Does Umbraco allow source generator pull requests?

[patrickdemooij9]

I was recently messing around with some source generators as they are new to me, but I think they can bring some performance improvements. However, the code can be a bit difficult to maintain sometimes. Before I start contributing, I just wanted to make sure: Does Umbraco allow pull requests with source generators to be accepted in the codebase?

[Luuk1983]

I dont know if there is an official policy, but personally I think you should at least understand the code and also recognize if the code is of good enough quality. Nothing wrong with AI and code generators per se, but I hate it when people just blindly accept rhe generated code if it works.

[patrickdemooij9]

Are we talking about the same thing? I was talking about: https://devblogs.microsoft.com/dotnet/introducing-c-source-generators/, not AI code generators.

Source generators are very powerful, but you also have to know what you are doing and as I haven't seen any source generators in the Umbraco source code, I just wanted to make sure they are fine about managing it

[Luuk1983]

Are we talking about the same thing? I was talking about: https://devblogs.microsoft.com/dotnet/introducing-c-source-generators/, not AI code generators.

Source generators are very powerful, but you also have to know what you are doing and as I haven't seen any source generators in the Umbraco source code, I just wanted to make sure they are fine about managing it

Ah I see. Never too old to learn I guess, never heard of it! My bad. What would be the benefit? Just curious what use case would benefit from this :)

[patrickdemooij9]

No problem! I don't think most developers would be using source generators as they are mostly used by Microsoft themselves to handle things like endpoint registration or regex attributes.

So, currently it is primarily used on places where you use reflection. I saw in the BaseMapper class, that we try to find the table name and the column name based on the NPoco attributes there. Source generators could move that logic from runtime to compile time, so that the result is already in the code the moment you run the website. So the performance improvement would be the time taken for that reflection to happen runtime.

I am starting with the BaseMapper class as it is quite a small isolated area in Umbraco, but after that I would like to take a look at areas where a lot more reflection is taking place.

[AndyButland]

It's not something I've used myself @patrickdemooij9 so I can't offer any strong opinions on pros and cons, but you'll appreciate that maintenance of the code is a key concern for us. With many contributors, a long history and a fairly complex repository, we have to consider that at a higher priority than you might otherwise in different situations (compared perhaps with a single web application managed by a small team). Of course performance, speed of delivery and other factors are all important too, but generally when it comes down to it, a maintainable solution to a problem is the most valued.

So given that, I'd say it's not something we are really looking for. As I said though, it's not something I'm personally familiar with or as a team we've used before. If you are interested and want to using as a learning exercise yourself with the Umbraco codebase, please feel free. Maybe you'll come up with something there's clearly value in. However you should see it more as a personal experimentation exercise that you might enjoy in and of itself, rather than a contribution we'll be likely to accept.

[patrickdemooij9]

That's fine! That's also the reason why I wanted to ask first. I know there are a lot of libraries out there nowadays that use them for performance (also dotnet themselves), but if there is no knowledge (yet) about in the Umbraco team then I don't think it will be a fit for it even if it brings performance improvements with it. Thanks for your response!