Similar to what the other pull requests have done, I've added a function to read .h5ad format for the expression matrices. This dramatically speeds up the load time during the _load_meta_counts step. The expression matrix is stored in the .X slot and gene symbols are extracted from the indices of the .var slot.
Also made a wrapper to read pandas created .h5 file formats if the users simply created an expression matrix with pandas' .to_hdf. It expects only a single pandas object in the .h5 file. This is not quite as fast as from .h5ad but still faster than .txt/.csv
Also co-opted the ability to read .mtx containg folders from #162
Also made corrections to try and catch/prevent errors/warnings about making edits on slices of views, dtype errors, indexing errors, concat errors, import numpy directly, rather than use the depreciated pd.np modules, mostly arising due to usage of different pandas versions (should now work for pandas 1.2).
Hi,
Similar to what the other pull requests have done, I've added a function to read
.h5adformat for the expression matrices. This dramatically speeds up the load time during the_load_meta_countsstep. The expression matrix is stored in the.Xslot and gene symbols are extracted from the indices of the.varslot.Also made a wrapper to read pandas created
.h5file formats if the users simply created an expression matrix with pandas'.to_hdf. It expects only a single pandas object in the.h5file. This is not quite as fast as from.h5adbut still faster than.txt/.csvAlso co-opted the ability to read
.mtxcontaing folders from #162Also made corrections to try and catch/prevent errors/warnings about making edits on slices of views, dtype errors, indexing errors, concat errors, import numpy directly, rather than use the depreciated
pd.npmodules, mostly arising due to usage of different pandas versions (should now work for pandas 1.2).