Alex Behm has posted comments on this change.

Change subject: IMPALA-3952: Clear scratch batch mem pool if Open() failed.
......................................................................


Patch Set 1:

(3 comments)

http://gerrit.cloudera.org:8080/#/c/3953/1//COMMIT_MSG
Commit Message:

PS1, Line 11: pts
> ptrs
Done


http://gerrit.cloudera.org:8080/#/c/3953/1/be/src/exec/hdfs-parquet-scanner.cc
File be/src/exec/hdfs-parquet-scanner.cc:

PS1, Line 210: else if (!FLAGS_enable_partitioned_hash_join ||
             :       !FLAGS_enable_partitioned_aggregation) {
> Not necessarily an issue, but it might be better to just
I thought about executing the branch in any case, but I felt it's better to be 
explicit about the conditions in which this code is required. Once we finally 
remove the legacy join/agg flags, then this code becomes dead and it should be 
obvious that we can delete it.

Also note that we need this extra branch even if the plan does not contain any 
legacy joins or aggs. The main difference in that if we enable a legacy joins 
or aggs, then the tuple ptrs of a row batch are allocated from the row batcht's 
mem pool (see c'tor of row-batch.cc). That's the main reason why need special 
handling here because if Open() fails the scratch batch will still have the 
tuple ptrs allocated.


PS1, Line 212: pts
> ptrs
Done


-- 
To view, visit http://gerrit.cloudera.org:8080/3953
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: I55f32ed698a5b6fed8c28af1391aa07e1560e782
Gerrit-PatchSet: 1
Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-Owner: Alex Behm <[email protected]>
Gerrit-Reviewer: Alex Behm <[email protected]>
Gerrit-Reviewer: Matthew Jacobs <[email protected]>
Gerrit-HasComments: Yes

Reply via email to