NumPy 2.0 migration guide#

This document contains a set of instructions on how to update your code to work with NumPy 2.0. It covers changes in NumPy’s Python and C APIs.


Note that NumPy 2.0 also breaks binary compatibility - if you are distributing binaries for a Python package that depends on NumPy’s C API, please see NumPy 2.0-specific advice.

Ruff plugin#

Many of the changes covered in the 2.0 release notes and in this migration guide can be automatically adapted in downstream code with a dedicated Ruff rule, namely rule NPY201.

You should install ruff>=0.4.8 and add the NPY201 rule to your pyproject.toml:

select = ["NPY201"]

You can also apply the NumPy 2.0 rule directly from the command line:

$ ruff check path/to/code/ --select NPY201

Changes to NumPy data type promotion#

NumPy 2.0 changes promotion (the result of combining dissimilar data types) as per NEP 50. Please see the NEP for details on this change. It includes a table of example changes and a backwards compatibility section.

The largest backwards compatibility change is that the precision of scalars is now preserved consistently. Two examples are:

  • np.float32(3) + 3. now returns a float32 when it previously returned a float64.

  • np.array([3], dtype=np.float32) + np.float64(3) will now return a float64 array. (The higher precision of the scalar is not ignored.)

For floating point values, this can lead to lower precision results when working with scalars. For integers, errors or overflows are possible.

To solve this, you may cast explicitly. Very often, it may also be a good solution to ensure you are working with Python scalars via int(), float(), or numpy_scalar.item().

To track down changes, you can enable emitting warnings for changed behavior (use warnings.simplefilter to raise it as an error for a traceback):


which is useful during testing. Unfortunately, running this may flag many changes that are irrelevant in practice.

Windows default integer#

The default integer used by NumPy is now 64bit on all 64bit systems (and 32bit on 32bit system). For historic reasons related to Python 2 it was previously equivalent to the C long type. The default integer is now equivalent to np.intp.

Most end-users should not be affected by this change. Some operations will use more memory, but some operations may actually become faster. If you experience issues due to calling a library written in a compiled language it may help to explicitly cast to a long, for example with: arr = arr.astype("long", copy=False).

Libraries interfacing with compiled code that are written in C, Cython, or a similar language may require updating to accommodate user input if they are using the long or equivalent type on the C-side. In this case, you may wish to use intp and cast user input or support both long and intp (to better support NumPy 1.x as well). When creating a new integer array in C or Cython, the new NPY_DEFAULT_INT macro will evaluate to either NPY_LONG or NPY_INTP depending on the NumPy version.

Note that the NumPy random API is not affected by this change.

C-API Changes#

Some definitions were removed or replaced due to being outdated or unmaintainable. Some new API definitions will evaluate differently at runtime between NumPy 2.0 and NumPy 1.x. Some are defined in numpy/_core/include/numpy/npy_2_compat.h (for example NPY_DEFAULT_INT) which can be vendored in full or part to have the definitions available when compiling against NumPy 1.x.

If necessary, PyArray_RUNTIME_VERSION >= NPY_2_0_API_VERSION can be used to explicitly implement different behavior on NumPy 1.x and 2.0. (The compat header defines it in a way compatible with such use.)

Please let us know if you require additional workarounds here.

The PyArray_Descr struct has been changed#

One of the most impactful C-API changes is that the PyArray_Descr struct is now more opaque to allow us to add additional flags and have itemsizes not limited by the size of int as well as allow improving structured dtypes in the future and not burden new dtypes with their fields.

Code which only uses the type number and other initial fields is unaffected. Most code will hopefully mainly access the ->elsize field, when the dtype/descriptor itself is attached to an array (e.g. arr->descr->elsize) this is best replaced with PyArray_ITEMSIZE(arr).

Where not possible, new accessor functions are required:

  • PyDataType_ELSIZE and PyDataType_SET_ELSIZE (note that the result is now npy_intp and not int).

  • PyDataType_ALIGNMENT

  • PyDataType_FIELDS, PyDataType_NAMES, PyDataType_SUBARRAY

  • PyDataType_C_METADATA

Cython code should use Cython 3, in which case the change is transparent. (Struct access is available for elsize and alignment when compiling only for NumPy 2.)

For compiling with both 1.x and 2.x if you use these new accessors it is unfortunately necessary to either define them locally via a macro like:

#if NPY_ABI_VERSION < 0x02000000
  #define PyDataType_ELSIZE(descr) ((descr)->elsize)

or adding npy2_compat.h into your code base and explicitly include it when compiling with NumPy 1.x (as they are new API). Including the file has no effect on NumPy 2.

Please do not hesitate to open a NumPy issue, if you require assistance or the provided functions are not sufficient.

Custom User DTypes: Existing user dtypes must now use PyArray_DescrProto to define their dtype and slightly modify the code. See note in PyArray_RegisterDataType.

