[SPARK-10151] [SQL] Support invocation of hive macro

[navis]

Macro in hive (which is GenericUDFMacro) contains real function inside of it but it's not conveyed to tasks, resulting null-pointer exception.

[SparkQA]

Test build #41360 has finished for PR 8354 at commit e1ef530.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[andrewor14]

@yhuai

[JoshRosen]

Is there a way to add a test for this?

[navis]

@JoshRosen It isn't because spark does not support "create macro". I've tried to support that in new patch.

[SparkQA]

Test build #42651 has finished for PR 8354 at commit e8983b9.

• This patch fails Spark unit tests.
• This patch merges cleanly.
• This patch adds no public classes.

[SparkQA]

Test build #42748 has finished for PR 8354 at commit 9d3dd3b.

• This patch fails Spark unit tests.
• This patch merges cleanly.
• This patch adds no public classes.

[SparkQA]

Test build #42811 has finished for PR 8354 at commit e074681.

• This patch fails Spark unit tests.
• This patch merges cleanly.
• This patch adds no public classes.

[SparkQA]

Test build #42887 has finished for PR 8354 at commit 3223621.

• This patch fails Spark unit tests.
• This patch merges cleanly.
• This patch adds no public classes.

[SparkQA]

Test build #43228 has finished for PR 8354 at commit fd799a1.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[JoshRosen]

Formatting nit: wrap these arguments one per line.

[JoshRosen]

Hey @chenghao-intel, could you take a look at this PR? I'd like to get your feedback/review because this PR largely undoes some of the changes that you made in HiveFunctionWrapper (in #3640) and I'm not familiar enough with this code to know whether this is safe.

[SparkQA]

Test build #43960 has finished for PR 8354 at commit 75dbc2e.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[SparkQA]

Test build #44021 has finished for PR 8354 at commit 3595ce7.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[chenghao-intel]

Nit: How about return boolean directly?

[JoshRosen]

+1, good catch.

[chenghao-intel]

@navis @JoshRosen this change is much cleaner now, and LGTM, except a Nit, we can leave it for the further improvement.

[JoshRosen]

Yep, LGTM as well. @navis, thanks for helping to simplify this!

[JoshRosen]

I've merged this to master. Thanks @navis!