GitHub stars + issues + PRs is going to favour languages developed on GitHub rather than mirrored on GitHub. It also favours languages undergoing rapid development compared to those that move at a slower pace.
Neither of these have anything to do with popularity, which means people actually using it. Here are some better proxies for popularity:
Actually, searches and SO questions would correlate with languages that are harder/less intuitive, rather than necessarily popular. It's like Distrowatch with Linux, which favours "flavour of the day" distributions people are looking for, rather than the well-established ones that actually dominate deployments - because nobody needs to search for those.
Developer surveys are good but they typically self-select by audience. JetBrains surveys will over-represent core JB audiences (e.g. Java developers), SO other constituencies (newbies, MS...), etc etc.
Right, which is why I said they're proxies for popularity. We don't actually know what people are running on their computers so we use these proxies. They're not perfect, but they're a lot better than number of stars on the Github repo (TFA) or number of Google results (Tiobe).
I also realise that StackOverflow has a bias towards newbies but that's not a bad thing. It just means that it's a forward looking prediction. The newbies of today are the grizzled veterans of 2030.
> SO questions would correlate with languages that are harder/less intuitive, rather than necessarily popular
People repeat this a lot because it sounds reasonable, but I don't see why it has to be the case.
I would expect popularity to still dominate the SO question signal even if a handful of experts aren't asking questions on SO because they can RTFM. Newcomers to a language seem to ask a lot of questions no matter what, so bad SO signal seems to mean decreasing popularity.
Also, the biggest questions on SO are super basic questions that not even the most intuitive, best documented language can escape. I don't see good evidence for the claim.
This is an extremely bad indicator for anything. What people search for does often not correlate to what they choose to use in their projects in the end. Take crypto/web3/blockchain for example. Lots of searches, but not many people who build a blockchain.
> How many questions they’re asking on StackOverflow
Not necessarily a good indicator for popularity (although it will play a role). I'd say it's more of an indicator how much developers struggle with a particular language.
> Developer surveys
Better, but even those have a strong bias based on their historic audience. StackOverflow is overrepresented with developers from 1-2 languages who have been historically the first/biggest community there.
Not to mention, commercial code is not going to be well reflected on github. This might show what is popular in OSS but not necessarily what industry trends are happening in production use.
> How many questions they’re asking on StackOverflow
Apart from already mentioned issues, another reason why this doesn't make sense as a metric is that not all language communities gravitate towards Stack Overflow. Julia, for example, has its own Discourse instance: https://discourse.julialang.org
> How many questions they’re asking on StackOverflow
StackOverflow is no more valid a metric than GitHub. Off the top of my head, Pascal, and all its dialects and variants, comes to mind. Many Pascal users and those of the many dialects that it has, just about never use StackOverFlow. They can use websites that correspond to their dialect or variant (Delphi, Object Pascal, Free Pascal, Oxygene...), books (which there are many), etc...
But, for the newer languages, I do think GitHub is much more representational. Not just the stars, but adding other GitHub metrics, can give a more wholesome picture.
I do think that each language, along with the ecosystem and culture that surrounds it, has to be examined by using various metrics and different sources in order to get a more complete picture. GitHub metrics are telling us a lot, just not the entire picture.
Sort of, but this metric can also be gamed. Look at this effort from 2008 to game the metric in favour of Delphi (http://delphi.org/2008/10/delphi-keeps-climbing/). That's why Delphi remains high in this useless, meaningless ranking even though no one uses Delphi.
issues + PRs probably is a better metric of language stability/maturity, and github stars is a (obvious?) metric for how many people on github like projects. Nothing in there about popularity. At all.
If anyone wants to see the most used languages on GitHub then we post that data regularly in our Octoverse report (https://octoverse.github.com/2022/top-programming-languages). This ranking is obtained by looking at the code pushed to GitHub and to Gists on GitHub with-in a 12 month period.
What I find most interesting is how consistent that top 10 list is with the Octoverse data. We have to look much further down the table to see the cool kids like Rust, Go or Lua. While growing very fast there is just so much code out there in Javascript, Typescript, Python, Java, C#, C++ etc that is takes a lot for a language to move up the charts.
That said, while I generally caution against an unhealthy interest in Stars - what the OSS Insight data does a decent job of showing is the activity and interest in the open source languages communities. Rust is clearly incredibly strong there along with things like TypeScript, Go, Python etc. But it's also heartening to see the strength in open source language ecosystems such as Swift, Java & .NET but also some surprises such as PowerShell or the more academic languages like OCAML.
I’m confused, this is actually a list of language repos that have the most stars, pull requests, and issues. Doesn’t seem like popular is the right word and especially for those metrics.
I agree with you. On GitHub, if someone is talking about the most popular language, I would assume that means that one that people are using the most. Not the one that's used in a project that received the most starts for a given time frame.
I am happy, however, to see Rust at the top. ^^
Purely just because Rust is named after fungi, which I like.
Very interesting, Vlang sounds like a cool project and kind of a dreamy language if it can become more stable.
However, it doesn't speak for the most popular languages but maybe rather to the most active ones? I really doubt that Vlang (as an example) is more popular than Python, PHP etc..
Well, such could be said about Rust, TypeScript, or Zig in relation to Python. Not just Vlang. It may point to other factors at play, such as user satisfaction or enthusiasm. Python has a large number of critics, as well as fans.
As Python is a older language, its users may tend to go other places and look to GitHub less. There are many sites, not to mention Python's own, where people can download releases or get information.
There are other GitHub metrics to look as well, such as contributors or forks. In both of these, Python has a huge number in comparison to other languages.
I think this is measuring '"stars" along a stability scale'. It's possible (but not proven) that stars *might* measure popularity to some degree, but PRs and issues is def a measure of stability -- more issues and PRs means it is changing a lot or has a lot of bugs or new features.
Does this dynamically load the data from the github API in frontend javascript on every page load?
Le sigh. Apart from the obvious UX issue of waiting for this data to load on every page navigation, a little caching might help avoid sending a DDOS at the GitHub API...
Yesterday it was taking 10+ seconds to load each dataset. Occams razor suggests missing caches in such a case (since the data doesn't change with any meaningful frequency).
Learning F# you learn dot net. And it's very similar to OCAML and other functional languages. The underlying principles are the same. So I'd not consider that a deadend
Neither of these have anything to do with popularity, which means people actually using it. Here are some better proxies for popularity:
- What people are searching for on search engines (https://trends.google.com/trends/explore?date=2012-07-31%202...)
- How many questions they’re asking on StackOverflow (https://insights.stackoverflow.com/trends?tags=java%2Cc%2B%2...)
- Developer surveys (https://survey.stackoverflow.co/2022/#technology-most-popula...).
But as long as we’re not measuring popularity based on number of results returned by Google, we’re probably fine. (https://blog.nindalf.com/posts/stop-citing-tiobe/)