Functionality moved to headers requiring import_array()#

If you previously included only ndarraytypes.h you may find that some functionality is not available anymore and requires the inclusion of ndarrayobject.h or similar. This include is also needed when vendoring npy_2_compat.h into your own codebase to allow use of the new definitions when compiling with NumPy 1.x.

Functionality which previously did not require import includes:

  • Functions to access dtype flags: PyDataType_FLAGCHK, PyDataType_REFCHK, and the related NPY_BEGIN_THREADS_DESCR.

  • PyArray_GETITEM and PyArray_SETITEM.


It is important that the import_array() mechanism is used to ensure that the full NumPy API is accessible when using the npy_2_compat.h header. In most cases your extension module probably already calls it. However, if not we have added PyArray_ImportNumPyAPI() as a preferable way to ensure the NumPy API is imported. This function is light-weight when called multiple times so that you may insert it wherever it may be needed (if you wish to avoid setting it up at module import).

Increased maximum number of dimensions#

The maximum number of dimensions (and arguments) was increased to 64. This affects the NPY_MAXDIMS and NPY_MAXARGS macros. It may be good to review their use, and we generally encourage you to not use these macros (especially NPY_MAXARGS), so that a future version of NumPy can remove this limitation on the number of dimensions.

NPY_MAXDIMS was also used to signal axis=None in the C-API, including the PyArray_AxisConverter. The latter will return -2147483648 as an axis (the smallest integer value). Other functions may error with AxisError: axis 64 is out of bounds for array of dimension in which case you need to pass NPY_RAVEL_AXIS instead of NPY_MAXDIMS. NPY_RAVEL_AXIS is defined in the npy_2_compat.h header and runtime dependent (mapping to 32 on NumPy 1.x and -2147483648 on NumPy 2.x).

Complex types - Underlying type changes#

The underlying C types for all of the complex types have been changed to use native C99 types. While the memory layout of those types remains identical to the types used in NumPy 1.x, the API is slightly different, since direct field access (like c.real or c.imag) is no longer possible.

It is recommended to use the functions npy_creal and npy_cimag (and the corresponding float and long double variants) to retrieve the real or imaginary part of a complex number, as these will work with both NumPy 1.x and with NumPy 2.x. New functions npy_csetreal and npy_csetimag, along with compatibility macros NPY_CSETREAL and NPY_CSETIMAG (and the corresponding float and long double variants), have been added for setting the real or imaginary part.

The underlying type remains a struct under C++ (all of the above still remains valid).

This has implications for Cython. It is recommened to always use the native typedefs cfloat_t, cdouble_t, clongdouble_t rather than the NumPy types npy_cfloat, etc, unless you have to interface with C code written using the NumPy types. You can still write cython code using the c.real and c.imag attributes (using the native typedefs), but you can no longer use in-place operators c.imag += 1 in Cython’s c++ mode.

Changes to namespaces#

In NumPy 2.0 certain functions, modules, and constants were moved or removed to make the NumPy namespace more user-friendly by removing unnecessary or outdated functionality and clarifying which parts of NumPy are considered private. Please see the tables below for guidance on migration. For most changes this means replacing it with a backwards compatible alternative.

Please refer to NEP 52 for more details.

Main namespace#

About 100 members of the main np namespace have been deprecated, removed, or moved to a new place. It was done to reduce clutter and establish only one way to access a given attribute. The table below shows members that have been removed:

removed member

migration guideline


It’s still available as np.lib.add_docstring.


It’s still available as np.lib.add_newdoc.


It’s an internal function and doesn’t have a replacement.


Use all instead.


Use np.asarray with a float dtype instead.


Now it’s available under np.lib.array_utils.byte_bounds


Use np.asarray(arr, dtype=dtype) instead.


Use np.complex128 instead.


Use np.clongdouble instead.


There’s no replacement, as Python 2 is no longer supported.


Use np.complex128 instead.


Use np.cumprod instead.


It’s still available as np.lib.npyio.DataSource.


Emit DeprecationWarning with warnings.warn directly, or use typing.deprecated.


Emit DeprecationWarning with warnings.warn directly, or use typing.deprecated.


Use your own printing function instead.


Use arr.T.copy() instead.


Use numpy.promote_types or numpy.result_type instead. To achieve semantics for the scalar_types argument, use numpy.result_type and pass the Python values 0, 0.0, or 0j.



Use np.float64 instead.


Use the np.errstate context manager instead.


Use np.inf instead.


Use np.inf instead.


Use np.inf instead.


Use issubclass(rep, np.generic) instead.


Use issubclass builtin instead.


Use np.issubdtype instead.


Use np.asmatrix instead.


Use a specific dtype instead. You should avoid relying on any implicit mechanism and select the largest dtype of a kind explicitly in the code.


Use np.nan instead.


Use np.dtype(<dtype>).itemsize instead.


Use -np.inf instead.


Use -0.0 instead.


Use np.clongdouble instead.


Use np.longdouble instead.


Search NumPy’s documentation directly.


