Scheduled weekly dependency update for week 40

[pyup-bot]

Update anyio from 4.6.2.post1 to 4.11.0.

Changelog

4.11.0

- Added support for cancellation reasons (the `reason` parameter to `CancelScope.cancel()`) ([\975](https://github.com/agronholm/anyio/pull/975))
- Bumped the minimum version of Trio to v0.31.0
- Added the ability to enter the event loop from foreign (non-worker) threads by passing the return value of `anyio.lowlevel.current_token()` to `anyio.from_thread.run()` and `anyio.from_thread.run_sync()` as the `token` keyword argument ([\256](https://github.com/agronholm/anyio/issues/256))
- Added pytest option (`anyio_mode = "auto"`) to make the pytest plugin automatically handle all async tests ([\971](https://github.com/agronholm/anyio/pull/971))
- Added the `anyio.Condition.wait_for()` method for feature parity with asyncio ([\974](https://github.com/agronholm/anyio/pull/974))
- Changed the default type argument of `anyio.abc.TaskStatus` from `Any` to `None` ([\964](https://github.com/agronholm/anyio/pull/964))
- Fixed TCP listener behavior to guarantee the same ephemeral port is used for all socket listeners when `local_port=0` ([\857](https://github.com/agronholm/anyio/issues/857); PR by 11kkw and agronholm)
- Fixed inconsistency between Trio and asyncio where a TCP stream that previously raised a `BrokenResourceError` on `send()` would still raise `BrokenResourceError` after the stream was closed on asyncio, but `ClosedResourceError` on Trio. They now both raise a `ClosedResourceError` in this scenario. ([\671](https://github.com/agronholm/anyio/issues/671))

4.10.0

-   Added the `feed_data()` method to the `BufferedByteReceiveStream` class, allowing users to inject data directly into the buffer
-   Added various class methods to wrap existing sockets as listeners or socket streams:
 -   `SocketListener.from_socket()`
 -   `SocketStream.from_socket()`
 -   `UNIXSocketStream.from_socket()`
 -   `UDPSocket.from_socket()`
 -   `ConnectedUDPSocket.from_socket()`
 -   `UNIXDatagramSocket.from_socket()`
 -   `ConnectedUNIXDatagramSocket.from_socket()`
-   Added a hierarchy of connectable stream classes for transparently connecting to various remote or local endpoints for exchanging bytes or objects
-   Added context manager mix-in classes (`anyio.ContextManagerMixin` and `anyio.AsyncContextManagerMixin`) to help write classes that embed other context managers, particularly cancel scopes or task groups ([\905](https://github.com/agronholm/anyio/pull/905); PR by agronholm and tapetersen)
-   Added the ability to specify the thread name in `start_blocking_portal()` ([\818](https://github.com/agronholm/anyio/issues/818); PR by davidbrochart)
-   Added `anyio.notify_closing` to allow waking `anyio.wait_readable` and `anyio.wait_writable` before closing a socket. Among other things, this prevents an OSError on the `ProactorEventLoop`. ([\896](https://github.com/agronholm/anyio/pull/896); PR by graingert)
-   Incorporated several documentation improvements from the EuroPython 2025 sprint (special thanks to the sprinters: Emmanuel Okedele, Jan Murre, Euxenia Miruna Goia and Christoffer Fjord)
-   Added a documentation page explaining why one might want to use AnyIO's APIs instead of asyncio's
-   Updated the `to_interpreters` module to use the public `concurrent.interpreters` API on Python 3.14 or later
-   Fixed `anyio.Path.copy()` and `anyio.Path.copy_into()` failing on Python 3.14.0a7
-   Fixed return annotation of `__aexit__` on async context managers. CMs which can suppress exceptions should return `bool`, or `None` otherwise. ([\913](https://github.com/agronholm/anyio/pull/913); PR by Enegg)
-   Fixed rollover boundary check in `SpooledTemporaryFile` so that rollover only occurs when the buffer size exceeds `max_size` ([\915](https://github.com/agronholm/anyio/pull/915); PR by 11kkw)
-   Migrated testing and documentation dependencies from extras to dependency groups
-   Fixed compatibility of `anyio.to_interpreter` with Python 3.14.0b2 ([\926](https://github.com/agronholm/anyio/issues/926); PR by hroncok)
-   Fixed `SyntaxWarning` on Python 3.14 about `return` in `finally` ([\816](https://github.com/agronholm/anyio/issues/816))
-   Fixed RunVar name conflicts. RunVar instances with the same name should not share storage ([\880](https://github.com/agronholm/anyio/issues/880); PR by vimfu)
-   Renamed the `BrokenWorkerIntepreter` exception to `BrokenWorkerInterpreter`. The old name is available as a deprecated alias. ([\938](https://github.com/agronholm/anyio/pull/938); PR by ayussh-verma)
-   Fixed an edge case in `CapacityLimiter` on asyncio where a task, waiting to acquire a limiter gets cancelled and is subsequently granted a token from the limiter, but before the cancellation is delivered, and then fails to notify the next waiting task ([\947](https://github.com/agronholm/anyio/issues/947))

4.9.0

- Added async support for temporary file handling ([\344](https://github.com/agronholm/anyio/issues/344); PR by 11kkw)
-   Added 4 new fixtures for the AnyIO `pytest` plugin:
 -   `free_tcp_port_factory`: session scoped fixture returning a callable that generates unused TCP port numbers
 -   `free_udp_port_factory`: session scoped fixture returning a callable that generates unused UDP port numbers
 -   `free_tcp_port`: function scoped fixture that invokes the `free_tcp_port_factory` fixture to generate a free TCP port number
 -   `free_udp_port`: function scoped fixture that invokes the `free_udp_port_factory` fixture to generate a free UDP port number
-   Added `stdin` argument to `anyio.run_process()` akin to what `anyio.open_process()`, `asyncio.create_subprocess()`, `trio.run_process()`, and `subprocess.run()` already accept (PR by jmehnle)
-   Added the `info` property to `anyio.Path` on Python 3.14
-   Changed `anyio.getaddrinfo()` to ignore (invalid) IPv6 name resolution results when IPv6 support is disabled in Python
-   Changed `EndOfStream` raised from `MemoryObjectReceiveStream.receive()` to leave out the `AttributeError` from the exception chain which was merely an implementation detail and caused some confusion
-   Fixed traceback formatting growing quadratically with level of `TaskGroup` nesting on asyncio due to exception chaining when raising `ExceptionGroups` in `TaskGroup.__aexit__` ([\863](https://github.com/agronholm/anyio/issues/863); PR by tapetersen)
-   Fixed `anyio.Path.iterdir()` making a blocking call in Python 3.13 ([\873](https://github.com/agronholm/anyio/issues/873); PR by cbornet and agronholm)
-   Fixed `connect_tcp()` producing cyclic references in tracebacks when raising exceptions ([\809](https://github.com/agronholm/anyio/pull/809); PR by graingert)
-   Fixed `anyio.to_thread.run_sync()` needlessly holding on to references of the context, function, arguments and others until the next work item on asyncio (PR by Wankupi)

4.8.0

-   Added **experimental** support for running functions in subinterpreters on Python 3.13 and later
-   Added support for the `copy()`, `copy_into()`, `move()` and `move_into()` methods in `anyio.Path`, available in Python 3.14
-   Changed `TaskGroup` on asyncio to always spawn tasks non-eagerly, even if using a task factory created via `asyncio.create_eager_task_factory()`, to preserve expected Trio-like task scheduling semantics (PR by agronholm and graingert)
-   Configure `SO_RCVBUF`, `SO_SNDBUF` and `TCP_NODELAY` on the selector thread waker socket pair (this should improve the performance of `wait_readable()` and `wait_writable()` when using the `ProactorEventLoop`) ([\836](https://github.com/agronholm/anyio/pull/836); PR by graingert)
-   Fixed `AssertionError` when using `nest-asyncio` ([\840](https://github.com/agronholm/anyio/issues/840))
-   Fixed return type annotation of various context managers' `__exit__` method ([\847](https://github.com/agronholm/anyio/issues/847); PR by Enegg)

