Hacker News Reader

Comment by nejsjsjsbsb | Hacker News Reader

A better metric until it becomes a target. Once it is a target, getting a billion clones is trivial.

Github should just stop showing star counts. Who cares about them.

I sort by most amount of stars quite frequently when I am learning a new language and want to know what the most popular package is for something. What do you think would be a better metric for a use case like that?

number of actual imports in code

CodeRank(tm)!

This might work but biases against languages whose package managers are not used in the rank. As well as code that is used alot but not referenced via code directly e.g. drop in dlls.

Goodheart's law - this would just cause imports in junk repos

I think it's a decent metric. I agree with the other comment that actual imports is probably a better metric, but that's not always as trivial to find.

That said, the package repositories for many popular languages list stats of either declared dependencies or package downloads, which helps.

rdeps are completely broken in github. I wrote a library that I have used in other projects of my own and it was always at 0 users.

Anyway if stuff is used by proprietary stuff it will also sit at 0.

I now moved to codeberg where there is less spam, although it does have stars

I wasn't suggesting using GitHub. In fact, I don't think I was aware that GitHub even listed reverse dependencies. Where is that in the UI?

https://github.com/ltworf/typedload/network/dependents

Ah yes. I forgot I actually use this feature in some of my projects. GitHub isn't always able to update this automatically but you can programmatically report dependencies.

Yeah not way to automatically detect "import xxx"

Right. I believe they rely on things like requirements.txt, Cargo.toml, etc. And of course not all languages are supported. So if you use an unsupported language or package manager, you're out of luck.

It doesn't even work with normal requirements.txt by the way. At least not all the time.

Don't count any of my stars then, I thought it was a bookmark feature. Every repo I've starred is only starred to find later, not an endorsement from me.

That's partly what I assume people star repos for, it really doesn't defeat my use case. It's a metric that indicates interest and popularity more so than approval or endorsements.

I think it's like a "upvote" thing which shows whether historically users have found the repo interesting. Even if you hide stars, there needs to be a way for the collective hivemind of github users to help each other with what repos are high quality or not right?

You don't need to crowdsource everything. I've never used stars as a good metric because it's literally zero effort. It's anybody who happens by just stars it, So all you can really conclude from star count is that this is interesting to this number of people.

Two metrics that I think correlate extremely highly with quality: The number of commits in the repository and the date of the most recent commit. I've used a metric based on those two inputs for the past 15 years to evaluate repos and I am not disappointed. Depending on the nature of the project, I weigh the two attributes differently. Some projects are arguably, 'done', and so the date of the most recent commit is not very important in that case.

I think "interesting to this number of people" is not a meaningless metric, but I would agree on the two other metrics you cite.

There is a big difference between “highest quality” and “most popular.” Online services constantly confuse the two because it’s easier to measure popularity.

I don't disagree. But I think there's at least some positive correlation. The highest quality is unlikely to be the least popular and vice versa.

Except that most people don't bother starring stuff, so the few who do are drowned by noise of fake stars.

I care. How do I know which is the better tool between 10 different tools?

You missed the point. Number of stars doesn't indicate anything about whether a particular tool is better (and never did). If you're using stars for that purpose then you're doing it wrong. Find another solution.

Have in your mind your requirements. E.g. license, community, how funded, bugs, try a few out and check the ergonomics, talk to others who have used it.

VC apparently.

Reply on news.ycombinator.com