separate projection and alias-projection

[jackwener]

Which issue does this PR close?

Part of #3927 #2212

We should separate projection and alias-projection.

In fact, alias-projection is subqueryAlias, after this job, we can easy remove the alias in projection, and replace it with subqueryAlias

Rationale for this change

What changes are included in this PR?

Are these changes tested?

Are there any user-facing changes?

[jackwener]

Problems like this indicate pushdown rule exist some bug.

[mingmwang]

Do we need to support such kind of usage?

select ('aa1', 'aa2' as AA1) as B

[mingmwang]

BTW, below is the Spark SQL's plan

== Optimized Logical Plan ==
Project [[aa1,aa2] AS B#941145], Statistics(sizeInBytes=6.0 B)
+- OneRowRelation, Statistics(sizeInBytes=1.0 B)

== Physical Plan ==
*(1) Project [[aa1,aa2] AS B#941145]
+- *(1) Scan OneRowRelation[]

Another example;

select ('aa1', 'aa2' as AA1) as B, ('xx1', 'xx2' as BB1) as X

[jackwener]

Spark doesn't have alias-projection. @mingmwang

We also can read sql standard 2003

Alias just contain

• expr-alias (like: projection contain is expr)
• table-ref-alias

select ('aa1', 'aa2' as AA1) as B

is 

Alias("B", Expr::List(aa1, Alias("AA1", aa2)))

Spark projection just contains Seq[NamedExpr] input

[jackwener]

https://imgse.com/i/znjM9I

So we don't need projection keep alias, just expr::Alias is enough

[mingmwang]

https://imgse.com/i/znjM9I

So we don't need projection keep alias, just expr::Alias is enough

I am OK with removing alias from Projection. Let get some advices from other members also.

[jackwener]

close. all job finish in #4333