fix: Encoding of List offsets was incorrect when slice offsets begin with zero #6805
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
…with zero When encoding offsets the code had an optimization to reuse the offsets if the first offset was zero assuming the slice already pointed to first element. But the offset can also be zero if all previous lists were empty. When this occured it mold make all lists in the slice as empty, even if they shouldn't be.
etseidl
reviewed
Nov 26, 2024
arrow-ipc/src/writer.rs
Outdated
| let offsets = match start_offset.as_usize() { | ||
| 0 => offsets.clone(), | ||
| let offsets: Buffer = match start_offset.as_usize() { | ||
| 0 => offset_slice.iter().copied().collect(), |
There was a problem hiding this comment.
Would something like 0 => Buffer::from_slice_ref(offset_slice) be better performing? Just curious as I'm still learning. 😄
There was a problem hiding this comment.
Good call, that would be more efficient than the iterating approach.
… through the slice.
tustvold
reviewed
Nov 27, 2024
arrow-ipc/src/writer.rs
Outdated
| let offsets = match start_offset.as_usize() { | ||
| 0 => offsets.clone(), | ||
| let offsets: Buffer = match start_offset.as_usize() { | ||
| 0 => Buffer::from_slice_ref(offset_slice), |
There was a problem hiding this comment.
Could we instead slice the Buffer passed in, this would avoid copying?
tustvold
reviewed
Nov 27, 2024
arrow-ipc/src/writer.rs
Outdated
| let offsets = match start_offset.as_usize() { | ||
| 0 => offsets.clone(), | ||
| let offsets: Buffer = match start_offset.as_usize() { | ||
| 0 => Buffer::from_slice_ref(offset_slice), |
There was a problem hiding this comment.
Something like
Suggested change
| 0 => Buffer::from_slice_ref(offset_slice), | |
| 0 => { | |
| let size = std::mem::sizeof::<O>(); | |
| offsets.slice_with_length(data.offset()*size, (data.offset() + data.length() + 1) * size) | |
| } |
There was a problem hiding this comment.
Good recommendation. code is changed.
|
CI seems to be legitimately failing |
missed that. fixed |
|
Looks good to me, thank you |
This was referenced Dec 16, 2024
alamb
pushed a commit
to alamb/arrow-rs
that referenced
this pull request
Jan 5, 2025
…with zero (apache#6805) * fix: Encoding of List offsets was incorrect when slice offsets begin with zero When encoding offsets the code had an optimization to reuse the offsets if the first offset was zero assuming the slice already pointed to first element. But the offset can also be zero if all previous lists were empty. When this occured it mold make all lists in the slice as empty, even if they shouldn't be. * Use Buffer::from_slice_ref which will be faster as it doesn't iterate through the slice. * Avoid copying * Explicitly reference std::mem::size_of
alexwilcoxson-rel
pushed a commit
to relativityone/arrow-rs
that referenced
this pull request
Jan 8, 2025
…with zero (apache#6805) * fix: Encoding of List offsets was incorrect when slice offsets begin with zero When encoding offsets the code had an optimization to reuse the offsets if the first offset was zero assuming the slice already pointed to first element. But the offset can also be zero if all previous lists were empty. When this occured it mold make all lists in the slice as empty, even if they shouldn't be. * Use Buffer::from_slice_ref which will be faster as it doesn't iterate through the slice. * Avoid copying * Explicitly reference std::mem::size_of (cherry picked from commit fe7e71a)
alamb
added a commit
that referenced
this pull request
Jan 8, 2025
…with zero (#6805) (#6943) * fix: Encoding of List offsets was incorrect when slice offsets begin with zero When encoding offsets the code had an optimization to reuse the offsets if the first offset was zero assuming the slice already pointed to first element. But the offset can also be zero if all previous lists were empty. When this occured it mold make all lists in the slice as empty, even if they shouldn't be. * Use Buffer::from_slice_ref which will be faster as it doesn't iterate through the slice. * Avoid copying * Explicitly reference std::mem::size_of Co-authored-by: Michael Maletich <[email protected]>
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.
When encoding offsets the code had an optimization to reuse the offsets if the first offset was zero assuming the slice already pointed to first element. But the offset can also be zero if all previous lists were empty. When this occured it mold make all lists in the slice as empty, even if they shouldn't be.
Which issue does this PR close?
Closes #6803 .
Rationale for this change
Fixing a bug that can lead to loss of information when encoding record batches.
What changes are included in this PR?
Are there any user-facing changes?
No, though some clients will experience smaller encoded messages since only the relevant slice will be sent.