Highlights#

• Interactive examples in the NumPy documentation.

• Building NumPy with OpenMP Parallelization.

• Preliminary support for Windows on ARM.

• Improved support for free threaded Python.

• Improved annotations.

New functions#

Deprecations#

• Thenumpy.typing.mypy_plugin has been deprecated in favor of platform-agnosticstatic type inference. Please removenumpy.typing.mypy_plugin from thepluginssection of your mypy configuration. If this change results in new errors beingreported, kindly open an issue.

  (gh-28129)

• Thenumpy.typing.NBitBase type has been deprecated and will be removed ina future version.

  This type was previously intended to be used as a generic upper bound fortype-parameters, for example:

  importnumpyasnpimportnumpy.typingasnptdeff[NT:npt.NBitBase](x:np.complexfloating[NT])->np.floating[NT]:...

  But in NumPy 2.2.0,float64 andcomplex128 were changed to concretesubtypes, causing static type-checkers to rejectx:np.float64=f(np.complex128(42j)).

  So instead, the better approach is to usetyping.overload:

  importnumpyasnpfromtypingimportoverload@overloaddeff(x:np.complex64)->np.float32:...@overloaddeff(x:np.complex128)->np.float64:...@overloaddeff(x:np.clongdouble)->np.longdouble:...

  (gh-28884)

Expired deprecations#

• Remove deprecated macros likeNPY_OWNDATA from Cython interfaces in favorofNPY_ARRAY_OWNDATA (deprecated since 1.7)

  (gh-28254)

• Removenumpy/npy_1_7_deprecated_api.h and C macros likeNPY_OWNDATAin favor ofNPY_ARRAY_OWNDATA (deprecated since 1.7)

  (gh-28254)

• Remove aliasgenerate_divbyzero_error tonpy_set_floatstatus_divbyzero andgenerate_overflow_error tonpy_set_floatstatus_overflow (deprecated since 1.10)

  (gh-28254)

• Removenp.tostring (deprecated since 1.19)

  (gh-28254)

• Raise onnp.conjugate of non-numeric types (deprecated since 1.13)

  (gh-28254)

• Raise when usingnp.bincount(...minlength=None), use 0 instead(deprecated since 1.14)

  (gh-28254)

• Passingshape=None to functions with a non-optional shape argumenterrors, use() instead (deprecated since 1.20)

  (gh-28254)

• Inexact matches formode andsearchside raise (deprecated since 1.20)

  (gh-28254)

• Setting__array_finalize__=None errors (deprecated since 1.23)

  (gh-28254)

• np.fromfile andnp.fromstring error on bad data, previously theywould guess (deprecated since 1.18)

  (gh-28254)

• datetime64 andtimedelta64 construction with a tuple no longeraccepts anevent value, either use a two-tuple of (unit, num) or a4-tuple of (unit, num, den, 1) (deprecated since 1.14)

  (gh-28254)

• When constructing adtype from a class with adtype attribute, thatattribute must be a dtype-instance rather than a thing that can be parsed asa dtype instance (deprecated in 1.19). At some point the whole construct ofusing a dtype attribute will be deprecated (see #25306)

  (gh-28254)

• Passing booleans as partition index errors (deprecated since 1.23)

  (gh-28254)

• Out-of-bounds indexes error even on empty arrays (deprecated since 1.20)

  (gh-28254)

• np.tostring has been removed, usetobytes instead (deprecated since 1.19)

  (gh-28254)

• Disallow make a non-writeable array writeable for arrays with a base that donot own their data (deprecated since 1.17)

  (gh-28254)

• concatenate() withaxis=None usessame-kind casting by default,notunsafe (deprecated since 1.20)

  (gh-28254)

• Unpickling a scalar with object dtype errors (deprecated since 1.20)

  (gh-28254)

• The binary mode offromstring now errors, usefrombuffer instead(deprecated since 1.14)

  (gh-28254)

• Convertingnp.inexact ornp.floating to a dtype errors (deprecatedsince 1.19)

