index.html - This HTML code outputs a documentation page fo…

/trunk/Examples/python/pointer/index.html

# · HTML · 171 lines · 137 code · 34 blank · 0 comment · 0 complexity · c125f91aa3cc78fc3e244d781e945ae5 MD5 · raw file


<html>
<head>
<title>SWIG:Examples:python:pointer</title>
</head>

<body bgcolor="#ffffff">

<tt>SWIG/Examples/python/pointer/</tt>
<hr>

<H2>Simple Pointer Handling</H2>

<p>
This example illustrates a couple of techniques for handling
simple pointers in SWIG.  The prototypical example is a C function
that operates on pointers such as this:

<blockquote>
<pre>
void add(int *x, int *y, int *r) { 
    *r = *x + *y;
}
</pre>
</blockquote>

By default, SWIG wraps this function exactly as specified and creates
an interface that expects pointer objects for arguments.  The only
problem is how does one go about creating these objects from a script?

<h2>Possible Solutions</h2>

<ul>
<li>Write some helper functions to explicitly create objects.  For
example:

<blockquote>
<pre>
int *new_int(int ivalue) {
  int *i = (int *) malloc(sizeof(ivalue));
  *i = ivalue;
  return i;
}
int get_int(int *i) {
  return *i;
}

void delete_int(int *i) {
  free(i);
}
</pre>
</blockquote>

Now, in a script you would do this:

<blockquote>
<pre>
a = new_int(37)
b = new_int(42)
c = new_int(0)
add(a,b,c)
r = get_int(c);
print "Result =",r
delete_int(a)
delete_int(b)
delete_int(c)
</pre>
</blockquote>

<p>
<li>Use the SWIG pointer library.  For example, in the interface file 
you would do this:

<blockquote>
<pre>
%include "pointer.i"
</pre>
</blockquote?

and in a script you would do this:

<blockquote>
<pre>
a = ptrcreate("int",37)
b = ptrcreate("int",42)
c = ptrcreate("int")
add(a,b,c)
r = ptrvalue(c)
print "Result =",r
ptrfree(a)
ptrfree(b)
ptrfree(c)
</pre>
</blockquote>

The advantage to using the pointer library is that it unifies some of the helper
functions behind a common set of names.  For example, the same set of functions work
with int, double, float, and other fundamental types.

<p>
<li>Use the SWIG typemap library.  This library allows you to completely
change the way arguments are processed by SWIG.  For example:

<blockquote>
<pre>
%include "typemaps.i"
void add(int *INPUT, int *INPUT, int *OUTPUT);
</pre>
</blockquote>

And in a script:

<blockquote>
<pre>
r = add(37,42)
print "Result =",r
</pre>
</blockquote>
Needless to say, this is substantially easier.

<p>
<li>A final alternative is to use the typemaps library in combination
with the %apply directive.  This allows you to change the names of parameters
that behave as input or output parameters. For example:

<blockquote>
<pre>
%include "typemaps.i"
%apply int *INPUT {int *x, int *y};
%apply int *OUTPUT {int *r};

void add(int *x, int *y, int *r);
void sub(int *x, int *y, int *r);
void mul(int *x, int *y, int *r);
... etc ...
</pre>
</blockquote>

</ul>

<h2>Example</h2>

The following example illustrates the use of these features for pointer
extraction.

<ul>
<li> <a href="example.c">example.c</a>  (C Source)
<li> <a href="example.i">example.i</a>  (SWIG interface)
<li> <a href="example.py">example.py</a> (Python Script)
</ul>

<h2>Notes</h2>

<ul>
<li>Since pointers are used for so many different things (arrays, output values,
etc...) the complexity of pointer handling can be as complicated as you want to
make it.

<p>
<li>More documentation on the typemaps.i and pointer.i library files can be
found in the SWIG user manual.  The files also contain documentation.

<p>
<li>The pointer.i library is designed primarily for convenience.  If you
are concerned about performance, you probably want to use a different
approach.

</ul>

<hr>
</body>
</html>