Skip to content

Release

Latest
Latest

Choose a tag to compare

View all tags
@snowflake-connectors-app snowflake-connectors-app released this 12 Nov 21:18
v1.57.0
abd294b

1.57.0 (2026-11-12)

Snowpark Python API Updates

New Features

  • Added Session.semantic_view() (PuPr) to query a semantic view. Pass the semantic view name plus any of dimensions, metrics, facts, and where; it returns a DataFrame.

  • Added Session.debug (Experimental) to turn the debug features on and off per session, with session.debug.set(...) or session.debug.enable_<feature>() / disable_<feature>(). The features are:

    • eager_schema_validation: validates the schema of a DataFrame when it is created, so an invalid operation fails at the line that defines it.
    • trace_sql_errors_to_dataframe: maps a SQL compilation error to the Python lines of the DataFrame operations that generated the failing SQL.
    • dataframe_trace_on_error: adds the most recent DataFrame operations of a failed DataFrame to the error.

    snowflake.snowpark.context.configure_development_features is now a mirror of Session.debug that applies to every session.

  • Added support for the following functions in functions.py:

    • try_to_boolean
    • try_to_date
    • try_to_decfloat
    • try_to_decimal (with aliases try_to_number and try_to_numeric)
    • try_to_double
    • try_to_time
    • try_to_timestamp
    • try_to_timestamp_ltz
    • try_to_timestamp_ntz
    • try_to_timestamp_tz

Improvements

  • The debug features of Session.debug (Experimental) and snowflake.snowpark.context.configure_development_features (Experimental) no longer depend on AST collection and no longer change the AST collection setting.
  • DataFrameReader.documents() (Private Preview) no longer takes the parse_mode, row_boundary and extract_images options. A read without a schema or prompt now always parses each file with AI_PARSE_DOCUMENT in LAYOUT mode and returns one row per file.
  • snowflake.snowpark.context.configure_development_features now changes only the features that are passed explicitly and keeps the others; calling it without arguments disables all of them.

Bug Fixes

  • Fixed a bug that Session.debug (Experimental) report wrong debug information when imported through a symlinked path.
  • Fixed a bug where trace_sql_errors_to_dataframe of Session.debug (Experimental) reported a wrong Python source position when CTE optimization was enabled.
  • Fixed a bug where trace_sql_errors_to_dataframe of Session.debug (Experimental) reported extra Python source lines when the failing DataFrame operation wraps earlier operations in a subquery.
  • Fixed a bug that caused invalid identifier errors when Session.reduce_describe_query_enabled is turned on and DataFrame.group_by is used on a case-sensitive quoted column.
  • Fixed a bug where importing Snowpark failed if the requests package was not installed. Event table telemetry now treats requests as an optional dependency of the opentelemetry extra.
  • Fixed a bug where DataFrameReader.documents() reads with a schema failed on every file in accounts that return structured types in the native Arrow format (for example, with ENABLE_STRUCTURED_TYPES_NATIVE_ARROW_FORMAT set).
  • Fixed a bug where DataFrameReader.documents() reads with a schema failed on every file, instead of falling back to the next model, when AI_COMPLETE reported the first model as unavailable (Model "<name>" is unavailable).
  • Fixed a bug where a UDF, UDTF or stored procedure with an imported module in packages failed with TypeError on a non-Anaconda artifact repository, such as PyPI. The module is now pinned to its local version, as Session.add_packages does. This also fixes snowflake.snowpark.dataframe.map with vectorized=True on such a repository when packages is not given.
  • Fixed a bug where trace_sql_errors_to_dataframe of Session.debug (Experimental) reported no Python source position for errors in a non-SELECT statement passed to Session.sql().

Dependency Updates

  • The [pandas] install extra now declares pandas and pyarrow directly instead of coupling them through snowflake-connector-python[pandas]. The version bounds are unchanged.

Snowpark Local Testing Updates

Bug Fixes

  • Fixed a bug where a column in local testing lost its name when it was copied or filtered.
  • Fixed a bug where arithmetic on columns with NULL values returned an extra row in local testing when it followed a filter.
  • Fixed a bug where Column.isin on a datetime column in local testing did not match any value when the subquery it was given also contained NULL or a value that cannot be parsed as a timestamp.
  • Fixed a bug where dateadd with the year, quarter or month part raised an error instead of returning NULL for a NULL date in local testing.
  • Fixed a bug where Session.create_dataframe raised OverflowError in local testing for a pandas DataFrame with a datetime column outside the nanosecond range (for example, year 3000).
  • Fixed a bug where DataFrame.pivot in local testing raised TypeError when an aggregated value came from date_part with the week, weekiso or yearofweekiso part and a pivot cell was empty.
  • Fixed a bug where DataFrameReader.csv in local testing accepted a string or a number for SKIP_BLANK_LINES and treated it as true. It now raises an error unless the value is True or False, as Snowflake does.
  • Fixed a bug where DataFrameReader.csv in local testing read strings such as NA and null as NaN in a column of a type it does not convert, such as VariantType.
  • Fixed a bug where Column.try_cast to TimeType failed instead of returning NULL for invalid input in local testing.
  • Fixed a bug where Column.try_cast to DoubleType returned NaN instead of NULL for invalid input in local testing.
Assets 8
Loading