex.rxjava3.exceptions.TestException: ex4
    at io.reactivex.rxjava3.exceptions.CompositeExceptionTest.nestedMultilineMessage(CompositeExceptionTest.java:342)
  |-- io.reactivex.rxjava3.exceptions.CompositeException: 2 exceptions occurred. 
      at io.reactivex.rxjava3.exceptions.CompositeExceptionTest.nestedMultilineMessage(CompositeExceptionTest.java:337)
    |-- io.reactivex.rxjava3.exceptions.CompositeException.ExceptionOverview: 
        Multiple exceptions (2)
        |-- io.reactivex.rxjava3.exceptions.TestException: ex1
            at io.reactivex.rxjava3.exceptions.CompositeExceptionTest.nestedMultilineMessage(CompositeExceptionTest.java:335)
        |-- io.reactivex.rxjava3.exceptions.TestException: ex2
            at io.reactivex.rxjava3.exceptions.CompositeExceptionTest.nestedMultilineMessage(CompositeExceptionTest.java:336)

There are a few formatting conveniences:

If there is only one inner exception, the CompositeException's cause will be simply that exception. This can happen when the very same exception is aggregated into the composite and get deduplicated.
If an inner exception's message is multi-lined, the message and cause traces should be indented properly.
Reoccurring causes are not expanded over and over:

Multiple exceptions (2)
|-- io.reactivex.rxjava3.exceptions.TestException: ex1
    at io.reactivex.rxjava3.exceptions.CompositeExceptionTest.reoccurringException(CompositeExceptionTest.java:316)
  |-- io.reactivex.rxjava3.exceptions.TestException: ex0
      at io.reactivex.rxjava3.exceptions.CompositeExceptionTest.reoccurringException(CompositeExceptionTest.java:315)
|-- io.reactivex.rxjava3.exceptions.TestException: ex2
    at io.reactivex.rxjava3.exceptions.CompositeExceptionTest.reoccurringException(CompositeExceptionTest.java:319)
  |-- io.reactivex.rxjava3.exceptions.TestException: ex1
      at io.reactivex.rxjava3.exceptions.CompositeExceptionTest.reoccurringException(CompositeExceptionTest.java:316)
      |-- (cause not expanded again) io.reactivex.rxjava3.exceptions.TestException: ex0

Currently, only the first line of the stacktraces are shown because it can get quite long (and thus memory consuming) to list them all. Maybe a system parameter can be introduced to control the verbosity.

Fixes #6747

codecov · 2019-12-06T10:45:44Z

Codecov Report

Merging #6748 into 3.x will decrease coverage by 0.06%.
The diff coverage is 100%.

@@             Coverage Diff              @@
##                3.x    #6748      +/-   ##
============================================
- Coverage     98.16%   98.09%   -0.07%     
  Complexity     6188     6188              
============================================
  Files           677      677              
  Lines         44619    44626       +7     
  Branches       6157     6161       +4     
============================================
- Hits          43800    43776      -24     
- Misses          292      309      +17     
- Partials        527      541      +14

Impacted Files	Coverage Δ	Complexity Δ
...activex/rxjava3/exceptions/CompositeException.java	`100% <100%> (ø)`	`35 <0> (-1)`	⬇️
...nternal/operators/observable/ObservableCreate.java	`79.48% <0%> (-16.24%)`	`2% <0%> (ø)`
...rnal/operators/flowable/FlowableFlatMapSingle.java	`92.44% <0%> (-2.91%)`	`2% <0%> (ø)`
...rnal/operators/observable/ObservableSwitchMap.java	`91.57% <0%> (-2.11%)`	`3% <0%> (ø)`
...ternal/operators/completable/CompletableMerge.java	`95.94% <0%> (-1.36%)`	`2% <0%> (ø)`
...rnal/operators/observable/ObservableObserveOn.java	`98.69% <0%> (-1.31%)`	`3% <0%> (ø)`
...perators/observable/ObservableMergeWithSingle.java	`99.05% <0%> (-0.95%)`	`2% <0%> (ø)`
.../operators/observable/ObservableCombineLatest.java	`96.17% <0%> (-0.64%)`	`6% <0%> (ø)`
...3/internal/operators/flowable/FlowableGroupBy.java	`95.59% <0%> (-0.63%)`	`3% <0%> (ø)`
...ternal/operators/observable/ObservableGroupBy.java	`95.08% <0%> (-0.55%)`	`2% <0%> (ø)`
... and 14 more

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 1e29958...ae3c299. Read the comment docs.

akarnokd added 2 commits December 6, 2019 11:09

3.x: Change how the cause of CompositeException is generated

1a19905

Fix reoccurrence formatting.

67074ae

akarnokd added Shortcoming 3.x labels Dec 6, 2019

akarnokd added this to the 3.0 milestone Dec 6, 2019

Fix a mistake in the unit test

ae3c299

vanniktech approved these changes Dec 7, 2019

View reviewed changes

akarnokd merged commit 0f0e219 into ReactiveX:3.x Dec 7, 2019

akarnokd deleted the CompositeExceptionCauseFix branch December 7, 2019 21:36

akarnokd mentioned this pull request Dec 11, 2019

What's Different in 3.x (tracking issue) #6515

Closed

70 tasks

akarnokd mentioned this pull request Dec 27, 2019

3.0.0-RC7 release preparations #6803

Closed

akarnokd mentioned this pull request Aug 18, 2020

[3.x] RxJavaAssemblyTracking stack trace is not printed akarnokd/RxJavaExtensions#76

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

3.x: Change how the cause of CompositeException is generated #6748

3.x: Change how the cause of CompositeException is generated #6748

Uh oh!

akarnokd commented Dec 6, 2019 •

edited

Loading

Uh oh!

codecov bot commented Dec 6, 2019

Uh oh!

Uh oh!

3.x: Change how the cause of CompositeException is generated #6748

3.x: Change how the cause of CompositeException is generated #6748

Uh oh!

Conversation

akarnokd commented Dec 6, 2019 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

codecov bot commented Dec 6, 2019

Codecov Report

Uh oh!

Uh oh!

akarnokd commented Dec 6, 2019 •

edited

Loading