Wrong way to install dependencies of plugins

[dwt]

Hi there,

I've got the problem, that I can't install a local version of the Lektor-atom plugin, as it has an extras_requireof test.

This leads to a egg.info/requires.txt of

MarkupSafe
feedgenerator

[test]
lxml
pytest

Which of course is no valid requirements.txt file that pip can use, which then leads to an explosion when this is installed here.

I am not entirely sure what the correct way to consume this information is, but I suspect that one should go about using the pkg_resources package. Quick googling has provided this stackoverflow which looks promising.

[dwt]

Thinking about this more - it seems that Lektor is doing quite a bit of package managing itself - wouldn't it be much more simple if it merely creates a virtualenv and then uses pip to install the dependencies there directly instead of generating the egg_info, fetching the dependencies out of that and then installing them by hand?

I mean, I could easily send a pull request that filters out everything in those requirements after the first line that starts with '[' - but what you guys do there seems a bit excessive.

[dwt]

After finding and documenting the original bugs that triggered @mitsuhiko to take this approach, I have decided that cleaning the requires.txt file to become nearer to a requirements.txt file to be the simplest approach. What do y'all think?