www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - ImportC + -H = bindings

reply Kapendev <alexandroskapretsos gmail.com> writes:
I recently made a script/library called 
[Clonoa](https://github.com/Kapendev/clonoa) that generates D 
bindings from C files using ImportC.
The way it works is really simple:

1. Runs: `compiler -o- -H file.c`
2. Parses the `.di` file from the command above and prints a 
cleaned version of it to stdout.

The generated output tries to avoid some redefinitions of Libc 
symbols and it assumes that names starting with `_` are always 
private, so they get skipped.
Overall it does a good job of including the right things.
For example, I was able to create bindings for SDL2 and use them.

To use the script via DUB, you can run:

```
dub run clonoa
```

Or even simpler, copy-paste the script into your own project.
Clonoa is just one D file.
Apr 23
next sibling parent reply Lance Bachmeier <no spam.net> writes:
On Friday, 24 April 2026 at 01:09:28 UTC, Kapendev wrote:

 I recently made a script/library called 
 [Clonoa](https://github.com/Kapendev/clonoa) that generates D 
 bindings from C files using ImportC.
 The way it works is really simple:

 1. Runs: `compiler -o- -H file.c`
 2. Parses the `.di` file from the command above and prints a 
 cleaned version of it to stdout.

 The generated output tries to avoid some redefinitions of Libc 
 symbols and it assumes that names starting with `_` are always 
 private, so they get skipped.
 Overall it does a good job of including the right things.
 For example, I was able to create bindings for SDL2 and use 
 them.

 To use the script via DUB, you can run:

 ```
 dub run clonoa
 ```

 Or even simpler, copy-paste the script into your own project.
 Clonoa is just one D file.


I'm not sure how it compares to clonoa, but I let ImportC do the 
translation of C headers in their entirety using the -Hf switch. 
Now, that isn't going to generate .di files that work, but I have 
a program that parses the files to remove duplicate definitions, 
functions with keywords as names, and whatever else doesn't work. 
It's at the point that it can automatically convert any C header 
I try.
Apr 23
parent reply Kapendev <alexandroskapretsos gmail.com> writes:
On Friday, 24 April 2026 at 02:09:00 UTC, Lance Bachmeier wrote:

 I'm not sure how it compares to clonoa, but I let ImportC do 
 the translation of C headers in their entirety using the -Hf 
 switch. Now, that isn't going to generate .di files that work, 
 but I have a program that parses the files to remove duplicate 
 definitions, functions with keywords as names, and whatever 
 else doesn't work. It's at the point that it can automatically 
 convert any C header I try.


It does the same thing! And yeah, by default the `.di` files will 
not always work because of D keywords or of some other weird 
translations like:

```
alias void foo_func_type(...)
```

The solution for this one is to change the order of things: name 
first and then the function type.
In general you have to clean the `.di` files.

Here are some things that I can control with my script (some 
options are not exposed in the CLI):

```d
struct ClonoaArgs {
     string compiler = defaultCompiler;
     string headerPath;
     string moduleName;
     string[] headerIncludes;
     string[] headerPrefixes;
     string[] opaqueStructs;
     bool removeRepeatedEnums;

     string moduleSymbolHeader = defaultModuleSymbolHeader;
     string indentation = defaultIndentation;
     string[string] typeMap;
     string[] typeSkipList;
     string[] funcSkipList;
     string[] lineSkipList;
}
```
Apr 23
parent reply Lance Bachmeier <no spam.net> writes:
On Friday, 24 April 2026 at 02:56:31 UTC, Kapendev wrote:

 On Friday, 24 April 2026 at 02:09:00 UTC, Lance Bachmeier wrote:

 I'm not sure how it compares to clonoa, but I let ImportC do 
 the translation of C headers in their entirety using the -Hf 
 switch. Now, that isn't going to generate .di files that work, 
 but I have a program that parses the files to remove duplicate 
 definitions, functions with keywords as names, and whatever 
 else doesn't work. It's at the point that it can automatically 
 convert any C header I try.


 It does the same thing! And yeah, by default the `.di` files 
 will not always work because of D keywords or of some other 
 weird translations like:

 ```
 alias void foo_func_type(...)
 ```

 The solution for this one is to change the order of things: 
 name first and then the function type.
 In general you have to clean the `.di` files.

