Dense reader: fix user buffer offset computation for multi-index queries. #3002
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
|
This pull request has been linked to Shortcut Story #16010: [CZI] Different query results between 2.6.4 and 2.7.1. |
ihnorton
reviewed
Mar 24, 2022
ypatia
reviewed
Mar 24, 2022
| // Compute the correct multipliers. | ||
| uint64_t mult = 1; | ||
| if (subarray.layout() == Layout::COL_MAJOR) { | ||
| for (int32_t d = 0; d < static_cast<int32_t>(dim_num); d++) { |
There was a problem hiding this comment.
why int32_t and not uint32_t ? auto d won't work here?
applies to next for as well
There was a problem hiding this comment.
In the one below, the condition is d >= 0, if this was a uint32_t, after 0, d-- would just wrap around, so the for loop would never exit. I could fix this one but left it as is for consistency.
| if (subarray.layout() == Layout::COL_MAJOR) { | ||
| for (int32_t d = 0; d < static_cast<int32_t>(dim_num); d++) { | ||
| auto saved = mult; | ||
| mult *= range_info[d].multiplier_; |
There was a problem hiding this comment.
I think we are fine here for dense.
…ies. The previous implementation of the dense reader would order the results in the user buffer by N-dimension range index, for example, if there are 2 ranges set per dimensions, the data would be ordered from N-dimension range 1 to 4. This implementation is incorrect as it doesn't really return the data in the expected row/column major. This change computes the correct offset for the data in the user buffers and returns the data in true row/column major order. --- TYPE: IMPROVEMENT DESC: Dense reader: fix user buffer offset computation for multi-index queries.
KiterLuc
force-pushed
the
lr/dense-multi-index-fix/ch16010
branch
from
2061051 to
976cb7b
Compare
|
The backport to To backport manually, run these commands in your terminal: # Fetch latest updates from GitHub
git fetch
# Create a new working tree
git worktree add .worktrees/backport-release-2.7 release-2.7
# Navigate to the new working tree
cd .worktrees/backport-release-2.7
# Create a new branch
git switch --create backport-3002-to-release-2.7
# Cherry-pick the merged commit of this pull request and resolve the conflicts
git cherry-pick --mainline 1 740c589c6d9df044b0197cd9fba08c061363231e
# Push it to GitHub
git push --set-upstream origin backport-3002-to-release-2.7
# Go back to the original working tree
cd ../..
# Delete the working tree
git worktree remove .worktrees/backport-release-2.7Then, create a pull request where the |
|
The backport to To backport manually, run these commands in your terminal: # Fetch latest updates from GitHub
git fetch
# Create a new working tree
git worktree add .worktrees/backport-release-2.8 release-2.8
# Navigate to the new working tree
cd .worktrees/backport-release-2.8
# Create a new branch
git switch --create backport-3002-to-release-2.8
# Cherry-pick the merged commit of this pull request and resolve the conflicts
git cherry-pick --mainline 1 740c589c6d9df044b0197cd9fba08c061363231e
# Push it to GitHub
git push --set-upstream origin backport-3002-to-release-2.8
# Go back to the original working tree
cd ../..
# Delete the working tree
git worktree remove .worktrees/backport-release-2.8Then, create a pull request where the |
ihnorton
pushed a commit
that referenced
this pull request
Mar 29, 2022
…x queries. (#3002) * Dense reader: fix user buffer offset computation for multi-index queries. The previous implementation of the dense reader would order the results in the user buffer by N-dimension range index, for example, if there are 2 ranges set per dimensions, the data would be ordered from N-dimension range 1 to 4. This implementation is incorrect as it doesn't really return the data in the expected row/column major. This change computes the correct offset for the data in the user buffers and returns the data in true row/column major order. * Addressing feedback from ihnorton. * Making sure ranges are sorted. --- TYPE: IMPROVEMENT DESC: Dense reader: fix user buffer offset computation for multi-index queries. (cherry picked from commit 740c589)
ihnorton
pushed a commit
that referenced
this pull request
Mar 29, 2022
…x queries. (#3002) * Dense reader: fix user buffer offset computation for multi-index queries. The previous implementation of the dense reader would order the results in the user buffer by N-dimension range index, for example, if there are 2 ranges set per dimensions, the data would be ordered from N-dimension range 1 to 4. This implementation is incorrect as it doesn't really return the data in the expected row/column major. This change computes the correct offset for the data in the user buffers and returns the data in true row/column major order. * Addressing feedback from ihnorton. * Making sure ranges are sorted. --- TYPE: IMPROVEMENT DESC: Dense reader: fix user buffer offset computation for multi-index queries. (cherry picked from commit 740c589)
