Mailing-List: contact reviews-help@impala.incubator.apache.org; run by ezmlm
Precedence: bulk
Message-Id: <201702180016.v1I0GNDW014950@ip-10-146-233-104.ec2.internal>
Date: Sat, 18 Feb 2017 00:16:23 +0000
From: "Alex Behm (Code Review)" <gerrit@cloudera.org>
To: Michael Ho <kwho@cloudera.com>, impala-cr@cloudera.com,
        reviews@impala.incubator.apache.org
CC: Dan Hecht <dhecht@cloudera.com>
Reply-To: alex.behm@cloudera.com
Subject: =?UTF-8?Q?=5BImpala-ASF-CR=5D_IMPALA-4821=3A_Update_AVG=28=29_for_DECIMAL_V2=0A?=
In-Reply-To: <gerrit.1487265008000.I28f5ef0370938440eb5b1c6d29b2f24e6f88499f@gerrit.cloudera.org>
References: <gerrit.1487265008000.I28f5ef0370938440eb5b1c6d29b2f24e6f88499f@gerrit.cloudera.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Content-Disposition: inline
User-Agent: Gerrit/2.12.7
archived-at: Sat, 18 Feb 2017 00:16:28 -0000

Alex Behm has posted comments on this change.

Change subject: IMPALA-4821: Update AVG() for DECIMAL_V2
......................................................................


Patch Set 3:

(3 comments)

http://gerrit.cloudera.org:8080/#/c/6038/3//COMMIT_MSG
Commit Message:

Line 12: 2. precision will be augmented if the scale is adjusted.
In addition to my other comment, I think we should mention some of the limitations here, for example, that it won't be possible to compute the AVG() of some decimal types at all.


http://gerrit.cloudera.org:8080/#/c/6038/3/fe/src/main/java/org/apache/impala/analysis/FunctionCallExpr.java
File fe/src/main/java/org/apache/impala/analysis/FunctionCallExpr.java:

Line 337:       // an implicit divide. This is similar to the MS SQL server's behavior which takes
Suggest these minor fixes and reordering:

AVG() gets at least MIN_ADJUSTED_SCALE decimal places since it performs an implicit divide. However, the output type is not always the same as SUM()/COUNT(). Our behavior is similar to MS SQL Server's which takes the max of the input's scale and MIN_ADJUSTED_SCALE.

The observations are ok, but at a high level:
* the type is not the same as SUM()/COUNT()
* the behavior is not the same as MS SQL Server's

It would be good to provide a justification/intuition for why we chose yet another behavior (also in the commit msg).


http://gerrit.cloudera.org:8080/#/c/6038/3/testdata/workloads/functional-query/queries/QueryTest/decimal-exprs.test
File testdata/workloads/functional-query/queries/QueryTest/decimal-exprs.test:

Line 208: ====
We should also add tests for those decimal types where AVG() cannot be computed (because we always overflow).


-- 
To view, visit http://gerrit.cloudera.org:8080/6038
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: I28f5ef0370938440eb5b1c6d29b2f24e6f88499f
Gerrit-PatchSet: 3
Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-Owner: Michael Ho <kwho@cloudera.com>
Gerrit-Reviewer: Alex Behm <alex.behm@cloudera.com>
Gerrit-Reviewer: Dan Hecht <dhecht@cloudera.com>
Gerrit-Reviewer: Michael Ho <kwho@cloudera.com>
Gerrit-HasComments: Yes