Problem:
Read_counts() was slow for large (scRNA matrices) mainly due to unneeded type conversion and checking. Up to 50 seconds for 16000x1800 matrix. While fread() reads in 2-3 seconds.
Solution:
Added as.matrix, as.char to read.as_matrix to skip type conversion when not needed.
Skipping pgx.checkINPUT check for read_counts
Integer64 are converted directly after fread to numeric. No need to convert to character.
Problem: Read_counts() was slow for large (scRNA matrices) mainly due to unneeded type conversion and checking. Up to 50 seconds for 16000x1800 matrix. While fread() reads in 2-3 seconds.
Solution: