Parquet reader: bug fix for a num_rows/skip_rows corner case, w/optimization for nested preprocessing #11752
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
…taining a mix of nested and non-nested types would result in incorrect row counts for the non-nested types. Also optimizes the preprocess path so that non-nested types do not end up getting visited by the kernel.
nvdbaranec
added
bug
…ists. Fixed an additional issue in the decoding where flat column types underneath structs could end up ignoring skip_rows/num_rows.
github-actions
bot
added
CMake
|
Wrongly switched branch. You may need to reverse changes from branch 22.12 and up merge with 22.10 instead. |
nvdbaranec
removed
gpuCI
Python
|
rerun tests |
hyperbolic2346
requested changes
Oct 3, 2022
Comment on lines
+1864
to
+1865
| // it would be better/safer to be checking (schema.max_repetition_level > 0) here, but there's | ||
| // no easy way to get at that info here. we'd have to move this function into reader_impl.cu |
There was a problem hiding this comment.
Should there be an issue for this?
There was a problem hiding this comment.
You were right. This ended up being a bug :) If you have a struct at the top of a nested hierarchy, this logic fails. The max_repetition_level check is the correct one.
PointKernel
reviewed
Oct 3, 2022
PointKernel
approved these changes
Oct 5, 2022
Codecov ReportBase: 87.40% // Head: 87.50% // Increases project coverage by
Additional details and impacted files@@ Coverage Diff @@
## branch-22.12 #11752 +/- ##
================================================
+ Coverage 87.40% 87.50% +0.09%
================================================
Files 133 133
Lines 21833 21826 -7
================================================
+ Hits 19084 19099 +15
+ Misses 2749 2727 -22
Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here. ☔ View full report at Codecov. |
ttnghia
approved these changes
Oct 5, 2022
hyperbolic2346
approved these changes
Oct 5, 2022
|
@gpucibot merge |
rapids-bot bot
pushed a commit
that referenced
this pull request
Oct 13, 2022
Fixes NVIDIA/spark-rapids#6718 There was a bug introduced recently #11752 where an insufficient check for whether an input column contained repetition information could cause incorrect results for column hierarchies with structs at the root. Authors: - https://github.com/nvdbaranec Approvers: - Jim Brennan (https://github.com/jbrennan333) - Nghia Truong (https://github.com/ttnghia) - Mike Wilson (https://github.com/hyperbolic2346) URL: #11910
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.
Fixes an issue where using user bounds with parquet files containing both nested and non-nested types could result in incorrect row counts for the non-nested columns. Originally reported by @etseidl
The nature of the fix also implements a longstanding desired optimization: when running the preprocess step for nested types, ignore pages for non-nested hierarchies. This can result in significant speedups for files containing only a few nested columns.
The tests added for this PR seem to tease a bug in the parquet writer into happening (#11748) so I will leave this as a draft until that issue is resolved.