##// END OF EJS Templates
Merge pull request #13402 from Kojoley/disable-coverage-on-pypy...
Merge pull request #13402 from Kojoley/disable-coverage-on-pypy Disable coverage collection on PyPy

File last commit:

r27307:93d13d63
r27334:02cb40e0 merge
Show More
version8.rst
428 lines | 14.8 KiB | text/x-rst | RstLexer

8.x Series

IPython 8.0

IPython 8.0 is still in alpha/beta stage. Please help us improve those release notes by sending PRs that modify docs/source/whatsnew/version8.rst

IPython 8.0 is bringing a large number of new features and improvements to both the user of the terminal and of the kernel via Jupyter. The removal of compatibility with older version of Python is also the opportunity to do a couple of performance improvement in particular with respect to startup time. The 8.x branch started diverging from its predecessor around IPython 7.12 (January 2020).

This release contains 250+ Pull Requests, in addition to many of the features and backports that have made it to the 7.x branch. All PRs that went into this released are properly tagged with the 8.0 milestone if you wish to have a more in depth look at the changes.

Please fell free to send pull-requests to updates those notes after release, I have likely forgotten a few things reviewing 250+ PRs.

Dependencies changes/downstream packaging

Note that most of our building step have been changes to be (mostly) declarative and follow PEP 517, we are trying to completely remove setup.py (:ghpull:`13238`) and are looking for help to do so.

  • Minimum supported traitlets version if now 5+
  • we now require stack_data
  • Minimal Python is now 3.8
  • nose is not a testing requirement anymore
  • pytest replaces nose.
  • iptest/iptest3 cli entrypoints do not exists anymore.
  • minimum officially support numpy version has been bumped, but this should not have much effect on packaging.

Deprecation and removal

We removed almost all features, arguments, functions, and modules that were marked as deprecated between IPython 1.0 and 5.0. As reminder 5.0 was released in 2016, and 1.0 in 2013. Last release of the 5 branch was 5.10.0, in may 2020. The few remaining deprecated features we left have better deprecation warnings or have been turned into explicit errors for better error messages.

I will use this occasion to add the following requests to anyone emitting a deprecation warning:

  • Please at at least stacklevel=2 so that the warning is emitted into the caller context, and not the callee one.
  • Please add since which version something is deprecated.

As a side note it is much easier to deal with conditional comparing to versions numbers than try/except when a functionality change with version.

I won't list all the removed features here, but modules like IPython.kernel, which was just a shim module around ipykernel for the past 8 years have been remove, and so many other similar things that pre-date the name Jupyter itself.

We no longer need to add IPyhton.extensions to the PYTHONPATH because that is being handled by load_extension.

We are also removing Cythonmagic, sympyprinting and rmagic as they are now in other packages and no longer need to be inside IPython.

Documentation

Majority of our docstrings have now been reformatted and automatically fixed by the experimental VĂ©lin project, to conform to numpydoc.

Type annotations

While IPython itself is highly dynamic and can't be completely typed, many of the function now have type annotation, and part of the codebase and now checked by mypy.

Numfocus Small Developer Grant

To prepare for Python 3.10 we have also started working on removing reliance and any dependency that is not Python 3.10 compatible; that include migrating our test suite to pytest, and starting to remove nose. This also mean that the iptest command is now gone, and all testing is via pytest.

This was in bog part thanks the NumFOCUS Small Developer grant, we were able to allocate 4000 to hire Nikita Kniazev (@Kojoley) who did a fantastic job at updating our code base, migrating to pytest, pushing our coverage, and fixing a large number of bugs. I highly recommend contacting them if you need help with C++ and Python projects

You can find all relevant issues and PRs with the SDG 2021 tag https://github.com/ipython/ipython/issues?q=label%3A%22Numfocus+SDG+2021%22+

Removing support for Older Python

We are also removing support for Python up to 3.7 allowing internal code to use more efficient pathlib, and make better use of type annotations.

"Meme image of Toy Story with Woody and Buzz, with the text 'pathlib, pathlib everywhere'"

We have about 34 PRs only to update some logic tu update some function from managing strings to using Pathlib.

The completer has also seen significant updates and make use of newer Jedi API offering faster and more reliable tab completion.