Kenneth Benoit
Kenneth Benoit
The first example is correct, although I can see how it would be confusing, because `context = "document"` and the explanation for the `ordered` argument states that 'Only makes sense...
The parent class of dfm is already CsparseMatrix...
True. And **Matrix** still does not seem to have implemented the sparse integer matrix variant.
OK great - I'll get to testing these asap.
It's not a quanteda issue, it's an issue for recipes on CRAN, when the macOS binary is built. We've issued a pull request for this, hoping it will be reviewed...
Probably a new function. The difference between `dfm_trim()` and what you are describing is that `dfm_trim()` trims the matrix dimensions by pruning features. In what you describe above, you are...
We could implement arguments similar to dfm_trim, for sure. "floor" however has connotations of rounding down to the nearest integer, which of course is not what we are doing. dfm_tozero()?...
I like that. `Matrix::drop0()` is generally designed for sparsifying the matrix by making small fractional values into zeros, but there's no reason that `tol` could not be an integer or...
Hopefully, 4.3.0 will be 15 or 16 May.
> I was almost finishing this PR but became not sure how the message should look if fcm is considered. "_m_ tokens (_n_ documents)" does not work for fcm because...