You are viewing a plain text version of this content. The canonical link for it is here.
Posted to github@arrow.apache.org by "tustvold (via GitHub)" <gi...@apache.org> on 2023/10/16 17:29:06 UTC
[PR] Assume Records Not Split Across Pages (#4921) [arrow-rs]
tustvold opened a new pull request, #4943:
URL: https://github.com/apache/arrow-rs/pull/4943
# Which issue does this PR close?
<!--
We generally require a GitHub issue to be filed for all bug fixes and enhancements and this helps us generate change logs for our releases. You can link an issue to this PR using the GitHub syntax. For example `Closes #123` indicates that this PR will close issue #123.
-->
Closes #4921
# Rationale for this change
<!--
Why are you proposing this change? If this is already explained clearly in the issue then this section is not needed.
Explaining clearly why changes are proposed helps reviewers understand your changes and offer better suggestions for fixes.
-->
The parquet specification, especially before version 2, did not make it especially clear that records should not be split across pages, and prior to https://github.com/apache/arrow-rs/pull/4327 the Rust writer would do this in certain situations.
As a result https://github.com/apache/arrow-rs/pull/4376 attempted to not assume page boundaries delimit records, and instead only relied on row group boundaries delimiting records.
Unfortunately this logic is incomplete:
* When reading or skipping the last record of a page, this record count will not be "counted", if the next page is then skipped entirely, the counts get out of sync with this record never actually accounted for
* When using the offset index to prune IO the reader tries to read pages beyond those necessary for the RowSelection, if a RowSelector ends on a page boundary
# What changes are included in this PR?
<!--
There is no need to duplicate the description in the issue here but it is sometimes worth providing a summary of the individual changes in this PR.
-->
This PR removes the workarounds, and simply assumes pages delimit records.
# Are there any user-facing changes?
<!--
If there are user-facing changes then we may require documentation to be updated before approving the PR.
-->
<!---
If there are any breaking changes to public APIs, please add the `breaking change` label.
-->
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] Assume Parquet Records Not Split Across Pages (#4921) [arrow-rs]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on code in PR #4943:
URL: https://github.com/apache/arrow-rs/pull/4943#discussion_r1361151497
##########
parquet/src/arrow/async_reader/mod.rs:
##########
@@ -1677,4 +1682,85 @@ mod tests {
assert!(sbbf.check(&"Hello"));
assert!(!sbbf.check(&"Hello_Not_Exists"));
}
+
+ #[tokio::test]
Review Comment:
I verified this test fails as follows without the code changes in this PR
```
called `Result::unwrap()` on an `Err` value: ArrowError("Parquet argument error: Parquet error: StructArrayReader out of sync in read_records, expected 1 read, got 0")
thread 'arrow::async_reader::tests::test_nested_skip' panicked at parquet/src/arrow/async_reader/mod.rs:1760:29:
called `Result::unwrap()` on an `Err` value: ArrowError("Parquet argument error: Parquet error: StructArrayReader out of sync in read_records, expected 1 read, got 0")
stack backtrace:
0: rust_begin_unwind
at /rustc/cc66ad468955717ab92600c770da8c1601a4ff33/library/std/src/panicking.rs:595:5
1: core::panicking::panic_fmt
at /rustc/cc66ad468955717ab92600c770da8c1601a4ff33/library/core/src/panicking.rs:67:14
2: core::result::unwrap_failed
at /rustc/cc66ad468955717ab92600c770da8c1601a4ff33/library/core/src/result.rs:1652:5
3: core::result::Result<T,E>::unwrap
at /rustc/cc66ad468955717ab92600c770da8c1601a4ff33/library/core/src/result.rs:1077:23
4: parquet::arrow::async_reader::tests::test_nested_skip::{{closure}}
at ./src/arrow/async_reader/mod.rs:1760:26
5: <core::pin::Pin<P> as core::future::future::Future>::poll
at /rustc/cc66ad468955717ab92600c770da8c1601a4ff33/library/core/src/future/future.rs:125:9
6: <core::pin::Pin<P> as core::future::future::Future>::poll
at /rustc/cc66ad468955717ab92600c770da8c1601a4ff33/library/core/src/future/future.rs:125:9
7: tokio::runtime::scheduler::current_thread::CoreGuard::block_on::{{closure}}::{{closure}}::{{closure}}
at /Users/alamb/.cargo/registry/src/index.crates.io-6f17d22bba15001f/tokio-1.32.0/src/runtime/scheduler/current_thread/mod.rs:665:57
8: tokio::runtime::coop::with_budget
at /Users/alamb/.cargo/registry/src/index.crates.io-6f17d22bba15001f/tokio-1.32.0/src/runtime/coop.rs:107:5
9: tokio::runtime::coop::budget
at /Users/alamb/.cargo/registry/src/index.crates.io-6f17d22bba15001f/tokio-1.32.0/src/runtime/coop.rs:73:5
10: tokio::runtime::scheduler::current_thread::CoreGuard::block_on::{{closure}}::{{closure}}
at /Users/alamb/.cargo/registry/src/index.crates.io-6f17d22bba15001f/tokio-1.32.0/src/runtime/scheduler/current_thread/mod.rs:665:25
