<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<!-- Hand crafted HTML -->
<html>
<head>
<title>SWIG and PHP</title>
<link rel="stylesheet" type="text/css" href="style.css">
</head>
<body bgcolor="#ffffff">
<H1><a name="Php"></a>32 SWIG and PHP</H1>
<!-- INDEX -->
<div class="sectiontoc">
<ul>
<li><a href="#Php_nn1">Generating PHP Extensions</a>
<ul>
<li><a href="#Php_nn1_1">Building a loadable extension</a>
<li><a href="#Php_nn1_3">Using PHP Extensions</a>
</ul>
<li><a href="#Php_nn2">Basic PHP interface</a>
<ul>
<li><a href="#Php_nn2_1">Constants</a>
<li><a href="#Php_nn2_2">Global Variables</a>
<li><a href="#Php_nn2_3">Functions</a>
<li><a href="#Php_nn2_4">Overloading</a>
<li><a href="#Php_nn2_5">Pointers and References</a>
<li><a href="#Php_nn2_6">Structures and C++ classes</a>
<ul>
<li><a href="#Php_nn2_6_1">Using <tt>-noproxy</tt></a>
<li><a href="#Php_nn2_6_2">Constructors and Destructors</a>
<li><a href="#Php_nn2_6_3">Static Member Variables</a>
<li><a href="#Php_nn2_6_4">Static Member Functions</a>
</ul>
<li><a href="#Php_nn2_7">PHP Pragmas, Startup and Shutdown code</a>
</ul>
<li><a href="#Php_nn3">Cross language polymorphism</a>
<ul>
<li><a href="#Php_nn3_1">Enabling directors</a>
<li><a href="#Php_nn3_2">Director classes</a>
<li><a href="#Php_nn3_3">Ownership and object destruction</a>
<li><a href="#Php_nn3_4">Exception unrolling</a>
<li><a href="#Php_nn3_5">Overhead and code bloat</a>
<li><a href="#Php_nn3_6">Typemaps</a>
<li><a href="#Php_nn3_7">Miscellaneous</a>
</ul>
</ul>
</div>
<!-- INDEX -->
<p>
SWIG supports generating wrappers for PHP5. Support for PHP4 was removed
in SWIG 1.3.37. The PHP developers are no longer making new PHP4 releases,
and won't even be patching critical security issues after 2008-08-08, so it
doesn't make much sense for SWIG to continue to support PHP4 now. If you
really need to continue to use PHP4, just stick with SWIG 1.3.36.
</p>
<p>
Currently any PHP5 release should work, but we don't regularly test with
PHP < 5.3.
</p>
<p>
In this chapter, we discuss SWIG's support of PHP. The PHP module
was extensively rewritten in release 1.3.26, and support for generating
OO wrappers for PHP5 was added in 1.3.30. The PHP module now supports most
of the features available in some of the other languages.
</p>
<p>
In order to use this module, you will need to have a copy of the PHP5
include files to compile the SWIG generated files. If you installed
PHP from a binary package, you may need to install a "php-dev" or "php-devel"
package for these to be installed. You can find out where these files are
by running <tt>php-config --includes</tt>. To use the built PHP module you
will need either the php binary or the Apache php module. If you want to build
your extension into php directly, you will need the complete PHP source tree
available.
</p>
<H2><a name="Php_nn1"></a>32.1 Generating PHP Extensions</H2>
<p>
To build a PHP extension, run swig using the <tt>-php</tt> option as
follows:
</p>
<div class="code"><pre>
swig -php example.i
</pre></div>
<p>
This will produce 3 files example_wrap.c, php_example.h and
example.php. The first file, <tt>example_wrap.c</tt> contains all of
the C code needed to build a PHP extension. The second file,
<tt>php_example.h</tt> contains the header information needed if
you wish to statically link the extension into the php interpreter.
The third file,
<tt>example.php</tt> can be included by PHP scripts. It attempts to
dynamically load the extension and contains extra php code specified
in the interface file. If wrapping C++ code with PHP classes, it will
also contain PHP5 class wrappers.
</p>
<p>
SWIG can generate PHP extensions from C++ libraries as well when
given the <tt>-c++</tt> option. The support for C++ is discussed in
more detail in <a href="#Php_nn2_6">section 27.2.6</a>.
</p>
<p>
The usual (and recommended) way is to build the extension as a separate
dynamically loaded module (which is supported by all modern operating
systems). You can then specify that this be loaded
automatically in <tt>php.ini</tt> or load it explicitly for any script which
needs it.
</p>
<p>
It is also possible to rebuild PHP from source so that your module is
statically linked into the php executable/library. This is a lot more
work, and also requires a full rebuild of PHP to update your module,
and it doesn't play nicely with package system. We don't recommend
this approach, or provide explicit support for it.
