(NOTE: This project is currently "sleeping" - I'm too optimistic to call it dead)
Cx - C' (C Prime) was unfortunately taken, and even though the project is dead, I don't want to steal the name
Vala - Coolest idea ever, but has some downsides: Unix-centric, Gnome-centric
I want to create yet another alternative to C as a systems programming language, so my idea is Cx:
- Fully Open Source
- Compiles to plain C (as terse yet readable as possible [ideas should be terse, identifiers readable])
- Generates the smallest and most equivalent C Code possible
- One thing we need for this is layered functionality. As an example, if a programmer declares an int, it should just be an "int" if we never do anything super fancy. Or we could use extension methods to extend basic types.
- Native support on Linux, Windows, and Mac OS X
- Small, efficient runtime built in to emitted C code
- Runtime can be disabled, resulting in reduced feature set but faster code
- Backward compatible with Vala and C. Can write C anywhere and it will Do The Right Thing - UPDATE: This may cause more trouble than it's worth
- Produces small executables. If using the V# Libraries, only pulls in the bare minimum it needs ("using System" doesn't pull in 8MB of crap)
- Originally written in vala but should be self-hosting.
- Should be able to compile V# and use the native libraries anywhere there's a C compiler and a POSIX subsystem. Windows will be treated specifically at the start, so there are no issues with that platform
- Doesn't throw out the baby with the bath water. We have enough "safe" languages. Cx is defined for real programmers who do real programming. Among the examples of this:
- statements can go in if statements. If you want to do if ((p = alloc_new_p()) != null) { // ... } you can. Note that if statements must still evaulate to bool, so you'll have to do the explicit comparison, but you don't have to put the assignment on another line
- case statements can fall through with no issues. If you miss a break, that's your fault
- you can drop into pointers if you need
- type punning will still be possible, because struct and class layouts will be well-defined and reliable
- the general guiding principle is: safe by default, but no straightjackets
Compactness:
I read an article today about D. It was a counterpoint to an earlier piece arguing in favor of the language. One of the best things out of the article was its argument against the assertion that D was the next "systems programming language." The point was made that you would never think of writing a kernel in D or using it for embedded systems, because the executables it produces are much too large.
This, to me, re-emphasizes one of the primary goals of Cx: stay close to C. The point of Cx is to bring a modern feel and modern capabilities to C, without all the baggage of C++. But, in general, it's extremely important to structure the runtime of Cx to be as small as possible, and to only pull in the bare minimum requested by the programmer.