Menu
▾
▴
Re: [Swig-user] how to wrap double ** in a struct ? - double * is ok
|
From: fmotif <fm...@ya...> - 2006-11-09 23:52:23
|
Hi William
I compiled the examples...
I saw the verbose from the makefile getting out the -s
option ...
My apologies, but I can´t undarstand how gifplot.h was
wrapped, I don´t see in the interface file
Examples/interface/gifplot.i how "double **" was been
wrapped.
I did:
make check-java-gifplot
The output:
-----------------------------
make -f ../../../Makefile SRCS=''
SWIG='/u/swig/swig-1.3.29/swig' \
INCLUDES='-I../../Include' LIBS='-L../.. -lgifplot'
SWIGOPT='-I../../Include -noproxy' \
TARGET='gifplot' INTERFACE='gifplot.i' java
/u/swig/swig-1.3.29/swig -java -I../../Include
-noproxy gifplot.i
cc -c gifplot_wrap.c -I../../Include
-I/usr/jdk/instances/jdk1.5.0/include
-I/usr/jdk/instances/jdk1.5.0/include/solaris
ld -G gifplot_wrap.o -L../.. -lgifplot -o
libgifplot.so
javac *.java
--------------------------
it looks like magic !!!
Fabio
--- William S Fulton <ws...@fu...>
escreveu:
> Run:
>
> make check-java-gifplot in the top level directory.
>
> Only works on unix.
>
> William
>
> fmotif wrote:
> > Hi William
> >
> > I saw in the gifplot.h example a struct that uses
> > double **, but I didn´t understand how it was
> wrapped
> > in the gifplot.i. It seems me it was bypassed. (At
> the
> > moment I have problems with examples building -
> may be
> > because I´m not administrator - just I can´t "make
> > install")
> >
> > ---------
> > Gifplot.h
> > ---------
> > typedef unsigned char Pixel;
> > typedef float Zvalue;
> >
> > typedef struct FrameBuffer {
> > Pixel **pixels;
> > Zvalue **zbuffer;
> > unsigned int height;
> > unsigned int width;
> > int xmin;
> > int ymin;
> > int xmax;
> > int ymax;
> > ...
> > }
> > ---------
> > Gifplot.i
> > ---------
> > typedef unsigned char Pixel;
> > typedef float Zvalue;
> >
> > typedef struct FrameBuffer {
> > unsigned int height;
> > unsigned int width;
> > int xmin;
> > int ymin;
> > int xmax;
> > int ymax;
> > ...
> > }
> >
> > This Zvalue seems have the same purpose of the
> mine Z.
> >
> > At the moment I already work with a struct that
> > defines a single pointer;
> >
> > This is my test code:
> > ------
> >
> > SWIGTYPE_p_StructGravacaoXyz xyz =
> > Baseproj.new_StructGravacaoXyz();
> >
> > SWIGTYPE_p_double p_x =
> Baseproj.new_doubleArray(2);
> >
> > Baseproj.doubleArray_setitem(p_x,0, 1.1);
> > Baseproj.doubleArray_setitem(p_x,1, 1.2);
> >
> > double a[] = {0,0};
> >
> > a[0] = Baseproj.doubleArray_getitem(p_x,0);
> > a[1] = Baseproj.doubleArray_getitem(p_x,1);
> > ----
> >
> > Thanks
> > Fabio
> > Rio de Janeiro
> >
> > --- William S Fulton <ws...@fu...>
> > escreveu:
> >
> >> For your short *reclass, you can use
> arrays_java.i
> >> as follows:
> >>
> >> %include <arrays_java.i>
> >>
> >> /*
> >> The default typemap which needs to be overridden
> as
> >> the interface does
> >> not indicate how big the C array is.
> >> %typemap(out) short[]
> >> %{$result = SWIG_JavaArrayOutShort(jenv, $1,
> >> FillMeInAsSizeCannotBeDeterminedAutomatically);
> %}
> >> */
> >>
> >> // replace 16 with whatever the array size is
> >> %typemap(out) short[]
> >> %{$result = SWIG_JavaArrayOutShort(jenv, $1, 16);
> %}
> >>
> >> // apply short[] typemaps from arrays_java.i
> >> %apply short[] { short *reclas }
> >>
> >>
> >> %inline %{
> >> typedef struct
> >> {
> >> short *reclas;
> >> } XYZ;
> >> %}
> >>
> >> Basically you are using the typemaps in
> >> arrays_java.i and tweaking a bit
> >> to fill in the missing information.
