thinkharderdev commented on PR #3616: URL: https://github.com/apache/arrow-datafusion/pull/3616#issuecomment-1258175145
> > pass the option to ParquetRecordBatchStreamBuilder > > Thanks for your advice! @thinkharderdev , IMOP, if we want to read `page_index` we must get its location from metadata, so it must read after metadata. i think put it directly behind read metadata is better. > > > if users would like to cache the indexes other than the file metadata > > the reason why i put the index in `metadata` is to reduce the code change in repo, if the user want to cache it anywhere else can use pub API in `page_index::index_reader` to read with metadata. > > If i am wrong or misunderstand plz correct me 😊? Correct, the advantage of what `ParquetRecordBatchStreamBuilder` does in arrow-rs is that it fill fetch all the indexes concurrently using `AsyncFileReader::get_ranges`. So if you don't read the metadata up front then it will fetch it automatically for you. If we do want to grab it up front in `ParquetExec` then we should copy the implementation in arrow-rs to grab everything concurrently. -- 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: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
