You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@orc.apache.org by "Xinyu Zeng (Jira)" <ji...@apache.org> on 2022/07/25 09:02:00 UTC

[jira] [Commented] (ORC-961) [C++] Expose metrics of the reader

    [ https://issues.apache.org/jira/browse/ORC-961?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17570786#comment-17570786 ] 

Xinyu Zeng commented on ORC-961:
--------------------------------

Could this feature be disabled as default? Currently by enabling it as default, reading time increased by more than 2x with orc-scan on my machine.

> [C++] Expose metrics of the reader
> ----------------------------------
>
>                 Key: ORC-961
>                 URL: https://issues.apache.org/jira/browse/ORC-961
>             Project: ORC
>          Issue Type: New Feature
>          Components: C++
>            Reporter: Quanlong Huang
>            Assignee: Quanlong Huang
>            Priority: Major
>             Fix For: 1.9.0
>
>
> It'd be useful for clients to collect metrics of the evaluation of the pushed down predicates. Currently SargsApplier keeps track of the selected RowGroups and evaluated RowGroups:
> [https://github.com/apache/orc/blob/40ee321b209cad1de73fc97b653636aa0fff28f6/c++/src/sargs/SargsApplier.hh#L79-L81]
> However, we are lack of an interface in RowReader to expose the metrics. Note that sargsApplier is a private member inside RowReaderImpl. We need an interface to expose it or only its stats.
> It'd also be useful if we expose other metrics of a RowReader, e.g. acutal mem usage, time elapsed in decoding/decompression, time elapsed in evaluating SearchArguments.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)