Compatibility Issue: Temporary Table Creation in save_as_table Conflicts with Snowflake Native Apps

[Synohara]

I've encountered a compatibility problem when using snowflake-ml-python within a Snowflake Native App environment. The issue stems from the internal use of snowpark.DataFrame.save_as_table within the library (e.g., in OrdinalEncoder).

This method appears to create a temporary table under the hood:

snowflake-ml-python/snowflake/ml/modeling/preprocessing/ordinal_encoder.py

Lines 250 to 258 in 2932445

	state_df.write.save_as_table( # type: ignore[call-overload]
	self._vocab_table_name,
	mode="overwrite",
	table_type="temporary",
	statement_params=telemetry.get_statement_params(base.PROJECT, base.SUBPROJECT, self.__class__.__name__),
	)
	self._state_pandas = state_df.to_pandas(
	statement_params=telemetry.get_statement_params(base.PROJECT, base.SUBPROJECT, self.__class__.__name__)
	)

However, Snowflake Native Apps have a restriction that prevents the creation of temporary tables. This incompatibility renders certain parts of snowflake-ml-python unusable within Native Apps.

Suggested Resolution:

I'd like to request that the team consider a modification to address this limitation.

[sfc-gh-sdas]

Thanks for reporting. We are aware of this issue. Native Apps team is considering some workaround for this.