pl$DataFrame() should accept a schema argument

[etiennebacher]

pl$DataFrame(a = 1:3, schema = list(a = pl$Int16))
#> Error in `as_polars_df()` at neo-r-polars/R/dataframe-frame.R:55:3:
#> ! Evaluation failed.
#> Caused by error in `FUN()`:
#> ! Unsupported class for `as_polars_series()`: polars_data_type, polars_object

[eitsupi]

To simplify the implementation, could you call the cast method later?
Since unlike Python lists, R vectors are guaranteed to have all elements of the same type and the type of the corresponding Polars is fixed, which seems more natural.
(The only exception is as_polars_series(<list>).)

[etiennebacher]

It would be easier to translate code from py-polars to r-polars if we accepted this argument (it would also save some time by not having to move code to $cast())

[eitsupi]

If we were to add an argument named schema to pl$DataFrame(), we would not be able to create a column named schema, so in the tidyverse world it is common to use an argument name like .schema.
Is that reasonable?

[etiennebacher]

Yes I think it would be ok

[eitsupi]

Hmmm, I looked at it a bit and it looks more complicated than I thought to implement.
I'm busy this week so I'll look at it next week.

[eitsupi]

Thinking about it again, isn't there a difference between schema and schema_overrides here, and what you really want is schema_overrides, not schema, I think.
See pola-rs/r-polars#897

[eitsupi]

Closed by 29ff4d2