Use Arrow PythonFile for remote CSV storage #9376
Merged
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
…zation (seems like it was wrong)
rjzamora
added
2 - In Progress
rjzamora
added
3 - Ready for Review
rjzamora
commented
Oct 5, 2021
Codecov Report
@@ Coverage Diff @@
## branch-21.12 #9376 +/- ##
================================================
- Coverage 10.79% 10.76% -0.03%
================================================
Files 116 116
Lines 18869 19476 +607
================================================
+ Hits 2036 2096 +60
- Misses 16833 17380 +547
Continue to review full report at Codecov.
|
brandon-b-miller
approved these changes
Oct 6, 2021
|
@gpucibot merge |
rapids-bot bot
pushed a commit
that referenced
this pull request
Oct 7, 2021
This is a follow-up to #9304, and is more-or-less the ORC version of #9376 These changes will enable partial IO to behave "correctly" for `cudf.read_orc` from remote storage. Simpe multi-stripe file example: ```python # After this PR %time gdf = cudf.read_orc(orc_path, stripes=[0], storage_options=storage_options) CPU times: user 579 ms, sys: 166 ms, total: 744 ms Wall time: 2.38 s # Before this PR %time gdf = cudf.read_orc(orc_path, stripes=[0], storage_options=storage_options) CPU times: user 3.9 s, sys: 1.47 s, total: 5.37 s Wall time: 8.5 s ``` Authors: - Richard (Rick) Zamora (https://github.com/rjzamora) Approvers: - Charles Blackmon-Luca (https://github.com/charlesbluca) URL: #9377
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Outdated suggestions cannot be applied.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.
This is a simple follow-up to #9304 and #9265 meant to achieve the following:
After this PR, the default behavior of
cudf.read_csvwill be to convert fsspec-basedAbstractBufferedFileobjects to ArrowPythonFileobjects for non-local file systems. SincePythonFileobjects inherit fromNativeFileobjects, libcudf can seek/read distinct byte ranges without requiring the entire file to be read into host memory (i.e. the default behavior enables proper partial IO from remote storage)Optimized fsspec data transfer for remote file-systems #9265 recently added an fsspec-based optimization for transfering csv byte ranges into local memory. That optimization already allowed us to avoid a full file transfer when a specific
byte_rangeis specified to thecudf.read_csvcall. However, the simpler approach introduced in this PR is (1) more general, (2) easier to maintain, and (3) demonstrates comparable performance. Therefore, this PR also rolls back one of the less-maintainable optimizations added in Optimized fsspec data transfer for remote file-systems #9265 (local buffer clipping).