v1.16.0

• drop wrangle now has handling for missing columns avoiding errors from trying to drop non-existent columns
• Added compute.case_when wrangle which allows users to assign values to a column based on conditional logic
• Fixed bug which prevented extract.regex from showing up in the schema
• .recipe has been added to the schema allowing .recipe files to be run in the same way as .wrgl.yaml files
• skip_empty parameter has been added to format.pad, format.prefix, format.suffix, merge.dictionaries, merge.key_value_pairs, and split.text wrangles.
• MSSQL version has been updated to allow compatibility with macOS > 13
• Fixed bug that caused errors when using a groupby within a matrix
• Fixed bug where sort was broken when sorting on numbers and strings
• Added handling for multiple defaults to be used in select.element
• sort parameter added to extract.custom allowing users to sort based on training order (default), input order, longest, shortest, alphabetical, reverse alphabetical, ascending, and descending
• Fixed bug where matrix was broken when passed integer variables
• Added handling to the s3 connector that allows directories to be created when they do not already exist, allowed AWS errors to proliferate through wrangles, changed, key param to file_key, and the file parameter to save_as
• Errors now indicate which batch and wrangle caused the error
• Wrangles that are not run do to an if statement are no longer being logged.
• first_element dropped from extract.address schema as it does not output a list
• Added drop_empty parameter to the file connector to drop empty columns when reading a file
• train.extract has now been updated to allow training with v3 extract columns
• compare.list wrangle added to compare lists with different methods, which include intersection, difference and union
• convert.to_json now outputs an array when run on a scalar
• New action parameter added to train.lookup connect which allows user to overwrite, update, insert or upsert into the existing lookup data
• Train connector now allows semantic lookups to be trained without the requirement of a Key column