Custom extraction

Whale supports custom metadata scrapers through plain python scripts, as long as these satisfy two criteria:

  • A list of all tables should be generated in ~/.whale/manifests/manifest.txt.

  • Table "stubs" should be stored as markdown files in ~/.whale/metadata/.

To register a build script, simply run wh init and follow the build-script workflow, manually add details (see Connection configuration).

This can be particularly useful for organizations where metadata is already collected and stored in an intermediary repository (e.g. Airbnb or Lyft), or if you want to get metadata from a source we do not support. For an example, see the databuilder/ file in the whale github repo.