> >>
> >> As for your double** type, I don't recall ever
> >> seeing any 2D array
> >> typemaps. You would need to take some of the
> >> typemaps in arrays_java.i
> >> and tweak for 2D arrays.
> >>
> >> Hope that helps.
> >>
> >> William
> >>
> >> fmotif wrote:
> >>> William
> >>>
> >>> Thanks for the answer...
> >>>
> >>> The pointers defined into the struct XYZ
> >> represent
> >>> arrays, the double pointer (double **) is a
> matrix
> >> 2D
> >>> and (char **) represents an array of strings.
> >>>
> >>> To wrap the array of strings i´m using the
> >> STRING
> >>> pattern, but about the others, any one have an
> >> example
> >>> with a case looks like this, but using only C
> >> (without
> >>> C++).
> >>>
> >>> I looked at arrays_java.i, but I need an easy
> >>> example ...
> >>>
> >>> Fabio
> >>>
> >>>
> >>>> fmotif wrote:
> >>>>> Hi
> >>>>>
> >>>>> at the moment i?m not expert with swig
> scripting
> >>>>> language...
> >>>>>
> >>>>> I?m trying to generate wrappers for JAVA to
> this
> >>>>> srtuct
> >>>>>
> >>>>> typedef struct
> >>>>> {
> >>>>> short *reclas;
> >>>>> char **siglas;
> >>>>> double *X;
> >>>>> double *Y;
> >>>>> double **Z;
> >>>>> char **descrZs;
> >>>>> long numZs;
> >>>>> long numSiglas;
> >>>>> } XYZ;
> >>>>>
> >>>>> the problems are the double *, double ** and
> >> short
> >>>> *,
> >>>>> that generates: SWIG_p..., I tryied resolve
> the
> >>>> short
> >>>>> * with:
> >>>>>
> >>>>> %include typemaps.i
> >>>>> %apply short *INPUT {short *};
> >>>>> typedef struct { ...
> >>>>> }
> >>>>>
> >>>>> but the java code doesn?t compile
> >>>>>
> >>>>> thanks
> >>>>>
> >>>> You need to first say how you want to marshal
> >> these
> >>>> types for Java, eg
> >>>> is reclas a pointer to a single element or an
> >> array?
> >>>> The same for all
> >>>> the pointers in your XYZ class above. The
> >> approach
> >>>> to wrap this would be
> >>>> different for each. Look at arrays_java.i for
> >>>> arrays.
> >>>>
> >>>> William
> >>>>
> >>>>
> >>>>
> >>>> ------------------------------
> >>>>
> >>>> Message: 6
> >>>> Date: Fri, 3 Nov 2006 15:57:50 -0600
> >>>> From: robert engels <re...@ix...>
> >>>> Subject: Re: [Swig-user] unsigned char pointer
> as
> >>>> byte[] in Java ?
> >>>> To: William S Fulton <ws...@fu...>
> >>>> Cc: swi...@li...
> >>>> Message-ID:
> >>>>
> >>
> <E63...@ix...>
> >>>> Content-Type: text/plain; charset=US-ASCII;
> >>>> delsp=yes; format=flowed
> >>>>
> >>>> Thank you very much. I will test it out.
> >>>>
> >>>> I confess I don't know SWIG at all (the mapping
> >>>> documents related to
> >>>> java were not very clear - at least to me). I
> am
> >>>> attempting to help
> >>>> out with some other projects and they used it
> for
> >>>> JNI integration. I
> >>>> have started to work on a custom JNI
> >> implementation
> >>>> because I could
> >>>> not get this to work.
> >>>>
> >>>> Robert
> >>>>
> >>>> On Nov 3, 2006, at 3:54 PM, William S Fulton
> >> wrote:
> >>>>> robert engels wrote:
> >>>>>> Hello all,
> >>>>>>
> >>>>>> I have a C struct that I am trying to use
> from
> >>>> Java, accessing the
> >>>>>> unsigned char * as a byte[]. (It seems that
> >>>> arrays_java.i wants to
> >>>>>> access this as a short[] for some reason).
