[POC][wip] faster DefaultEngine parquet reads
#595
There are no files selected for viewing
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 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 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 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 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
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -202,6 +202,7 @@ impl<E: TaskExecutor> ParquetHandler for DefaultParquetHandler<E> { | ||
| physical_schema.clone(), | ||
| predicate, | ||
| self.store.clone(), | ||
| None, | ||
| )), | ||
| }; | ||
| FileStream::new_async_read_iterator( | ||
| @@ -215,13 +216,14 @@ impl<E: TaskExecutor> ParquetHandler for DefaultParquetHandler<E> { | ||
| } | ||
|
|
||
| /// Implements [`FileOpener`] for a parquet file | ||
| pub(crate) struct ParquetOpener { | ||
| // projection: Arc<[usize]>, | ||
| batch_size: usize, | ||
| table_schema: SchemaRef, | ||
| predicate: Option<ExpressionRef>, | ||
| limit: Option<usize>, | ||
| store: Arc<DynObjectStore>, | ||
| runtime: Option<Arc<tokio::runtime::Runtime>>, | ||
| } | ||
|
|
||
| impl ParquetOpener { | ||
| @@ -230,13 +232,15 @@ impl ParquetOpener { | ||
| table_schema: SchemaRef, | ||
| predicate: Option<ExpressionRef>, | ||
| store: Arc<DynObjectStore>, | ||
| runtime: Option<Arc<tokio::runtime::Runtime>>, | ||
| ) -> Self { | ||
| Self { | ||
| batch_size, | ||
| table_schema, | ||
| predicate, | ||
| limit: None, | ||
| store, | ||
| runtime, | ||
| } | ||
| } | ||
| } | ||
| @@ -251,11 +255,18 @@ impl FileOpener for ParquetOpener { | ||
| let table_schema = self.table_schema.clone(); | ||
| let predicate = self.predicate.clone(); | ||
| let limit = self.limit; | ||
| let handle = match &self.runtime { | ||
| Some(runtime) => Some(runtime.handle().clone()), | ||
| None => None, | ||
| }; | ||
|
|
||
| Ok(Box::pin(async move { | ||
| // TODO avoid IO by converting passed file meta to ObjectMeta | ||
| let meta = store.head(&path).await?; | ||
| let mut reader = ParquetObjectReader::new(store, meta); | ||
| if let Some(handle) = handle { | ||
| reader = reader.with_runtime(handle); | ||
|
There was a problem hiding this comment. what does setting this do? There was a problem hiding this comment. new in arrow 53.3 i think - lets you push down a runtime for them to schedule their IO on. This has gotten me thinking about various ways to enable this sort of 'runtime passthrough' ourselves..
see There was a problem hiding this comment. Thanks. Yeah, seems similar to what you're doing. |
||
| } | ||
| let metadata = ArrowReaderMetadata::load_async(&mut reader, Default::default()).await?; | ||
| let parquet_schema = metadata.schema(); | ||
| let (indicies, requested_ordering) = | ||
| @@ -281,6 +292,9 @@ impl FileOpener for ParquetOpener { | ||
|
|
||
| let stream = builder.with_batch_size(batch_size).build()?; | ||
|
|
||
| // println!("read IO"); | ||
| // tokio::time::sleep(std::time::Duration::from_millis(10000)).await; // simulate IO delay | ||
|
|
||
| let stream = stream.map(move |rbr| { | ||
| // re-order each batch if needed | ||
| rbr.map_err(Error::Parquet).and_then(|rb| { | ||
| @@ -293,7 +307,7 @@ impl FileOpener for ParquetOpener { | ||
| } | ||
|
|
||
| /// Implements [`FileOpener`] for a opening a parquet file from a presigned URL | ||
| pub(crate) struct PresignedUrlOpener { | ||
| batch_size: usize, | ||
| predicate: Option<ExpressionRef>, | ||
| limit: Option<usize>, | ||
Oops, something went wrong.
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.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
for experimenting, i'd suggest using the multi-threaded reader. although i guess this does help determine how much a single call can read. regardless,
read-table-multi-threadedhas a--limitoption for this case so you can see that some data got returned but not print it all, but it does tell you the total row count. maybe add that as an option here too :)There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
yep thanks I ended up playing with both but yea the
--limitis nicer :)