[SPARK-17673] [SQL] Incorrect exchange reuse with RowDataSourceScan#15273
[SPARK-17673] [SQL] Incorrect exchange reuse with RowDataSourceScan#15273ericl wants to merge 3 commits intoapache:masterfrom
Conversation
|
Test build #66014 has finished for PR 15273 at commit
|
|
Test build #66017 has finished for PR 15273 at commit
|
|
Test build #66018 has finished for PR 15273 at commit
|
|
Did the unit test fail without the fix? |
|
Yep, it returns 1 row only. On Wed, Sep 28, 2016, 12:59 AM Reynold Xin notifications@github.com wrote:
|
|
Thanks - merging in master/2.0. I think the test is actually in a non-ideal location, but I'm going to merge first in order to make 2.0.1 rc4. |
…ackport) This backports #15273 to branch-2.0 Also verified the test passes after the patch was applied. rxin Author: Eric Liang <ekl@databricks.com> Closes #15282 from ericl/spark-17673-2.
## What changes were proposed in this pull request? As a followup for #15273 we should move non-JDBC specific tests out of that suite. ## How was this patch tested? Ran the test. Author: Eric Liang <ekl@databricks.com> Closes #15287 from ericl/spark-17713.
What changes were proposed in this pull request?
It seems the equality check for reuse of
RowDataSourceScanExecnodes doesn't respect the output schema. This can cause self-joins or unions over the same underlying data source to return incorrect results if they select different fields.How was this patch tested?
New unit test passes after the fix.