4.7.0

-   Updated `TaskGroup` to work with asyncio's eager task factories ([\764](https://github.com/agronholm/anyio/issues/764))
-   Added the `wait_readable()` and `wait_writable()` functions which will accept an object with a `.fileno()` method or an integer handle, and deprecated their now obsolete versions (`wait_socket_readable()` and `wait_socket_writable()`) (PR by davidbrochart)
-   Changed `EventAdapter` (an `Event` with no bound async backend) to allow `set()` to work even before an async backend is bound to it ([\819](https://github.com/agronholm/anyio/issues/819))
-   Added support for `wait_readable()` and `wait_writable()` on `ProactorEventLoop` (used on asyncio + Windows by default)
-   Fixed a misleading `ValueError` in the context of DNS failures ([\815](https://github.com/agronholm/anyio/issues/815); PR by graingert)
-   Fixed the return type annotations of `readinto()` and `readinto1()` methods in the `anyio.AsyncFile` class ([\825](https://github.com/agronholm/anyio/issues/825))
-   Fixed `TaskInfo.has_pending_cancellation()` on asyncio returning false positives in cleanup code on Python \>= 3.11 ([\832](https://github.com/agronholm/anyio/issues/832); PR by gschaffner)
-   Fixed cancelled cancel scopes on asyncio calling `asyncio.Task.uncancel` when propagating a `CancelledError` on exit to a cancelled parent scope ([\790](https://github.com/agronholm/anyio/pull/790); PR by gschaffner)

Links

• PyPI: https://pypi.org/project/anyio
• Changelog: https://data.safetycli.com/changelogs/anyio/

Update argon2-cffi from 23.1.0 to 25.1.0.

The bot wasn't able to find a changelog for this release. Got an idea?

Links

• PyPI: https://pypi.org/project/argon2-cffi

Update asgiref from 3.8.1 to 3.10.0.

Changelog

3.10.0

-------------------

* Added AsyncSingleThreadContext context manager to ensure multiple AsyncToSync
invocations use the same thread. (511)

3.9.2

------------------

* Adds support for Python 3.14.

* Fixes wsgi.errors file descriptor in WsgiToAsgi adapter.

3.9.1

------------------

* Fixed deletion of Local values affecting other contexts. (523)

* Skip CPython specific garbage collection test on pypy. (521)

3.9.0

------------------

* Adds support for Python 3.13.

* Drops support for (end-of-life) Python 3.8.

* Fixes an error with conflicting kwargs between AsyncToSync and the wrapped
function. (471)

* Fixes Local isolation between asyncio Tasks. (478)

* Fixes a reference cycle in Local (508)

* Fixes a deadlock in CurrentThreadExecutor with nested async_to_sync →
sync_to_async → async_to_sync → create_task calls. (494)

* The ApplicationCommunicator testing utility will now return the task result
if it's already completed on send_input and receive_nothing. You may need to
catch (e.g.) the asyncio.exceptions.CancelledError if sending messages to
already finished consumers in your tests. (505)

Links

• PyPI: https://pypi.org/project/asgiref
• Changelog: https://data.safetycli.com/changelogs/asgiref/
• Repo: https://github.com/django/asgiref/

Update astroid from 3.3.5 to 4.0.0.

The bot wasn't able to find a changelog for this release. Got an idea?

Links

• PyPI: https://pypi.org/project/astroid

Update astropy from 6.1.6 to 7.1.0.

Changelog

7.1.0

==========================


New Features
------------

astropy.coordinates
^^^^^^^^^^^^^^^^^^^

- ``search_around_sky`` and ``search_around_3D`` now accept separations/distlimits
broadcastable to the same shape as ``coords1``. [17824]

astropy.io.ascii
^^^^^^^^^^^^^^^^

- Add functionality to read and write to a Table from the TDAT format as part of
the Unified File Read/Write Interface. [16780]

- ``io.ascii`` now supports on-the-fly decompression of LZW-compressed files
(typically ".Z" extension) via the optional package uncompresspy. [17960]

astropy.io.fits
^^^^^^^^^^^^^^^

- Astropy can now not only read but also write headers that have ``HIERARCH``
keys with long values, by allowing the use of ``CONTINUE`` cards for those
(as was already the case for regular FITS keys). [17748]

- Add ``strip_spaces`` option to ``Table.read`` to strip trailing whitespaces in
string columns.  This will be activated by default in the next major release. [17777]

- ``io.fits`` now supports on-the-fly decompression of LZW-compressed files
(typically ".Z" extension) via the optional package uncompresspy. [17960]

- ``io.fits`` now supports on-the-fly decompression of LZMA-compressed files
(typically ".xz" extension) if the lzma module is provided by the Python
installation. [17968]

astropy.io.misc
^^^^^^^^^^^^^^^

- Add a fast ``Table`` CSV reader that uses the PyArrow ``read_csv()`` function. This can
be significantly faster and more memory-efficient than the ``astropy.io.ascii`` fast
reader. This new reader can be used with ``Table.read()`` by setting
``format="pyarrow.csv"``. [17706]

astropy.io.votable
^^^^^^^^^^^^^^^^^^

- New module ``astropy.io.votable.dataorigin`` to extract Data Origin information from INFO in VOTable. [17839]

- ``CooSys`` VOTable elements now have a method ``to_astropy_frame`` that returns the
corresponding astropy built-in frame, when possible. [17999]

astropy.modeling
^^^^^^^^^^^^^^^^

- Added a ``fit_info=`` keyword argument to ``parallel_fit_dask`` to allow users to preserve fit information from each individual fit. [17538]

astropy.nddata
^^^^^^^^^^^^^^

- Adds a utility class, ``astropy.nddata.Covariance``, used to construct, access,
and store covariance matrices.  The class depends on use of the ``scipy.sparse``
module. [16690]

- Add the ``limit_rounding_method`` parameter to `~astropy.nddata.Cutout2D`,
`~astropy.nddata.overlap_slices`, `~astropy.nddata.extract_array`, and
`~astropy.nddata.add_array` to allow users to specify the rounding method
used when calculating the pixel limits of the cutout. The default method
is to use `~numpy.ceil`. [17876]

astropy.table
^^^^^^^^^^^^^

- Document that ``Table.group_by``'s underlying sorting algorithm is guaranteed
to be stable. This reflects behavior that was already present but undocumented,
at least since astropy 6.0 . [17676]

astropy.timeseries
^^^^^^^^^^^^^^^^^^

- Downsampling now works correctly also on ``MaskedColumn`` and
``MaskedQuantity`` with possibly masked elements.  Furthermore, the type of
(Masked) column will now be properly preserved in downsampling. [18023]

astropy.units
^^^^^^^^^^^^^

- Units with the "micro" prefix can now be imported using ``"μ"`` in the name.
For example, the microgram can now be imported with
``from astropy.units import μg``. [17651]

- It is now possible to import angström, litre and ohm from ``astropy.units``
using the ``Å``, ``ℓ`` and ``Ω`` symbols. [17829]

- Unit conversions between kelvins and degrees Rankine no longer require the
``temperature`` equivalency. [17985]

astropy.utils
^^^^^^^^^^^^^

- Make commonly used Masked subclasses importable for ASDF support.

Registered types associated with ASDF converters must be importable by
their fully qualified name. Masked classes are dynamically created and have
apparent names like ``astropy.utils.masked.core.MaskedQuantity`` although
they aren't actually attributes of this module. Customize module attribute
lookup so that certain commonly used Masked classes are importable.

