You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Add a clear_cache function which automatically cleans scipy.datasets cache directory based on the users' platform. This will help avoid the tedious manual way for a user to empty cache if required. (PR ENH: Add clear_cache utility for scipy.datasets #17478)
When running the complete SciPy test suite offline, we should intelligently skip tests given there is no network connection. Also when pooch is not installed, all dataset tests can be skipped. Maybe they also need an XSLOW marker. BLD: Allow disabling network requiring tests #17965
Move other dataset files (eg. scipy.stats has its own test datasets like nist within the repo) to their respective new repos (like https://github.com/scipy/dataset-ascent) and utilize the functionality from scipy.datasets and pooch to reduce the wheel size by moving these datasets to scipy.datasets. Since the data files are only used in testing, it will NOT introduce any new dependency (like pooch) on the stats module.
As mentioned in the comment about the restrictions and a packager needing to adhere to Debian rules, add a separate script or package (Eg: pip install scipy-datasets) to download everything at once before SciPy is built or tested. (Handled in PR ENH: Add download_all utility method & script #17163)
About more dataset: in general, it would be good to think about datasets in cooperation with Scikit-learn and Scikit-image (non exhaustive list at all). Since these are outside of SciPy, it would make sense to have a wide community effort.
Maintaining below a list of TODO tasks and follow-ups compile after the recently added submodule
scipy.datasets
.download_all
utility method to fetch and cache all the datasets at once. (PR ENH: Adddownload_all
utility method & script #17163)clear_cache
function which automatically cleansscipy.datasets
cache directory based on the users' platform. This will help avoid the tedious manual way for a user to empty cache if required. (PR ENH: Add clear_cache utility forscipy.datasets
#17478)pooch
is not installed, all dataset tests can be skipped. Maybe they also need anXSLOW
marker. BLD: Allow disabling network requiring tests #17965scipy.stats
has its own test datasets like nist within the repo) to their respective new repos (like https://github.com/scipy/dataset-ascent) and utilize the functionality fromscipy.datasets
andpooch
to reduce the wheel size by moving these datasets toscipy.datasets
. Since the data files are only used in testing, it will NOT introduce any new dependency (like pooch) on the stats module.pip install scipy-datasets
) to download everything at once before SciPy is built or tested. (Handled in PR ENH: Adddownload_all
utility method & script #17163)cerifi
release.certifi
had issues withpy3.11
which are now fixed in master. Pending release of Fix deprecation warning on Python 3.11 certifi/python-certifi#199. (PR MAINT: remove certifi py3.11 warning filter #17149)scipy.datasets
submodule:I'll start working on the tasks above, feel free to pick up one of the tasks if it interests you (comment below before you start working).
cc @rgommers
The text was updated successfully, but these errors were encountered: