Patrick Hoefler
Patrick Hoefler
Having a smaller environment because you don’t have to install arrow is the main benefit of fastparquet as far as I am aware of. Pandas will require arrow as a...
> Probably we need to maintain a filesystem= keyword in case someone has an odd filesystem and needs to use fsspec, but this would be the exception I think Just...
We should probably catch the warnings, but I wouldn't worry too much about this (it seems like a non-standard use case for dask)
Hi, there are two different issues at play: - converting to bool won't work because bool(pd.NA) raises - we are converting your values to arrow strings before you do the...
Right now we are converting object dtype to string dtype, we can't look into the data unfortunately
The dataset has some columns that are lists, which we convert to PyArrow strings in a post processing step (which is kind of a known bug because we shouldn't be...
This is fixed with dask-expr
yeah there was definitely something in the cache, it should be gone now
#8618 tries this
did that on my revert pr