[IMPALA-11452] Improve alias handling in CollectionTableRef

• Type: Improvement
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: Impala 4.1.0
• Component/s: Frontend
• Epic Color: ghx-label-5
• Assignee: Unassigned
• Reporter: Csaba Ringhofer
• Created: 22/Jul/22 17:36
• Updated: 09/Aug/22 02:37

Description

Currently a TableRef is expected to have 1 or 2 aliases:
1: it is an explicit alias, like "t" in "select * from db.tbl t"
2: the first is fully qualified name ("db.tbl"), the second is an implicit name ("tbl") in "select * from db.tbl"

https://github.com/apache/impala/blob/02043744beee8c57ee876cb8d94eba28660927cb/fe/src/main/java/org/apache/impala/analysis/TableRef.java#L83

The case with 2 aliases is not enough to properly express naming of collection tables, e.g. "select item from db.tbl.arr_col".

The issue is that while the fully qualified alias is always enough to point to the right collection, the second implicit alias is often not really usable, for example for the inner collection of a 2d table it will be always "item" (while the fully qualified alias will be "db.table.arr_col.item". If there are two such collections in a query, only the fully qualified form becomes usable, which is a bit too verbose. It could be useful to add an alias that starts at the column name, e.g "arr_col.item" it the 2d example above.

Issue Links

• relates to: IMPALA-11148 Unnesting from views work only when alias is provided