Spawned off https://github.com/onnx/onnx/pull/2772/files/7ab93cc1b635eada330dae7424d4ff7e8c22c295#r440422245, opening issue to track resolution

If you join Dask DataFrame on a categorical column, then the outputted Dask DataFrame column is still category dtype. However, the moment you .compute() the outputted Dask DataFrame, then the column is the wrong dtype, not categorical.

Tested on Dask 2.14.0 and Pandas 1.0.3
This example where the category type looks like a float, so after .compute(), the dtype is float.

import dask.d

For example, if there is a relationship transaction.session_id -> sessions.id and we are calculating a feature transactions: sessions.SUM(transactions.value) any rows for which there is no corresponding session should be given the default value of 0 instead of NaN.

Of course this should not normally occur, but when it does it seems more reasonable to use the default_value.

`DirectF

with the Power Transformer.

I see the code

device = ‘cuda’ if torch.cuda.is_available() else ‘cpu’

repeated often in user code. Maybe we should introduce device='auto' exactly for this case?

Yes

I think it could be useful, when one wants to plot only e.g. class 1, to have an option to produce consistent plots for both plot_cumulative_gain and plot_roc

At the moment, instead, only plot_roc supports such option.

Thanks a lot

Support DataFrame.select_dtypes