我想修改自己项目使用的一个依赖。这个依赖是 kivy(https://kivy.org/)，它有一些可选依赖，并通过 cython 编译部分原生模块，我的项目需要特定的组合，实际上是空集，而 kivy 自己的测试需要 [dev,full]。
两个项目需要不同配置，所以不能共用 venv，但两边都需要以 editable 模式安装 kivy，才能方便地在两个项目中测试改动。
顺便说一下，我觉得逐步测试多个 python 版本，并分别使用 editable 模式的 venv，也会遇到同样的限制。
有人知道怎样让这种配置工作吗？
I'd like to make changes to a dependency used by one of my projects. This dependency (kivy (https://kivy.org/)) offers several optional dependencies and compiles some native modules using cython, with my project needing a particular selection (in fact none), while kivy's own tests need [dev,full].
The projects need different configurations, so sharing a venv between them won't work, but each still needs kivy installed in editable mode so that I can easily test a change in both projects.
As an aside, I think the same constraints apply when gradually testing several python versions, using separate editable-mode venvs for for that purpose.
Does anyone know how to make this arrangement work?
应该用一个能为每个环境分别声明依赖的环境管理工具。如果用 tox(https://tox.wiki/)，大致可以这样配置。
pyproject.toml:
dependencies = [
"kivy",
...
]
dependency-groups = [
"widgetdev" = "kivy[dev,full]",
]tox.ini:
[tox]
envlist =
py314
py314-widget
[testenv]
dependency_groups =
widget: widgetdev
...要完成配置，还得看 tox 文档里的具体说明。pixi(https://pixi.prefix.dev/)也能定义具有不同功能的环境。其他工具大概也支持。
An environment manager that lets you specify dependencies separately for each environment should do it. With tox (https://tox.wiki/), something along these lines could work:
pyproject.toml:
dependencies = [
"kivy",
...
]
dependency-groups = [
"widgetdev" = "kivy[dev,full]",
]tox.ini:
[tox]
envlist =
py314
py314-widget
[testenv]
dependency_groups =
widget: widgetdev
...Check the tox documentation for the details needed to finish your configuration. pixi (https://pixi.prefix.dev/) can also define differently featured environments. Other tools probably support this too.
啊，你的意思是调整我的项目设置，让它接受兼容的 kivy 配置，而不是尝试同时使用两种不同的配置。这个思路不错，我会仔细看看 uv 在这里能做什么，毕竟目前我用的就是它，不过 tox 看起来也确实有些实用功能。谢谢！
Ah, so adapt my project's settings to accept a compatible kivy configuration, instead of trying to run both configurations together. That sounds like a sensible approach, I'll look more closely at what uv can do here since that's the tool I'm usng at the moment, though tox certainly seems to have some useful features. Thanks!
在开发环境里用 docker-compose up 启动所有服务时，我的几个容器会输出几百行又几百行内容。有没有方便的办法过滤这些输出，或者更好一些，让这些服务在 docker-compose up 运行时不要写到标准输出？
看来 D 能与 C 链接。我想用 D 编写 Python 扩展模块。我是不是漏掉了什么原因，导致这种方法根本无法运行？
python 可以这样写。 Python.withPackages (q: with q; [ fastapi ]) Micro 编辑器有类似的机制吗？ 能不能把 micro 编辑器的软件包和插件一起指定？:thinking: 那样会方便很多。
