jorisvandenbossche commented on issue #38432: URL: https://github.com/apache/arrow/issues/38432#issuecomment-1777012321
Although I have to add that we also have "dataset-read" benchmarks (https://conbench.ursa.dev/c-benchmarks/dataset-read), which also benchmark Parquet reading, and there I don't directly see any visible regressions (although those are reading from S3, and might be less sensitive to small regressions in the reading part) Also, the "file-read" benchmark is actually using `pq.read_table`, which already did set `pre_buffer=True` for the default for some time. So it can't be that change in default (on the dataset side) that causes the slowdown. The other change in that PR is changing `CacheOptions::Defaults()` to `LazyDefaults()` -- 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]
