daf.storage formats¶
For in-memory storage use the memory storage.
For disk storage, use either files or h5fs storage. The trade-off is that files storage allows for (much) better
performance for large data sets (even allowing for accessing data larger than the available RAM); and it stores each
data (array) in a different file in a single directory, allowing the use a multitude of file-based tools, most notably
build tools like make for automating reproducible computation pipelines. In contrast, h5fs storage places
everything in a single file, and provides advanced storage features such as compression (though these aren’t easy to
apply through the daf API).
Finally, if you need to read or write AnnData, use anndata, and to disable caching of data, use none.