[SPARK-14460] [SQL] properly handling of column name contains space

[bomeng]

What changes were proposed in this pull request?

Although it is not recommended, table can be created with column name containing space. For example,

create table test.people1 (name TEXT(32) NOT NULL,the id INTEGER NOT NULL)

When we query the table via JDBC, we also need to quote the column name in the same way if the column name contains space.

I am leveraging the quote method in the dialect, it will add the proper quote to the column name based on its data source.

How was this patch tested?

I've updated the test case to test this scenario.

[SparkQA]

Test build #55312 has finished for PR 12252 at commit 51b7f86.

• This patch fails Scala style tests.
• This patch merges cleanly.
• This patch adds no public classes.

[SparkQA]

Test build #55309 has finished for PR 12252 at commit 88935c5.

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

[SparkQA]

Test build #55330 has finished for PR 12252 at commit d0c86ea.

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

[SparkQA]

Test build #55328 has finished for PR 12252 at commit 9ffb462.

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

[bomeng]

@andrewor14 Could you please review my codes?

[SparkQA]

Test build #55851 has finished for PR 12252 at commit ecd52bf.

• This patch fails Scala style tests.
• This patch merges cleanly.
• This patch adds no public classes.

[SparkQA]

Test build #55854 has finished for PR 12252 at commit b70f21f.

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

[ash211]

Is the new dialect you're passing in different from this one in some way?

[bomeng]

The purpose to pass in dialect is to get proper quote for columns based on its data source. Any suggestion?

[ash211]

You pass in a parameter named dialect to the schemaString method, but there's also the dialect that comes from JdbcDialects.get(url) --- that's the duplicate I was trying to point out

[bomeng]

Thanks for pointing out. I've modified the codes. Please check it out.

[SparkQA]

Test build #55889 has finished for PR 12252 at commit 715777a.

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

[SparkQA]

Test build #55891 has finished for PR 12252 at commit 675700f.

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

[SparkQA]

Test build #56384 has finished for PR 12252 at commit b151065.

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

[cloud-fan]

seems it's fixed in #15662 ?

[HyukjinKwon]

ping @bomeng for ^.