Add from_dict classmethod

[MarcoGorelli]

as discussed on the latest call

[MarcoGorelli]

Thinking about this again, I'm not sure how this would work

In order for this to be possible, I think the following would probably be necessary:

• constructors for Column
• a way to generate a given DataFrame from a dict of Columns, which would probably require being able to get the underlying data from the columns

Let's talk about this more next time, but I'm not sure from_dict is the solution

[rgommers]

constructors for Column

There should be a lot of functions that one can apply directly to columns, so you can get new columns from ops on existing ones. But agreed - probably also need to be able to construct columns from sequences and arrays for many use cases.

a way to generate a given DataFrame from a dict of Columns, which would probably require being able to get the underlying data from the columns

This one I don't understand - this should be a completely trivial operation, and no access of underlying data should be needed (a dataframe is simply an ordered collection of columns). Or am I missing your point?

[MarcoGorelli]

Since we agreed in #118 that each library can have private methods to access the underlying data of its own columns, you're right that it is indeed trivial - e.g. in pandas we could do

    @classmethod
    def from_dict(cls, data: dict[str, PandasColumn]) -> PandasDataFrame:
        return cls(pd.DataFrame({label: column._series for label, column in data.items()}))

[MarcoGorelli]

is this one OK, or does anything else need doing?

[rgommers]

This LGTM indeed, in it goes. Thanks Marco.