[SPARK-14907][MLLIB] Use repartition in GLMRegressionModel.save

[dongjoon-hyun]

What changes were proposed in this pull request?

This PR changes GLMRegressionModel.save function like the following code that is similar to other algorithms' parquet write.

- val dataRDD: DataFrame = sc.parallelize(Seq(data), 1).toDF()
- // TODO: repartition with 1 partition after SPARK-5532 gets fixed
- dataRDD.write.parquet(Loader.dataPath(path))
+ sqlContext.createDataFrame(Seq(data)).repartition(1).write.parquet(Loader.dataPath(path))

How was this patch tested?

Manual.

[SparkQA]

Test build #56947 has finished for PR 12676 at commit b237877.

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

[dongjoon-hyun]

Hi, @jkbradley .
Could you review this PR when you have some time?

[MLnick]

LGTM.

("save/load" test case in LinearRegressionSuite covers this)

[dongjoon-hyun]

Thank you for review, @MLnick !

[dongjoon-hyun]

Hi, @mengxr .
Could you review this too?

[jkbradley]

This isn't really changing anything, and I actually think this makes the DF creation less similar to most spark.mllib save methods in terms of code style. I'd prefer to close this issue.

[dongjoon-hyun]

If you think so, it's okay, @jkbradley .
But, if you don't mind, could you remove those TODO by yourself.
Do you have any reason to maintain that?

TODO comments always mislead community developer like me.

[dongjoon-hyun]

In fact, I didn't try to change that if it's just a style problem.

[jkbradley]

Oh, I apologize; I missed the TODO removal. You're right; that should have been removed previously when the sc.parallelize call was changed to use 1 partition.

It looks like that's the only remaining mention of SPARK-5532 in the codebase, so your fix be it.

[jkbradley]

LGTM
I'll merge this with master
Thanks (and thanks for pushing back)!

[dongjoon-hyun]

Oh, thank YOU, @jkbradley .