 Here are some things that I can control with my script (some 
 options are not exposed in the CLI):

 ```d
 struct ClonoaArgs {
     string compiler = defaultCompiler;
     string headerPath;
     string moduleName;
     string[] headerIncludes;
     string[] headerPrefixes;
     string[] opaqueStructs;
     bool removeRepeatedEnums;

     string moduleSymbolHeader = defaultModuleSymbolHeader;
     string indentation = defaultIndentation;
     string[string] typeMap;
     string[] typeSkipList;
     string[] funcSkipList;
     string[] lineSkipList;
 }
 ```


This would be a good addition to the compiler if the two of us 
working independently were able to solve the problem. In an 
alternative world where it's worth one's time to propose such 
things, it would make a great addition to the compiler.
Apr 24
parent reply Serg Gini <kornburn yandex.ru> writes:
On Friday, 24 April 2026 at 13:52:48 UTC, Lance Bachmeier wrote:

 This would be a good addition to the compiler if the two of us 
 working independently were able to solve the problem. In an 
 alternative world where it's worth one's time to propose such 
 things, it would make a great addition to the compiler.


I agree. This addition to the compiler functionality of ImportC 
could be useful.
Apr 24
parent reply jmh530 <john.michael.hall gmail.com> writes:
On Friday, 24 April 2026 at 14:28:14 UTC, Serg Gini wrote:

 On Friday, 24 April 2026 at 13:52:48 UTC, Lance Bachmeier wrote:

 This would be a good addition to the compiler if the two of us 
 working independently were able to solve the problem. In an 
 alternative world where it's worth one's time to propose such 
 things, it would make a great addition to the compiler.


 I agree. This addition to the compiler functionality of ImportC 
 could be useful.


If all the kinks get worked out and people start making a lot of 
use of it, then it's an easier sell to get in compiler.
Apr 24
next sibling parent reply Kapendev <alexandroskapretsos gmail.com> writes:
On Friday, 24 April 2026 at 14:30:20 UTC, jmh530 wrote:

 On Friday, 24 April 2026 at 14:28:14 UTC, Serg Gini wrote:

 On Friday, 24 April 2026 at 13:52:48 UTC, Lance Bachmeier 
 wrote:

 This would be a good addition to the compiler if the two of 
 us working independently were able to solve the problem. In 
 an alternative world where it's worth one's time to propose 
 such things, it would make a great addition to the compiler.


 I agree. This addition to the compiler functionality of 
 ImportC could be useful.


 If all the kinks get worked out and people start making a lot 
 of use of it, then it's an easier sell to get in compiler.


Yeah, and I think that the compiler doesn't need to do everything 
the script does. One QoL things would be to automatically change 
names like `alias` to `alias_` when using the `-H` flag so it's 
easier for people to modify the output.
Apr 24
parent Kapendev <alexandroskapretsos gmail.com> writes:
On Friday, 24 April 2026 at 15:40:13 UTC, Kapendev wrote:

 Yeah, and I think that the compiler doesn't need to do 
 everything the script does. One QoL things would be to 
 automatically change names like `alias` to `alias_` when using 
 the `-H` flag so it's easier for people to modify the output.


Or maybe not. C is weird sometimes lol:

```c
struct debug {
     int value;
};

void debug(struct debug debug) {
     debug.value += 1;
}

void main() {
     struct debug d = {0};
     debug(d);
}
```
Apr 24
prev sibling parent Lance Bachmeier <no spam.net> writes:
On Friday, 24 April 2026 at 14:30:20 UTC, jmh530 wrote:

 On Friday, 24 April 2026 at 14:28:14 UTC, Serg Gini wrote:

 On Friday, 24 April 2026 at 13:52:48 UTC, Lance Bachmeier 
 wrote:

 This would be a good addition to the compiler if the two of 
 us working independently were able to solve the problem. In 
 an alternative world where it's worth one's time to propose 
 such things, it would make a great addition to the compiler.


 I agree. This addition to the compiler functionality of 
 ImportC could be useful.


 If all the kinks get worked out and people start making a lot 
 of use of it, then it's an easier sell to get in compiler.


The issue with this is that the compiler currently gives 
incorrect output that you cannot use without further 
transformation. It's not about adding a feature to the compiler, 
it's about fixing an existing feature that doesn't work. The 
truth is that it will be added if the right person makes the 
request, not if there is evidence that it is being used.
Apr 24
prev sibling parent reply Serg Gini <kornburn yandex.ru> writes:
On Friday, 24 April 2026 at 01:09:28 UTC, Kapendev wrote:

 I recently made a script/library called 
 [Clonoa](https://github.com/Kapendev/clonoa) that generates D


Have you fixed all the bugs?
Apr 24
parent Kapendev <alexandroskapretsos gmail.com> writes:
On Friday, 24 April 2026 at 09:54:55 UTC, Serg Gini wrote:

 On Friday, 24 April 2026 at 01:09:28 UTC, Kapendev wrote:

 I recently made a script/library called 
 [Clonoa](https://github.com/Kapendev/clonoa) that generates D


 Have you fixed all the bugs?


There will be some, but most of the bad and easy to find ones are 
fixed.
For example, the type map now doesn't replace the name `PROFILE` 
with `PROstdc.stdio.FILE` lol.
Apr 24