aeon-toolkit
[ENH] Storage of TSC datasets when loading from a package using import
Describe the feature or idea you want to propose
writing labs on machine learning using aeon for data loading.
from aeon.datasets import load_gunpoint, load_from_ts_file
X, y = load_gunpoint(split="train", return_type="numpy2D")
print(X.shape, y.shape)
I have only ever done it within aeon, and wondered where the actual ts file gets placed. Turns out it goes in the environment.
I dont think it should go there!
Describe your proposed solution
all to do with
Describe alternatives you've considered, if relevant
No response
Additional context
No response
Hi, I’d like to work on this enhancement.
From my understanding, the issue is that TSC datasets are currently being downloaded into the package directory inside the environment (e.g., site-packages/aeon/datasets), which is not ideal since package directories should be read-only.
I plan to update the default storage location to a user-writable data directory (e.g., via platformdirs), while still supporting custom paths through extract_path. I’ll also ensure backward compatibility and update the documentation.
Please let me know if this approach aligns with what you expect before I proceed.
Feel free to make a PR. We can discuss where the best default is there, the code to create a dir should not change that much regardless of what is chosen.
