| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
We found that large applications that have undergone a 2 -> 3 migration
and wound up with a lot of six.ensure_str and six.ensure_binary calls
could save 1-2% CPU usage by optimizing these for the common case.
Further optimization could be done by replacing them with extension
module implementations - assumed out of scope for the pure Python six
project itself.
Ideally all of these calls and use of six in people's code would be
removed after there all need for any Python 2 compatibility is gone.
But completing that kind of type cleanup requires a lot of human
engineering time. This lowers the ongoing costs in the interim.
Contributed by YouTube.
|
|
|
| |
Co-authored-by: Benjamin Peterson <benjamin@python.org>
|
| |
|
| |
|
| |
|
|
|
|
|
|
| |
Fixes #288.
Co-authored-by: Benjamin Peterson <benjamin@python.org>
|
| |
|
| |
|
|
|
|
| |
Fixes #317.
|
| |
|
| |
|
|
|
|
|
|
|
| |
Fixes #308.
Stops testing them on the CI, update python_requries and remove some code specifically for those versions.
Not done anything to remove any six functionality that's only a benefit on those versions, that should be in a separate PR and would be a breaking change that should ideally deprecate first.
|
| |
|
|
|
|
|
|
|
|
|
| |
This is pretty-much a straight backport of Py3 implementations of update_wrapper and (privately) wraps.
Fixes #250
Fixes #165
Co-authored-by: Benjamin Peterson <benjamin@python.org>
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
| |
Fixes #155.
Closes #241.
|
| |
|
| |
|
|
|
|
|
| |
- Add Python 3.8.
- Unpin pypy2.7 and pypy3.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This reverts commit d5efa74e2dfde8d4ddba13e127cd85c687e6016b.
The original intention of providing a distutils fallback has not gone
away. These three extra lines will almost never be used, as most people
have setuptools or install with pip anyway, and at the same time it
serves as a very low-maintenance solution for people who *do* want to
bootstrap setuptools.
Bootstrapping setuptools without vendored six (and pyparsing) is a valid
use case. Although setuptools by default ships with vendored
dependencies, it supports removing the contents of the _vendored/
directory and falling back to an installed version of six.py -- and
thirdparty vendors of setuptools, such as Arch Linux and OpenSUSE,
actually make use of this and need to be able to bootstrap the
dependency tree of setuptools itself.
|
|
|
| |
We are obliged to use a very old version that supports Python 3.2.
|
| |
|
| |
|
| |
|
|
|
| |
There was `six.io` if and only if py3, which was error-prone.
|
| |
|
|
|
|
| |
Also, [remove the __sudo__ tag per recommendation of Travis](https://blog.travis-ci.com/2018-11-19-required-linux-infrastructure-migration).
|
|
|
|
| |
Small update to the `ensure_str` function's documentation formatting to
match `ensure_binary` and `ensure_text`'s formatting.
|
|
|
|
|
|
|
|
|
|
|
|
| |
As documented in the Python 2.6 release notes:
https://docs.python.org/2/whatsnew/2.6.html#pep-3112-byte-literals
> For future compatibility, Python 2.6 adds bytes as a synonym for the
> str type ...
To encourage more forward compatible code bases, inform users of this
builtin alias. This addition is similar in spirit to the note for the
b() function.
|
|
|
|
|
|
|
| |
setuptools includes a vendored version of six (and other dependencies).
They are not installed through traditional tools. Therefore, distutils
is not required as a fallback to facilitate setuptools.
https://github.com/pypa/setuptools/blob/v40.6.3/setuptools/_vendor/six.py
|
| |
|
| |
|
| |
|
| |
|
|
|
| |
The docs say that the Python 2 builtin __reload()__ was moved into the __imp__ module (which _used_ to be true) but in all currently supported versions of CPython, __reload()__ is found in the __importlib__ module: https://docs.python.org/3/library/importlib.html#importlib.reload
|
| |
|
|
|
|
| |
Fixes #259.
|
| |
|
| |
|
|
|
|
|
|
|
| |
https://tox.readthedocs.io/en/latest/config.html#confval-indexserver
> DEPRECATED, will be removed in a future version
Unnecessary anyway as all deps come from PyPI through pip.
|
|
|
| |
Avoids the need for users to look this up by experimentation or code inspection.
|
|
|
|
|
| |
For details on the new PyPI, see the blog post:
https://pythoninsider.blogspot.ca/2018/04/new-pypi-launched-legacy-pypi-shutting.html
|