bug: AttributeError: 'NoneType' object has no attribute 'pop'

[yazad3]

What happened?

Context

Not sure if this is an sqlglot bug or truely an ibis bug - but either ways would need a fix in ibis code or dependency on a certain version of sqlglot. The error AttributeError: 'NoneType' object has no attribute 'pop' is coming from deep inside sqlglot's add_within_group_for_percentiles function.

Error

Traceback (most recent call last):
  File "/Users/xyz/abc/abc_ds/trial/tests_trial/sanity/ibis/test_ibis_sanity.py", line 129, in test_sanity_snowflake
    .to_pandas()
  File "/Users/xyz/opt/anaconda3/envs/trial_env/lib/python3.10/site-packages/ibis/expr/types/relations.py", line 3332, in to_pandas
    return self.execute(**kwargs)
  File "/Users/xyz/opt/anaconda3/envs/trial_env/lib/python3.10/site-packages/ibis/expr/types/core.py", line 359, in execute
    return self._find_backend(use_default=True).execute(
  File "/Users/xyz/opt/anaconda3/envs/trial_env/lib/python3.10/site-packages/ibis/backends/sql/__init__.py", line 242, in execute
    sql = self.compile(table, params=params, limit=limit, **kwargs)
  File "/Users/xyz/opt/anaconda3/envs/trial_env/lib/python3.10/site-packages/ibis/backends/sql/__init__.py", line 127, in compile
    sql = query.sql(dialect=self.dialect, pretty=pretty, copy=False)
  File "/Users/xyz/opt/anaconda3/envs/trial_env/lib/python3.10/site-packages/sqlglot/expressions.py", line 561, in sql
    return Dialect.get_or_raise(dialect).generate(self, **opts)
  File "/Users/xyz/opt/anaconda3/envs/trial_env/lib/python3.10/site-packages/sqlglot/dialects/dialect.py", line 498, in generate
    return self.generator(**opts).generate(expression, copy=copy)
  File "/Users/xyz/opt/anaconda3/envs/trial_env/lib/python3.10/site-packages/sqlglot/generator.py", line 583, in generate
    sql = self.sql(expression).strip()
  File "/Users/xyz/opt/anaconda3/envs/trial_env/lib/python3.10/site-packages/sqlglot/generator.py", line 734, in sql
    sql = transform(self, expression)
  File "/Users/xyz/opt/anaconda3/envs/trial_env/lib/python3.10/site-packages/sqlglot/transforms.py", line 601, in _to_sql
    return _sql_handler(expression)
  File "/Users/xyz/opt/anaconda3/envs/trial_env/lib/python3.10/site-packages/sqlglot/generator.py", line 2199, in select_sql
    expressions = self.expressions(expression)
  File "/Users/xyz/opt/anaconda3/envs/trial_env/lib/python3.10/site-packages/sqlglot/generator.py", line 3229, in expressions
    sql = self.sql(e, comment=False)
  File "/Users/xyz/opt/anaconda3/envs/trial_env/lib/python3.10/site-packages/sqlglot/generator.py", line 739, in sql
    sql = getattr(self, exp_handler_name)(expression)
  File "/Users/xyz/opt/anaconda3/envs/trial_env/lib/python3.10/site-packages/sqlglot/generator.py", line 2725, in alias_sql
    return f"{self.sql(expression, 'this')}{alias}"
  File "/Users/xyz/opt/anaconda3/envs/trial_env/lib/python3.10/site-packages/sqlglot/generator.py", line 728, in sql
    return self.sql(value)
  File "/Users/xyz/opt/anaconda3/envs/trial_env/lib/python3.10/site-packages/sqlglot/generator.py", line 734, in sql
    sql = transform(self, expression)
  File "/Users/xyz/opt/anaconda3/envs/trial_env/lib/python3.10/site-packages/sqlglot/transforms.py", line 595, in _to_sql
    expression = transforms[0](expression)
  File "/Users/xyz/opt/anaconda3/envs/trial_env/lib/python3.10/site-packages/sqlglot/transforms.py", line 329, in add_within_group_for_percentiles
    column = expression.this.pop()
AttributeError: 'NoneType' object has no attribute 'pop'

Snippet that cuased the error

# Using the following import to avoid confusion with ignore variables (`_`)
from ibis import _ as c

# INFORMATION_SCHEMA
tables_table: Table = snowflake_backend.table("TABLES")

filtered_table: Table = tables_table.filter(
	tables_table.TABLE_SCHEMA == "TESTING"
).select(
	"TABLE_CATALOG",
	"TABLE_SCHEMA",
	"TABLE_NAME",
	"TABLE_OWNER",
	"ROW_COUNT",
	"BYTES",
)

row_count_stats: dict[str, int | float] = (
	filtered_table.aggregate(
		min_row_count=c.ROW_COUNT.min(),
		max_row_count=c.ROW_COUNT.max(),
		sum_row_count=c.ROW_COUNT.sum(),
		mean_row_count=c.ROW_COUNT.mean(),
		median_row_count=c.ROW_COUNT.median(),
		quantile_0_25_row_count=c.ROW_COUNT.quantile(0.25),
		quantile_0_75_row_count=c.ROW_COUNT.quantile(0.75),
	)
	.to_pandas()
	.to_dict("records")[0]
)

Environment Information

Python

Anaconda Python 3.10 on Mac

Installed requirements

sqlglot==23.0.5
ibis-framework[pandas,snowflake,duckdb]==9.0.0.dev511

What version of ibis are you using?

latest dev veriosn - 9.0.0.dev511

What backend(s) are you using, if any?

Snowflake

Relevant log output

Stacktrace in "What happened?"

Code of Conduct

• I agree to follow this project's Code of Conduct

[gforsyth]

Hey @yazad3! Thanks for reporting this!
I can reproduce the issue and I've got a proximate cause identified, just working on tracking down a simpler reproducer and then we should be able to push up a fix.

[yazad3]

Thank you! Let me know if I can be of any help with reproduction and / or fixing.

[cpcloud]

@gforsyth Can we close this out?

[cpcloud]

Whoops, didn't see the PR, reviewing now!

[yazad3]

Thank you for the fix!