Tim Armstrong has uploaded a new patch set (#15).

Change subject: IMPALA-3567: Part 1: groundwork to make Join build sides 
DataSinks
......................................................................

IMPALA-3567: Part 1: groundwork to make Join build sides DataSinks

Refactor DataSink interface to be more generic. We need more flexibility
in setting up MemTrackers, so that memory is accounted against the right
ExecNode. Also removes some redundancy between DataSink subclasses in
setting up RuntimeProfiles, etc.

Remove the redundancy in the DataSink between passing eos to GetNext()
and FlushFinal(). This simplifies HdfsTableSink quite a bit and makes
handling empty batches simpler.

Partially refactor join nodes that so control flow between
BlockingJoinNode::Open() and its subclasses is easier to follow.
BlockingJoinNode now only calls one virtual function on its
subclasses: ConstructBuildSide(). Once we convert all join nodes
to use the DataSink interface, we will also be able to remove that
as well.

As a minor optimisation, avoid updating a timer that is ignored for
non-async builds.

As a proof of concept, this patch separates out the build side of
NestedLoopJoinNode into a class that implements the DataSink
interface. The main challenge here is that NestedLoopJoinNode
recycles row batches to avoid reallocations and copies of row
batches in subplans. The solution to this is:
- Retain the special-case optimisation for SingularRowSrc
- Use the row batch cache and std::swap() to efficiently recycle
  and copy the state of row batches passed to Send()

Refactoring the partitioned hash join is left for Part 2.

Change-Id: I9d7608181eeacfe706a09c1e153d0a3e1ee9b475
---
M be/src/exec/CMakeLists.txt
M be/src/exec/blocking-join-node.cc
M be/src/exec/blocking-join-node.h
M be/src/exec/data-sink.cc
M be/src/exec/data-sink.h
M be/src/exec/hash-join-node.cc
M be/src/exec/hash-join-node.h
M be/src/exec/hbase-table-sink.cc
M be/src/exec/hbase-table-sink.h
M be/src/exec/hdfs-table-sink.cc
M be/src/exec/hdfs-table-sink.h
M be/src/exec/kudu-table-sink-test.cc
M be/src/exec/kudu-table-sink.cc
M be/src/exec/kudu-table-sink.h
A be/src/exec/nested-loop-join-builder.cc
A be/src/exec/nested-loop-join-builder.h
M be/src/exec/nested-loop-join-node.cc
M be/src/exec/nested-loop-join-node.h
M be/src/exec/partitioned-hash-join-node.cc
M be/src/exec/partitioned-hash-join-node.h
M be/src/exec/row-batch-cache.h
M be/src/runtime/data-stream-sender.cc
M be/src/runtime/data-stream-sender.h
M be/src/runtime/data-stream-test.cc
M be/src/runtime/plan-fragment-executor.cc
M be/src/runtime/plan-fragment-executor.h
M be/src/runtime/row-batch.cc
M be/src/runtime/row-batch.h
M be/src/util/stopwatch.h
29 files changed, 701 insertions(+), 459 deletions(-)


  git pull ssh://gerrit.cloudera.org:29418/Impala refs/changes/82/3282/15
-- 
To view, visit http://gerrit.cloudera.org:8080/3282
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-MessageType: newpatchset
Gerrit-Change-Id: I9d7608181eeacfe706a09c1e153d0a3e1ee9b475
Gerrit-PatchSet: 15
Gerrit-Project: Impala
Gerrit-Branch: cdh5-trunk
Gerrit-Owner: Tim Armstrong <[email protected]>
Gerrit-Reviewer: Alex Behm <[email protected]>
Gerrit-Reviewer: Marcel Kornacker <[email protected]>
Gerrit-Reviewer: Michael Ho <[email protected]>
Gerrit-Reviewer: Tim Armstrong <[email protected]>

Reply via email to