See:

- https://asdf.readthedocs.io/en/latest/asdf/extending/converters.html#entry-point-performance-considerations
- https://github.com/astropy/asdf-astropy/pull/253 [#17685]

- ``astropy.utils.data.download_file`` can now recover from a ``TimeoutError``
when given a list of alternative source URLs. Previously, only ``URLError``
exceptions were recoverable. An exception is still being raised after trying all
URLs provided if none of them could be reached. [17691]

- ``utils.data`` now supports on-the-fly decompression of LZW-compressed files
(typically ".Z" extension) via the optional package uncompresspy. [17960]


API Changes
-----------

astropy.coordinates
^^^^^^^^^^^^^^^^^^^

- On representations the method ``get_name`` has been deprecated in favor of the class-level
attribute ``name``. The method will be removed in a future release. [17503]

astropy.cosmology
^^^^^^^^^^^^^^^^^

- A new public module, ``astropy.cosmology.io``, has been added to provide support
for reading, writing, and converting cosmology instances.

The private modules ``astropy.cosmology.funcs``,
``astropy.cosmology.parameter``, ``astropy.cosmology.connect``,
``astropy.cosmology.core``, and ``astropy.cosmology.flrw`` have been deprecated.
Their functionality remains accessible in the `astropy.cosmology` module or in
the new ``astropy.cosmology.io`` module. [17543]

- Comoving distances now accept an optional 2nd argument, where the two-argument form is
the comoving distance between two redshifts. The one-argument form is the comoving
distance from redshift 0 to the input redshift. [17701]

- A new public module, ``astropy.cosmology.traits``, has been added to provide building
blocks for custom cosmologies. The currently available traits are:
- ``astropy.cosmology.traits.ScaleFactor``
- ``astropy.cosmology.traits.TemperatureCMB`` [17702]

astropy.extern
^^^^^^^^^^^^^^

- Astropy used to bundle the javascript libraries jQuery and DataTables for
interactive (e.g. sorting by column values) tables using the ``show_in_browser()``
method.
This bundling requires relatively large files in astropy itself, for a relatively minor feature.
Furthermore, the astropy developers are not experts in javascript development, and
javascript libraries many need updates to improve on security vulnerabilities.
This change removes the bundled versions of jQuery and DataTables from astropy,
updates the default version of the remote URLs to version 2.1.8 of DataTables, and
sets the default for ``show_in_browser(use_local_files=False)`` to use the remote versions
in all cases. If the method is called with ``use_local_files=True``, a warning is
displayed and remote version are used anyway.
This may break the use of the method when working offline, unless the javascript
files are cached by the browser from a previous online session. [17521]

astropy.table
^^^^^^^^^^^^^

- ``showtable`` CLI is now deprecated to avoid a name clash on Debian; use ``showtable-astropy`` instead. [18047]

- Fix issues in the handling of a call like ``tbl.loc[item]`` or ``tbl.loc_indices[item]``
and make the behavior consistent with pandas. Here ``tbl`` is a ``Table`` or ``QTable``
with an index defined.

If ``item`` is an empty list or zero-length ``np.ndarray`` or an empty slice, then
previously ``tbl.loc[item]`` would raise a ``KeyError`` exception. Now it returns the
zero-length table ``tbl[[]]``.

If ``item`` is a one-element list like ``["foo"]``, then previously
``tbl.loc[item]`` would return either a ``Row`` or a ``Table`` with multiple row,
depending on whether the index was unique. Now it always returns a ``Table``, consistent
with behavior for ``tbl.loc[[]]`` and ``tbl.loc[["foo", "bar"]]``.

See https://github.com/astropy/astropy/pull/18051 for more details. [#18051]

astropy.units
^^^^^^^^^^^^^

- Passing ``fraction='multiline'`` to ``unit.to_string()`` will no longer raise
an exception if the given format does not support multiline fractions, but
rather give a warning and use an inline fraction. [17374]

- Automatic conversion of a ``str`` or ``bytes`` instance to a unit when it is
multiplied or divided with an existing unit or quantity is deprecated. [17586]

- Accessing the contents of the ``units.deprecated`` module now emits deprecation
warnings.
The module may be removed in a future version. [17929]

astropy.visualization
^^^^^^^^^^^^^^^^^^^^^

- All arguments from ``simple_norm`` are marked as future keyword-only, with the
exception of the first two (``data`` and ``stretch``).
A warning is now displayed if any other arguments are passed positionally. [17489]


Bug Fixes
---------

astropy.io.fits
^^^^^^^^^^^^^^^

- Fix possible int overflow in the tile compression C code. [17995]

astropy.io.votable
^^^^^^^^^^^^^^^^^^

- In ``CooSys`` elements, the system was not checked for votable version 1.5 [17999]

astropy.samp
^^^^^^^^^^^^

- Fix setting logging level from the ``samp_hub`` command line.
Previously, ``samp_hub --log-level OFF`` was documented as supported but actually caused an exception to be raised.
The patch infers valid choices from the standard library's ``logging`` module.
A ``CRITICAL`` level will closely emulate the intended ``OFF`` setting. [17673]

astropy.table
^^^^^^^^^^^^^

- Initializing a Table with ``rows`` or ``data`` set to ``[]`` or a numpy array with
zero size (e.g., ``np.array([[], []])``) is now equivalent to
``Table(data=None, ...)`` and creates a table with no data values. This allows
defining the table names and/or dtype when creating the table, for instance:
``Table(rows=[], names=["a", "b"], dtype=[int, float])``. Previously this
raised an exception. [17717]

- Fix issues in the handling of a call like ``tbl.loc[item]`` or ``tbl.loc_indices[item]``
and make the behavior consistent with pandas. Here ``tbl`` is a ``Table`` or ``QTable``
with an index defined.

If ``item`` is an empty list or zero-length ``np.ndarray`` or an empty slice, then
previously ``tbl.loc[item]`` would raise a ``KeyError`` exception. Now it returns the
zero-length table ``tbl[[]]``.

If ``item`` is a one-element list like ``["foo"]``, then previously
``tbl.loc[item]`` would return either a ``Row`` or a ``Table`` with multiple row,
depending on whether the index was unique. Now it always returns a ``Table``, consistent
with behavior for ``tbl.loc[[]]`` and ``tbl.loc[["foo", "bar"]]``.

See https://github.com/astropy/astropy/pull/18051 for more details. [#18051]

astropy.timeseries
^^^^^^^^^^^^^^^^^^

- Made ``TimeSeries.from_pandas`` and ``BinnedTimeSeries.read`` more robust to
subclassing. [17351]

astropy.units
^^^^^^^^^^^^^

- For the Angstrom unit in the CDS module, ``u.cds.Angstrom``, the string
representation is now "Angstrom" (instead of "AA"), consistent with what was
always the case for ``u.Angstrom``, and conformant with the CDS standard. [17536]

- Previously the string representation of the ``solMass`` unit in the ``"cds"``
format depended on whether the unit was imported directly from ``units`` or
from ``units.cds``.
Although both representations were valid according to the CDS standard, the
inconsistency was nonetheless needlessly surprising.
The representation of ``units.cds.solMass`` has been changed to match the
representation of ``units.solMass``. [17560]

- The degrees Rankine is now represented as "$\mathrm{{}^{\circ}R}$" in the
``"latex"`` and ``"latex_inline"`` formats and as "°R" in the ``"unicode"``
format. [18049]

astropy.utils
^^^^^^^^^^^^^

- Properly detect invalid LZMA files in ``utils.data``. [17984]

astropy.visualization
^^^^^^^^^^^^^^^^^^^^^

- Fixed an issue when using ``plot_coord`` after slicing the ``WCS`` object coordinates. [18005]


Performance Improvements
------------------------

astropy.timeseries
^^^^^^^^^^^^^^^^^^

- Improved the ``aggregate_downsample`` performance using a new default ``aggregate_func``. [17574]

astropy.units
^^^^^^^^^^^^^

- Converting strings to units with ``Unit()`` is now up to 225% faster. [17399]

- ``UnitBase.compose()`` is now 20% faster. [17425]


