How to change cache location? about neuropyxels HOT 5 CLOSED

Hi Chris,

Yes I realized this could become a problem, I am in the process of completely getting rid of my 'home-made caching' inherited form a time when I didn't know better to replace it with joblib - I will implement this feature asap, and have the default be in the same directory of the dataset. It would distribute the data more.

from neuropyxels.

Sounds exciting! In the meantime a hacky way for me to get around this would be to change the path in cachedir = Path(op.expanduser("~")) / ".NeuroPyxels" in the import section of a bunch of functions?

from neuropyxels.

Yes that is the quick hack! Just search all the repo for ".NeuroPyxels", I am not sure in how many places I have set it (it will be in a single place in the future, I promise!). It seems it is at the top of the following modules: behav.py, coor.py, spk_t.py and spk_wvf.py.

from neuropyxels.

Implemented in the last PR (not yet available on pip, only on the git master branch - I want to make sure it doesn't break anything for a bit longer)!

Now, you can now simply edit the directory in .../npyx/CONFIG.py, which is __cachedir__ = "~/.NeuroPyxels" by default, with any path.

I would stick to a fast SSD though!

(It isn't actually possible to have the joblib cache directory sit in individual datasets, because this directory needs to be defined at module initialization time. However, some of the caching already relies on dataset-wise directories, this is what 'npyxMemory' is! But it is 'hand-made caching' using numpy, which is more error-prone and a pain to implement everywhere. If it causes more issues in the future, I will work harder on it.)

from neuropyxels.

FYI, I implemented a solution (turned it into a python package because it turned out to be much more work than anticipated): https://github.com/m-beau/cachecache

Testing it for a couple of weeks before pushing it to the master branch.

from neuropyxels.