At the moment, each dataset has a specific way to create the folder, and sometimes we have situations where if we change the config the path, the moabb will try to access the initial name.
Example: Weibo has the complete MNE-weibo-2014, but MOABB wants MNE-weibo-2014-data.
We need to standardize this. To do so, we first need to identify where the issue occurs by running an integration test across all datasets. In this test, we should mock the download mechanisms for the 160+ datasets and check whether a configuration change allows continued access or if the download attempts to revert to the old location.
If that happens, we must adjust the download function for each problematic dataset without deprecating any paths or anything of the sort.
This is a two-step process: first, we dedicate time to the integration test—which must iterate through all datasets—to identify the problematic ones; then, we apply the necessary patches to those datasets.
At the moment, each dataset has a specific way to create the folder, and sometimes we have situations where if we change the config the path, the moabb will try to access the initial name.
Example: Weibo has the complete MNE-weibo-2014, but MOABB wants MNE-weibo-2014-data.
We need to standardize this. To do so, we first need to identify where the issue occurs by running an integration test across all datasets. In this test, we should mock the download mechanisms for the 160+ datasets and check whether a configuration change allows continued access or if the download attempts to revert to the old location.
If that happens, we must adjust the download function for each problematic dataset without deprecating any paths or anything of the sort.
This is a two-step process: first, we dedicate time to the integration test—which must iterate through all datasets—to identify the problematic ones; then, we apply the necessary patches to those datasets.