Databricks `SET VAR` / `SET VARIABLE` syntax is not parsed correctly

[baruchoxman]

>>> query1 = "SET VAR a = 5"
>>> query2 = "SET VARIABLE a = 5"
>>> sqlglot.parse_one(query1, "databricks")
'SET VAR a = 5' contains unsupported syntax. Falling back to parsing as a 'Command'.
Command(this=SET, expression=VAR a = 5)
>>> sqlglot.parse_one(query2, "databricks")
'SET VARIABLE a = 5' contains unsupported syntax. Falling back to parsing as a 'Command'.
Command(this=SET, expression=VARIABLE a = 5)
>>> sqlglot.parse_one(query2, "duckdb")
Set(
  expressions=[
    SetItem(
      this=EQ(
        this=Column(
          this=Identifier(this=a, quoted=False)),
        expression=Literal(this=5, is_string=False)),
      kind=VARIABLE)],
  unset=False,
  tag=False)

For comparison, DuckDB already supports similar parsing (see example above).
Documentation: https://docs.databricks.com/aws/en/sql/language-manual/sql-ref-syntax-aux-set-variable

[georgesittas]

This is low priority for the core team right now, but we'll be happy to accept a well-crafted PR. There should be existing logic in sqlglot for handling SET statements already, so shouldn't be a big effort to tackle this. Thanks for the report.

[baruchoxman]

This appears to have been fixed as a side effect of PR #7235 (Spark parser extraction refactor), which added "VAR" and "VARIABLE" to SparkParser.SET_PARSERS. Since DatabricksParser extends SparkParser, it inherits this behavior automatically.

>>> sqlglot.parse_one('SET VAR a = 5', 'databricks')
Set(expressions=[SetItem(this=EQ(...), kind=VARIABLE)], ...)
>>> sqlglot.parse_one('SET VARIABLE a = 5', 'databricks').sql('databricks')
'SET VARIABLE a = 5'

Tests for Spark dialect covering the Databricks output already exist in tests/dialects/test_spark.py:1338. I'll submit a PR adding explicit Databricks-specific tests.

[baruchoxman]

Submitted PR #7490 with explicit Databricks-specific tests for all SET VAR / SET VARIABLE variants.