Shelnutt2
pushed a commit
that referenced
this pull request
Mar 30, 2022
…on for multi-inde… (#3017) * [2.7] Dense reader: fix user buffer offset computation for multi-index queries. (#3002) * Dense reader: fix user buffer offset computation for multi-index queries. The previous implementation of the dense reader would order the results in the user buffer by N-dimension range index, for example, if there are 2 ranges set per dimensions, the data would be ordered from N-dimension range 1 to 4. This implementation is incorrect as it doesn't really return the data in the expected row/column major. This change computes the correct offset for the data in the user buffers and returns the data in true row/column major order. * Addressing feedback from ihnorton. * Making sure ranges are sorted. --- TYPE: IMPROVEMENT DESC: Dense reader: fix user buffer offset computation for multi-index queries. (cherry picked from commit 740c589) * Fix docs CI breakage due to sphinx/jinja2 (new) incompatible release (#3003) * Fix docs CI breakage due to sphinx/jinja2 (new) incompatible release * Pin jinja for GCS emulator installation Co-authored-by: Seth Shelnutt <[email protected]> (cherry picked from commit 2c61486) * [ci] Pin werkzeug to fix GCS emulator startup (cherry picked from commit f584fb3) Co-authored-by: KiterLuc <[email protected]>
Shelnutt2
pushed a commit
that referenced
this pull request
Mar 30, 2022
* [2.8] Dense reader: fix user buffer offset computation for multi-index queries. (#3002) * Dense reader: fix user buffer offset computation for multi-index queries. The previous implementation of the dense reader would order the results in the user buffer by N-dimension range index, for example, if there are 2 ranges set per dimensions, the data would be ordered from N-dimension range 1 to 4. This implementation is incorrect as it doesn't really return the data in the expected row/column major. This change computes the correct offset for the data in the user buffers and returns the data in true row/column major order. * Addressing feedback from ihnorton. * Making sure ranges are sorted. --- TYPE: IMPROVEMENT DESC: Dense reader: fix user buffer offset computation for multi-index queries. (cherry picked from commit 740c589) * [ci] Pin werkzeug to fix GCS emulator startup (cherry picked from commit f584fb3) Co-authored-by: KiterLuc <[email protected]>
KiterLuc
added a commit
that referenced
this pull request
Mar 31, 2022
Following up on #3002, it was determined we should not sort the input ranges. If the user has a dense vector of domain 1-10 and requests ranges 3-4 then 1-2, we should return the ranges in that order. --- TYPE: IMPROVEMENT DESC: Dense reader: do not sort input ranges.
KiterLuc
added a commit
that referenced
this pull request
Mar 31, 2022
Following up on #3002, it was determined we should not sort the input ranges. If the user has a dense vector of domain 1-10 and requests ranges 3-4 then 1-2, we should return the ranges in that order. --- TYPE: IMPROVEMENT DESC: Dense reader: do not sort input ranges.
ihnorton
pushed a commit
that referenced
this pull request
Apr 5, 2022
Following up on #3002, it was determined we should not sort the input ranges. If the user has a dense vector of domain 1-10 and requests ranges 3-4 then 1-2, we should return the ranges in that order. --- TYPE: IMPROVEMENT DESC: Dense reader: do not sort input ranges.
ihnorton
pushed a commit
that referenced
this pull request
Apr 6, 2022
Following up on #3002, it was determined we should not sort the input ranges. If the user has a dense vector of domain 1-10 and requests ranges 3-4 then 1-2, we should return the ranges in that order. --- TYPE: IMPROVEMENT DESC: Dense reader: do not sort input ranges. (cherry picked from commit 8aa752c)
ihnorton
pushed a commit
that referenced
this pull request
Apr 7, 2022
Following up on #3002, it was determined we should not sort the input ranges. If the user has a dense vector of domain 1-10 and requests ranges 3-4 then 1-2, we should return the ranges in that order. --- TYPE: IMPROVEMENT DESC: Dense reader: do not sort input ranges. (cherry picked from commit 8aa752c)
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.
The previous implementation of the dense reader would order the results
in the user buffer by N-dimension range index, for example, if there are
2 ranges set per dimensions, the data would be ordered from N-dimension
range 1 to 4. This implementation is incorrect as it doesn't really
return the data in the expected row/column major.
This change computes the correct offset for the data in the user buffers
and returns the data in true row/column major order.
TYPE: IMPROVEMENT
DESC: Dense reader: fix user buffer offset computation for multi-index queries.