NumPy 1.19.0 Release Notes#
This NumPy release is marked by the removal of much technical debt: support for Python 2 has been removed, many deprecations have been expired, and documentation has been improved. The polishing of the random module continues apace with bug fixes and better usability from Cython.
The Python versions supported for this release are 3.6-3.8. Downstream developers should use Cython >= 0.29.16 for Python 3.8 support and OpenBLAS >= 3.7 to avoid problems on the Skylake architecture.
Code compatibility with Python versions < 3.6 (including Python 2) was dropped from both the python and C code. The shims in
numpy.compatwill remain to support third-party packages, but they may be deprecated in a future release. Note that 1.19.x will not compile with earlier versions of Python due to the use of f-strings.
numpy.delete can no longer be passed an axis on 0d arrays#
This concludes a deprecation from 1.9, where when an
axis argument was
passed to a call to
~numpy.delete on a 0d array, the
obj argument and indices would be completely ignored.
In these cases,
insert(arr, "nonsense", 42, axis=0) would actually overwrite the
entire array, while
delete(arr, "nonsense", axis=0) would be
axis on a 0d array raises
numpy.delete no longer ignores out-of-bounds indices#
This concludes deprecations from 1.8 and 1.9, where
np.delete would ignore
both negative and out-of-bounds items in a sequence of indices. This was at
odds with its behavior when passed a single index.
Now out-of-bounds items throw
IndexError, and negative items index from the
numpy.delete no longer accept non-integral indices#
This concludes a deprecation from 1.9, where sequences of non-integers indices
were allowed and cast to integers. Now passing sequences of non-integral
IndexError, just like it does when passing a single
numpy.delete no longer casts boolean indices to integers#
This concludes a deprecation from 1.8, where
np.delete would cast boolean
arrays and scalars passed as an index argument into integer indices. The
behavior now is to treat boolean arrays as a mask, and to raise an error
on boolean scalars.
Changed random variate stream from
A bug in the generation of random variates for the Dirichlet distribution
with small ‘alpha’ values was fixed by using a different algorithm when
max(alpha) < 0.1. Because of the change, the stream of variates
dirichlet in this case will be different from previous
Scalar promotion in
The promotion of mixed scalars and arrays in
has been changed to adhere to those used by
This means that input such as
(1000, np.array(, dtype=np.uint8)))
will now return
uint16 dtypes. In most cases the behaviour is unchanged.
Note that the use of this C-API function is generally discouraged.
This also fixes
np.choose to behave the same way as the rest of NumPy
in this respect.
Fasttake and fastputmask slots are deprecated and NULL’ed#
The fasttake and fastputmask slots are now never used and must always be set to NULL. This will result in no change in behaviour. However, if a user dtype should set one of these a DeprecationWarning will be given.
np.ediff1d casting behaviour with
np.ediff1d now uses the
"same_kind" casting rule for
to_begin arguments. This
ensures type safety except when the input array has a smaller
integer type than
In rare cases, the behaviour will be more strict than it was
previously in 1.16 and 1.17. This is necessary to solve issues
with floating point NaN.
Converting of empty array-like objects to NumPy arrays#
len(obj) == 0 which implement an “array-like” interface,
meaning an object implementing
obj.__array_struct__, or the python
buffer interface and which are also sequences (i.e. Pandas objects)
will now always retain there shape correctly when converted to an array.
If such an object has a shape of
(0, 1) previously, it could
be converted into an array of shape
(0,) (losing all dimensions
after the first 0).
As part of the continued removal of Python 2 compatibility,
multiarray.int_asbuffer was removed. On Python 3, it threw a
NotImplementedError and was unused internally. It is expected that there
are no downstream use cases for this method with Python 3.
numpy.distutils.compat has been removed#
This module contained only the function
get_exception(), which was used as:
try: ... except Exception: e = get_exception()
Its purpose was to handle the change in syntax introduced in Python 2.6, from
except Exception, e: to
except Exception as e:, meaning it was only
necessary for codebases supporting Python 2.5 and older.
issubdtype no longer interprets
numpy.issubdtype had a FutureWarning since NumPy 1.14 which
has expired now. This means that certain input where the second
argument was neither a datatype nor a NumPy scalar type
(such as a string or a python type like
will now be consistent with passing in
This makes the result consistent with expectations and leads to
a false result in some cases which previously returned true.
Change output of
round on scalars to be consistent with Python#
Output of the
__round__ dunder method and consequently the Python
round has been changed to be a Python
int to be consistent
with calling it on Python
float objects when called with no arguments.
Previously, it would return a scalar of the
np.dtype that was passed in.
numpy.ndarray constructor no longer interprets
The former has changed to have the expected meaning of setting
(), while the latter continues to result in
strides being chosen automatically.
C-Level string to datetime casts changed#
The C-level casts from strings were simplified. This changed
also fixes string to datetime and timedelta casts to behave
correctly (i.e. like Python casts using
while previously the cast would behave like
This only affects code using low-level C-API to do manual casts
(not full array casts) of single scalar values or using e.g.