Other Changes and Additions
---------------------------

- After ``import astropy``, ``dir(astropy)`` will now list all subpackages,
including those that have not yet been loaded. This also means tab
completion will work as expected (e.g., ``from astropy.coo<TAB>`` will
expand to ``from astropy.coordinates``). [17598]

- Updated bundled WCSLIB version to 8.4, fixing issues in ``wcs_chksum``
and ``wcs_fletcher32``. For a full list of changes - see
``astropy/cextern/wcslib/CHANGES``. [17886]

7.0.2

==========================

Bug Fixes
---------

astropy.config
^^^^^^^^^^^^^^

- Fix a bug where config file generation did not parse nested subclasses of ``astropy.config.ConfigNamespace``. [18107]

astropy.io.fits
^^^^^^^^^^^^^^^

- Fix a bug in ``nddata.Cutout2D`` when creating partial cutouts of ``Section`` objects by adding a ``dtype`` property to the ``Section`` class. [17611]

- Fixed a bug so that now the scaling state from the source HDU to the new appended HDU is copied on the
destination file, when the HDU is read with ``do_not_scale_image_data=True``. [17642]

- Fix setting a slice on table rows (``FITS_record``). [17737]

- Fix checksum computation for tables with VLA columns, when table is loaded in
memory. [17806]

- Fix ``.fileinfo()`` for compressed HDUs. [17815]

- Fix FITS_rec repr when a column has scaling factors, leading to a crash with
numpy>=2.0. [17933]

- Fixed a bug that caused THEAP, ZBLANK, ZSCALE, and ZZERO to not be correctly
removed during decompression of tile-compressed FITS files. [18072]

astropy.io.votable
^^^^^^^^^^^^^^^^^^

- ``astropy`` v7.0.0 erroneously refused to write a VOTable if it contained units that
could not be represented in the CDS format.
Now ``astropy`` correctly chooses the unit format based on the VOTable version.
The bug in question did not cause any corruption in tables that were successfully
written because the newer VOUnit format is backwards compatible with the CDS format.
Furthermore, any unit that is in neither formats would still be written out
but would issue a warning. [17570]

- ``unicodeChar`` fields can now be of bounded variable size (``arraysize="10*``). [18075]

astropy.modeling
^^^^^^^^^^^^^^^^

- Fixed an issue where the ``filter_non_finite`` option was not working
for 2D models. An error is raised when the ``filter_non_finite`` option
is set to ``True`` and all values are non-finite. [17869]

astropy.stats
^^^^^^^^^^^^^

- Now ``bayesian_blocks(t, x, fitness="events")`` correctly handles the case
when the input data ``x`` contains zeros. [17800]

astropy.table
^^^^^^^^^^^^^

- Prevent corrupting a column by mutating its name to an invalid type.
A ``TypeError`` is now raised when a name is set to anything other than a
string. [17450]

- Fix a bug in creating a ``Table`` from a list of rows that dropped the units
of non-scalar Quantity, e.g., ``Table(rows=[([1] * u.m,), ([2] * u.m,)])``. [17936]

astropy.units
^^^^^^^^^^^^^

- Ensured that the units of ``yp``, ``refa`` and ``refb`` are properly
taken into account when calling ``erfa.apio`` (previously, the
conversion required for ``xp`` was applied to those inputs too). [17742]

- The machinery that injects units into a namespace (used e.g. by ``def_unit()``)
now applies NFKC normalization to unit names when checking for name collisions.
This prevents name collisions if the namespace belongs to a module and the unit
is accessed as an attribute of that module. [17853]

- The string representations of the prefixed versions of ``solLum``, ``solMass``
and ``solRad`` units can now be parsed by default.
Previously they could only be parsed if the ``required_by_vounit`` module had
been imported, possibly indirectly by using the ``"vounit"`` format. [17868]

astropy.utils
^^^^^^^^^^^^^

- Prevent corrupting a mixin column's ``info`` attribute by mutating its name to
an invalid type. A ``TypeError`` is now raised when a name is set to anything
other than a string. [17450]

astropy.visualization
^^^^^^^^^^^^^^^^^^^^^

- Ensure that the ``astropy.visualization.wcsaxes.custom_ucd_coord_meta_mapping``
context manager performs a (correct) cleanup. [17749]

- Fixed interval classes for masked input (``MaskedArray`` and ``MaskedNDArray``). [17927]

- Fixed the limits of ``a`` parameter in the ``PowerDistStretch``
and ``InvertedPowerDistStretch`` classes so that a value of
0 in no longer allowed. That value gives infinity values in
``InvertedPowerDistStretch`` and it makes the ``PowerDistStretch``
results independent of the input data. [17941]

- Fixed an issue where LinearStretch values were not being clipped to
[0:1] when ``clip=True``. [17943]

astropy.wcs
^^^^^^^^^^^

- Fix UCD for air wavelengths, following the IVOA recommendation that ``'em.wl'``
be reserved for vacuum wavelengths. ``'em.wl;obs.atmos'`` is now used to
represent air wavelengths instead. [17769]


Other Changes and Additions
---------------------------

- Updated the bundled CFITSIO library to 4.6.0. [17904]

7.0.1

==========================

API Changes
-----------

astropy.table
^^^^^^^^^^^^^

- The use of the keyword ``use_local_files`` for the js viewer in
``astropy.table.Table.show_in_browser`` is now deprecated. Starting in Astropy
7.1 this keyword will be ignored and use of it will issue a warning. The
default behavior will be to use the remote versions of jQuery and DataTables
from a CDN. [17480]

Bug Fixes
---------

astropy.config
^^^^^^^^^^^^^^

- With ``astropy`` v7.0.0 the cache directory cannot be customized with the
``XDG_CACHE_HOME`` environment variable.
Instead, ``XDG_CONFIG_HOME`` erroneously controls both configuration and cache
directories.
The correct pre-v7.0.0 behaviour has been restored, but it is possible that
``astropy`` v7.0.0 has written cache files to surprising locations.
Concerned users can use the ``get_cache_dir_path()`` function to check where
the cache files are written.

The bug in question does not affect systems where the ``XDG_CACHE_HOME`` and
``XDG_CONFIG_HOME`` environment variables are unset. [17514]

astropy.coordinates
^^^^^^^^^^^^^^^^^^^

- Fixed a numerical-precision bug with the calculation of the ``theta``
component when converting from ``CylindricalRepresentation`` to
``PhysicsSphericalRepresentation`` for vectors very close to the Z axis (within
milliarcseconds). [17693]

astropy.io.ascii
^^^^^^^^^^^^^^^^

- Fixed parsing ASCII table with data that starts with a tilda. [17565]

- Find and read ASCII tables even if there is white space before
``\begin{tabular}``, ``\tablehead``, and similar markers. [17624]

astropy.io.fits
^^^^^^^^^^^^^^^

- Fix memory leak in BinTableHDU.copy() [16143]

- Fix overflow error with Numpy 2 and VLA columns using P format. [17328]

- Fix ``ImageHDU.scale`` with float. [17458]

- Fixed ``Table.write(..., format="fits", overwrite=True)`` when filename is
provided as ``pathlib.Path``. [17552]

astropy.io.votable
^^^^^^^^^^^^^^^^^^

- Updated XML writer for ``VOTableFile`` element to include or drop
``coordinate_systems`` regardless of version. [17356]

astropy.modeling
^^^^^^^^^^^^^^^^

- Fix fitting of compound models when inputs has more than one dimension, and
specifically fix fitting of compound models with Polynomial2D components
[17618]

astropy.table
^^^^^^^^^^^^^

- Ensure that representations and differentials, like SkyCoord, can be used in
table join operations, by making use of the fact that they can now be masked. [17381]

- Fix a crash in ``Table.show_in_browser`` due to an internal type inconsistency. [17513]

