impala-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Tim Armstrong (Code Review)" <>
Subject [Impala-ASF-CR] IMPALA-6187: Fix missing conjuncts evaluation with empty projection
Date Tue, 28 Nov 2017 19:52:25 GMT
Tim Armstrong has posted comments on this change. ( )

Change subject: IMPALA-6187: Fix missing conjuncts evaluation with empty projection

Patch Set 2:

File be/src/exec/
PS2, Line 226:     DCHECK_EQ(scan_node_->tuple_desc()->byte_size(), 0);
This seems to violate the invariant that the scan node only produces non-NULL tuples (the
previous version could too). I'm not sure if the planner can produce plans that have a TupleIsNull()
predicate on a scan producing zero slots, because as far as I understand, TupleIsNull() predicates
are only emitted when a slot from the scan is referenced, but I could be wrong - I feel like
there could be some edge case if you get creative with nested loop joins, etc (which pass
the tuple pointers through unmodified).

I think we should either write a dummy non-NULL pointer or have a comment explaining why that's
not necessary.
File be/src/runtime/
PS2, Line 60:   tuple_ptrs_ = reinterpret_cast<Tuple**>(calloc(1, tuple_ptrs_size_));
I don't think we should zero-initialise this. We can't depend on the tuple_ptrs_ being zero-initialised
since Reset() doesn't zero-initialise them, so it's just unnecessary overhead.

To view, visit
To unsubscribe, visit

Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: Ib530f1fdcd2c6de699977db163b3f6eb38481517
Gerrit-Change-Number: 8623
Gerrit-PatchSet: 2
Gerrit-Owner: Michael Ho <>
Gerrit-Reviewer: Alex Behm <>
Gerrit-Reviewer: Dan Hecht <>
Gerrit-Reviewer: Michael Ho <>
Gerrit-Reviewer: Tim Armstrong <>
Gerrit-Comment-Date: Tue, 28 Nov 2017 19:52:25 +0000
Gerrit-HasComments: Yes

  • Unnamed multipart/alternative (inline, 8-Bit, 0 bytes)
View raw message