mapleFU commented on issue #34510: URL: https://github.com/apache/arrow/issues/34510#issuecomment-1463884481
> and it could be optimized to use a single static value, right? Yes, in the future, developer may optimize it. If `FixedSizeArray` is non-nullable, Parquet can have a single static value, but if `FixedSizeArray` is non-nullable, it cannot. > other reasons or replevels cascade somehow into even worse perf I've profile the C++ part, in my MacOS with release (O2): <img width="1305" alt="C811842B-4DDA-423A-B8A1-EC6A7E4ADE33" src="https://user-images.githubusercontent.com/24351052/224341042-8d2cb020-2aa6-41f4-8c66-a913436ea1c2.png";> 1. Decoding double is fast 2. Decoding levels use nearly same time as Decoding double 3. Constructing List cost a little time The benchmark uses list rather than FixedSizeList, but I think the benchmark is similiar. I'm not so familiar with Python part, maybe someone can profile that path -- 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]
![https://user-images.githubusercontent.com/24351052/224341042-8d2cb020-2aa6-41f4-8c66-a913436ea1c2.png"](https://user-images.githubusercontent.com/24351052/224341042-8d2cb020-2aa6-41f4-8c66-a913436ea1c2.png"){kind=link}