Upgrade to arrow/parquet 55, and object_store to 0.12.0 and pyo3 to 0.24.0
#15466
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
alamb
changed the title
8 tasks
github-actions
bot
added
core
alamb
commented
Apr 7, 2025
| @@ -166,7 +166,7 @@ impl ObjectStore for BlockingObjectStore { | |||
| fn list( | |||
| &self, | |||
| prefix: Option<&Path>, | |||
| ) -> BoxStream<'_, object_store::Result<ObjectMeta>> { | |||
| ) -> BoxStream<'static, object_store::Result<ObjectMeta>> { | |||
There was a problem hiding this comment.
| @@ -305,7 +305,7 @@ impl FileOpener for ArrowOpener { | |||
| )?; | |||
| // read footer according to footer_len | |||
| let get_option = GetOptions { | |||
| range: Some(GetRange::Suffix(10 + footer_len)), | |||
| range: Some(GetRange::Suffix(10 + (footer_len as u64))), | |||
There was a problem hiding this comment.
The changes to usize/u64 are for better wasm support, see
| self.inner.get_metadata() | ||
| fn get_metadata<'a>( | ||
| &'a mut self, | ||
| options: Option<&'a ArrowReaderOptions>, |
There was a problem hiding this comment.
| @@ -44,7 +44,7 @@ datafusion-common = { workspace = true, default-features = true } | |||
| datafusion-expr = { workspace = true } | |||
| futures = { workspace = true } | |||
| log = { workspace = true } | |||
| object_store = { workspace = true } | |||
| object_store = { workspace = true, features = ["fs"] } | |||
There was a problem hiding this comment.
| @@ -884,7 +884,7 @@ SELECT extract(day from arrow_cast('14400 minutes', 'Interval(DayTime)')) | |||
| query I | |||
| SELECT extract(minute from arrow_cast('14400 minutes', 'Interval(DayTime)')) | |||
| ---- | |||
| 14400 | |||
| 0 | |||
There was a problem hiding this comment.
alamb
commented
Apr 8, 2025
| use rand::Rng; | ||
| use rand::SeedableRng; | ||
|
|
There was a problem hiding this comment.
I inlined the small amount of code from bench_util so this benchmark is standalone and easier to understand what is tested
38 tasks
|
🤖 |
|
🤖: Benchmark completed Details
|
|
I tried briefly to reproduce the performance improvements reported above and it seems like I can: andrewlamb@Andrews-MacBook-Pro-2:~/Downloads$ for i in `seq 1 5`; do datafusion-cli -f q14.sql ; done | grep seconds
Elapsed 0.374 seconds.
Elapsed 0.386 seconds.
Elapsed 0.378 seconds.
Elapsed 0.366 seconds.
Elapsed 0.366 seconds.I poked around and I can't quite figure out if the improvement is related to concat batches or maybe reducing the number of IOs for reading the parquet metadata 🤔 |
|
Still seeing if this is just noise, but here are flame graphs for Q14 from my machine if anyone else wants to stare at them: This PR: main: |
My theory is that the improvement is due to @rluvaton 's PR to improve concat I thought it might be related to improved pre-fetching / fewer IOs due to However, I tried an experiment on main to reduce IO and it doesn't seem to have changed anything |
alamb
changed the title
object_store to 0.12.0 and pyo3 to 0.24.0
This should be easy to confirm with |
|
Happy to improve performance 😄 I got more in my chamber |
|
Thanks everyone! |
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 PR upgrades to the latest arrow/parquet release and all the dependencies.
I used this PR to test the arrow/parquet 55 release prior to creating an RC.
Some benchmark results show performance is as good or better (see below)
I havn't figured out exactly why but I theorize it is at least in part due to @rluvaton 's PR to improve
concatconcatperformance, and addappend_arrayfor some array builder implementations arrow-rs#7309