`csr_row_norms` for scikit-learn, using fused types and typed memory views.

csr_row_norms.pyx

from libc.math cimport sqrt
cimport cython
cimport numpy as np
import numpy as np

ctypedef fused my_fused_type:
    cython.short
    cython.int
    cython.long
    cython.float
    cython.double

def _csr_row_norms(my_fused_type[:] data, int[:] indptr, bint squared=False):
    cdef:
        int num_rows = indptr.shape[0] - 1
        double[:] norms
        double sum, x
        int i, j
    norms = out = np.zeros(num_rows, dtype=np.float64)
    for i in range(num_rows):
        sum = 0.
        for j in range(indptr[i], indptr[i + 1]):
            x = <double>data[j]
            sum += x * x
        if not squared:
            sum = sqrt(sum)
        norms[i] = sum
    return out

def csr_row_norms(X, bint squared=False):
    return _csr_row_norms(X.data, X.indptr, squared)