generate-table-partitions fails for Spanner with Unknown dialect spanner

[sundar-mudupalli-work]

Hi

A typical generate-table-partitions command fails because we are not able to convert the IBIS where clause (i.e. an IBIS table expression) into SQL text. This works for 8 other databases BigQuery, Hive, MySQL, Oracle, Postgres, Snowflake, SQL Server and Teradata (we have test cases for all those databases).

The command is

data-validation generate-table-partitions --secret-manager-type gcp --secret-manager-project-id span-cloud-ck-testing-external --source-conn my_mysql -tc my_spanner --tables-list person.person1=test-db.person1 --hash 'first_name2' --primary-keys ID -cdir gs://dvt-ck-cloud-run-testing/mudupalli/MySp8GPart16 -pn 16

Replacing my_spanner (which points to a Spanner database) with my_mysql - i.e. validate against same database succeeds.

The specific error is below:

  File "/home/mudupalli/professional-services-data-validator/env/lib/python3.9/site-packages/data_validation/partition_builder.py", line 73, in partition_configs
    partition_filters = self._get_partition_key_filters()
  File "/home/mudupalli/professional-services-data-validator/env/lib/python3.9/site-packages/data_validation/partition_builder.py", line 229, in _get_partition_key_filters
    self._extract_where(target_table.filter(filter_target_clause))
  File "/home/mudupalli/professional-services-data-validator/env/lib/python3.9/site-packages/data_validation/partition_builder.py", line 85, in _extract_where
    return re.sub(r"\s\s+", " ", ibis.to_sql(x).sql.split("WHERE")[1]).replace(
  File "/home/mudupalli/professional-services-data-validator/env/lib/python3.9/site-packages/ibis/expr/sql.py", line 381, in to_sql
    (pretty,) = sg.transpile(sql, read=read, write=write, pretty=True)
  File "/home/mudupalli/professional-services-data-validator/env/lib/python3.9/site-packages/sqlglot/__init__.py", line 187, in transpile
    for expression in parse(sql, read, error_level=error_level)
  File "/home/mudupalli/professional-services-data-validator/env/lib/python3.9/site-packages/sqlglot/__init__.py", line 74, in parse
    dialect = Dialect.get_or_raise(read)()
  File "/home/mudupalli/professional-services-data-validator/env/lib/python3.9/site-packages/sqlglot/dialects/dialect.py", line 136, in get_or_raise
    raise ValueError(f"Unknown dialect '{dialect}'")
ValueError: Unknown dialect 'spanner'

[sundar-mudupalli-work]

Hi,

The problem occurs because the Spanner backend behaves differently from the BigQuery backend - perhaps the Spanner backend is not fully implemented. I can work around this issue - however it might be a very simple fix in the Spanner backend. Here is a Python interpreter session demonstrating the issue. BigQuery works fine with the to_sql function returning the sql string. When Spanner is presented with a table expression that execute() correctly, it still fails to render with the to_sql function.

(env) mudupalli@dvt-mudupalli-test001:~/gcsdir$ python3
Python 3.9.2 (default, Feb 28 2021, 17:03:44) 
[GCC 10.2.1 20210110] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import ibis
>>> from data_validation.clients import get_data_client
>>> bq_c = get_data_client({"source_type": "BigQuery", "project_id": "span-cloud-ck-testing-external"})
>>> result_tbl = bq_c.table('span-cloud-ck-testing-external.pso_data_validator.results')
>>> ibis.to_sql(result_tbl)
SQLString(sql='SELECT\n  t0.*\nFROM `span-cloud-ck-testing-external.pso_data_validator.results` AS t0')
>>> span_c = get_data_client({"source_type": "Spanner", "instance_id": "df-test", "database_id": "test-db"})
>>> person_tbl = span_c.table('person1')
>>> person_100 = person_tbl.filter(person_tbl.ID == 100)
>>> person_100.execute()
                                         first_name1  ...   ID
0  dshlfiuadhsfkudashgfkuadshgkaudgshkuagykfuydag...  ...  100

[1 rows x 7 columns]
>>> ibis.to_sql(person_100)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/mudupalli/professional-services-data-validator/env/lib/python3.9/site-packages/ibis/expr/sql.py", line 381, in to_sql
    (pretty,) = sg.transpile(sql, read=read, write=write, pretty=True)
  File "/home/mudupalli/professional-services-data-validator/env/lib/python3.9/site-packages/sqlglot/__init__.py", line 187, in transpile
    for expression in parse(sql, read, error_level=error_level)
  File "/home/mudupalli/professional-services-data-validator/env/lib/python3.9/site-packages/sqlglot/__init__.py", line 74, in parse
    dialect = Dialect.get_or_raise(read)()
  File "/home/mudupalli/professional-services-data-validator/env/lib/python3.9/site-packages/sqlglot/dialects/dialect.py", line 136, in get_or_raise
    raise ValueError(f"Unknown dialect '{dialect}'")
ValueError: Unknown dialect 'spanner'

The issue seems to be that the dialect spanner is not registered with sqlglot. Is there a way to fix this in the backend?

Thanks.

Sundar Mudupalli

[nehanene15]

The easiest way to do this is to pass in the BQ dialect for Spanner clients in the to_sql() call we make in partition_builder. BQ and Spanner have the same SQL syntax so this should be interchangeable.

This is a separate issue, but Spanner doesn't support ROW_NUMBER() as shown in this error when Spanner is the source system:
Command:
data-validation -v generate-table-partitions -sc spanner -tc spanner -tbls pso_data_validator.dvt_core_types -pk id -hash '*' -pn 2 -cdir /home/user/professional-services-data-validator/cdir

Error:
google.api_core.exceptions.MethodNotImplemented: 501 Unsupported built-in function: row_number.