ubuntu-latest is in the process of being updated from ubuntu-22.04 to
ubuntu-24.04, which leads to issues with py37:

```
Version 3.7 was not found in the local cache
Error: The version '3.7' with architecture 'x64' was not found for Ubuntu 24.04.
```

This commit hard-codes the ubuntu-version for py37 to ubuntu-22.04,
similar to what we've done for py36.

See https://github.com/actions/runner-images/issues/10636 for more
information.

The flake8 section is the only one left, so might as well rename the
file.

To keep build-compatibility with Python 3.6, I've switched the build
backend from `setuptools` to `hatchling`. `setuptools` only supports
`pyproject.toml` from version 61 onwards, which itself isn't compatible
with Python 3.6. This does not mean we have to fully use Hatch as we can
still create build using `python -m build`. Hatchling automatically
includes all non-gitignored files, so the `MANIFEST.in` file has been
removed.

Renaming `types.py` to `_types.py` to avoid conflicts with Python's
built-in modules. This change addresses some `ImportErrors` such as:

```
ImportError: cannot import name 'MappingProxyType' from partially initialized module 'types' (most likely due to a circular import)
```

As far as I can tell this is not a problem when installing the library
using pip or running tests in tox, only local execution when adding the
source-root to the `PYTHONPATH`. This is e.g. the case in the default
IntelliJ/Pycharm run configuration when testing in scratch files.

See https://github.com/mwclient/mwclient/pull/336#discussion_r1733422778

Previously only `mwclient.errors.InsufficientPermission` was included

The logic for this was removed in cd4830a5, seems like the error was
forgotten at the time.

Non-integers in these positions would lead to issues when comparing
versions, e.g. `self.version[:2] > (1, 27)`.

In the examples in the MediaWiki documentation, versions are compared
using the `version_compare` function from the PHP standard library:

https://www.mediawiki.org/wiki/Manual:$wgVersion#Example_code

Before comparison, this version canonicalizes version strings according
to the following rules:

- Replace _, - and + with a dot .
- Insert dots . before and after any non number

This commit updates `Site.version_tuple_from_generator` to parse
versions according to those rules.

100%, baby

Signed-off-by: Adam Williamson <awilliam@redhat.com>

This fallback seems broken and weird. I can't see why:

1) We do it at all
2) Given that we do it, we do it only for namespace IDs in
   site.namespaces

It's existed since the first version of mwclient, but I really
can't see how it's valid. A `Site` instance's `self.namespaces`
starts out as `default_namespaces`, then in `site_init`, we read
in the real namespace configuration of the specific wiki. If the
wiki changes the default namespaces, we should respect that. So
it seems to me we should just trust `self.namespaces`. I don't
see how any kind of fallback to `self.default_namespaces` is ever
going to give a more useful result than not doing it.

Signed-off-by: Adam Williamson <awilliam@redhat.com>

This covers missing statements in the PageList class.

Signed-off-by: Adam Williamson <awilliam@redhat.com>

This covers some uncovered statements in the Category class.

Signed-off-by: Adam Williamson <awilliam@redhat.com>

I had to invent the API response because this is only used for
the CheckUser extension and I don't have enough privileges to
actually use that extension on any wiki with it installed, but
I'm pretty sure this more or less captures what it's meant to do.

Signed-off-by: Adam Williamson <awilliam@redhat.com>

This covers the `List.__repr__` method.

Signed-off-by: Adam Williamson <awilliam@redhat.com>

This covers the `if self.result_member not in data['query']`
branch of `set_iter`.

Signed-off-by: Adam Williamson <awilliam@redhat.com>

new-style continuation is available since 1.21, and we are
already forcing it for mw 1.21-1.25 in `api` by adding `continue`
to kwargs if it's not there. Old-style continuation is only
available by passing `rawcontinue=` and is not recommended.
I don't think we need to maintain support for it in the List
class. Theoretically it could maybe still be used if someone is
passing `rawcontinue=""` as an arg to `List` or a subclass of it
(and that overrides the `continue=` that we force in - I'm not
sure which wins), but I can't imagine why anyone would want to do
that.

This improves test coverage because old-style continuation is
uncovered at present. I'd rather remove it than write a test for
it.

See: https://www.mediawiki.org/wiki/API:Continue



Signed-off-by: Adam Williamson <awilliam@redhat.com>

This covers the `if not data: raise StopIteration` bit of
`load_chunk`.

Signed-off-by: Adam Williamson <awilliam@redhat.com>

Signed-off-by: Adam Williamson <awilliam@redhat.com>

made 429 Too Many Request HTTP error in raw_call() retryable

Bumps [actions/setup-python](https://github.com/actions/setup-python) from 5.1.1 to 5.2.0.
- [Release notes](https://github.com/actions/setup-python/releases)
- [Commits](https://github.com/actions/setup-python/compare/v5.1.1...v5.2.0

)

---
updated-dependencies:
- dependency-name: actions/setup-python
  dependency-type: direct:production
  update-type: version-update:semver-minor
...

Signed-off-by: dependabot[bot] <support@github.com>

This adds an integration test that runs a set of mediawiki
containers, configures them enough that the API works, then runs
a few simple tests of logging in, creating, moving and deleting
pages. We also enable this in CI. We can extend this with more
tests later, I think this is a reasonable initial set.

Signed-off-by: Adam Williamson <awilliam@redhat.com>

3.12 is current now.

Signed-off-by: Adam Williamson <awilliam@redhat.com>

That's the current stable Python.

Signed-off-by: Adam Williamson <awilliam@redhat.com>

The setup.py test command and pytest-runner have been deprecated for
several years and are no longer recommended. The pytest documentation
no longer recommends pytest-runner and their own documentation
recommends against it.

- https://github.com/pypa/setuptools/issues/1684
- https://github.com/pytest-dev/pytest/issues/5534
- https://github.com/pytest-dev/pytest-runner/issues/50

Changes made:

- Removed pytest-runner: Eliminated the use of pytest-runner from
  `setup.py`.
- Moved test dependencies to the `extra-dependencies`, which can be used
  by tox as well as for manual installation.
- Updated the documentation to recommend using tox as the default test
  runner. Alternatively, users can manually run tests with the `py.test`
  command.

These changes will also make an eventual migration to `pyproject.toml`
easier, as the `extras_require` can be copied to the
`project.optional-dependencies` section.

I found this while writing integration tests. If you try to login
with an incorrect username and password, mwclient correctly raises
mwclient.errors.LoginError...but if you then check logged_in, it
is True. Clearly that's wrong.

The logic for setting logged_in will basically consider us logged
in any time we process an API call and there is no 'anon' key in
the userinfo dict *or there is no userinfo dict at all*. The last
API call processed on a failed login attempt has no userinfo
dict.

This tweaks it so we only changed the logged_in value (in either
direction) if the API call we're handling has a userinfo dict.

We also fix the type of the empty userinfo we create if there
isn't one in the call (no idea why it was a tuple, it doesn't
really matter but it's weird) and explicitly set self.logged_in
to False in Site initialization. I don't think this is strictly
necessary as we can expect the init process to result in at least
one API call with a userinfo dict, but it feels more correct.

Signed-off-by: Adam Williamson <awilliam@redhat.com>