A tool that does the boring parts of translating C to D.

Try it in your browser!

YouTube: DConf Online '22 - Translating C to D

From the root of this repository:

dub run --arch=x86_64 -- ../test/main.c

Arguments are .c or .h files, for which it outputs a translated .d file next to it. This .d file may not compile yet, but it's supposed to get you 90% there. The remaining errors are usually because of non-trivial macros, initializers, or because of D's stricter type system.

You can copy the built binary to a folder in your PATH, for example on linux:

dub build
sudo cp build/ctod /usr/local/bin/

This allows you to simply invoke ctod from anywhere.

This was written before ImportC. Statically linking C libraries with dub is clumsy: while there's usually C bindings in the package repository, you still need to grab release binaries for each build target / configration, include them in the repo, and fiddle with dub flags until there are no linker errors anymore.

That's why I like to translate C code to D. While this requires some intelligence, a lot of the work is simple syntactic find-and-replace actions, and converting declaration syntax. Even with VIM macros it was getting tedious, so I wrote this tool. I used it for translating glfw and libsoundio, as well as some single-file C libs.

The tree-sitter parser for C is used to parse C code, including macros. tree-sitter is a parser generator with a focus on incremental parsing and error recovery, useful for IDE tools. Error recovery is particularly useful for this tool: it always gives a best-effort translated output, even when it encounters what it thinks is a parse error. If you want, you can convert broken C code to similarly broken D code.

This tool makes use of the fact that in general, if D code looks like C, it either behaves the same as C or doesn't compile. Also, D has almost all features that C has. The tool doesn't need a deep understanding of C code, and a lot of things are passed through without any change:

// Identical in both C and D
float fabs(float x) {
    if (x < 0.0)
        return -x;
    return x;
}

However, there's also a lot of things that are different between C and D. This ranges from simple syntax changes:

• x->y => x.y
• (float) 1lu => cast(float) 1Lu
• "con" "cat" => "con" ~ "cat"

To more difficult translations, such as declarations with complex types:

char *(*(*fun)[5])(int);

D:

char* function(int)[5]* fun;

It's being developed like this:

• Run ctod on a C file
• Look at the output, notice something that needs to be adjusted to be valid D
• Enter the C code in the tree-sitter playground to see what it looks like in the AST
• Add code to recognize and translate the pattern
• Repeat for as many patterns as possible

So this is really just a big find-and-replace tool that operates on an AST provided by tree-sitter.

You can take a look at testcases.md to see the capabilities of ctod.

If you only need to translate C headers and not actual code, there are also these tools available:

• dpp
• dstep
• ohmygentool

Those tools use libclang (the LLVM C parser api) to parse C instead of tree-sitter.

• Programming in D for C Programmers