Lars Volker has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/11694 )

Change subject: IMPALA-7644: Hide Parquet page index writing with feature flag
......................................................................


Patch Set 2:

(2 comments)

http://gerrit.cloudera.org:8080/#/c/11694/2//COMMIT_MSG
Commit Message:

http://gerrit.cloudera.org:8080/#/c/11694/2//COMMIT_MSG@18
PS2, Line 18: custom_cluset
nit: typo


http://gerrit.cloudera.org:8080/#/c/11694/2/be/src/exec/hdfs-parquet-table-writer.cc
File be/src/exec/hdfs-parquet-table-writer.cc:

http://gerrit.cloudera.org:8080/#/c/11694/2/be/src/exec/hdfs-parquet-table-writer.cc@705
PS2, Line 705: num_data_pages_
> Maybe we could just drop 'num_data_pages_' and call std::vector::resize(0)
I agree, that would simplify things substantially. When you do this, can you 
either do a perf run or inspect the assembly to make sure that the additional 
dtor ctor calls for the pages are constant time, i.e. the DataPage structs are 
all PODs?

It might be better to do all of that in a subsequent change, but given we can 
into this issue, we should improve the current code. If you don't have time to 
push the additional change soon, please add a comment here.

The SO discussion linked in the cppreference reads as if clear() should be 
safe. If you're unsure, you can add a DCHECK on the capacity right after 
calling clear().



--
To view, visit http://gerrit.cloudera.org:8080/11694
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: Ib4a9098a2085a385351477c715ae245d83bf1c72
Gerrit-Change-Number: 11694
Gerrit-PatchSet: 2
Gerrit-Owner: Zoltan Borok-Nagy <[email protected]>
Gerrit-Reviewer: Csaba Ringhofer <[email protected]>
Gerrit-Reviewer: Impala Public Jenkins <[email protected]>
Gerrit-Reviewer: Lars Volker <[email protected]>
Gerrit-Reviewer: Zoltan Borok-Nagy <[email protected]>
Gerrit-Comment-Date: Tue, 16 Oct 2018 22:16:38 +0000
Gerrit-HasComments: Yes

Reply via email to