REF: Use more cython memoryviews (#58271)

* fix some cython 3 todos * Another memoryview * Add another memoryview * Use more memoryview * another memoryview * more memoryview syntax in groupby * Add a few more memoryviews * Use Py_ssize_t

Author: mroeschke

pandas/_libs/algos.pyx

@@ -351,7 +351,7 @@ def nancorr(const float64_t[:, :] mat, bint cov=False, minp=None):
  Py_ssize_t i, xi, yi, N, K
  int64_t minpv
  float64_t[:, ::1] result
- ndarray[uint8_t, ndim=2] mask
+ uint8_t[:, :] mask
  int64_t nobs = 0
  float64_t vx, vy, dx, dy, meanx, meany, divisor, ssqdmx, ssqdmy, covxy
 
@@ -407,7 +407,7 @@ def nancorr_spearman(ndarray[float64_t, ndim=2] mat, Py_ssize_t minp=1) -> ndarr
  Py_ssize_t i, xi, yi, N, K
  ndarray[float64_t, ndim=2] result
  ndarray[float64_t, ndim=2] ranked_mat
- ndarray[float64_t, ndim=1] rankedx, rankedy
+ float64_t[::1] rankedx, rankedy
  float64_t[::1] maskedx, maskedy
  ndarray[uint8_t, ndim=2] mask
  int64_t nobs = 0
@@ -566,8 +566,8 @@ def get_fill_indexer(const uint8_t[:] mask, limit=None):
 @cython.boundscheck(False)
 @cython.wraparound(False)
 def pad(
- ndarray[numeric_object_t] old,
- ndarray[numeric_object_t] new,
+ const numeric_object_t[:] old,
+ const numeric_object_t[:] new,
  limit=None
 ) -> ndarray:
  # -> ndarray[intp_t, ndim=1]
@@ -691,8 +691,8 @@ def pad_2d_inplace(numeric_object_t[:, :] values, uint8_t[:, :] mask, limit=None
 @cython.boundscheck(False)
 @cython.wraparound(False)
 def backfill(
- ndarray[numeric_object_t] old,
- ndarray[numeric_object_t] new,
+ const numeric_object_t[:] old,
+ const numeric_object_t[:] new,
  limit=None
 ) -> ndarray: # -> ndarray[intp_t, ndim=1]
  """
@@ -786,7 +786,7 @@ def backfill_2d_inplace(numeric_object_t[:, :] values,
 
 @cython.boundscheck(False)
 @cython.wraparound(False)
-def is_monotonic(ndarray[numeric_object_t, ndim=1] arr, bint timelike):
+def is_monotonic(const numeric_object_t[:] arr, bint timelike):
  """
  Returns
  -------
@@ -1089,8 +1089,7 @@ cdef void rank_sorted_1d(
  float64_t[::1] out,
  int64_t[::1] grp_sizes,
  const intp_t[:] sort_indexer,
- # TODO(cython3): make const (https://github.com/cython/cython/issues/3222)
- numeric_object_t[:] masked_vals,
+ const numeric_object_t[:] masked_vals,
  const uint8_t[:] mask,
  bint check_mask,
  Py_ssize_t N,
@@ -1378,16 +1377,15 @@ ctypedef fused out_t:
 @cython.boundscheck(False)
 @cython.wraparound(False)
 def diff_2d(
- ndarray[diff_t, ndim=2] arr, # TODO(cython3) update to "const diff_t[:, :] arr"
+ const diff_t[:, :] arr,
  ndarray[out_t, ndim=2] out,
  Py_ssize_t periods,
  int axis,
  bint datetimelike=False,
 ):
  cdef:
  Py_ssize_t i, j, sx, sy, start, stop
- bint f_contig = arr.flags.f_contiguous
- # bint f_contig = arr.is_f_contig() # TODO(cython3) once arr is memoryview
+ bint f_contig = arr.is_f_contig()
  diff_t left, right
 
  # Disable for unsupported dtype combinations,

pandas/_libs/groupby.pyx

@@ -511,9 +511,9 @@ def group_shift_indexer(
 @cython.wraparound(False)
 @cython.boundscheck(False)
 def group_fillna_indexer(
- ndarray[intp_t] out,
- ndarray[intp_t] labels,
- ndarray[uint8_t] mask,
+ Py_ssize_t[::1] out,
+ const intp_t[::1] labels,
+ const uint8_t[:] mask,
  int64_t limit,
  bint compute_ffill,
  int ngroups,
@@ -1179,13 +1179,13 @@ def group_ohlc(
 @cython.boundscheck(False)
 @cython.wraparound(False)
 def group_quantile(
- ndarray[float64_t, ndim=2] out,
+ float64_t[:, ::1] out,
  ndarray[numeric_t, ndim=1] values,
- ndarray[intp_t] labels,
+ const intp_t[::1] labels,
  const uint8_t[:] mask,
  const float64_t[:] qs,
- ndarray[int64_t] starts,
- ndarray[int64_t] ends,
+ const int64_t[::1] starts,
+ const int64_t[::1] ends,
  str interpolation,
  uint8_t[:, ::1] result_mask,
  bint is_datetimelike,
@@ -1388,7 +1388,7 @@ cdef inline void _check_below_mincount(
  uint8_t[:, ::1] result_mask,
  Py_ssize_t ncounts,
  Py_ssize_t K,
- int64_t[:, ::1] nobs,
+ const int64_t[:, ::1] nobs,
  int64_t min_count,
  mincount_t[:, ::1] resx,
 ) noexcept:
@@ -1435,14 +1435,12 @@ cdef inline void _check_below_mincount(
  out[i, j] = 0
 
 
-# TODO(cython3): GH#31710 use memorviews once cython 0.30 is released so we can
-# use `const numeric_object_t[:, :] values`
 @cython.wraparound(False)
 @cython.boundscheck(False)
 def group_last(
  numeric_object_t[:, ::1] out,
  int64_t[::1] counts,
- ndarray[numeric_object_t, ndim=2] values,
+ const numeric_object_t[:, :] values,
  const intp_t[::1] labels,
  const uint8_t[:, :] mask,
  uint8_t[:, ::1] result_mask=None,
@@ -1502,14 +1500,12 @@ def group_last(
  )
 
 
-# TODO(cython3): GH#31710 use memorviews once cython 0.30 is released so we can
-# use `const numeric_object_t[:, :] values`
 @cython.wraparound(False)
 @cython.boundscheck(False)
 def group_nth(
  numeric_object_t[:, ::1] out,
  int64_t[::1] counts,
- ndarray[numeric_object_t, ndim=2] values,
+ const numeric_object_t[:, :] values,
  const intp_t[::1] labels,
  const uint8_t[:, :] mask,
  uint8_t[:, ::1] result_mask=None,

pandas/_libs/hashing.pyx

@@ -11,7 +11,6 @@ import numpy as np
 
 from numpy cimport (
  import_array,
- ndarray,
  uint8_t,
  uint64_t,
 )
@@ -23,7 +22,7 @@ from pandas._libs.util cimport is_nan
 
 @cython.boundscheck(False)
 def hash_object_array(
- ndarray[object] arr, str key, str encoding="utf8"
+ object[:] arr, str key, str encoding="utf8"
 ) -> np.ndarray[np.uint64]:
  """
  Parameters