Support aliases in PARTITION BY and GROUP BY

[big-andy-coates]

Description

fixes: #4881
fixes: #4813

The commit adds support for aliases in GROUP BY and PARTITION BY clauses. For example:

-- input schema: A -> B, C
CREATE STREAM OUTPUT AS 
   SELECT A, B FROM INPUT 
   PARTITION BY C AS D;
-- output schema: D -> A, B.  (where D contains same data as C from input)

-- input schema: A -> B, C
CREATE STREAM TABLE AS 
   SELECT COUNT(1) AS COUNT FROM INPUT
   GROUP BY C AS D;
-- output schema: D -> COUNT.  (where D contains same data as C from input)

This is particularly useful where the new key column name would be generated, e.g.

-- without alias:
-- input schema: A -> B, C
CREATE STREAM OUTPUT AS 
   SELECT A, B FROM INPUT 
   PARTITION BY UDF(C);
-- output schema: KSQL_COL_0 -> A, B

-- with alias:
-- input schema: A -> B, C
CREATE STREAM OUTPUT AS 
   SELECT A, B FROM INPUT 
   PARTITION BY UDF(C) AS D;
-- output schema: D -> A, B

PARTITION BY only supports a single value expression, so aliasing on this is as you'd expect PARTITION BY X as Y.

GROUP BY supports multiple value expressions, e.g. GROUP BY X, Y. This will currently create a new STRING KEY with a generated name e.g. KSQL_COL_0. The commit allows the full set of columns to be aliased, e.g. GROUP BY (X, Y) AS Z. This is an interim solution. Once ksqlDB supports multiple key columns, the ability to alias the set of columns will be removed, and instead support will be added to alias each expression in the GROUP BY e.g. GROUP BY W as X, Y as Z.

Testing done

usual

How to review.

Review the first commit which:

• updates the grammar to support aliases on GROUP BY and PARTITION BY
• adds a PartitionBy much like the existing GroupBy AstNode.

Review the second commit which wires in the aliases. Flow is:

• tweaks the grammar to support aliasing of single GROUP BY.
• Adds an optional alias to GroupBy and PartitionBy.
• Updates SqlFormatter to know about aliases.
• Changes Analysis to take the whole GroupBy and PartitionBy nodes, rather than extract info out of them.
• Changes the LogicalPlanner to use the alias when building schemas and to pass it along where needed.
• Changes the existing StreamsSelectKey query plan step, used for PARTITION BY, to have an optional alias. (Backwards compatible as its optional)
• Changes the existing StreamsGroupBy and TableGroupBy query plan steps, used by GROUP BY, to have an optional alias. (Backwards compatible as its optional)
• Updates the builders for the above steps to do the right thing with regards to the new optional alias.

Reviewer checklist

• Ensure docs are updated if necessary. (eg. if a user visible feature is being added or changed).
• Ensure relevant issues are linked (description should include text like "Fixes #")

[agavra]

LGTM - I did a pretty quick scan as it's a large PR but it seems pretty striaghtforward. If there's anything specific you want me to look at let me know

Let's not forget the docs PR after this :)

[agavra]

just wondering, why not just go through with it?

[big-andy-coates]

Docs will be covered by: #4686 in a single pass.