Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/sachaa-thanasius/experimental
Personal __future__-esque imports.
https://github.com/sachaa-thanasius/experimental
python
Last synced: 9 days ago
JSON representation
Personal __future__-esque imports.
- Host: GitHub
- URL: https://github.com/sachaa-thanasius/experimental
- Owner: Sachaa-Thanasius
- License: mit
- Created: 2024-03-31T22:34:41.000Z (9 months ago)
- Default Branch: master
- Last Pushed: 2024-07-23T13:41:43.000Z (5 months ago)
- Last Synced: 2024-07-23T15:55:30.949Z (5 months ago)
- Topics: python
- Language: Python
- Homepage:
- Size: 286 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# experimental
Yet another way of messing with non-existent features in Python.
This is a personal `__future__`-like collection of features that will only be active if imported from `__experimental__` near the top of a Python file. Some of those features include:
- Late-bound function argument defaults
- Based on [PEP 671](https://peps.python.org/pep-0671/).
- Currently, the late-bound expressions must be surrounded in parentheses, unlike the syntax proposed in the relevant PEP.
- Module-level and context manager-level lazy imports
- Based on [PEP 690](https://peps.python.org/pep-0690/) and several other attempts at lazy importing.
- Currently, `from` imports are evaluated eagerly.
- Inline import expressions
- Based on [import-expression](https://github.com/ioistired/import-expression-parser).
- Elision of `typing.cast`/`typing_extensions.cast`
- Based on past discussions had about possibly eliminating the cost of `cast`.
They shouldn't be mutually exclusive syntax-wise.
## Installation
```shell
python -m pip install https://github.com/Sachaa-Thanasius/experimental
```
I don't see a reason to put this on PyPI as of now.
## Examples
```py
"""example1.py"""
from __experimental__ import late_bound_arg_defaults
def bisect_right(a, x, lo=0, hi=>(len(a)), *, key=None):
# If nothing is passed in for hi, only then will it evaluate as len(a).
# Late-bound defaults will be evaluated in left-to-right order.
...
"""example2.py"""
from __experimental__ import inline_import
# This will import the collections module and use it without having to place
# an import statement at the top of the file.
assert collections!.Counter("bccdddeeee") == {'e': 4, 'd': 3, 'c': 2, 'b': 1}
"""example3.py"""
from __experimental__ import lazy_import
# All module imports within this file henceforth should be performed lazily,
# i.e. delaying module load until accessing an attribute from the module.
...
```
## Caveats
### Registration
This package uses a `.pth` file to register an import hook on interpreter startup. The hook replaces the built-in file finder's [`path hook`](https://docs.python.org/3/library/importlib.html#importlib.machinery.FileFinder.path_hook) on [`sys.path_hooks`](https://docs.python.org/3/library/sys.html#sys.path_hooks). That should work fine if you're using a regular setup with site packages, where that `.pth` file should end up.
However, if your environment is atypical, you might need to manually register that finder to have your code be processed by this package. Do so in a file away from the rest of your code, before any of it executes. For example:
```py
import __experimental__
__experimental__.install()
import your_actual_module
```
## Documentation
See the docstrings and comments in the code. They're currently a little bare, but they will improve over time.
## Why?
It's fun. It's also a learning experience in manipulating Python via import hooks, AST transformations, token stream transformations, and more.
## Acknowledgements
- aroberge's [ideas](https://github.com/aroberge/ideas) package, which introduced me to the idea of modifying Python like this.
- asottile's future packages, including things like [future-fstrings](https://github.com/asottile-archive/future-fstrings), that are practical applications showcasing the usefulness of these kinds of transformations ("memes" though they may be).
- RocketRace's [hoopy](https://github.com/RocketRace/hoopy) and [custom-literals](https://github.com/RocketRace/custom-literals) packages, which are much more interesting, complex, and featured Python manipulations.
- Those who wrote the original PEPs and packages that this attempts to (re)implement (see the links in the overview).
- Addendum: aroberge had the idea (and the code) for an `__experimental__` package well before I did, but I didn't know until well after making this.