Show HN: Tagging all GitHub repos | Hacker News Reader

Show HN: Tagging all GitHub repos

github.com

68 pointsadieu30 comments

If anyone is interested in a more personal tagging service, there is Astral (http://astralapp.com/). Shame it isn't open source though.

I guess the point is, whether you want to tag everything yourself, most of the tags are really standardized anyways. Probably a standardized shared tagging system that allows some personal touch would be the best.

Yes, astral is a great tool. But our goal is different. We're trying to crowd source the tagging work so that most people don't have to tag repos any more. And the tagged data could be used by any Github related services or anyone who wants to study open source projects.

Author of Astral here, it is currently open source (http://github.com/astralapp/astral), but a bit of a mess since I'm trying to upgrade the stack and fix some bower dependencies. Hoping to get it tidied and ready for real contributions within the next month or so.

And this is why I love open source and HN.

Oh, I guess maybe this open-sourced a little after launch? I looked around for the source when it was first launched and couldn't find it.

You're correct, the decision to open source it was quite recent. I'm excited to see where the dev community will take it once the repo is in good shape :]

The first thing I noticed is you hijacked my back button. After landing on your homepage I was unable to go back.

We use pjax to make users navigating through our site easier. It might be a bug in the frontend. Will fix it asap. Thank you for your feedback.

I totally agree that this is a major issue at the moment with github. I even tried to build something to organize repositories http://hora.reggi.com/, the thing is most package managers (npm, and bower for instance) have a keywords field, can't we just use those?

Hora looks very nice. I see many people using different tools and methods to organize their favorite repos. If we could put those efforts together and standard things a bit (so no more angular and angularjs). The end result will help many others. We have plans to crawl npm, pypi and other platforms for tags data and make all those data available for everyone. Hope this could attract others with similar needs to join.

The number of directories under repos/ is far too long for GitHub or most browsers to handle.

That bothers us too. We choose this structure to make commit message more meaningful. Any suggestions to improve this?

alphabetical ? 26 entries or 2626 (aa-zz) entires ? maybe ?

But I fear, one particular element would be very much populated ? like repos like angular-, or rails-* or react-*. But still, this would at least be better than having the entire list in a single directory.

hmm... Sounds doable. I'll try to gather more ideas about the structure, but we definitely should put a README file in the repos folder to explain things a little bit. Is it possible that you could open an issue in the repo and put your ideas there? Thank you so much.

https://github.com/porter-io/tag-github/issues/4 :+1:

Github trucatees it to 1000, but I use a browser extension that adds a directory tree to github repos for easier navigation[0] that freezes Firefox when it tries to build the list.

0. https://github.com/buunguyen/octotree#

oops, sorry. will fix it.

That's not really your fault... I've opened an octotree issue here: https://github.com/buunguyen/octotree/issues/163

Very nice! I was actually wondering lately how I could find out about github repositories that are of interest to me (without reading HN every day).

I'm wondering, what kind of recommendation engine is this project using?

You mean our service Porter.io ? We analyze starred repo of our user and recommend news based on similarity and popularity . And we have bunch of ideas to improve this.

Well, given that there is already a lot of research in the area of collaborative filtering (see for example [1]), I would expect that there exist also some open source engines for this problem.

[1] http://en.wikipedia.org/wiki/Netflix_Prize

How about the Python classifiers as tags? Should cover everything and save re-inventing the wheel:

http://pythonhosted.org/an_example_pypi_project/setuptools.h...

when I click the [Rails] tag, oh-my-zsh appears as first in the list. I realize it has a rails plugin but I'm not sure it's one of the most relevant rails projects.

I think there should be some kind of "sort by relevance" feature.

The demo page we built is a simple showcase of what could be done with the tagging data. Hope it could inspire more people to play with this dataset. "sort by relevance" sounds cool. Might add it to the demo in the future.

...and the idea is actually awesome! (sorry i'm not sandwich approach-ing you, I just forgot to mention it :)

This is awesome I love it, I wonder what are the API limitations?

Is there a way to recommend more repos depending on what I've already starred? Since most of them are tagged now.

The coolest part was what's hot on HN section!

The demo seems down: https://porter.io/explore/tags/

Just visited it. Looks ok.

Reminds me of Fluidinfo... (general purpose tagging of everything by everyone with an API and built in query language.)

Reply on news.ycombinator.com