sligh VS Cwerg

But, you still seem hung up on this, so here’s actual code: https://github.com/amw-zero/sligh/blob/main/lib/codegen.ml.

Of course. Here was the first incarnation: https://github.com/amw-zero/sligh. It has a decent overview of the idea in the readme. To sum it up here, the idea is: have a language built around model-driven development and model-based testing, where you write a simple model of an application, and the implementation and model-based tests are compiled for you. I wrote about the overall model-based testing strategy here. This idea comes from self-certifying compilers that produce proofs of their correctness such as Cogent, but we drop the formality requirement and use property-based testing to compare the implementation and model.

Honorable mention - I’m working on a language with similar goals: Sligh, and I’ve written about why I think it’s such a compelling idea before as well too.

In Sligh, I spent most of the last month introducing a new intermediate representation to make tier splitting (choosing if code should live on the client or server) easier. My goal was to enable derived data, as in a model that queries other models for its data and combines them by processing them in memory. I've been using the example of a personal finance application, so imagine:

The language that I work on is Sligh, and it's out of the bulleted list because it's nowhere near as mature as any of those that I listed, and I'm more of a verification enthusiast vs. expert. Almost all of the ideas in it are borrowed from somewhere else, but I think the one quasi-unique idea is it allows you to write a pure logical description / specification of an application, and it generates full-stack web application code from that.

I’m currently using Pest, though I wouldn’t exactly recommend my compiler as a ‘good example’ just yet because I’m prototyping and just churning code out.

Sligh

So tactically it’s currently a source-to-source compiler, where in the source language (my language) you denote the system state transitions, i.e. by writing create!, update!, etc, and those get compiled to corresponding client and server code in the target language (JS for now, but hoping to support WebAssembly in the future). Heres an example program. The compiler source is there too. I’m hacking it together right now, so it’s not my finest work :D

Perhaps these have already been dealt with and I'm missing critical information. If so, my apologies. Great work, in any case.

[1] https://github.com/robertmuth/Cwerg/tree/master/FrontEnd#dis...

Cwerg has backend that can be used as JIT and is written with readability in mind. Additional documentation can be found here: https://github.com/robertmuth/Cwerg/tree/master/Docs

I defined a basic ISA-like IR for Cwerg. It has unlimited registers and no constraints on immediates.

entire compiler front end ast nodes

I was also going down the path of bike shedding concrete syntax for my language Cwerg before pulling the plug on that effort and just using s-exprs. I managed to make the s-expr quite succinct by carefully choosing the order of arguments so I can omit optional ones. Also very helpful was to use square brackets for list, e.g. (call fun-name [arg1 arg2]). This simplifies parsing a little bit and is easier on the eye. Here are some Code Examples

I am iterating over the languages features for Cwerg's Frontend which aims to be a low level language with about the complexity of C but with some of the comforts of modern languages. I am especially happy with the choice of adding sum types. Relative to C the current feature set looks like this: Removed: * arrays decay to pointers * bitfields * separate compilation (more of a backend issue) * pre-processor * varargs * implcit type conversions * (untagged) unions * ++/-- * comma operator * implicitly nullable pointers * goto