Software can still rely on components written in C++ but new code is largely written in another language.
C++ can also be superseded not by one but by a fragmented ecosystem of several programming languages. There doesn't have to be a single successor.
Let's say 40% of systems are still written in C++, but the remaining 60% are split equally between language A and language B. Even though neither A or B surpass C++ individually, they would still count as superseding C++ together if both are clear alternatives.
The overall resolution criteria is that among new code written towards system workloads, C++ usage falls below 50% according to GitHub statistics on open source contributions or the annual StackOverflow survey. Note that the source for this metric can be subject to change if the numbers provided by GitHub or StackOverflow become less representative by the time the market closes.
1,000
3.00There is no way to restrict either Github pull request data or StackOverflow survey data to greenfield projects, aka "new code" that the question is about. This makes it exceedingly unclear what data the resolution is going to be based on.
If you are going to look at Github pull request statistics, then this is a pretty obvious "no" given the current trends. It does not reflect the "new code" part of the resolution criteria, and neither does StackOverflow survey, so the resolution criteria contradict themselves.
This contradiction prevents me from trading. I also suspect that the market hovers at around 50% because the traders have no idea what the resolution is going to be.
@vberlier to make this market actually predict something useful, you need to either drop the "new code" part of the criteria (and then I'll immediately bet the market down), or specify how exactly you're going to measure "new code" because the currently listed sources do not consider it.
In the meantime, there's this market with clear and objective resolution criteria:
https://manifold.markets/BoydKane/will-more-professional-devs-use-rus
I can't think of any source where this isn't already the case. A handful of sources show Go, C, or Rust on par or above C++. Add in Swift or Objective-C if you want some margin, and maybe newcomers like Zig will make a dent. Do you classify cppfront/carbon/etc as C++ ?
What's your source of github stats, there are big variations between for example languish and githut. How would you resolve if one source says yes and the other no ?
@moltonel3xCombo Github pull request statistics show both C and C++ far above Rust: https://madnight.github.io/githut/#/pull_requests/2024/1
@SergeyDavidoff That graph also shows Go pull requests well above C and C++. Or check https://tjpalmer.github.io/languish/#y=pulls&names=c%2B%2B%2Cc%2Crust instead, which shows Rust pull requests above C and pretty close to C++. Note that neither site is an official Github site, and there are some heuristics involved in identifying repos, deduplicating forks, etc, From what I can tell, Languish is better at identifying new repos than Githut,
But this market explicitly allows multiple languages (including C, AFAICT) to "team up" against C++. Pretty unfair and unwinable by C++ if you ask me, but that's how the question is worded.
My other gripe with this market is that it doesn't try to define "systems programming language" (opinions on this are more diverse than you might think).
As for identifying "new code", I think taking the dev stats of the latest period is a decent approximation. New code is not just greenfield projects, the hundreds of thousands of lines of C added to each Linux kernel release is still new code. I do agree that the criteria is too loosely defined though: choose either SO survey or github, and if you choose github specify push/pull/users and the source of truth.
@MrLuke255 Most likely GitHub stats on open source contributions combined with the annual stackoverflow survey. But deciding on how to get the most representative numbers will only be possible once we get closer to market resolution.
@vberlier I feel like this really should have a nailed down resolution criteria from the outset. Depending on what metrics you pick one could make the case it’s already been superseded, or that it won’t be for at least a decade.
@spider It's possible that the criteria we would settle on today may not be as relevant when the market closes. For example stackoverflow may fade in popularity by 2027 and make the survey results less representative. I can update the market with a preliminary criteria but we would have to be able to change it for a better metric if it declines in accuracy in the future.
@vberlier You should be able to describe your method for resolution nonetheless. Instead of "stackoverflow" you could also write "the website where most programming questions are answered" or something along those lines.
Without a clear resolution criterion you could just update how you are going to resolve it by choosing a method which you subjectively think is more accurate.
@TimKuipers I did end up updating the description to specify the criteria. There will be a poll in case the criteria needs to change to stay relevant.
@vberlier What is "new code written towards system workloads"? Is that a category you can choose for GitHub statistics? Not sure if I don't understand or the resolution criterion isn't 100% clearly defined.
@TimKuipers It just means code written in languages that advertise themselves as suitable for systems programming. We'll aggregate the numbers for each language from GitHub.
@vberlier Aren’t most systems written in C++ proprietary? Is it possible to easily retrieve stats for code added to GitHub only in the recent year?
@MrLuke255 GitHub stats are still regarded as pretty representative of the industry as a whole. I'm open to suggestions if you have a better criteria in mind.
@MrLuke255
> Aren’t most systems written in C++ proprietary
big part of why I'm not betting no, even though I would guess a probability in the 8-15% range. It seems open source stuff could be a lot faster to switch.