  (gh-28254)

• Convertingnp.complex,np.integer,np.signedinteger,np.unsignedinteger,np.generic to a dtype errors (deprecated since1.19)

  (gh-28254)

• The Python built-inround errors for complex scalars. Usenp.round orscalar.round instead (deprecated since 1.19)

  (gh-28254)

• ‘np.bool’ scalars can no longer be interpreted as an index (deprecated since 1.19)

  (gh-28254)

• Parsing an integer via a float string is no longer supported. (deprecatedsince 1.23) To avoid this error you can* make sure the original data is stored as integers.* use theconverters=float keyword argument.* Usenp.loadtxt(...).astype(np.int64)

  (gh-28254)

• The use of a length 1 tuple for the ufuncsignature errors. Usedtypeor fill the tuple withNone (deprecated since 1.19)

  (gh-28254)

• Special handling of matrix is in np.outer is removed. Convert to a ndarrayviamatrix.A (deprecated since 1.20)

  (gh-28254)

• Removed thenp.compat package source code (removed in 2.0)

  (gh-28961)

C API changes#

• NpyIter_GetTransferFlags is now available to check ifthe iterator needs the Python API or if casts may cause floating pointerrors (FPE). FPEs can for example be set when castingfloat64(1e300)tofloat32 (overflow to infinity) or a NaN to an integer (invalid value).

  (gh-27883)

• NpyIter now has no limit on the number of operands it supports.

  (gh-28080)

New Features#

• The type parameter ofnp.dtype now defaults totyping.Any.This way, static type-checkers will inferdtype:np.dtype asdtype:np.dtype[Any], without reporting an error.

  (gh-28669)

• Static type-checkers now interpret:

  • _:np.ndarray as_:npt.NDArray[typing.Any].

  • _:np.flatiter as_:np.flatiter[np.ndarray].

  This is because their type parameters now have default values.

  (gh-28940)

Improvements#

• Scalar comparisons between non-comparable dtypes such asnp.array(1)==np.array('s') now return a NumPy bool instead ofa Python bool.

  (gh-27288)

• np.nditer now has no limit on the number of supported operands(C-integer).

  (gh-28080)

• No-copy pickling is now supported for anyarray that can be transposed to a C-contiguous array.

  (gh-28105)

• The__repr__ for user-defined dtypes now prefers the__name__ of thecustom dtype over a more generic name constructed from itskind anditemsize.

  (gh-28250)

• np.dot now reports floating point exceptions.

  (gh-28442)

• np.dtypes.StringDType is now ageneric type whichaccepts a type argument forna_object that defaults totyping.Never.For example,StringDType(na_object=None) returns aStringDType[None],andStringDType() returns aStringDType[typing.Never].

  (gh-28856)

Performance improvements and changes#

Changes#

• The vector normord=inf and the matrix normsord={1,2,inf,'nuc'}now always returns zero for empty arrays. Empty arrays have at least one axisof size zero. This affectsnp.linalg.norm,np.linalg.vector_norm, andnp.linalg.matrix_norm. Previously, NumPy would raises errors or returnzero depending on the shape of the array.

  (gh-28343)

• A spelling error in the error message returned when converting a string to afloat with the methodnp.format_float_positional has been fixed.

  (gh-28569)

• NumPy’s__array_api_version__ was upgraded from2023.12 to2024.12.

• numpy.count_nonzero foraxis=None (default) now returns a NumPy scalarinstead of a Python integer.

• The parameteraxis innumpy.take_along_axis function has now a defaultvalue of-1.

  (gh-28615)

• Printing ofnp.float16 andnp.float32 scalars and arrays have been improved byadjusting the transition to scientific notation based on the floating point precision.A new legacynp.printoptions mode'2.2' has been added for backwards compatibility.

  (gh-28703)

• Multiplication between a string and integer now raises OverflowError insteadof MemoryError if the result of the multiplication would create a string thatis too large to be represented. This follows Python’s behavior.

  (gh-29060)