Merge pull request numpy#27291 from DimitriPapadopoulos/codespell

DOC, MAINT: Fix new typos found by codespell

Author: charris

doc/release/upcoming_changes/27147.performance.rst

@@ -2,7 +2,7 @@
   benchmarking, there are 5 clusters of performance around these kernels:
   ``PRESCOTT NEHALEM SANDYBRIDGE HASWELL SKYLAKEX``.
 
-* OpenBLAS on windows is linked without quadmath, simplfying licensing
+* OpenBLAS on windows is linked without quadmath, simplifying licensing
 
 * Due to a regression in OpenBLAS on windows, the performance improvements
   when using multiple threads for OpenBLAS 0.3.26 were reverted.

doc/source/reference/arrays.promotion.rst

@@ -172,7 +172,7 @@ would give.
 
 Behavior of ``sum`` and ``prod``
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-**``np.sum`` and ``np.prod``:** Will alway return the default integer type
+**``np.sum`` and ``np.prod``:** Will always return the default integer type
 when summing over integer values (or booleans).  This is usually an ``int64``.
 The reason for this is that integer summations are otherwise very likely
 to overflow and give confusing results.
@@ -214,7 +214,7 @@ The following rules apply:
   non-ascii characters.
 * For some purposes NumPy will promote almost any other datatype to strings.
   This applies to array creation or concatenation.
-* The array constructers like ``np.array()`` will use ``object`` dtype when
+* The array constructors like ``np.array()`` will use ``object`` dtype when
   there is no viable promotion.
 * Structured dtypes can promote when their field names and order matches.
   In that case all fields are promoted individually.

numpy/_core/src/multiarray/dlpack.c

@@ -57,7 +57,7 @@ array_dlpack_deleter_unversioned(DLManagedTensor *self)
 
 
 /*
- * Deleter for a DLPack capsule wrapping a DLManagedTensor(Versioed).
+ * Deleter for a DLPack capsule wrapping a DLManagedTensor(Versioned).
  *
  * This is exactly as mandated by dlpack
  */

numpy/_core/src/multiarray/scalartypes.c.src

@@ -253,7 +253,7 @@ find_binary_operation_path(
      *
      * However, NumPy (historically) made this often work magically because
      * ufuncs for object dtype end up casting to object with `.item()`. This in
-     * turn ofthen returns a Python type (e.g. float for float32, float64)!
+     * turn often returns a Python type (e.g. float for float32, float64)!
      * Retrying then succeeds. So if (and only if) `self.item()` returns a new
      * type, we can safely attempt the operation (again) with that.
      */

numpy/_core/tests/test_array_coercion.py

@@ -763,7 +763,7 @@ def __getitem__(self):
             np.array(BadSequence())
 
     def test_array_interface_descr_optional(self):
-        # The descr should be optional regresion test for gh-27249
+        # The descr should be optional regression test for gh-27249
         arr = np.ones(10, dtype="V10")
         iface = arr.__array_interface__
         iface.pop("descr")