Databricks: Add support for UPDATE SET * and INSERT * in MERGE statements

[finchxxia]

Add parsing support for the Databricks star shorthand syntax in MERGE statements, allowing UPDATE SET * and INSERT * to reference all columns from the source.

Syntax

[ common_table_expression ]
  MERGE [ WITH SCHEMA EVOLUTION ] INTO target_table_name [target_alias]
     USING source_table_reference [source_alias]
     ON merge_condition
     { WHEN MATCHED [ AND matched_condition ] THEN matched_action |
       WHEN NOT MATCHED [BY TARGET] [ AND not_matched_condition ] THEN not_matched_action |
       WHEN NOT MATCHED BY SOURCE [ AND not_matched_by_source_condition ] THEN not_matched_by_source_action } [...]

matched_action
 { DELETE |
   UPDATE SET * |
   UPDATE SET { column = { expr | DEFAULT } } [, ...] }

not_matched_action
 { INSERT * |
   INSERT (column1 [, ...] ) VALUES ( expr | DEFAULT ] [, ...] )

not_matched_by_source_action
 { DELETE |
   UPDATE SET { column = { expr | DEFAULT } } [, ...] }

Changes

• Add MergeUpdateKind enum (Set(Vec<Assignment>) | Star) to replace the previous assignments field on MergeUpdateExpr, enforcing valid states at the type level
• Add MergeInsertKind::Star variant for INSERT * shorthand
• Add supports_merge_star_syntax() dialect method for Databricks and Generic

[iffyio]

I looks like we should be able to drop the dialect method and let the parser accept the * syntax anytime it shows up?

[finchxxia]

I looks like we should be able to drop the dialect method and let the parser accept the * syntax anytime it shows up?

done

[iffyio]

the tests can be merged into a single function? (maybe into the existing MERGE tests we have in common)

[finchxxia]

the tests can be merged into a single function? (maybe into the existing MERGE tests we have in common)

done. Thanks for your suggestions.

[iffyio]

LGTM! Thanks @finchxxia!