Support for CSRMatrix

[FFroehlich]

I am currently working on code that computes matrix-matrix products of large, sparse matrices (>1k rows/columns, similar number of non-zero entries), which is relatively slow (multiple hours per product). I assume that this could be computed much faster by storing these matrices in a sparse representation and computing the products using respective sparse routines. However, I could not find any support for the sparse matrix class CSRMatrix in this wrapper.

Is this already implemented and I simply did not look in the right place? Are there any plans to support CSRMatrix in this wrapper in the near future?

[bjodah]

You're right that it is not yet exposed in this python wrapper. I believe it's simply due to no one having had the time / need for it up till now.

[FFroehlich]

Do I correctly understand that implementation would simply require replicating

symengine.py/symengine/lib/symengine.pxd

Line 793 in 5496ffc

cdef cppclass DenseMatrix(MatrixBase):

and following and adapting it to the CSRMatrix class?

[bjodah]

Yes, together with the corresponding adaptions in symengine_wrapper.pyx and adding a few tests to the test suite.

[laisnegautier]

Hi, will we see a bind for CSRMatrix soon?