Le ven. 29 janv. 2021 à 16:58, luca regini luca.regini@gmail.com a écrit :
I would also be willing to contribute more to Gambit but the hard fact is that this requires a big investment of time upfront.
Getting into the internals without specific documentation probably requires to invest at least 2 weeks full-time upfront.
Which is not impossible if one is fully dedicated to this project ( as a student can be ) but for people with a job and a family this is impossible. I can dedicate time in bursts of 1 hours maybe 2 hours max.
An idea that I could suggest is to write a commercial book describing the implementation so to sponsor some additional work on Gambit. I would be willing to buy such a book but have no idea how many other would be willing to spend money on that.
I agree I would buy a book about gambit compiler, and other internal aspects (like how macro works).
FWIW, my plan WAS to build my own scheme that targets javascript with Chez as host. Among other things because it is difficult, but also because gambit is great, and I want to take part in something bigger than myself.
I do my best to document what I am doing with gambit (https://github.com/pre-srfi/webui/), and try to workaround some behaviors that are not perfect or at least could be improved upon. I can work around those.
But more important than all of the above. I agree that the compiler or compilers must be documented. It might be the first and most important priority to help gambit grow as a community. nanopass is overrated. It seems to me based on a few couple hours of hacking that it is possible to build something simpler and easier to use. Again, I am not a compiler specialist. Maybe a nanopass is a dead end for gambit, or maybe gambit approach is better. Who knows?
Marc, you are in the best position to describe the compiler.