cached loading code

[le-ander]

I cannot see any code for actually loading a cached dataset in _load_cached()
am I missing something?

[le-ander]

also I think the funciton breaks if self.cache_path is not set. or is this handled somewehre outside?

[davidsebfischer]

• Loading code was missing, I added this now.
• I am throwing explicit caching errors now if these functionalities are used and cache_path is not set. The data zoo can still be used without caching (ie if data is only ever loaded raw), but I think that cache_path should be almost always set for our puropses.

[le-ander]

Cool, thanks!
Some thoughts on that:

This will throw an error if self.cache_path is None

sfaira/sfaira/data/base.py

Lines 149 to 153 in 8e8756b

	fn_cache = os.path.join(
	self.cache_path,
	self._directory_formatted_doi,
	self._directory_formatted_id + ".h5ad"
	)

Should we really throw an error if load_raw=False and the cached file is not found? I'd rather have interpreted load_raw as "don't use the cache file if I set this to true. When setting it to false, just load the raw one anyway if you cannot find the cached file"

sfaira/sfaira/data/base.py

Line 159 in 8e8756b

assert os.path.exists(fn_cache), f"did not find cache file {fn_cache}, consider caching first"

[davidsebfischer]

I implemented what you suggested, @le-ander!