| I like it. On python-ideas the idea of a curated list of pypi packages comes up from time-to-time. The most recent incarnation was started just last month and you can read it here: https://mail.python.org/pipermail/python-ideas/2016-March/03... Such discussions usually die out after some people give their opinions on what such a package curation/voting system should be like, whilst others state why we should not have such a system. For my money, most of the objections are really just people pointing out that it is a very difficult to design problem. Rather than objecting to the actual idea. So anyway, it's really nice to see some actual working code. A few of comments: 1. You describe this as the "go-to Python Toolbox", but I'm not sure that really describes what it is well. It's really a summary of github statistics. Even forgetting how the packages are rated, you have a package rating system rather than a single toolbox. I think what you have is more interesting than a single python toolbox. 2. You have not quite solved the major complaint that usually comes up in discussions of such package voting systems. That is, that a package is not generally just "good" or "bad", but appropriate or not appropriate for given uses. Furthermore, I think the 'Popularity' metric is not quite obvious from a first look at, say, a package category. In other words, it's not quite obvious that this is a measure of how the github repository is starred. At least then the user can decide themselves whether or not this is a useful metric. 3. Github stars is not a bad measure. A better one would include the ability to downvote. 4. Pretty much all repulation/voting systems have the problem that for any "votes", or even full-blown reviews, it's not clear which version of the library they are about. This is really difficult to solve. Since you only have positive votes the problem is not so pronounced as it is in other systems (or proposed systems). But the basic problem is that a bunch of users downvote a library because it is difficult to use, or doesn't have tests, or they cannot install it etc. The developer takes that into consideration and produces a new version of the library, what should the reputation system do? It's really not a trivial question. You still have that issue in that a bunch of users may have decided not to star a library because of an earlier version. They may even star a library which later becomes less useful, for example because it is superceded by a better library, or because the code degrades. I understand users can unstar a repository, or flock to a new one, but still, the general problem is not entirely solved. That said, using GitHub stars probably suffers from this problem less than other measures. 5. I really like the look of it. I think the colour scheme for Python is the best of the three so far. |