v1.17.0

Bug Fixes

• Accordion now runs with use of where
• Set all dictionary parameters equal to None default to avoid shared mutable default
• Allowed accordion to let empty dataframes pass through unaffected
• Allowed empty columns in dataframes to be passed through
• Added handling to ensure that Key or MatchingColumns is passed when training a key or semantic lookup, ensuring that models are not broken when trained.
• Fixed bug in accordion when using a where clause that filtered out the first row
• Updated variable handling in python wrangle to properly interpret variables while having delayed interpretation.

Enhancements

• Added input connector to init file which will allow it to be shown in schema
• Brackets now allowed in select.list_element so that something like '[:1]' can be passed.
• auto_rename_columns param added to select.groupby allowing users to opt out of the renaming of columns in groupby. (ie column1 -> column1.list)
• Refactored merge.coalesce recipe wrangle to run quicker
• Updated pytest version
• Set all dictionary defaults to None instead of an empty dictionary to avoid shared mutable defaults
• Added handling for the use of brackets in select.list_element. Ie element: '[0:2]'
• lookup_mode param added to lookup wrangle to expediate lookups. Values include by_row (default), by_dataframe, and by_matrix.
• Added log wrangle to list of wrangles with delayed variable interpretation. row_count, column_count, columns, and df variables are now available in log.
• Added system variable called recipe_variables that contains a dictionary of all variables.