PyArray_GetCastFunc, and should thus not affect the vast majority
SeedSequence with small seeds no longer conflicts with spawning#
Small seeds (less than
2**96) were previously implicitly 0-padded out to
128 bits, the size of the internal entropy pool. When spawned, the spawn key
was concatenated before the 0-padding. Since the first spawn key is
small seeds before the spawn created the same states as the first spawned
SeedSequence. Now, the seed is explicitly 0-padded out to the internal
pool size before concatenating the spawn key. Spawned
produce different results than in the previous release. Unspawned
SeedSequences will still produce the same results.
dtype=object for ragged input#
np.array([[1, [1, 2, 3]]) will issue a
per NEP 34. Users should explicitly use
dtype=object to avoid the
shape=0 to factory functions in
numpy.rec is deprecated#
0 is treated as a special case and is aliased to
None in the functions:
0 will not be special cased, and will be treated as an array
length like any other integer.
Deprecation of probably unused C-API functions#
The following C-API functions are probably unused and have been deprecated:
In most cases
PyArray_GetArrayParamsFromObject should be replaced
by converting to an array, while
PyUFunc_GenericFunction can be
PyObject_Call (see documentation for details).
Converting certain types to dtypes is Deprecated#
The super classes of scalar types, such as
np.inexact will now give a deprecation warning when converted
to a dtype (or used in a dtype keyword argument).
The reason for this is that
np.integer is converted to
while it would be expected to represent any integer (e.g. also
dtype=np.floating is currently identical to
dtype=np.float64, even though also
np.float32 is a subclass of
Output of the
__round__ dunder method and consequently the Python built-in
round has been deprecated on complex scalars. This does not affect
numpy.ndarray.tostring() is deprecated in favor of
~numpy.ndarray.tobytes has existed since the 1.9 release, but until this
~numpy.ndarray.tostring emitted no warning. The change to emit a
warning brings NumPy in line with the builtin
array.array methods of the
C API changes#
Better support for
const dimensions in API functions#
The following functions now accept a constant array of
Previously the caller would have to cast away the const-ness to call these functions.
Const qualify UFunc inner loops#
UFuncGenericFunction now expects pointers to const
strides as arguments. This means inner loops may no longer modify
strides. This change leads to an
incompatible-pointer-types warning forcing users to either ignore
the compiler warnings or to const qualify their own loop signatures.
numpy.frompyfunc now accepts an identity argument#
This allows the
numpy.ufunc.identity attribute to be set on the
resulting ufunc, meaning it can be used for empty and multi-dimensional
np.str_ scalars now support the buffer protocol#
np.str_ arrays are always stored as UCS4, so the corresponding scalars
now expose this through the buffer interface, meaning
memoryview(np.str_('test')) now works.
subok option for
A new kwarg,
subok, was added to
numpy.copy to allow users to toggle
the behavior of
numpy.copy with respect to array subclasses. The default
False which is consistent with the behavior of
previous numpy versions. To create a copy that preserves an array subclass with
np.copy(arr, subok=True). This addition better
documents that the default behavior of
numpy.copy differs from the
numpy.ndarray.copy method which respects array subclasses by default.
numpy.linalg.multi_dot now accepts an
out can be used to avoid creating unnecessary copies of the final product
keepdims parameter for
keepdims was added to
parameter has the same meaning as it does in reduction functions such
equal_nan parameter for
The keyword argument
equal_nan was added to
equal_nan is a boolean value that toggles whether or not
nan values are
considered equal in comparison (default is
False). This matches API used in
related functions such as
Improve detection of CPU features#
npy_cpu_supports which was a gcc specific mechanism to test support
of AVX with more general functions
expose the results via a
NPY_CPU_HAVE c-macro as well as a python-level
Use 64-bit integer size on 64-bit platforms in fallback lapack_lite#
Use 64-bit integer size on 64-bit platforms in the fallback LAPACK library, which is used when the system has no LAPACK installed, allowing it to deal with linear algebra for large arrays.
Use AVX512 intrinsic to implement
np.exp when input is
Use AVX512 intrinsic to implement
np.exp when input is
which can improve the performance of
np.float64 input 5-7x
faster than before. The
_multiarray_umath.so module has grown about 63 KB
Ability to disable madvise hugepages#
On Linux NumPy has previously added support for madavise hugepages which can improve performance for very large arrays. Unfortunately, on older Kernel versions this led to performance regressions, thus by default the support has been disabled on kernels before version 4.6. To override the default, you can use the environment variable:
or set it to 1 to force enabling support. Note that this only makes a difference if the operating system is set up to use madvise transparent hugepage.
numpy.einsum accepts NumPy
int64 type in subscript list#
There is no longer a type error thrown when
numpy.einsum is passed
int64 array as its subscript list.
np.logaddexp2.identity changed to
~numpy.logaddexp2 now has an identity of
-inf, allowing it to
be called on empty sequences. This matches the identity of
Remove handling of extra argument to
A code path and test have been in the code since NumPy 0.4 for a two-argument
__array__(dtype=None, context=None). It was activated when
However that variant is not documented, and it is not clear what the intention
was for its use. It has been removed.
numpy.random._bit_generator moved to
In order to expose
numpy.random.SeedSequence to Cython, the
_bitgenerator module is now
Cython access to the random distributions is provided via a
c_distributions.pxd provides access to the c functions behind many of the
random distributions from Cython, making it convenient to use and extend them.
cholesky methods in
Previously, when passing
numpy.random.multivariate_normal produced samples from the wrong
distribution. This is now fixed.
Fixed the jumping implementation in
This fix changes the stream produced from jumped MT19937 generators. It does
not affect the stream produced using
are directly seeded.
The translation of the jumping code for the MT19937 contained a reversed loop
MT19937.jumped matches the Makoto Matsumoto’s original
implementation of the Horner and Sliding Window jump methods.