Allow subqueries without aliases #2418
Conversation
|
@alamb @yjshen @xudong963 @houqp How do you think we should handle this? Should I consider enabling this or not based on the SQL dialect being used? |
There was a problem hiding this comment.
I don't honestly know the deeper background and understanding of why postgres requires an alias or if there are deeper implications to allowing unaliased subqueries.
However, I think the additional test coverage is 👍 And the fact there are no additional errors is a good sign to me.
| assert_batches_eq!(expected, &actual); | ||
| Ok(()) | ||
| } | ||
|
|
There was a problem hiding this comment.
Maybe we could also add a test here that had two unaliased subqueries (should result in a cross join) as a potential other case that would be good to cover
There was a problem hiding this comment.
Thanks for the review @alamb. I added the suggested test and did not run into problems so will go ahead and merge this one.
Which issue does this PR close?
Closes #2417
Rationale for this change
I am trying to run some ANSI SQL that works fine in Apache Spark but fails in DataFusion because we currently require subqueries to have aliases.
As described in the issue though, Postgres does require subqueries to be aliased, so that is probably what determined the current behavior?
Should we make this behavior configurable?
There was a discussion about this in #1067
What changes are included in this PR?
Remove the check that caused subqueries without aliases to fail. Note that there were no tests for this functionality.
Are there any user-facing changes?
No