- Fix incorrect description of the ``unique`` parameter in ``Table.add_index``'s
docstring. Add missing Raises section. [17677]

astropy.units
^^^^^^^^^^^^^

- Ensure that ``Unit.to`` allows as ``value`` argument all array types that
follow the array API standard and define ``__array_namespace__``. Furthermore,
for backwards compatibility, generally pass through arguments that define a
``.dtype``, independent of whether that is a numpy data type. [17469]

- The zebi (Zi, 2^70) and yobi (Yi, 2^80) binary prefixes are now supported. [17692]

astropy.visualization
^^^^^^^^^^^^^^^^^^^^^

- Fix ``CoordinateHelper.ticklabels``. The getter was incorrectly returning
the helper's ticks rather than the labels. [17444]

- The following private classes from ``astropy.visualization.lupton_rgb``, that
were dropped without deprecation in astropy 7.0.0, were re-introduced following
a report that they were used downstream. The following classes are now
considered public:

- ``Mapping``
- ``AsinhMapping``
- ``LinearMapping``
- ``AsinhZScaleMapping`` [17531]

Other Changes and Additions
---------------------------

- Update bundled js library datatables to version 2.1.8, which is current at the time of this PR. [17480]

7.0.0

==========================


New Features
------------

astropy.config
^^^^^^^^^^^^^^

- Added ``get_config_dir_path`` (and ``get_cache_dir_path``) which is equivalent
to ``get_config_dir`` (respectively ``get_cache_dir``) except that it returns a
``pathlib.Path`` object instead of ``str``. [17118]

astropy.coordinates
^^^^^^^^^^^^^^^^^^^

- ``BaseCoordinateFrame`` instances such as ``ICRS``, ``SkyOffsetFrame``, etc.,
can now be stored directly in tables (previously, they were stored as
``object`` type columns). Furthermore, storage in tables is now also possible
for frames that have no data (but which have attributes with the correct shape
to fit in the table). [16831]

- ``BaseCoordinateFrame`` now has a ``to_table()`` method, which converts the
frame to a ``QTable``, analogously to the ``SkyCoord.to_table()`` method. [17009]

- ``SkyCoord``, coordinate frames, and representations have all have gained the
ability to deal with ``Masked`` data. In general, the procedure is similar to
that of ``Time``, except that different representation components do not share
the mask, to enable holding, e.g., a catalogue of objects in which only some
have associated distances. [17016]

astropy.io.ascii
^^^^^^^^^^^^^^^^

- Add support for ``pathlib.Path`` objects in
``astropy.io.ascii.core.BaseInputter.get_lines``. [16930]

astropy.io.fits
^^^^^^^^^^^^^^^

- Expanded ``FITSDiff`` output for ``PrimaryHDU`` and ``ImageHDU`` to include the
maximum relative and absolute differences in the data. [17097]

astropy.io.misc
^^^^^^^^^^^^^^^

- The HDF5 writer, ``write_table_hdf5()``, now accepts ``os.PathLike`` objects
as ``output``. [16955]

astropy.io.votable
^^^^^^^^^^^^^^^^^^

- Support reading and writing of VOTable version 1.5, including the new
``refposition`` attribute of ``COOSYS``. [16856]

astropy.modeling
^^^^^^^^^^^^^^^^

- Added ``Model.has_tied``, ``Model.has_fixed``, and ``Model.has_bounds`` attributes to make
it easy to check whether models have various kinds of constraints set without having to
inspect ``Model.tied``, ``Model.fixed``, and ``Model.bounds`` in detail. [16677]

- Added a new ``parallel_fit_dask`` function that can be used to fit models to
many sections (e.g. spectra, image slices) on an N-dimensional array in
parallel. [16696]

- Added a ``Lorentz2D`` model. [16800]

- Added ``inplace=False/True`` keyword argument to the ``__call__`` method of most fitters,
to optionally allow the original model passed to the fitter to be modified with the fitted
values of the parameters, rather than return a copy. This can improve performance if users
don't need to keep hold of the initial parameter values. [17033]

astropy.stats
^^^^^^^^^^^^^

- Added a ``SigmaClippedStats`` convenience class for computing sigma-clipped
statistics. [17221]

astropy.table
^^^^^^^^^^^^^

- Changed a number of dict-like containers in ``io.ascii`` from ``OrderedDict`` to
``dict``. The ``dict`` class maintains key order since Python 3.8 so ``OrderedDict`` is
no longer needed. The changes are largely internal and should not affect users in any
way. See also the API change log entry for this PR. [16250]

- Add a ``keep_order`` argument to the ``astropy.table.join`` function which specifies to
maintain the original order of the key table in the joined output. This applies for
inner, left, and right joins. The default is ``False`` in which case the output is
ordered by the join keys, consistent with prior behavior. [16361]

astropy.units
^^^^^^^^^^^^^

- Add a ``formatter`` argument to the ``to_string`` method of the ``Quantity``
class. Enables custom number formatting with a callable formatter or
format_spec, especially useful for consistent notation. [16087]

- Add the unit foe (or Bethe, equivalent to 1e51 erg), which is often used to
express the energy emitted by a supernova explosion. [16441]

- Add ``magnetic_flux_field`` equivalency to convert magnetic field between
magnetic field strength (H) and magnetic flux density (B). [16516]

- Added SI-units ``sievert``, ``gray``, ``katal``, and ``hectare`` in ``astropy.units.si``. [16729]

- When parsing invalid unit strings with ``u.Unit(..., parse_strict="warn")`` or
``u.Unit(..., parse_strict="silent")``, a normal unit may be returned if the
problem is not too serious.
If parsing the string fails completely then an ``UnrecognizedUnit`` instance is
returned, just as before. [16892]

- Added a ``np.arange`` dispatch for ``Quantity`` (requires one to use
``like=<some_quantity>``). [17059]

- Added support for calling numpy array constructors (``np.empty``, ``np.ones``,
``np.zeros`` and ``np.full``) with ``like=Quantity(...)`` . [17120]

- Added support for calling numpy array constructors (``np.array``,
``np.asarray``, ``np.asanyarray``, ``np.ascontiguousarray`` and
``np.asfortranarray``) with ``like=Quantity(...)`` . [17125]

- Added support for calling numpy array constructors (``np.frombuffer``,
``np.fromfile``, ``np.fromiter``, ``np.fromstring`` and ``np.fromfunction``)
with ``like=Quantity(...))`` . [17128]

- Added support for calling numpy array constructors (``np.require``,
``np.identity``, ``np.eye``, ``np.tri``, ``np.genfromtxt`` and ``np.loadtxt``)
with ``like=Quantity(...))`` . [17130]

astropy.utils
^^^^^^^^^^^^^

- Added the ``astropy.system_info`` function to help document runtime systems in
bug reports. [16335]

- Add support for specifying files as ``pathlib.Path`` objects in ``IERS_A.read``
and ``IERS_B.read``. [16931]

astropy.visualization
^^^^^^^^^^^^^^^^^^^^^

- Add ``make_rgb()``, a convenience
function for creating RGB images with independent scaling on each filter.
Refactors ``make_lupton_rgb()`` to work with instances of subclasses of
``BaseStretch``, including the new Lupton-specific classes
``LuptonAsinhStretch`` and ``LuptonAsinhZscaleStretch``. [15081]

- Add support for custom coordinate frames for ``WCSAxes`` through a context
manager ``astropy.visualization.wcsaxes.custom_ucd_coord_meta_mapping``. [16347]

- Added ``get_ticks_position``, ``get_ticklabel_position``, and
``get_axislabel_position`` methods on ``CoordinateHelper`` in WCSAxes. [16686]

- Added the ability to disable the automatic simplification of WCSAxes tick labels
by specifying ``simplify=False`` to ``set_ticklabel()`` for a coordinate axis. [16938]

- Added the ability to specify that WCSAxes tick labels always include the sign
(namely for positive values) by starting the format string with a ``+``
character. [16985]

- Allow ``astropy.visualization.units.quantity_support`` to be used as a
decorator in addition to the already supported use as a context manager. [17006]

