Add tumble windowing function

[dai-chen]

Description

Add TUMBLE windowing function which is required by materialized view support.

The function generates a new column called window. It's a struct field consist of start and end field inside. The implementation is actually delegated to Spark existing window() function. Here is an example:

spark-sql>
... SELECT window, COUNT(*)
... FROM stream.lineitem_tiny
... GROUP BY TUMBLE(l_shipdate, '1 week')
... ORDER BY window.start
... LIMIT 10;

window	count(1)
{"start":1992-01-02 00:00:00,"end":1992-01-09 00:00:00}	1390
{"start":1992-06-04 00:00:00,"end":1992-06-11 00:00:00}	24980
{"start":1992-07-09 00:00:00,"end":1992-07-16 00:00:00}	24964
{"start":1992-07-16 00:00:00,"end":1992-07-23 00:00:00}	24957
{"start":1992-07-30 00:00:00,"end":1992-08-06 00:00:00}	25012
{"start":1992-08-06 00:00:00,"end":1992-08-13 00:00:00}	24929
{"start":1992-08-20 00:00:00,"end":1992-08-27 00:00:00}	24968
{"start":1992-09-03 00:00:00,"end":1992-09-10 00:00:00}	24972
{"start":1992-11-12 00:00:00,"end":1992-11-19 00:00:00}	24984
{"start":1993-01-07 00:00:00,"end":1993-01-14 00:00:00}	24753
Time taken: 5.095 seconds, Fetched 10 row(s)

Issues Resolved

#25

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.
For more information on following Developer Certificate of Origin and signing off your commits, please check here.

[penghuo]

why not using window function directly? tumble function is more align with streaming sql grammer?

[dai-chen]

Yes, currently tumble is just an alias of window. The reason for doing this is Streaming SQL has a family of windowing function, ex. tumbling, sliding, session etc. Tumbling window is one of them that happen to be supported by Spark window() function directly.