Python interaction when using memory views as attributes on Cython classes

mahalanobis.pyx

%%cython --annotate
#cython: boundscheck=False
#cython: wraparound=False
#cython: cdivision=True

## Adapted from sklearn.neighbors.Mahalanobis
# https://github.com/scikit-learn/scikit-learn/blob/main/sklearn/neighbors/_dist_metrics.pyx#L669

import numpy as np
cimport numpy as np
np.import_array()

ctypedef np.float64_t DTYPE_t
ctypedef np.intp_t ITYPE_t

from libc.math cimport sqrt

DTYPE = np.float64
ITYPE = np.intp

cdef class SnippedMahalanobisDistance():
    cdef:
        DTYPE_t[:, ::1] mat
        DTYPE_t[::1] vec
        DTYPE_t p
        ITYPE_t size
        
    def __cinit__(self):
        self.p = 2
        self.vec = np.zeros(1, dtype=DTYPE, order='c')
        self.mat = np.zeros((1, 1), dtype=DTYPE, order='c')
        self.size = 1
    
    def __init__(self, V=None, VI=None):
        if VI is None:
            if V is None:
                raise ValueError("Must provide either V or VI "
                                 "for Mahalanobis distance")
            VI = np.linalg.inv(V)
        if VI.ndim != 2 or VI.shape[0] != VI.shape[1]:
            raise ValueError("V/VI must be square")

        self.mat = np.asarray(VI, dtype=float, order='C')
        self.size = self.mat.shape[0]
        # we need vec as a work buffer
        self.vec = np.zeros(self.size, dtype=DTYPE)
    
    cdef inline DTYPE_t rdist(self,
                              const DTYPE_t* x1,
                              const DTYPE_t* x2,
                              ITYPE_t size) nogil except -1:
        cdef DTYPE_t tmp, d = 0
        cdef np.intp_t i, j

        # compute (x1 - x2).T * VI * (x1 - x2)

        # There is interaction with CPython from here...
        for i in range(size):
            self.vec[i] = x1[i] - x2[i]

        for i in range(size):
            tmp = 0
            for j in range(size):
                tmp += self.mat[i, j] * self.vec[j]
            d += tmp * self.vec[i]
            
        # ... to there
        return d
    
cdef inline DTYPE_t snipped_mahalanobis(self,
                          DTYPE_t[:, ::1] mat,
                          DTYPE_t[::1] vec,
                          const DTYPE_t* x1,
                          const DTYPE_t* x2,
                          ITYPE_t size) nogil except -1:
    # A variant which does not have any interaction
    # with CPython
    cdef DTYPE_t tmp, d = 0
    cdef np.intp_t i, j

    # compute (x1 - x2).T * VI * (x1 - x2)
    for i in range(size):
        vec[i] = x1[i] - x2[i]

    for i in range(size):
        tmp = 0
        for j in range(size):
            tmp += mat[i, j] * vec[j]
        d += tmp * vec[i]
    return d

Generated code for SnippedMahalanobisDistance.rdist:

+47:     cdef inline DTYPE_t rdist(self,
 48:                               const DTYPE_t* x1,
 49:                               const DTYPE_t* x2,
 50:                               ITYPE_t size) nogil except -1:
+51:         cdef DTYPE_t tmp, d = 0
 52:         cdef np.intp_t i, j
 53: 
 54:         # compute (x1 - x2).T * VI * (x1 - x2)
 55: 
 56:         # There is interaction with CPython from here...
+57:         for i in range(size):
+58:             self.vec[i] = x1[i] - x2[i]
    if (unlikely(!__pyx_v_self->vec.memview)) {PyErr_SetString(PyExc_AttributeError,"Memoryview is not initialized");__PYX_ERR(0, 58, __pyx_L1_error)}
    __pyx_t_4 = __pyx_v_i;
    *((__pyx_t_54_cython_magic_a64c371eaaa29cbf2ef07aea5d70c4ef7f1ec7b9_DTYPE_t *) ( /* dim=0 */ ((char *) (((__pyx_t_54_cython_magic_a64c371eaaa29cbf2ef07aea5d70c4ef7f1ec7b9_DTYPE_t *) __pyx_v_self->vec.data) + __pyx_t_4)) )) = ((__pyx_v_x1[__pyx_v_i]) - (__pyx_v_x2[__pyx_v_i]));
  }
 59: 
+60:         for i in range(size):
+61:             tmp = 0
+62:             for j in range(size):
+63:                 tmp += self.mat[i, j] * self.vec[j]
      if (unlikely(!__pyx_v_self->mat.memview)) {PyErr_SetString(PyExc_AttributeError,"Memoryview is not initialized");__PYX_ERR(0, 63, __pyx_L1_error)}
      __pyx_t_7 = __pyx_v_i;
      __pyx_t_8 = __pyx_v_j;
      if (unlikely(!__pyx_v_self->vec.memview)) {PyErr_SetString(PyExc_AttributeError,"Memoryview is not initialized");__PYX_ERR(0, 63, __pyx_L1_error)}
      __pyx_t_9 = __pyx_v_j;
      __pyx_v_tmp = (__pyx_v_tmp + ((*((__pyx_t_54_cython_magic_a64c371eaaa29cbf2ef07aea5d70c4ef7f1ec7b9_DTYPE_t *) ( /* dim=1 */ ((char *) (((__pyx_t_54_cython_magic_a64c371eaaa29cbf2ef07aea5d70c4ef7f1ec7b9_DTYPE_t *) ( /* dim=0 */ (__pyx_v_self->mat.data + __pyx_t_7 * __pyx_v_self->mat.strides[0]) )) + __pyx_t_8)) ))) * (*((__pyx_t_54_cython_magic_a64c371eaaa29cbf2ef07aea5d70c4ef7f1ec7b9_DTYPE_t *) ( /* dim=0 */ ((char *) (((__pyx_t_54_cython_magic_a64c371eaaa29cbf2ef07aea5d70c4ef7f1ec7b9_DTYPE_t *) __pyx_v_self->vec.data) + __pyx_t_9)) )))));
    }
+64:             d += tmp * self.vec[i]
    if (unlikely(!__pyx_v_self->vec.memview)) {PyErr_SetString(PyExc_AttributeError,"Memoryview is not initialized");__PYX_ERR(0, 64, __pyx_L1_error)}
    __pyx_t_4 = __pyx_v_i;
    __pyx_v_d = (__pyx_v_d + (__pyx_v_tmp * (*((__pyx_t_54_cython_magic_a64c371eaaa29cbf2ef07aea5d70c4ef7f1ec7b9_DTYPE_t *) ( /* dim=0 */ ((char *) (((__pyx_t_54_cython_magic_a64c371eaaa29cbf2ef07aea5d70c4ef7f1ec7b9_DTYPE_t *) __pyx_v_self->vec.data) + __pyx_t_4)) )))));
  }
 65: 
 66:         # ... to there
+67:         return d