Parallel table transformer

[lars-reimann]

Is your feature request related to a problem?

The sequential table transformer (#802) is great if later transformations depend on prior ones. Often, however, columns are transformed independently, which gives us opportunity to parallelize.

Desired solution

New class ParallelTableTransformer.

• Superclass: InvertibleTableTransformer
• Constructor parameters:
  • transformers: list[TableTransformer].
  • Raise if multiple transformers are applied to the same column.
• fit:
  • Fit the individual transformers on the given table in parallel. Keep in mind that multithreading is usually not helpful for CPU-bound tasks in Python due to the Global Interpreter lock. Multiprocessing is likely more efficient. Profile your code!
• transform:
  • Transform the table with the individual transformers in parallel (see fit). Gather the results of the transformers and merge them to a single table.
• inverse_transform:
  • Raise an error if any transformer is not invertible
  • Inverse-transform the table with the individual transformers in parallel (see fit). Gather the results of the transformers and merge them to a single table.

Possible alternatives (optional)

No response

Screenshots (optional)

No response

Additional Context (optional)

No response

[wastedareas]

Story points were decided by assignees