[feature] self-contained steel program executables about steel

[mattwparas]

Whipped up a proof of concept in #113 - I will continue to iterate on it and update the progress here

[mattwparas]

I do have a work in progress statically linked program representation that can be serialized to bytes - I guess the workflow could be something like this:

Rather than calling something like steel main.scm - you'd be able to do something like:

steel build main.scm (or something like that)

That would then generate a main.rs file, with the whole program embedded statically in the file alongside an engine instance to run it - we then just compile this to a binary.

I think that would get the job done?

Another approach that I've experimented with is generating the entire VM + handlers dynamically specific to the program that you've given, but that requires some more work :) (but would be a cool future goal)

[kskarthik]

That would then generate a main.rs file, with the whole program embedded statically in the file alongside an engine instance to run it - we then just compile this to a binary.

This is interesting too!

My idea is more like https://deno.com/[email protected]/tools/compiler#compiling-executables

[mattwparas]

I'll have to go look at the deno implementation, my hunch is its doing something more or less like what I've described though. Of course, I could be totally wrong

[kskarthik]

a quick search showed https://github.com/denoland/deno/blob/main/cli/tools/compile.rs

[mattwparas]

Nice, looks like they had discussed my approach but didn't end up landing on it. Would it be a deal breaker if it required the rust tool chain? It would certainly make the implementation trivial

[kskarthik]

Would it be a deal breaker if it required the rust tool chain? It would certainly make the implementation trivial

At this point, I am not sure 😄