Conversion to R matrix wreaks havoc in case of mixed type DataFrames

[lbeltrame]

When converting a mixed_type DataFrame to a R matrix (not a data.frame), everything is casted to string vectors, because a matrix can't be heterogeneously typed, to my knowledge.

I have two possible solutions I can send as pull request, and I would like to ask which is better:

1. raise ValueError for mixed type dataframes
2. raise a warning

Given that most matrices needed in R are the numeric ones, I would go for option 1.

[lbeltrame]

Fixing this with a warning is opening a can of worms. Setting with the exception for now.

[wesm]

That seems like the right move. As long as heterogeneous data.frames are not an issue