Zach Amsden has posted comments on this change.

Change subject: IMPALA-5003: Constant propagation in scan nodes
......................................................................


Patch Set 16:

(4 comments)

http://gerrit.cloudera.org:8080/#/c/6389/16/testdata/workloads/functional-planner/queries/PlannerTest/constant-propagation.test
File 
testdata/workloads/functional-planner/queries/PlannerTest/constant-propagation.test:

Line 159:  select count(*) from
> There's still a question as to whether we should impose a limit on the size
I'll probably just re-use the same rule.


Line 245: (coalesce(NULL, T.int_col) + random() * T.tinyint_col = 100 OR
> This is the current intended behavior. See the bottom of hdfs.test planner 
It is fixable.  I think this is because the resulting Expr ends up with no 
SlotRefs, and as a result conjunct.isBoundBySlotIds(partitionSlots_) returns 
true, which is questionable.  Changing that would have deep consequences, but 
we can always build something to collect all the SlotRefs from an Expr and 
validate they are a non-empty subset of partitionSlots_


Line 289: where l_partkey < l_suppkey and c.c_nationkey = 10 and o_orderkey = 
o_shippriority and l_suppkey = 10 and o_shippriority = c_nationkey
> long line
Done


Line 335: 00:SCAN HDFS [functional.alltypes]
> Why not an EmptySetNode?
bool_col = null is stripped from conjunct evaluation.  bool_col IS null is not. 
 I am guessing because 10 = null => null and 10 IS NULL => false.

Is this behavior incorrect?  It certainly is suspect, but there are some 
comments about how conjunct evaluation treats null conditions specially and I 
wasn't sure which was the correct behavior here.


-- 
To view, visit http://gerrit.cloudera.org:8080/6389
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: I79750a8edb945effee2a519fa3b8192b77042cb4
Gerrit-PatchSet: 16
Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-Owner: Zach Amsden <[email protected]>
Gerrit-Reviewer: Alex Behm <[email protected]>
Gerrit-Reviewer: Dan Hecht <[email protected]>
Gerrit-Reviewer: Marcel Kornacker <[email protected]>
Gerrit-Reviewer: Zach Amsden <[email protected]>
Gerrit-HasComments: Yes

Reply via email to