- Added the ability to specify a callable function in ``CoordinateHelper.set_major_formatter`` [17020]

- Added a ``SimpleNorm`` class to create a matplotlib normalization object. [17217]

- WCSAxes will now select which axis to draw which tick labels and axis labels on based on the number of drawn tick labels, rather than picking them in the order they are listed in the WCS. This means that axes may be swapped in comparison with previous versions of Astropy by default. [17243]


API Changes
-----------

astropy.coordinates
^^^^^^^^^^^^^^^^^^^

- For non-scalar frames without data, ``len(frame)`` will now return the first
element of its ``shape``, just like for frames with data (or arrays more
generally).  For scalar frames, a ``TypeError`` will be raised.  Both these
instead of raising a ``ValueError`` stating the frame has no data. [16833]

- The deprecated ``coordinates.get_moon()`` function has been removed. Use
``coordinates.get_body("moon")`` instead. [17046]

- The deprecated ``BaseCoordinateFrame.get_frame_attr_names()`` is removed.
Use ``get_frame_attr_defaults()`` instead. [17252]

astropy.cosmology
^^^^^^^^^^^^^^^^^

- Passing redshift arguments as keywords is deprecated in many methods. [16597]

- Deprecated ``cosmology.utils`` module has been removed. Any public API may
be imported directly from the ``cosmology`` module instead. [16730]

- Setting ``Ob0 = None`` in FLRW cosmologies has been deprecated in favor of ``Ob0 =
0.0``. Conceptually this is a change in that baryons are now always a component of the
cosmology. Practically, the only change (besides that ``Ob0`` is never ``None``) is that
methods relying on ``Ob0`` always work, rather than sometimes raising an exception,
instead by default taking the contribution of the baryons to be negligible. [16847]

astropy.io.ascii
^^^^^^^^^^^^^^^^

- Remove all deprecated arguments from functions within ``astropy.io.ascii``.

``read()``:
- ``Reader`` is removed. Instead supply the equivalent ``format`` argument.
- Use ``inputter_cls`` instead of ``Inputter``.
- Use ``outputter_cls`` instead of ``Outputter``.

``get_reader()``:
- Use ``reader_cls`` instead of ``Reader``.
- Use ``inputter_cls`` instead of ``Inputter``.
- Use ``outputter_cls`` instead of ``Outputter``.

``write()``:
- ``Writer`` is removed. Instead supply the equivalent ``format`` argument.

``get_writer()``:
- Use ``writer_cls`` instead of ``Writer``. [15758]

astropy.io.fits
^^^^^^^^^^^^^^^

- The ``CompImageHDU`` class has been refactored to inherit from ``ImageHDU``
instead of ``BinTableHDU``. This change should be for the most part preserve the
API, but any calls to ``isinstance(hdu, BinTableHDU)`` will now return ``False``
if ``hdu`` is a ``CompImageHDU`` whereas before it would have returned ``True``.
In addition, the ``uint`` keyword argument to ``CompImageHDU`` now defaults to
``True`` for consistency with ``ImageHDU``. [15474]

- Remove many unintended exports from ``astropy.io.fits.hdu.compressed``.
The low-level functions ``compress_image_data`` and ``decompress_image_data_section``
are now only available at the qualified names
``astropy.io.fits.hdu.compressed._tiled_compression.compress_image_data``
and ``astropy.io.fits.hdu.compressed._tiled_compression.decompress_image_data_section``.
The rest of the removed exports are external modules or properly exported
elsewhere in astropy. May break imports in rare cases that relied
on these exports. [15781]

- The ``CompImageHeader`` class is now deprecated, and headers on ``CompImageHDU``
instances are now plain ``Header`` instances. If a reserved keyword is set on
``CompImageHDU.header``, a warning will now be emitted at the point where the
file is written rather than at the point where the keyword is set. [17100]

- - Remove code that was deprecated in previous versions: ``_ExtensionHDU`` and
 ``_NonstandardExtHDU``, ``(Bin)Table.update``, ``tile_size`` argument for
 ``CompImageHDU``. Also specifying an invalid ``tile_shape`` now raises an
 error. [17155]

astropy.io.misc
^^^^^^^^^^^^^^^

- New format ``"parquet.votable"`` is added to read and write a parquet file
with a votable metadata included. [16375]

astropy.io.votable
^^^^^^^^^^^^^^^^^^

- ``Table.read(..., format='votable')``, ``votable.parse`` and
``votable.parse_single_table`` now respect the ``columns`` argument and will only output
selected columns. Previously, unselected columns would just be masked (and unallocated).
``astropy.io.votable.tree.TableElement.create_arrays`` also gained a ``colnumbers``
keyword argument to allow column selection. [15959]

astropy.modeling
^^^^^^^^^^^^^^^^

- Subclasses of ``_NonLinearLSQFitter``, so any subclasses of the public ``LevMarLSQFitter``, ``TRFLSQFitter``, ``LMLSQFitter`` or ``DogBoxLSQFitter``, should now accept an additional ``fit_param_indices`` kwarg in the function signature of their ``objective_function`` methods.
Nothing is needed to be done with this kwarg, and it might not be set, but it can optionally be passed through to ``fitter_to_model_params_array`` for a performance improvement.
We also recommended accepting all kwargs (with ``**kwargs``) in this method so that future additional kwargs do not cause breakage. [16673]

- Exception message for when broadcast shapes mismatch has changed.
Previously, it used complicated regex to maintain backward compatibility.
To ease maintenance, this regex has been removed and now directly
passes exception from ``numpy.broadcast_shapes`` function. [16770]

- Using the ``LMLSQFitter`` fitter with models that have bounds is now deprecated,
as support for bounds was very basic. Instead, non-linear fitters with more
sophisticated support for bounds should be used instead. [16994]

- The optional ``use_min_max_bounds`` keyword argument in ``TRFLSQFitter`` and
``DogBoxLSQFitter`` has now been deprecated and should not be used. These
fitters handle bounds correctly by default and this keyword argument was only
provided to opt-in to a more basic form of bounds handling. [16995]

- The deprecated ``comb()`` function has been removed.
Use ``math.comb()`` from the Python standard library instead. [17248]

astropy.stats
^^^^^^^^^^^^^

- Integer inputs to ``sigma_clip`` and ``SigmaClip`` are not converted to
``np.float32`` instead of ``float`` if necessary. [17116]

astropy.table
^^^^^^^^^^^^^

- Change the default type for the ``meta`` attribute in ``Table`` and ``Column`` (and
subclasses) from ``OrderedDict`` to ``dict``. Since Python 3.8 the ``dict`` class is
ordered by default, so there is no need to use ``OrderedDict``.

In addition the ECSV table writer in ``astropy.io.ascii`` was updated to consistently
write the ``meta`` attribute as an ordered map using the  ``!!omap`` tag. This
convention conforms to the ECSV specification and is supported by existing ECSV readers.
Previously the ``meta`` attribute could be written as an ordinary YAML map, which is not
guaranteed to preserve the order of the keys. [16250]

- An exception is now raised when trying to add a multi-dimensional column as an
index via ``Table.add_index``. [16360]

- Aggregating table groups for ``MaskedColumn`` no longer converts
fully masked groups to ``NaN``, but instead returns a masked element. [16498]

- Always use ``MaskedQuantity`` in ``QTable`` to represent masked ``Quantity``
data or when the ``QTable`` is created with ``masked=True``.  Previously the
default was to use a normal ``Quantity`` with a ``mask`` attribute of type
``FalseArray`` as a stub to allow a minimal level of compatibility for certain
operations. This update brings more consistent behavior and fixes functions
like reading of table data from a list of dict that includes quantities with
missing entries, and aggregation of ``MaskedQuantity`` in table groups. [16500]

- Setting an empty table to a scalar no longer raises an exception, but
creates an empty column. This is to support cases where the number of
elements in a table is not known in advance, and could be zero. [17102]

