SparkSQL: TABLESAMPLE with aliased database.table_name fails

[CDonnerer]

Fully reproducible code snippet

1/ using transpile:

import sqlglot

query = "SELECT * FROM my_database.my_table TABLESAMPLE (1 PERCENT) foo"
sqlglot.transpile(query, read="spark")

SELECT * FROM my_table TABLESAMPLE (1 PERCENT) AS foo

here my_database is missing, expected:

SELECT * FROM my_database.my_table TABLESAMPLE (1 PERCENT) AS foo

2/ using parse_one:

import sqlglot

query = "SELECT * FROM my_database.my_table TABLESAMPLE (1 PERCENT) foo"
sqlglot.parse_one(query, read="spark")

SELECT * FROM my_database.my_table AS foo TABLESAMPLE (1 PERCENT)

here the alias is misplaced, expected:

SELECT * FROM my_database.my_table TABLESAMPLE (1 PERCENT) AS foo

Official Documentation

https://spark.apache.org/docs/latest/sql-ref-syntax-qry-select-sampling.html

See also here for example snippets https://github.com/sqlfluff/sqlfluff/blob/main/test/fixtures/dialects/hive/select_sampling_table.sql

[tobymao]

in the second case, this is expected because you didn't specify an output dialect, try it with sql(dialect="spark")

[CDonnerer]

Awesome, thanks a lot for the super quick fix!