Re: Directors and passing args by value.

The director method generator needs to cast c_result back to
the enum type in the return statement.  This problem arises
because swig maps enums to type 'int' when using them as an
lvalue.  Anyhow, the fix is easy.  As I wrote it though it
isn't enum-specific, so I want to test it on a wide array of 
different return types just to be safe.  Another possibility
is to only add the cast if the return type is an enum.  It
might be better to have the cast in all cases though, even
if it turns out to be mostly redundant.

Mark

On Wed, Nov 19, 2003 at 12:33:31PM +0300, Kerim Borchaev wrote:
> Hello Mark,
> 
> MR> Ok.  I had to change the (python) director method generator slightly
> MR> to handle enum returns, and I want to run some tests before checking
> MR> in the new code.  Good motivation to finally write that typemap 
> MR> regression test I guess.
> 
> I tried to create typemap by myself:
>      
> %typemap(directorout) enum SWIGTYPE
>    "$result = ($1_type) PyInt_AsLong($input);"
> 
> This produces code like this:
> 
> E SwigDirector_C::method() {
>     int c_result ;
>     PyObject *result;
>     
>     if (swig_get_up()) {
>         return C::method();
>     }
>     result = PyObject_CallMethod(swig_get_self(), "method", NULL);
>     c_result = (E) PyInt_AsLong(result);
>     Py_XDECREF(result);
>     return c_result;
> }
> 
> That doesn't compile:
> ... : error C2440: 'return' : cannot convert from 'int' to 'E'
>         Conversion to enumeration type requires an explicit cast
>         (static_cast, C-style cast or function-style cast)
> 
> Could you tell me what your solution is?
>    
> Best regards,
>  Kerim                            mailto:warkid@xxxxxxxxx
> 
> Tuesday, November 18, 2003, 7:44:44 PM, you wrote:
> 
> MR> Cheers,
> MR> Mark
> 
> 
> MR> On Tue, Nov 18, 2003 at 05:10:43PM +0300, Kerim Borchaev wrote:
> >>   I'm sorry - I've mistaken. Passing enum to function is ok.
> >>   Returning - this is what doesn't work.
> >> 
> >> Best regards,
> >>  Kerim                            mailto:warkid@xxxxxxxxx
> >> 
> >> Tuesday, November 18, 2003, 3:46:36 PM, you wrote:
> >> 
> >> KB> Hello.
> >> 
> >> KB>      I've uncommented that "dangerous" typemap;-) It works well except
> >> KB>      it doesn't work with enums.
> >> KB>      What could I do to make it handle them?
> >> 
> >> KB> Thanks.
> >> KB>  Kerim                            mailto:warkid@xxxxxxxxx
> >> 
> >> KB> Tuesday, November 18, 2003, 10:28:21 AM, you wrote:
> >> 
> >> MR>> I actually added the typemap to do this to python.swg many
> >> MR>> months ago, but commented it out with the note "this is rather
> >> MR>> dangerous." ;-)  I probably should have been more explicit at
> >> MR>> that time, since I can't remember now exactly what concerned 
> >> MR>> me about it (other than the possibility of the python method 
> >> MR>> hanging on to references to its arguments).  Maybe there's 
> >> MR>> nothing terribly wrong with that typemap as it stands, or 
> >> MR>> maybe there's another approach that would be safer.  Any 
> >> MR>> typemap/object ownership gurus care to comment?
> >> 
> >> MR>> Mark
> >> 
> >> 
> >> MR>> On Mon, Nov 17, 2003 at 12:08:20PM +0300, Kerim Borchaev wrote:
> >> >>> Hello.
> >> >>> 
> >> >>>    I've checked CVS version and now it seems that directors aren't