Use np.dtype(obj).type instead.


Use np.inf instead.


Use instead.


Use 0.0 instead.


Use np.genfromtxt with comma delimiter instead.


Use np.genfromtxt instead.


Use np.round instead.


Use ast.literal_eval instead.


Use np.dtype(obj).char instead.


Access dtypes explicitly instead.


Use the np.errstate context manager instead.


For the general case, use PyUFunc_ReplaceLoopBySignature. For ndarray subclasses, define the __array_ufunc__ method and override the relevant ufunc.


Use np.set_printoptions instead with a formatter for custom printing of NumPy objects.


Use np.complex64 instead.


Use np.bytes_ instead.


Use any instead.


Use inspect.getsource instead.


It’s now available from np.lib.


Use np.str_ instead.


Use an IDE variable explorer or locals() instead.

If the table doesn’t contain an item that you were using but was removed in 2.0, then it means it was a private member. You should either use the existing API or, in case it’s infeasible, reach out to us with a request to restore the removed entry.

The next table presents deprecated members, which will be removed in a release after 2.0:

deprecated member

migration guideline


Use np.isin instead.


Use np.vstack instead (row_stack was an alias for vstack).


Use np.trapezoid or a scipy.integrate function instead.

Finally, a set of internal enums has been removed. As they weren’t used in downstream libraries we don’t provide any information on how to replace them:


numpy.lib namespace#

Most of the functions available within np.lib are also present in the main namespace, which is their primary location. To make it unambiguous how to access each public function, np.lib is now empty and contains only a handful of specialized submodules, classes and functions:

  • array_utils, format, introspect, mixins, npyio and stride_tricks submodules,

  • Arrayterator and NumpyVersion classes,

  • add_docstring and add_newdoc functions,

  • tracemalloc_domain constant.

If you get an AttributeError when accessing an attribute from np.lib you should try accessing it from the main np namespace then. If an item is also missing from the main namespace, then you’re using a private member. You should either use the existing API or, in case it’s infeasible, reach out to us with a request to restore the removed entry.

numpy.core namespace#

The np.core namespace is now officially private and has been renamed to np._core. The user should never fetch members from the _core directly - instead the main namespace should be used to access the attribute in question. The layout of the _core module might change in the future without notice, contrary to public modules which adhere to the deprecation period policy. If an item is also missing from the main namespace, then you should either use the existing API or, in case it’s infeasible, reach out to us with a request to restore the removed entry.

ndarray and scalar methods#

A few methods from np.ndarray and np.generic scalar classes have been removed. The table below provides replacements for the removed members:

expired member

migration guideline


Use arr.view(arr.dtype.newbyteorder(order)) instead.


Use np.ptp(arr, ...) instead.


Use arr[index] = value instead.

numpy.strings namespace#

A new numpy.strings namespace has been created, where most of the string operations are implemented as ufuncs. The old numpy.char namespace still is available, and, wherever possible, uses the new ufuncs for greater performance. We recommend using the strings functions going forward. The char namespace may be deprecated in the future.

Other changes#

Note about pickled files#

NumPy 2.0 is designed to load pickle files created with NumPy 1.26, and vice versa. For versions 1.25 and earlier loading NumPy 2.0 pickle file will throw an exception.

Adapting to changes in the copy keyword#

The copy keyword behavior changes in asarray, array and ndarray.__array__ may require these changes:

  • Code using np.array(..., copy=False) can in most cases be changed to np.asarray(...). Older code tended to use np.array like this because it had less overhead than the default np.asarray copy-if-needed behavior. This is no longer true, and np.asarray is the preferred function.

  • For code that explicitly needs to pass None/False meaning “copy if needed” in a way that’s compatible with NumPy 1.x and 2.x, see scipy#20172 for an example of how to do so.

  • For any __array__ method on a non-NumPy array-like object, dtype=None and copy=None keywords must be added to the signature - this will work with older NumPy versions as well (although older numpy versions will never pass in copy keyword). If the keywords are added to the __array__ signature, then for:

    • copy=True and any dtype value always return a new copy,

    • copy=None create a copy if required (for example by dtype),

    • copy=False a copy must never be made. If a copy is needed to return a numpy array or satisfy dtype, then raise an exception (ValueError).

Writing numpy-version-dependent code#

It should be fairly rare to have to write code that explicitly branches on the numpy version - in most cases, code can be rewritten to be compatible with 1.x and 2.0 at the same time. However, if it is necessary, here is a suggested code pattern to use, using numpy.lib.NumpyVersion:

# example with AxisError, which is no longer available in
# the main namespace in 2.0, and not available in the
# `exceptions` namespace in <1.25.0 (example uses <2.0.0b1
# for illustrative purposes):
if np.lib.NumpyVersion(np.__version__) >= '2.0.0b1':
    from numpy.exceptions import AxisError
    from numpy import AxisError

This pattern will work correctly including with NumPy release candidates, which is important during the 2.0.0 release period.