www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.learn - problems with swig generated code

reply Martin DeMello <martindemello gmail.com> writes:
I'm trying to add a D binding to a C++ project that already has a 
swig definition file; you can see my attempt here:

https://github.com/martindemello/quackle/tree/d/bindings

but trying to run dmd over the generated file errors with:

$ dmd quackle.d
quackle.d(105): Error: undefined identifier SwigPendingException 
in module quackle_im, did you mean struct SwigPendingException?


Looking at quackle_im.d, SwigPendingException is defined via

package struct SwigPendingException {
   ...
}

and used in quackle.d via

  if (quackle_im.SwigPendingException.isPending) throw 
quackle_im.SwigPendingException.retrieve();
Aug 28 2019
parent reply DanielG <simpletangent gmail.com> writes:
Do you know whether SWIG's D generator is even being maintained?

I've searched for it on the forums in the past and got the 
impression that it's outdated.
Sep 01 2019
parent reply Martin DeMello <martindemello gmail.com> writes:
On Sunday, 1 September 2019 at 11:19:11 UTC, DanielG wrote:

 Do you know whether SWIG's D generator is even being maintained?

 I've searched for it on the forums in the past and got the 
 impression that it's outdated.


I didn't realise that :( It was included in the current release 
of swig, so I figured it was maintained.

It's pretty sad if it's not, because trying to access C++ 
libraries directly from D has some limitations (most notably not 
being able to create new C++ objects from D) and swig would have 
let things just work.
Sep 03 2019
next sibling parent Laeeth Isharc <laeeth kaleidic.io> writes:
On Tuesday, 3 September 2019 at 20:03:37 UTC, Martin DeMello 
wrote:

 On Sunday, 1 September 2019 at 11:19:11 UTC, DanielG wrote:

 Do you know whether SWIG's D generator is even being 
 maintained?

 I've searched for it on the forums in the past and got the 
 impression that it's outdated.


 I didn't realise that :( It was included in the current release 
 of swig, so I figured it was maintained.

 It's pretty sad if it's not, because trying to access C++ 
 libraries directly from D has some limitations (most notably 
 not being able to create new C++ objects from D) and swig would 
 have let things just work.


I think DPP can call constructors. YMMV.

We are working on a little project I started as another step in 
the eternal personal hackathon.

Libclang isn't my cup of tea.  It's almost very cool but they put 
in whatever the guy needed and so it's inconsistent but your 
alternative is code that breaks things between breakfast and 
teatime.

Cling is used at CERN and I found libcling more pleasant.  I only 
wrapped the cppyy fork but that actually allows you to reflect at 
runtime on a lot.  From our DSL at work I can include a header, 
instantiate a templated type, create an instance of the class and 
call a method on it.  Can, but it's not what I would call fun.

I didn't yet get time to wrap the rest of the interpreter.

But the idea is to make a tool for wrapping c++ via simpler 
extern (C++).  If one isn't quite sure upfront what will work and 
not then it's much more pragmatic.  I can't see how it won't work 
but we will know in a week or two.

My first version is here.

https://github.com/kaleidicassociates/cpp-reflect-d

There is another route people don't think of.  Calypso can 
introspect on C++ too.  You might not want to use it in 
production but you don't need to.  Either I guess you could use 
it to generate wrappers or you can use it to replace a chunk of 
cpp code.

The surface area of a method must be larger than thr surface area 
of a chunk.  Well chosen then it's easy to write glue code semi 
automatically.  But it's nice to be able to replace a method at a 
time and have working code at every step.

You don't need to trust Calypso in production to be able to use 
it for this purpose.
Sep 03 2019
prev sibling parent Ernesto Castellotti <erny.castell gmail.com> writes:
On Tuesday, 3 September 2019 at 20:03:37 UTC, Martin DeMello 
wrote:

 On Sunday, 1 September 2019 at 11:19:11 UTC, DanielG wrote:

 Do you know whether SWIG's D generator is even being 
 maintained?

 I've searched for it on the forums in the past and got the 
 impression that it's outdated.


 I didn't realise that :( It was included in the current release 
 of swig, so I figured it was maintained.

 It's pretty sad if it's not, because trying to access C++ 
 libraries directly from D has some limitations (most notably 
 not being able to create new C++ objects from D) and swig would 
 have let things just work.


Not true, with the core.stdcpp.new_ module it is possible to 
allocate the memory to create an object directly from D!
Just allocate the memory (YourCPPClass.sizeof) and call 
std.conv.emplace to use the constructor

If you want there is also a wrapper that simplifies the operation 
https://github.com/ErnyTech/CPPNew


A small example:
// test.cpp
#include <iostream>

class Test {
   public:
     Test(int);
     ~Test();
     void set(int);
     void print();
   private:
     int a;
};

Test::Test(int number) {
   this->set(number);
}

Test::~Test() {
   std::cout << "Destructor called" << std::endl;
}

void Test::set(int number) {
   this->a = number;
}

void Test::print() {
   std::cout << this->a << std::endl;
}

// test.d
extern(C++) {
   class Test {
     final this(int);
     final ~this();
     final void set(int);
     final void print();
   }
}

void main() {
   import cppnew : CPPNew;
   import cppnew : CPPDelete;

   auto test = CPPNew!Test(67);
   test.print(); // will print 67
   test.setNumber(12);
   test.print(); // will print 12
   CPPDelete(test); // will print Destructor called
}
Sep 12 2019