> >> >>>    generated for methods with args passed by value. Any plans to
> >> >>>    support it?
> >> >>> 
> >> >>> Best regards,
> >> >>>  Kerim                            mailto:warkid@xxxxxxxxx
> >> >>> 
> >> >>> Thursday, November 13, 2003, 8:16:02 PM, you wrote:
> >> >>> 
> >> >>> MR> Hi Kerim,
> >> >>> 
> >> >>> MR> What version of SWIG are you using?   SWIG cvs parses your
> >> >>> MR> interface without any warnings on my machine, and does the
> >> >>> MR> right thing in the python test.   Note that if you ever get 
> >> >>> MR> warning 461, SWIG will not generate any director stubs for 
> >> >>> MR> the method.  So the result you get in your Python test is 
> >> >>> MR> understandable, and the only question is what is causing the
> >> >>> MR> warning.  
> >> >>> 
> >> >>> MR> Cheers,
> >> >>> MR> Mark
> >> >>> 
> >> >>> 
> >> >>> MR> On Thu, Nov 13, 2003 at 05:55:55PM +0300, Kerim Borchaev wrote:
> >> >>> >> Hello swig,
> >> >>> >> 
> >> >>> >>   SWIG doesn't wrap for directors methods that return objects by
> >> >>> >>   value?
> >> >>> >> 
> >> >>> >> This code:
> >> >>> >> //main.i
> >> >>> >> %module(directors="1") test
> >> >>> >> 
> >> >>> >> %feature("director");
> >> >>> >> %inline %{
> >> >>> >> struct D{
> >> >>> >>     int value;
> >> >>> >>     };
> >> >>> >>     
> >> >>> >> struct C{
> >> >>> >>     virtual ~C(){}
> >> >>> >>     virtual D method(){
> >> >>> >>         D d;
> >> >>> >>         d.value = 7;
> >> >>> >>         return d;
> >> >>> >>         }
> >> >>> >>     };
> >> >>> >>     
> >> >>> >> int getValue(C&o){
> >> >>> >>     D d = o.method();
> >> >>> >>     return d.value;
> >> >>> >>     }
> >> >>> >> %}
> >> >>> >> //////////////////////////
> >> >>> >> 
> >> >>> >> Produces this warning:
> >> >>> >> 
> >> >>> >> main.i:29: Warning(461): Unable to return type D in director method 
> >> >>> >> C::method (s
> >> >>> >> kipping method).
> >> >>> >> 
> >> >>> >> And works like this:
> >> >>> >> 
> >> >>> >> Python 2.3.2 (#49, Oct  2 2003, 20:02:00) [MSC v.1200 32 bit 
> >> >>> >> (Intel)] on win32
> >> >>> >> Type "help", "copyright", "credits" or "license" for more 
> >> >>> >> information.
> >> >>> >> >>> import test
> >> >>> >> >>> from test import *
> >> >>> >> >>> class Derived(C):
> >> >>> >> ...     def method(self):
> >> >>> >> ...             d = D()
> >> >>> >> ...             d.value = 13
> >> >>> >> ...             return d
> >> >>> >> ...
> >> >>> >> >>> getValue(Derived())#should return 13
> >> >>> >> 7
> >> >>> >> 
> >> >>> >> 
> >> >>> >> 
> >> >>> >> Best regards,
> >> >>> >>  Kerim                          mailto:warkid@xxxxxxxxx
> >> >>> >> 
> >> >>> >> 
> >> >>> >> _______________________________________________
> >> >>> >> Swig maillist  -  Swig@xxxxxxxxxxxxxxx
> >> >>> >> http://mailman.cs.uchicago.edu/mailman/listinfo/swig
> >> >>> 
> >> >>> 
> >> 
> >> 
> >> KB> _______________________________________________
> >> KB> Swig maillist  -  Swig@xxxxxxxxxxxxxxx
> >> KB> http://mailman.cs.uchicago.edu/mailman/listinfo/swig
> >> 
> >> 
> >> _______________________________________________
> >> Swig maillist  -  Swig@xxxxxxxxxxxxxxx
> >> http://mailman.cs.uchicago.edu/mailman/listinfo/swig
> 
> 
_______________________________________________
Swig maillist  -  Swig@xxxxxxxxxxxxxxx
http://mailman.cs.uchicago.edu/mailman/listinfo/swig