11: tokio::runtime::scheduler::current_thread::Context::enter
at /Users/alamb/.cargo/registry/src/index.crates.io-6f17d22bba15001f/tokio-1.32.0/src/runtime/scheduler/current_thread/mod.rs:410:19
12: tokio::runtime::scheduler::current_thread::CoreGuard::block_on::{{closure}}
at /Users/alamb/.cargo/registry/src/index.crates.io-6f17d22bba15001f/tokio-1.32.0/src/runtime/scheduler/current_thread/mod.rs:664:36
13: tokio::runtime::scheduler::current_thread::CoreGuard::enter::{{closure}}
at /Users/alamb/.cargo/registry/src/index.crates.io-6f17d22bba15001f/tokio-1.32.0/src/runtime/scheduler/current_thread/mod.rs:743:68
14: tokio::runtime::context::scoped::Scoped<T>::set
at /Users/alamb/.cargo/registry/src/index.crates.io-6f17d22bba15001f/tokio-1.32.0/src/runtime/context/scoped.rs:40:9
15: tokio::runtime::context::set_scheduler::{{closure}}
at /Users/alamb/.cargo/registry/src/index.crates.io-6f17d22bba15001f/tokio-1.32.0/src/runtime/context.rs:176:26
16: std::thread::local::LocalKey<T>::try_with
at /rustc/cc66ad468955717ab92600c770da8c1601a4ff33/library/std/src/thread/local.rs:270:16
17: std::thread::local::LocalKey<T>::with
at /rustc/cc66ad468955717ab92600c770da8c1601a4ff33/library/std/src/thread/local.rs:246:9
18: tokio::runtime::context::set_scheduler
at /Users/alamb/.cargo/registry/src/index.crates.io-6f17d22bba15001f/tokio-1.32.0/src/runtime/context.rs:176:9
19: tokio::runtime::scheduler::current_thread::CoreGuard::enter
at /Users/alamb/.cargo/registry/src/index.crates.io-6f17d22bba15001f/tokio-1.32.0/src/runtime/scheduler/current_thread/mod.rs:743:27
20: tokio::runtime::scheduler::current_thread::CoreGuard::block_on
at /Users/alamb/.cargo/registry/src/index.crates.io-6f17d22bba15001f/tokio-1.32.0/src/runtime/scheduler/current_thread/mod.rs:652:19
21: tokio::runtime::scheduler::current_thread::CurrentThread::block_on::{{closure}}
at /Users/alamb/.cargo/registry/src/index.crates.io-6f17d22bba15001f/tokio-1.32.0/src/runtime/scheduler/current_thread/mod.rs:175:28
22: tokio::runtime::context::runtime::enter_runtime
at /Users/alamb/.cargo/registry/src/index.crates.io-6f17d22bba15001f/tokio-1.32.0/src/runtime/context/runtime.rs:65:16
23: tokio::runtime::scheduler::current_thread::CurrentThread::block_on
at /Users/alamb/.cargo/registry/src/index.crates.io-6f17d22bba15001f/tokio-1.32.0/src/runtime/scheduler/current_thread/mod.rs:167:9
24: tokio::runtime::runtime::Runtime::block_on
at /Users/alamb/.cargo/registry/src/index.crates.io-6f17d22bba15001f/tokio-1.32.0/src/runtime/runtime.rs:347:47
25: parquet::arrow::async_reader::tests::test_nested_skip
at ./src/arrow/async_reader/mod.rs:1744:9
26: parquet::arrow::async_reader::tests::test_nested_skip::{{closure}}
at ./src/arrow/async_reader/mod.rs:1687:33
27: core::ops::function::FnOnce::call_once
at /rustc/cc66ad468955717ab92600c770da8c1601a4ff33/library/core/src/ops/function.rs:250:5
28: core::ops::function::FnOnce::call_once
at /rustc/cc66ad468955717ab92600c770da8c1601a4ff33/library/core/src/ops/function.rs:250:5
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.
```
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] Assume Pages Delimit Records When Offset Index Loaded (#4921) [arrow-rs]
Posted by "tustvold (via GitHub)" <gi...@apache.org>.
tustvold merged PR #4943:
URL: https://github.com/apache/arrow-rs/pull/4943
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] Assume Parquet Records Not Split Across Pages (#4921) [arrow-rs]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #4943:
URL: https://github.com/apache/arrow-rs/pull/4943#issuecomment-1765103217
> This PR removes the workarounds, and simply assumes pages delimit records.
What will happen if the reader gets a file that does split a record across a page boundary?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] Assume Parquet Records Not Split Across Pages (#4921) [arrow-rs]
Posted by "tustvold (via GitHub)" <gi...@apache.org>.
tustvold commented on PR #4943:
URL: https://github.com/apache/arrow-rs/pull/4943#issuecomment-1765107541
> What will happen if the reader gets a file that does split a record across a page boundary?
Most likely it will error
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] Assume Pages Delimit Records When Offset Index Loaded (#4921) [arrow-rs]
Posted by "tustvold (via GitHub)" <gi...@apache.org>.
tustvold commented on PR #4943:
URL: https://github.com/apache/arrow-rs/pull/4943#issuecomment-1765366187
I reworked this to only assume pages delimit records if the offset index is loaded, this will ensure this change is restricted to the use-case where the behaviour was previously incorrect / buggy and avoid regressing the case of no offset index - which will continue to work as before. This will also serve as an opt-out mechanism should the new behaviour be problematic for any reason.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org