[SPARK-29622][SQL] do not allow leading 'interval' in the interval string format

[cloud-fan]

What changes were proposed in this pull request?

When parsing a string to interval, do not allow the leading "interval".

This PR also removes the leading "interval" from CalendarInterval.toString, to be consistent.

Why are the changes needed?

When parsing string to interval, we already know it's an interval and the leading "interval" in the string is kind of duplicated. The behavior is also consistent with pgsql

cloud0fan=# select 'interval 1 day'::interval;
ERROR:  invalid input syntax for type interval: "interval 1 day"
LINE 1: select 'interval 1 day'::interval;
               ^
cloud0fan=# select '1 day'::interval;
 interval 
----------
 1 day
(1 row)

cloud0fan=# select interval'interval 1 year';
ERROR:  invalid input syntax for type interval: " interval 1 year"
LINE 1: select interval' interval 1 year';

cloud0fan=# select interval'1 year';
 interval 
----------
 1 year
(1 row)

Does this PR introduce any user-facing change?

Yes. The INTERVAL'...' syntax is newly introduced in the master branch, so it's OK to change. The cast behavior in 2.4 is pretty weird: the leading "interval" in the string is required. I think it's better to change it (with a migration guide and legacy config).

How was this patch tested?

updated tests

[cloud-fan]

This is mostly the original visitSingleInterval, with some methods inlined.

[cloud-fan]

these are copied from the original visitIntervalField.

[MaxGekk]

I think this should be case insensitive check. You can use regionMatches() for that.

[MaxGekk]

If we don't allow 'interval' in an input strings, shouldn't we output it too:

spark/common/unsafe/src/main/java/org/apache/spark/unsafe/types/CalendarInterval.java

Line 282 in cdea520

StringBuilder sb = new StringBuilder("interval");

Just in case, could you write a round trip test if such test doesn't exist. Otherwise it looks strange that Spark cannot parse back its output.

[SparkQA]

Test build #112773 has finished for PR 26283 at commit ce43b9a.

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

[cloud-fan]

The parser cleanup has been extracted to #26285

[dongjoon-hyun]

Thank you, @cloud-fan !

[MaxGekk]

@cloud-fan Please, rebase this on the master.

[SparkQA]

Test build #113214 has finished for PR 26283 at commit 03e2a53.

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

[SparkQA]

Test build #113236 has finished for PR 26283 at commit 135f8e9.

• This patch fails to build.
• This patch merges cleanly.
• This patch adds no public classes.

[SparkQA]

Test build #113235 has finished for PR 26283 at commit 14e0239.

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

[SparkQA]

Test build #113239 has finished for PR 26283 at commit a4ffd00.

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

[cloud-fan]

This has bigger impacts than I expect, e.g. it breaks functions.window which is a streaming-related public API.

I'm closing it to avoid breaking public APIs.