How can I get stripe level statistics without reading row group Index

[wushap]

I found that getStripeStatistics function read stripe level statistics and row group Index. How can I get stripe level statistics without read row group Index? Did I miss something?
If the file is stored in a remote storage, these read IO will have a overhead.

[dongjoon-hyun]

Could you describe your example more specifically? Are you using Java API or C++ API?

[wushap]

Could you describe your example more specifically? Are you using Java API or C++ API?

C++API getStripeStatistics

orc/c++/src/Reader.cc

Line 768 in da6cfc1

getRowIndexStatistics(currentStripeInfo, stripeIndex, currentStripeFooter, &indexStats);

I need to read only target stripe level statistics , but the internal implementation seems to read the whole row group Index of this stripe as well.
Is there a C++API that just gets target stripe level statistics?

[wgtmac]

I think you're right. It is a valid use case to get only the stripe stats. Would you like to create a PR for this?

[wushap]

I think you're right. It is a valid use case to get only the stripe stats. Would you like to create a PR for this?

I am not sure whether to add a lazy read parameter to the original interface or add new one.

[dongjoon-hyun]

For the record, Apache ORC community started to discuss to release Apache ORC 2.2.0 earlier than the pre-defined schedule in order to deliver this.

• https://lists.apache.org/thread/8wftl03kknm28c7f9rdxg2ckb7nn1q66