Open Zethson opened 18 hours ago
The problem is that reticulate isn't used for .load()
and .cache()
.
If reticulate was used, this would all be resolved.
Hence, the fix should be using reticulate for these two methods and it's going to work.
I think we should have a discussion about whether it is worth using the API at all. We have about reached the limit of what the API can do currently and if we have to use {reticulate} for some things anyway, maybe it's better to use it for everything? It would mean a fairly big refactor but after that development might be quicker.
Yes, we should have that discussion next week or so, I agree.
But what's indeed much better with the API is that you're not relying on 20 Python packages that Django needs to map all the schema modules.
So, it's not a clear decision pro reticulate for querying say bionty. That's likely more elegant through the REST API.
We observed that when data from any non-default instance such as cellxgene is used as input during a
Run
, it does not show up in the linage graph.@lazappi had suggested that this is probably because of our usage of the API to get the data and not the Python code which probably does more magic.