fix(incrementalDelivery): fix null bubbling with async iterables #3760
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
When a non-null list is streamed and completeValue for a list item fails, the error should bubble up to the list itself and no further payloads should be sent. Addition of tests within this change demonstrates that this is currently the case only when the field resolver for the list returns an iterable, but not when it returns an async iterable. In the latter case, the null correctly bubbles to the list itself for the given payload (`items: null` rather than `items: [null`), but further payloads are sent.
when a null bubbles up, no further payloads should be sent.
✅ Deploy Preview for compassionate-pike-271cb3 ready!
To edit notification comments on pull requests, go to your Netlify site settings. |
robrichard
approved these changes
Oct 18, 2022
yaacovCR
added a commit
to yaacovCR/graphql-js
that referenced
this pull request
Jan 5, 2023
I introduced use of expect.fail() in graphql#3760 mistakenly believing that expect.fail() will always cause the test to fail. Actually, expect.fail() just throws an error, and that error is wrapped is wrapped by GraphQL and then filtered, so the lines are reached. Fortunately, it's not a problem that the line is reached, as long as it is filtered correctly. The number of calls to next() may be more than 1 (in our case it is 2) depending on the # of ticks that it takes for the deferred fragment to resolve. So the test as written is intending to be overly strict, but is broken and so functions as desires. This commit makes no change to the test semantics, but changes the comment, the coverage ignore statement, and removes the confusing use of expect.fail, so that the test semantics are more clearly apparent.
yaacovCR
added a commit
that referenced
this pull request
Jan 6, 2023
`expect.fail()` was mistakenly introduced in #3760 under the assumption that expect.fail() will always cause the test to fail, and that `.next()` should be called at most once. Actually, thought, `expect.fail()` just throws an error, `.next()` is called more than once, the error is produced and wrapped by GraphQL, but then it is filtered, so the test ultimately passes. Fortunately, that's actually the desired behavior! It's ok if the number of calls to `.next()` is more than 1 (in our case it is 2). The exact number of calls to `.next()` depends on the # of ticks that it takes for the erroring deferred fragment to resolve, which should be as low as possible, but that is a separate objective (with other PRs in the mix already aimed at reducing). So the test as written is intending to be overly strict, but is not actually meeting that goal and so functions as desires. This PR makes no change to the test semantics, but changes the comment, the coverage ignore statement, and removes the potentially confusing use of `expect.fail`, so that the test semantics are more clearly apparent.
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.
extracted from #3754
This is actually more than a simple refactoring, as it includes a bug fix.
When a non-null list is streamed and completeValue for a list item fails, the error should bubble up to the list itself and no further payloads should be sent.
Previously, this was case only when the field resolver for the list returned an iterable, but not when it returned an async iterable. In the latter case, the null correctly bubbles to the list itself for the given payload (
items: nullrather thanitems: [null), but further payloads would still be sent. This PR aligns async iterables to the behavior of iterables where further payloads will not be sent.