Export.py for sparse pandas data frames[results]

[cbravo93]

In export.py these lines cause an error when using a sparse data frame:

# Calculate the number of genes per cell.
binary_mtx = ex_mtx.copy()
binary_mtx[binary_mtx != 0] = 1.0
ngenes = binary_mtx.sum(axis=1).astype(int)

Something like this (not-tested) could be used for sparse pandas:

ngenes = ex_mtx.values().count_nonzero().astype(int)

The question is, is a dense matrix required at any point for creating the loom file? Otherwise changing these would be more efficient for very large data sets.

Cheers!

C

[ghuls]

The binarisation can also be optimized if the input is a sparse matrix.

[cflerin]

Fixed with the 0.11.1 release.