as_dm() accepts non-dataframe objects which can lead to unexpected errors

[DanChaltiel]

Hi,

I recently discovered this great package, thank you very much for your work!

I'm working with a package that outputs a list of dataframe from a database.
However, the list also contains some metadata members of various types along with the dataframes.
Using as_dm() on the list throws no warning regarding those members and their presence causes some bugs in functions such as dm_draw().

Here is a reprex:

library(dm)
a = list(a=iris, b=mtcars, x=1)
x = as_dm(a)
x
#> -- Metadata --------------------------------------------------------------------
#> Tables: `a`, `b`, `x`
#> Columns: 16
#> Primary keys: 0
#> Foreign keys: 0

dm_draw(x, rankdir = "TB", view_type = "all")
#> Error in `vec_rbind()`:
#> ! Can't combine `..1$column` <character> and `..3$column` <logical>.

^{Created on 2024-02-21 with reprex v2.0.2}

Note that x is considered as a table in the printed Metadata.

The error points to unnest.R#L47, and is caused by non-dataframes yielding a logical (and empty) column column in col_data.

I would expect either an early error in as_dm() forbidding the list to contain any non-dataframe, or an automatic removal of non-dataframes, maybe with a warning.
A hint to use purr::keep(my_list, is.data.frame) might be useful too.

I'd be happy to provide a PR if that can help you.

[krlmlr]

Thanks, I agree we should be testing inputs more carefully here. Would you like to contribute?