GH-34474: [C++] Detect and raise an error if a join will need too much key data #35087
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
…which is not supported by the RowArray. We return an invalid status. We cannot support that large of a join without spilling.
|
|
github-actions
bot
added
Component: C++
awaiting committer review
zeroshade
approved these changes
Apr 12, 2023
github-actions
bot
added
awaiting merge
jorisvandenbossche
approved these changes
Apr 13, 2023
|
Benchmark runs are scheduled for baseline = 6432a23 and contender = a1d1373. a1d1373 is a master commit associated with this PR. Results will be available as each benchmark for each run completes. |
|
['Python', 'R'] benchmarks have high level of regressions. |
raulcd
pushed a commit
that referenced
this pull request
Apr 17, 2023
…h key data (#35087) ### Rationale for this change This fixes the test in #34474 though there are likely still other bad scenarios with large joins. I've fixed this one since the behavior (invalid data) is particularly bad. Most of the time if there is too much data I'm guessing we probably just crash. Still, I think a test suite of some kind stressing large joins would be good to have. Perhaps this could be added if someone finds time to work on join spilling. ### What changes are included in this PR? If the join will require more than 4GiB of key data it should now return an invalid status instead of invalid data. ### Are these changes tested? No. I created a unit test but it requires over 16GiB of RAM (Besides the input data itself (4GiB), by the time you get 4GiB of key data there are various other join state buffers that also grow. The test also took nearly a minute to run. I think investigation and creation of a test suite for large joins is probably a standalone effort. ### Are there any user-facing changes? No. * Closes: #34474 Authored-by: Weston Pace <[email protected]> Signed-off-by: Joris Van den Bossche <[email protected]>
liujiacheng777
pushed a commit
to LoongArch-Python/arrow
that referenced
this pull request
May 11, 2023
…oo much key data (apache#35087) ### Rationale for this change This fixes the test in apache#34474 though there are likely still other bad scenarios with large joins. I've fixed this one since the behavior (invalid data) is particularly bad. Most of the time if there is too much data I'm guessing we probably just crash. Still, I think a test suite of some kind stressing large joins would be good to have. Perhaps this could be added if someone finds time to work on join spilling. ### What changes are included in this PR? If the join will require more than 4GiB of key data it should now return an invalid status instead of invalid data. ### Are these changes tested? No. I created a unit test but it requires over 16GiB of RAM (Besides the input data itself (4GiB), by the time you get 4GiB of key data there are various other join state buffers that also grow. The test also took nearly a minute to run. I think investigation and creation of a test suite for large joins is probably a standalone effort. ### Are there any user-facing changes? No. * Closes: apache#34474 Authored-by: Weston Pace <[email protected]> Signed-off-by: Joris Van den Bossche <[email protected]>
ArgusLi
pushed a commit
to Bit-Quill/arrow
that referenced
this pull request
May 15, 2023
…oo much key data (apache#35087) ### Rationale for this change This fixes the test in apache#34474 though there are likely still other bad scenarios with large joins. I've fixed this one since the behavior (invalid data) is particularly bad. Most of the time if there is too much data I'm guessing we probably just crash. Still, I think a test suite of some kind stressing large joins would be good to have. Perhaps this could be added if someone finds time to work on join spilling. ### What changes are included in this PR? If the join will require more than 4GiB of key data it should now return an invalid status instead of invalid data. ### Are these changes tested? No. I created a unit test but it requires over 16GiB of RAM (Besides the input data itself (4GiB), by the time you get 4GiB of key data there are various other join state buffers that also grow. The test also took nearly a minute to run. I think investigation and creation of a test suite for large joins is probably a standalone effort. ### Are there any user-facing changes? No. * Closes: apache#34474 Authored-by: Weston Pace <[email protected]> Signed-off-by: Joris Van den Bossche <[email protected]>
rtpsw
pushed a commit
to rtpsw/arrow
that referenced
this pull request
May 16, 2023
…oo much key data (apache#35087) ### Rationale for this change This fixes the test in apache#34474 though there are likely still other bad scenarios with large joins. I've fixed this one since the behavior (invalid data) is particularly bad. Most of the time if there is too much data I'm guessing we probably just crash. Still, I think a test suite of some kind stressing large joins would be good to have. Perhaps this could be added if someone finds time to work on join spilling. ### What changes are included in this PR? If the join will require more than 4GiB of key data it should now return an invalid status instead of invalid data. ### Are these changes tested? No. I created a unit test but it requires over 16GiB of RAM (Besides the input data itself (4GiB), by the time you get 4GiB of key data there are various other join state buffers that also grow. The test also took nearly a minute to run. I think investigation and creation of a test suite for large joins is probably a standalone effort. ### Are there any user-facing changes? No. * Closes: apache#34474 Authored-by: Weston Pace <[email protected]> Signed-off-by: Joris Van den Bossche <[email protected]>
oliviermeslin
added a commit
to oliviermeslin/arrow
that referenced
this pull request
Sep 13, 2023
[PR 35087](apache#35087) introduced an explicit fail in large joins with Acero when key data is larger than 4GB (solving the problem reported by [issue 34474](apache#34474)). However, I think (but I'm not sure) that this quick fix is too restrictive because the total size condition is applied to the total size of tables to be joined, rather than to the size of keys. As a consequence, Acero fails when trying to merge large tables, even when the size of key data is well below 4 GB. This PR modifies the source code so that the logical test only verifies whether the total size of _key variable_ is below 4 GB.
This was referenced Sep 13, 2023
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.
Rationale for this change
This fixes the test in #34474 though there are likely still other bad scenarios with large joins. I've fixed this one since the behavior (invalid data) is particularly bad. Most of the time if there is too much data I'm guessing we probably just crash. Still, I think a test suite of some kind stressing large joins would be good to have. Perhaps this could be added if someone finds time to work on join spilling.
What changes are included in this PR?
If the join will require more than 4GiB of key data it should now return an invalid status instead of invalid data.
Are these changes tested?
No. I created a unit test but it requires over 16GiB of RAM (Besides the input data itself (4GiB), by the time you get 4GiB of key data there are various other join state buffers that also grow. The test also took nearly a minute to run. I think investigation and creation of a test suite for large joins is probably a standalone effort.
Are there any user-facing changes?
No.