- ``show_in_notebook`` method for Astropy tables has been un-deprecated and the API has
been updated to accept a ``backend`` keyword and require only keyword arguments. The new
default ``backend="ipydatagrid"`` relies on an optional dependency, ``ipydatagrid``. The
previous default table viewer (prior to v7.0) is still available as
``backend="classic"``, but it has been deprecated since v6.1 and will be removed in a future release. [17165]

- The default behavior of ``Table.pformat`` was changed to include all rows and columns
instead of truncating the outputs to fit the current terminal.  The new default
keyword arguments ``max_width=-1`` and ``max_lines=-1`` now match those in
``Table.pformat_all``. Since the ``Table.pformat_all`` method is now redundant, it is
pending deprecation. Similarly, the default behavior of ``Column.pformat`` was changed
to include all rows instead of truncating the outputs to fit the current terminal. [17184]

astropy.time
^^^^^^^^^^^^

- ``Time.ptp`` now properly emits a deprecation warning independently of NumPy's
version. This method was previously deprecated in astropy 6.1, but the warning
was not visible for users that had NumPy 1.x installed. Because of this, the
warning message was updated to state that ``Time.ptp`` is deprecated since
version 7.0 instead. [17212]

astropy.units
^^^^^^^^^^^^^

- The deprecated ``Quantity.nansum()`` method has been removed.  Use
``np.nansum`` instead. [15642]

- The ``factor`` parameter of the ``spectral_density`` equivalency, the use of
which has been discouraged in the documentation since version 0.3, is now
deprecated.
Use the ``wav`` parameter as a ``Quantity``, not as a bare unit. [16343]

- The ``format.Fits`` formatter class has been renamed to ``format.FITS`` and the
old name is deprecated.
Specifying the FITS format for converting ``Quantity`` and ``UnitBase``
instances to and from strings is not affected by this change. [16455]

- Conversion from one unit to another using ``old_unit.to(new_unit, value)`` no longer
converts  ``value`` automatically to a numpy array, but passes through array duck types
such as ``dask`` arrays, with equivalencies properly accounted for. [16613]

- The ``format_exponential_notation()`` method of the ``Base`` unit formatter has
changed.
Any unit formatters that inherit directly from ``Base`` but have not
implemented their own ``format_exponential_notation()`` and wish to retain
previous behavior should implement it as:

.. code-block:: python

   def format_exponential_notation(cls, val, format_spec):
       return format(val, format_spec)

Any formatters that inherit directly from ``Base`` and call
``super().format_exponential_notation(val, format_spec)`` should instead call
``format(val, format_spec)``
The specific unit formatters in ``astropy.units`` and custom formatters that
inherit from any of them are not affected. [16676]

- The deprecated ``units.format.Unscaled`` has been removed. Use ``units.format.Generic``
instead. [16707]

- Added a __round__() dunder method to ``Quantity``
in order to support the built-in round() function. [16784]

- For ``Masked`` initialization in which a mask is passed in, ensure that that
mask is combined with any mask present on the input. [16875]

- The ``get_format_name()`` method of ``NamedUnit`` and its subclasses is
deprecated.
The ``to_string()`` method can be used instead. [16958]

- The ``UnitBase.in_units()`` method is deprecated.
The ``to()`` method can be used as a drop-in replacement. [17121]

- Unit conversions to a given system with ``unit.to_system()``,
``unit.si``, and ``unit.cgs``, will now prefer the simplest unit if it
is in the given system, rather than prioritizing more complicated
units if those had a base unit component.  E.g., ``u.Pa.si`` will now
simply return ``Unit("Pa")`` rather than ``Unit("N / m2")``.  However,
the case where a unit can be simply described in base units remains
unchanged: ``u.Gal.cgs`` will still give ``Unit("cm / s2")``. [17122]

- The ``CDS``, ``OGIP`` and ``VOUnit`` unit formatters are now subclasses of the
``FITS`` unit formatter. [17178]

- The ``eV`` and ``rydberg`` units were moved to ``astropy.units.misc`` (from
``astropy.units.si`` and ``astropy.units.astrophys``, respectively).
Practically, this means that ``Unit.to_system(u.si)`` no longer includes
``eV`` as a SI-compatible unit. [17246]

astropy.utils
^^^^^^^^^^^^^

- ``IERS_Auto.open()`` now always returns a table of type ``IERS_Auto`` that
contains the combination of IERS-A and IERS-B data, even if automatic
updating of the IERS-A file is disabled or if downloading the new file fails.
Previously, under those conditions, it would return a table of a different type
(``IERS_B``) with only IERS-B data. [16187]

- ``astropy.utils.check_broadcast`` is now deprecated in favor of
``numpy.broadcast_shapes`` [16346]

- Added a new keyword ``pending_warning_type`` to ``deprecated`` decorator so downstream developers could customize the type of warning for pending deprecation state. [16463]

- The ``introspection.resolve_name()`` function is deprecated.
It is better to use the standard library ``importlib`` instead. [16479]

- ``format_exception()`` is deprecated because it provides little benefit, if
any, over normal Python tracebacks. [16807]

- The ``utils.masked`` module has gained a mixin class, ``MaskableShapedLikeNDArray``,
as well as two utility functions, ``get_data_and_mask`` and ``combine_masks``,
that can help make a container classes carry masked data. Within astropy, these
are now used in the implementation of masks for ``Time``. [16844]

- The deprecated ``compat.override__dir__()`` utility has been removed. [17190]

astropy.visualization
^^^^^^^^^^^^^^^^^^^^^

- Removed deprecated ``exp`` attribute in the ``LogStretch``,
``InvertedLogStretch``, ``PowerDistStretch``, and
``InvertedPowerDistStretch`` stretch classes, and the ``power``
attribute in the ``PowerStretch``. Instead, use the ``a`` attribute,
which matches the input keyword. [15751]

- Removes the unintended NumPy export previously at ``astropy.visualization.np``. [15781]

- Accessing or setting the following attributes on ``CoordinateHelper`` has been deprecated:

* ``ticks``
* ``ticklabels``
* ``axislabels``

Setting the following attributes on ``CoordinateHelper`` directly has been deprecated:

* ``parent_axes``
* ``parent_map``
* ``transform``
* ``coord_index``
* ``coord_unit``
* ``coord_type`` (use ``set_coord_type`` instead)
* ``coord_wrap`` (use ``set_coord_type`` instead)
* ``frame``
* ``default_label``

Accessing or setting the following attributes on ``CoordinateHelper`` has been
removed (without deprecation, as these were clearly internal variables):

* ``grid_lines_kwargs``
* ``grid_lines``
* ``lblinfo``
* ``lbl_world``
* ``minor_frequency`` (there were already public methods to set/get this) [16685]

- The deprecated ``nsamples`` parameter of ``ZScaleInterval`` is removed. [17186]

astropy.wcs
^^^^^^^^^^^

- Errors may now occur if a ``BaseLowLevelWCS`` class defines
``world_axis_object_components`` which returns values that are not scalars or
plain Numpy arrays as per APE 14. [16287]

- ``WCS.pixel_to_world_values``, ``WCS.world_to_pixel_values``,
``WCS.pixel_to_world`` and ``WCS.world_to_pixel`` now correctly return NaN values for
pixel positions that are outside of ``pixel_bounds``. [16328]


Bug Fixes
---------

astropy.io.ascii
^^^^^^^^^^^^^^^^

- Fix the broken behavior of reading an ASCII table and filling values using column names.
This PR addresses the issue and improves the functionality. [15774]

astropy.io.fits
^^^^^^^^^^^^^^^

- Fix a number of bugs in ``CompImageHDU``:

* Fix the ability to pickle ``CompImageHDU`` objects
* Ensure that compression settings are not lost if initializing ``CompImageHDU``
 without data but with compression settings and setting the data later
* Make sure that keywords are properly updated when setting the header of a
 ``CompImageHDU`` to an existing image header.
* Fix the ability to use ``CompImageHDU.section`` on instances that have not yet
 been written to disk
* Fix the image checksum/datasum in ``CompImageHDU.header`` to be those for the
 image HDU instead of for the underlying binary table. [15474]

