I'm not necessarily saying this is good code to merge as it is right now, but I think the implementation details here are useful for starting a structured discussion about what this should look like. And once we settle on something and implement it, we can merge that.
In particular, there are two details to notice. Firstly, there's the lustre/dev/tools.{run} entrypoint: is this the API we want for specifying hooks in some sort of /test/dev.gleam code? Possibly, but it definitely deserves discussion. Secondly, there's the architectural point of the build step hook, which serves as a nice problem-case for future hooks: trying to be as noninvasive as possible, you can see I still had to touch many different things to add it, and included some mildly concerning code duplication. We should discuss whether (and how) to change the library architecture with hooks in mind.
I'm not necessarily saying this is good code to merge as it is right now, but I think the implementation details here are useful for starting a structured discussion about what this should look like. And once we settle on something and implement it, we can merge that.
In particular, there are two details to notice. Firstly, there's the
lustre/dev/tools.{run}entrypoint: is this the API we want for specifying hooks in some sort of/test/dev.gleamcode? Possibly, but it definitely deserves discussion. Secondly, there's the architectural point of the build step hook, which serves as a nice problem-case for future hooks: trying to be as noninvasive as possible, you can see I still had to touch many different things to add it, and included some mildly concerning code duplication. We should discuss whether (and how) to change the library architecture with hooks in mind.