Jérôme Dockès

more generally I would really like to avoid going overboard with the precommit and linting stuff and only enforce coding style rules that actually make the codebase easier to maintain...

how often is this needed? because this is fairly easy to do by modifying the matplotlib figure produced by nilearn, so if only a few users need it they can...

here is one way to do it by the way: ``` from nilearn import datasets, plotting img = datasets.fetch_neurovault_motor_task()['images'][0] coords = plotting.find_xyz_cut_coords(img) display = plotting.plot_stat_map(img, cut_coords=coords) coords = dict(zip('xyz', coords))...

> if the decimal is non-zero, actually show it automatically selected slice coordinates will almost never be integers

I'll ping @michellewang and @mtorabi59 who might be interested :)

that's true but regardless of cluster failure I think it might be nice to have some utilities for sanity check / pipeline inspection report / integration test that shows how...

> > I think CircleCI does all this but it takes way too long. Maybe a faster simpler test to quickly generate plots and reports to visually examine will be...

I think you can close this one for now

> i've tried to reproduce with adhd-data, but not successful. I also tried and failed to reproduce it with the development dataset: ``` import numpy as np from matplotlib import...

> It looks like some functions of this module are either not tested correctly, or not tested at all. > For example: `generate_labeled_regions_large`, `write_fake_bold_img`, `generate_signals_from_precisions`.... > I think it would...