Bharath Vissapragada has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/12438 )

Change subject: IMPALA-8181: Abbreviate row counts in EXPLAIN
......................................................................


Patch Set 5:

(2 comments)

http://gerrit.cloudera.org:8080/#/c/12438/5/fe/src/main/java/org/apache/impala/common/PrintUtils.java
File fe/src/main/java/org/apache/impala/common/PrintUtils.java:

http://gerrit.cloudera.org:8080/#/c/12438/5/fe/src/main/java/org/apache/impala/common/PrintUtils.java@64
PS5, Line 64: if (value == -1) return "unavailable";
Shouldn't we instead use printCardinality() or some other helper instead of 
doing it here?


http://gerrit.cloudera.org:8080/#/c/12438/5/fe/src/main/java/org/apache/impala/common/PrintUtils.java@66
PS5, Line 66:     if (value >= TERA) return new 
DecimalFormat(".00T").format(result / TERA);
Curious if we should be consistent with the backend metrics which print with 
the following format -  pretty-printed-val (actual val)

Ex:

 PeakMemoryUsage: 149.50 KB (153088)
       - PeakReservation: 8.00 KB (8192)
       - PeakUsedReservation: 0
       - PerHostPeakMemUsage: 169.50 KB (173568)
       - RowsProduced: 1 (1)

I think we probably want exact values in some cases, ex: off-by-one cases in 
cardinality etc.



--
To view, visit http://gerrit.cloudera.org:8080/12438
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: I08faaa9ad7b5ed42dcd7a15a333e8734bb45f10c
Gerrit-Change-Number: 12438
Gerrit-PatchSet: 5
Gerrit-Owner: Paul Rogers <[email protected]>
Gerrit-Reviewer: Bharath Vissapragada <[email protected]>
Gerrit-Reviewer: Impala Public Jenkins <[email protected]>
Gerrit-Reviewer: Paul Rogers <[email protected]>
Gerrit-Comment-Date: Tue, 19 Feb 2019 18:35:32 +0000
Gerrit-HasComments: Yes

Reply via email to