> >>>>>>
> >>>>> It is the most appropriate type that covers
> all
> >>>> possible values of an
> >>>>> unsigned char array.
> >>>>>
> >>>>>> struct mystruct {
> >>>>>> unsigned char *mydata;
> >>>>>> int mydatalen;
> >>>>>> };
> >>>>>>
> >>>>>> I looked at various.i and saw the 'accessing
> a
> >>>> char* as a byte[], but
> >>>>>> when I create the following for an unsigned
> >> char
> >>>> *
> >>>>>> %typemap(jni) (unsigned char *BYTE)
> >> "jbyteArray"
> >>>>>> %typemap(jtype) (unsigned char *BYTE)
> "byte[]"
> >>>>>> %typemap(jstype) (unsigned char *BYTE)
> "byte[]"
> >>>>>> %typemap(in) unsigned char *BYTE { $1 =
> >> (unsigned
> >>>> char *)
> >>>>>> JCALL2(GetByteArrayElements, jenv, $input,
> 0);
> >>>>>> }
> >>>>>> %typemap(argout) unsigned char *BYTE {
> >>>>>> JCALL3(ReleaseByteArrayElements, jenv,
> >>>> $input, (jbyte *) $1, 0);
> >>>>>> }
> >>>>>> %typemap(javain) (unsigned char *BYTE)
> >>>> "$javainput"
> >>>>>> /* Prevent default freearg typemap from being
> >>>> used */
> >>>>>> %typemap(freearg) (unsigned char *BYTE) ""
> >>>>>>
> >>>>>> and then try to apply it using:
> >>>>>>
> >>>>>> %apply unsigned char *BYTE { (unsigned char
> >>>> *mydata) };
> >>>>>> it generates the following code
> >>>>>>
> >>>>>> public void setMydata(byte[] value) {
> >>>>>>
> MySwigTestJNI.mystruct_mydata_set(swigCPtr,
> >>>> value);
> >>>>>> }
> >>>>>>
> >>>>>> public byte[] getMydata() {
> >>>>>> long cPtr =
> >>>> MySwigTestJNI.mystruct_mydata_get(swigCPtr);
> >>>>>> return (cPtr == 0) ? null : new
> >>>> SWIGTYPE_p_unsigned_char(cPtr,
> >>>>>> false);
> >>>>>> }
> >>>>>>
> >>>>>> which does not compile
> >>>>>>
> >>>>>> mystruct.java:41: incompatible types
> >>>>>> found : byte[]
> >>>>>> required: long
> >>>>>> long cPtr =
> >>>> MySwigTestJNI.mystruct_mydata_get(swigCPtr);
> >>>>>>
>
> >>>> ^
> >>>>>> mystruct.java:42: incompatible types
> >>>>>> found :
> >> net.sf.libusb.SWIGTYPE_p_unsigned_char
> >>>>>> required: byte[]
> >>>>>> return (cPtr == 0) ? null : new
> >>>> SWIGTYPE_p_unsigned_char(cPtr,
> >>>>>> false);
> >>>>>> ^
> >>>>>> I have attached a complete testcase with
> >>>> makefile. I am sure it is
> >>>>>> something simple, but I can't seem to find
> it.
> >>>>>>
> >>>>> You are missing the javaout typemap. You
> >> probably
> >>>> need:
> >>>>> %typemap(javaout) (unsigned char *BYTE) {
> >>>>> return $jnicall;
> >>>>> }
> >>>>>
> >>>>> Also your freearg typemap has a typo for the
> >> type,
> >>>> it should be
> >>>>> "unsigned char" not "unsignedchar".
> >>>>>
> >
> >
> >
> > __________________________________________________
> > Fale com seus amigos de graça com o novo Yahoo!
> Messenger
> > http://br.messenger.yahoo.com/
> >
>
>
__________________________________________________
Fale com seus amigos de graça com o novo Yahoo! Messenger
http://br.messenger.yahoo.com/
|