Re: [Swig-devel] Output typemaps not properly applied in Java? (cvs)

[William S F. <ws...@fu...>]

John Koleszar wrote:
> Hi all,
> 
> I have a function GetBool(t_bool *) that swig is mapping (in java) to 
> GetBool(short []). This same typemap file works for perl, python, and 
> php, so it seems like it's a java thing. Seems like Swig should really 
> be generating some sort of mutable integer wrapper class and using that 
> instead. Currently I have to do something like:
>    short x[] = {1};
>    test.GetBool(x);
>    if(x[0]) do_stuff();
> 
> which isn't very nice syntactically. Does anyone have a workaround for 
> this? Does this problem exist in 1.3.27?
> 
You are specifically asking for this approach by using typemaps.i and 
%apply. If you can think of a better way to map a primitive output using 
classes in the JDK, let us know. Or you can write your own mutable 
number wrapper class as per this in the docs:

http://www.swig.org/Doc1.3/Java.html#using_typemaps_return_arguments


> On another syntax note, and this is more of a feature request than a bug 
> report, if I have a typedef like:
>    typedef void* handle;
> it'd be nice to be able to reference that type as "handle" or similar 
> rather than "SWIGTYPE_p_void". This would provide a little extra type 
> safety if you've got multpile types that alias void*.
> 
> John
> 
> === test.i ===
> %include "typemaps.i"
> %module test
> %{
> typedef unsigned char  t_u8;
> typedef t_u8           t_bool;
> static t_bool priv_bool = 0;
> 
> int GetBool(t_bool *b) {
>   *b=priv_bool;
>   printf("C: Returning %d\n",*b);
>   return 0;
> };
> 
> int SetBool(t_bool b) {
>   printf("C: Setting %d\n",b);
>   priv_bool = b;
>   return 0;
> };
> 
> %}
> 
> %apply unsigned char *OUTPUT { unsigned char * }
> 
> typedef unsigned char  t_u8;
> typedef t_u8           t_bool;
> int GetBool(t_bool *b);
> int SetBool(t_bool b);
> 
> === Generated test.java (1.3.28) ===
> public class test {
>  public static int GetBool(short[] b) {
>    return testJNI.GetBool(b);
>  }
> 
>  public static int SetBool(short b) {
>    return testJNI.SetBool(b);
>  }
> }
> 

With regard to the void* typedef, all types are fully resolved as they 
are in C++. The implementation is basically doing what the coder 
originally intended, eg

void h(handle x);
void v(void *x);

void foo() {
   void* p = new int(10);
   h(p);
   v(p);
}

That is a void* is exactly the same as a handle. Typedefs don't add in 
extra type safety. If you want extra type safety, it should be addressed 
at the C++ level. If you want to add in extra typesafety in your Java 
wrappers that is not present in your C++ layer, you'd need to mess with 
some typemaps. Try this:

// modified slightly from the SWIGTYPE* typemaps in java.swg
%typemap(jstype) handle "handle"
%typemap(javain) handle "handle.getCPtr($javainput)"
%typemap(javaout) handle {
     long cPtr = $jnicall;
     return (cPtr == 0) ? null : new handle(cPtr, $owner);
   }

and copy the SWIGTYPE_p_void class to handle.java and change all 
'SWIGTYPE_p_void' to 'handle' within this class only. If you have other 
code which uses void*, then from Java you'll have to use 
SWIGTYPE_p_void, which is what you want, or you can add in more typemaps 
like above.

William