issue_comments

Reconsidering: #89 was a feature request that relates to this, so maybe this is worth implementing after all.

I'm not going to implement this. I'll leave extract=... as it is right now, suitable for quick simple single-column operations on input, but if users want to do something more complicated involving multiple columns they should use the table.extract() method after the initial insert instead.

But this begins to feel too complicated, given that table.extract() can already be used to achieve the same thing.

If I want to support that most complicated example, I think the option to pass a Extracts() object to extracts= is the best way to do it: python fresh_db.table("tree", extracts=[Extract( columns=("CommonName", "LatinName"), table="Species", fk_column="species_id", rename={"CommonName": "name", "LatinName": "latin"} )])

Here's the most complex example of .extracts(): python db["Trees"].extract( ["CommonName", "LatinName"], table="Species", fk_column="species_id", rename={"CommonName": "name", "LatinName": "latin"} ) Resulting in: sql CREATE TABLE [Species] ( [id] INTEGER PRIMARY KEY, [name] TEXT, [latin] TEXT ) From https://sqlite-utils.readthedocs.io/en/stable/python-api.html#extracting-columns-into-a-separate-table

I think there is. It's a nice existing feature, and I don't think adding tuple support to it would be a huge lift.

I wonder if there's value in extending the extracts= option at all given the existence of table.extract().

Using a tuple would work: python fresh_db.table("tree", extracts=[("common_name", "latin_name")]) Or to define a custom name: python fresh_db.table("tree", extracts={("common_name", "latin_name"): "names"})

The new .extract() method can handle multiple columns:

https://github.com/simonw/sqlite-utils/blob/2c541fac352632e23e40b0d21e3f233f7a744a57/tests/test_extract.py#L70-L87

This syntax should be shared with #42 as much as possible.

Maybe something based on a namedtuple would work, since those are already used in the library.

python workouts = db.table("workouts", extracts=[Extract( columns=["source", "source_version"], table="Sources" )]) Since namedtuples cannot have default values this should probably be a class instead. Actually it looks like there is a trick for defaults here: https://stackoverflow.com/a/18348004