- Fix a spurious exception when reading integer compressed images with blanks. [17099]

- Fix creating ``CompImageHDU`` from header with BSCALE/BZERO: keywords are now
ignored, as done in ``ImageHDU``. [17237]

astropy.io.votable
^^^^^^^^^^^^^^^^^^

- Making the "votable.parquet" format available as a reader format to ensure
consistency with the writer formats, even though the format it recognised
automatically by "votable". [16488]

- Explicitly set ``usedforsecurity=False`` when using ``hashlib.md5``. Without this, ``hashlib.md5`` will be blocked in FIPS mode.
FIPS (Federal Information Processing Standards) is a set of standards created by NIST (National Institute of Standards and Technology) for US government agencies regarding computer security and interoperability.
This affects validation results ingestion. [17156]

astropy.modeling
^^^^^^^^^^^^^^^^

- Fixed the output representation of models with parameters that have
units of ``dimensionless_unscaled``. [16829]

astropy.stats
^^^^^^^^^^^^^

- Fixed accuracy of sigma clipping for large ``float32`` arrays when
``bottleneck`` is installed. Performance may be impacted for computations
involving arrays with dtype other than ``float64``. This change has no impact
for environments that do not have ``bottleneck`` installed. [17204]

- Fix an issue in sigma-clipping where the use of ``np.copy()`` was causing
the input data mask to be discarded in cases where ``grow`` was set. [17402]

astropy.table
^^^^^^^^^^^^^

- Fix a bug where column names would be lost when instantiating ``Table`` from a list of ``Row`` objects. [15735]

- Aggregating table groups for ``MaskedColumn`` now ensures that fully-masked
groups result in masked elements rather than ``NaN``. [16498]

- Ensure that tables holding coordinates or representations can also be stacked
if they have zero length. This fix also ensures that the ``insert`` method
works correctly with a zero-length table holding a coordinate object. [17380]

- Fixed table aggregate with empty columns when float is present. [17385]

astropy.units
^^^^^^^^^^^^^

- Allow SI-prefixes for radioactivity units ``becquerel`` and ``curie`` in ``astropy.units.si``, conforming to BIPM's guidelines for SI units. [16529]

- The OGIP unit parser no longer accepts strings where a component unit is
followed by a parenthesized unit without a separator in between, such as
``'m(s)'`` or ``'m(s)**2'``.
Such strings are not allowed by the OGIP standard. [16749]

- A few edge cases that could result in a power of a unit to be a numerical value
from ``numpy``, instead of the intended Python ``int``, ``float`` or
``fractions.Fraction`` instance, have been fixed. [16779]

- The OGIP unit parser now detects negative powers that are not enclosed in
parenthesis.
For example, ``u.Unit("s**-1", format="ogip")`` now raises an error because the
OGIP standard expects the string to be written as ``"s**(-1)"`` instead, but it
is still possible to parse the unit with
``u.Unit("s**-1", format="ogip", parse_strict="warn")`` or
``parse_strict="silent"``. [16788]

- ``UnitScaleError`` can now be imported from the ``astropy.units`` namespace. [16861]

- Parsing custom units with ``u.Unit()`` using the ``"vounit"`` format now obeys
the ``parse_strict`` parameter, unless the custom units are made explicit with
quotation marks.
For example, ``u.Unit("custom_unit", format="vounit")`` now raises an error,
but ``u.Unit("custom_unit", format="vounit", parse_strict="silent")`` or
``u.Unit("'custom_unit'", format="vounit")`` do not. [17232]

- It is now possible to use ``Unit`` to create dimensionless units with a scale
factor that is a complex number or a ``fractions.Fraction`` instance.
It was already possible to create such units directly with ``CompositeUnit``. [17355]

astropy.utils
^^^^^^^^^^^^^

- Fixed the unintended behavior where the IERS-A file bundled in ``astropy-iers-data`` would be ignored if automatic updating of the IERS-A file were disabled or if downloading the new file failed. [16187]

- Ensure ``MaskedQuantity`` can be initialized with a list of masked
quantities (as long as their shapes match), just like regular
``Quantity`` and ``ndarray``. [16503]

- For ``Masked`` instances, ``np.put``, ``np.putmask``, ``np.place`` and
``np.copyto`` can now handle putting/copying not just ``np.ma.masked`` but
also ``np.ma.nomask``; for both cases, only the mask of the relevant entries
will be set. [17014]

- Explicitly set ``usedforsecurity=False`` when using ``hashlib.md5``. Without this, ``hashlib.md5`` will be blocked in FIPS mode.
FIPS (Federal Information Processing Standards) is a set of standards created by NIST (National Institute of Standards and Technology) for US government agencies regarding computer security and interoperability.
This affects download caching. [17156]

- Fixed a bug where an old IERS-A table with stale predictive values could trigger
the download of a new IERS-A table even if automatic downloading was disabled. [17387]

astropy.wcs
^^^^^^^^^^^

- Avoid a ``RuntimeWarning`` in ``WCS.world_to_array_index`` by converting
NaN inputs to int. [17236]


Performance Improvements
------------------------

astropy.io.ascii
^^^^^^^^^^^^^^^^

- The performance of guessing the table format when reading large files with
``astropy.io.ascii`` has been improved. Now the process uses at most
10000 lines of the file to check if it matches the format. This behavior can
be configured using the ``astropy.io.ascii.conf.guess_limit_lines``
configuration item, including disabling the limit entirely. [16840]

astropy.io.fits
^^^^^^^^^^^^^^^

- Optimize checksum computation. [17209]

astropy.modeling
^^^^^^^^^^^^^^^^

- Improved the performance of 1D models, models with scalar parameters, and models
without units, when evaluating them with scalar or small arrays of inputs. For
models that satisfy all of the conditions above, the improvement can be on the
order of 30-40% in execution time. [16670]

- Performance of most non-linear fitters has been significantly improved by reducing the overhead in evaluating models inside the objective function. [16673]

- Improved the performance of ``parallel_fit_dask`` by avoiding unnecessary copies of the
model inside the fitter. [17033]

- ``CompoundModel`` now implements numerical derivatives of parameters when using the +, -, * or / operators. This improves the speed of fitting these models because numerical derivatives of the parameters are not calculated. [17034]

astropy.stats
^^^^^^^^^^^^^

- The performance of biweight_location, biweight_scale,
biweight_midvariance, and median_absolute_deviation has been improved by
using the bottleneck nan* functions when available. This requires the
bottleneck optional dependency to be installed. [16967]

astropy.units
^^^^^^^^^^^^^

- The ``units.quantity_input`` decorator has been optimized, especially in the case that no equivalencies are provided to the decorator, and the speed-up is very noticeable when wrapping short functions. [16742]

- Parsing composite units with the OGIP formatter is now up to 25% faster. [16761]

- Parsing units with scale factors is now up to 50% faster. [16813]

- Parsing strings representing non-composite units with ``Unit`` is now up to 25%
faster. [17004]

- Converting composite units to strings with the ``"cds"``, ``"fits"``,
``"ogip"`` and ``"vounit"`` formatters is now at least twice as fast. [17043]

astropy.visualization
^^^^^^^^^^^^^^^^^^^^^

- Removed redundant transformations when WCSAxes determines the coordinate ranges
for ticks/gridlines, which speeds up typical plot generation by ~10%, and by
much more if ``astropy.visualization.wcsaxes.conf.coordinate_range_samples`` is
set to a large value [16366]


Other Changes and Additions
---------------------------

- Updated minimum supported Python version to 3.11. As a result, minimum
requirements were updated to compatible versions.
Astropy now requires
- ``numpy>=1.23.2``
- ``PyYAML>=6.0.0``
- ``packaging>=22.0.0`` [16903]