Many of us write opensource code in a void: nobody ever looks at it, uses it nor reviews it. We are the only users and authors.
In order to improve, where can we get our code reviewed? I don’t mean professionally, just from like-minded individuals.
This is not an answer to your question but it’s tangentially related.
Someone I greatly respected ran an open-source project with the policy of merge everything. Completely flip this idea of carefully review, debate and revise every PR. His theory was that it helps to build an open community, and if something breaks someone else will revert that commit. He says that the main branch was almost always stable, a massive improvement to how it was run previously. He passed several years ago and for some reason this reminded me of him.
I guess what I’m trying to say is if you get something out there that people find useful, the code will be looked at. It doesn’t help you if you’re looking for someone to collaborate sorry.
This is how you get supply chain attacks.
Yes it is. The code is in Bitcoin Core so there’s a non-zero chance you win at life if you want to try.
There is a very effective approach (34:00), that big companies like cloudflare use, to ship a product in a fast and quality way. It bears parallels to what you are describing. In essence engineers should not get hung up in the details to trying to solve everything.
- Just build a proof of concept
- Discard the prototype no matter what and start from scratch keeping the initial feedback in mind
- Build something internally that you yourself will use
- Only once something is good enough and is used internally, then release it to beta.
So that tedious process in trying to flush out all the details before seeing a product (or open source effort) working end to end, might be premature before having the full picture.
That’s Wikipedia’s approach, arguably one of the most successful “open source” projects in history - certainly not without its problems, but overall it’s pretty great
I’ve actually found his blog where he talks about this “optimistic merge”
There’s a number of them as the idea grows. See also the C4 process RFC
mbin (fork of kbin) is currently trying to implement the process.
It’s great to see the attempt and also an example of what the C4 guidelines are made to avoid.
Notice how many comments are little nitpicks about this and that. Completely stalling the commit and getting further away from the original point of C4 which is to reduce contributor friction and avoid these kind of endless discussions on PRs.
I don’t want to be too critical because some of that is a clear lack of understanding of the motivations of C4 which is explained more thoroughly in Pieter’s blog posts. You don’t want to adopt a contributor guidelines that you don’t understand of course.
IMO it’s better just to implement it as-is and start using it in practice rather than bikeshedding.
Feel free to make that comment there.
I would have but I don’t want to tie my Lemmy account to my actual identity :/
This is for 0mq right? I remember reading Pieter Hintjens about this realization he had over a long time of developing 0mq.
Yes that’s right. I was only just transitioning into adulthood and Pieter mentored me and profoundly changed how I view many things. It wasn’t just zeromq but that was the main thing. I still keep his books at hand on my bookshelf. His death impacted me greatly.
Let’s get a community going on programming.dev?
Just created a community request post for it https://programming.dev/post/6295144
If anyone is interested upvote the request and if anyone wants to mod it let me know
Edit: community has been made at [email protected]
Would love something like this, but it’s very hard to get such a community working. Most of the time there will be way more people wanting a review of their code than people wanting to put in the work and review something.
Maybe some kind of point system could work. Like one needs to review at least two times before you can get your work reviewed.
One issue I can see with the points system is that people could just approve it with a “Looks good to me!” without even looking at the code. Or just looking at a small portion of the code.
Also bike shedding. Simple stuff may get many people willing to review / argue about how you should have used A over B, or even C
Big / complicated projects - forget it. Especially if if they would need major refactoring.
Yeah that’s definitely a possibility. It’s really a tricky topic to build a review community. One could try other approaches like dedicated volunteer reviewers. But if you look at exorcism, that will become hard to manage for languages with huge demand.
A system where everyone also needs to review is definitely better. Especially since everyone also learns reading other peoples code.
Could there be a way to award the points proportionally, based on how useful the feedback is?
But who decides that? If the person who requested it can do this, you only need a friend to abuse the system and give each other a few extra points.
Maybe a open point history could help with this and users moderate themselves with this.
with that there would need some way for people to have things to initially review to get the community started. Could do something like after the first 3 posts the point system goes into effect
True, it’s not a fleshed out idea haha. But starting with a few points would work.
Could actually be a cool in it self, wish I had time for something like that.
Something like c sharp acedemy would be kinda cool. Basically do some beginner projects with good code reviews you get a “belt”. That belt allows you to review the beginner projects as you work on intermediate projects. Then you complete those and get another belt that let’s you do advanced projects and review the intermediate projects others submitted.
Kinda sounds like a format similar to Stack Exchange might work
Just remove everything as a duplicate?
There actually is a Stack Exchange community for code reviews: https://codereview.stackexchange.com/
deleted by creator
For those who wants to vote, ^^
There is a Stack Exchange community for code reviews: https://codereview.stackexchange.com/
However they have some strict rules on what is allowed so it may not be suitable for open source projects. Two problematic ones are: 1) The poster must be the author/owner of the code. You, for example, can’t post code that someone else has written. 2) All code that is to be reviewed must be included in the post. You can’t just link to a repository.
I would totally get in on that. Seems almost like a fun OS project itself. Like a volunteer hub type site to just seek reviews
There was something like that on Reddit but I can’t remember the sub. I haven’t seen one here. The best way to get reviews of open source is via contributions to larger projects. That of course doesn’t answer your question directly but is worth noting.
I’d recommend joining some discord or matrix servers for the language you work in or likeminded folk. They tend to have channels for that, and some have really helpful communities.
eta: I just checked my local instance and saw a community with very few subs and zero posts called /c/reviewmycode. Somebody has to post first.
It is quite empty indeed 🤔 [email protected]
don’t put us on the rack and post the link to the git
It ain’t near ready mate. Let me cook.
Not exactly what you want, but plug-ins like sonarlint can nudge you in the right direction.
Removed by mod
Removed by mod
Removed by mod
Removed by mod