</p>
<H3><a name="Php_nn1_1"></a>32.1.1 Building a loadable extension</H3>
<p>
To build your module as a dynamically loadable extension, use compilation
commands like these (if you aren't using GCC, the commands will be different,
and there may be some variation between platforms - these commands should at
least work for Linux though):
</p>
<div class="code"><pre>
gcc `php-config --includes` -fpic -c example_wrap.c
gcc -shared example_wrap.o -o example.so
</pre></div>
<H3><a name="Php_nn1_3"></a>32.1.2 Using PHP Extensions</H3>
<p>
To test the extension from a PHP script, you need to load it first. You
can load it for every script by adding this line the <tt>[PHP]</tt> section of
<tt>php.ini</tt>:
</p>
<div class="code"><pre>
extension=/path/to/modulename.so
</pre></div>
<p>
Alternatively, you can load it explicitly only for scripts which need it
by adding this line:
</p>
<div class="code"><pre>
dl("/path/to/modulename.so"); // Load the module
</pre></div>
<p>
to the start of each PHP file. SWIG also generates a php module, which
attempts to do the <tt>dl()</tt> call for you:
</p>
<div class="code"><pre>
include("example.php");
</pre></div>
<H2><a name="Php_nn2"></a>32.2 Basic PHP interface</H2>
<p>
It is important to understand that PHP uses a single global namespace
into which all symbols from extension modules are loaded. It is quite
possible for names of symbols in one extension module to clash with
other symbols unless care is taken to <tt>%rename</tt> them. At present
SWIG doesn't have support for the namespace feature added in PHP 5.3.
</p>
<H3><a name="Php_nn2_1"></a>32.2.1 Constants</H3>
<p>
These work in much the same way as in C/C++. Constants can be defined
by using either the normal C pre-processor declarations, or the
<tt>%constant</tt> SWIG directive. These will then be available from
your PHP script as a PHP constant, (i.e. no dollar sign is needed to
access them.) For example, with a swig interface file like this,
</p>
<div class="code"><pre>
%module example
#define PI 3.14159
%constant int E = 2.71828
</pre>
</div>
<p>
you can access the constants in your PHP script like this,
</p>
<div class="code"><pre>
include("example.php");
echo "PI = " . PI . "\n";
echo "E = " . E . "\n";
</pre>
</div>
<p>
There's one peculiarity of how constants work in PHP which it is useful
to note (this is not specific to SWIG though) - if you try to use an undeclared
constant, PHP will issue a warning and then expand the constant to a string
version of the constant's name. The warning will often be missed though as
if you're using PHP in a webserver, it will probably end up in error.log or
similar.
</p>
<p>
For example,
</p>
<div class="code"><pre>
%module example
#define EASY_TO_MISPELL 0
</pre>
</div>
<p>
accessed incorrectly in PHP,
</p>
<div class="code">
<pre>
include("example.php");
if(EASY_TO_MISPEL) {
....
} else {
....
}
</pre>
</div>
<p>
The mis-spelled constant will become the string 'EASY_TO_MISPEL', which
is treated as true by the if test, when the value of the intended constant
would be treated as false!
</p>
<H3><a name="Php_nn2_2"></a>32.2.2 Global Variables</H3>
<p>
Because PHP does not provide a mechanism to intercept access and
assignment of global variables, global variables are supported through
the use of automatically generated accessor functions.
</p>
<div class="code"><pre>
%module example;
%inline %{
double seki = 2;
void print_seki() {
zend_printf("seki is now %f\n",seki);
}
%}
</pre></div>
<p>
is accessed as follows:
</p>
<div class="code"><pre>
include("example.php");
print seki_get();
seki_set( seki_get() * 2); # The C variable is now 4.
print seki_get();
</pre></div>
<p>
SWIG supports global variables of all C datatypes including pointers
and complex objects. Additional types can be supported by using the
<tt>varinit</tt> typemap.
</p>
<p>
SWIG honors the <tt>%immutable</tt> modifier by not generating code
for the <tt>_set</tt> method. This provides read-only access to the
variable from the php script. Attempting to access the <tt>_set</tt>
method will result in a php fatal error because the function is
undefined.
</p>
<p>
At this time SWIG does not support custom accessor methods.
</p>
<H3><a name="Php_nn2_3"></a>32.2.3 Functions</H3>
<p>
C functions are converted into PHP functions. Default/optional arguments are
also allowed. An interface file like this :
</p>
<div class="code"><pre>
%module example
int foo(int a);
double bar(double, double b = 3.0);
...
</pre></div>
<p>
Will be accessed in PHP like this :
</p>
<div class="code"><pre>
include("example.php");
$a = foo(2);
$b = bar(3.5, -1.5);
$c = bar(3.5); # Use default argument for 2nd parameter
</pre></div>
<!-- This isn't correct for 1.3.30 and needs rewriting to reflect reality
<p>
Because PHP is a dynamically typed language, the default typemaps
used for simple types will attempt to coerce the arguments into the appropriate type. That is the following invocations are equivalent:
</p>
<div class="code"><pre>
$a = foo(2);
$a = foo("2");
$a = foo(2.0);
</pre></div>
<p>
Functions are invoked using pass by value semantics like all of PHP.
This means the conversion which automatically takes place when
invoking a swig wrapped method does not change the native type of the
argument variable.
</p>
<div class="code"><pre>
$s = "2 A string representing two";
$a = foo($s); # invokes 'foo(2)';
print $s; # The value of $s was not changed.
</pre></div>
-->
<H3><a name="Php_nn2_4"></a>32.2.4 Overloading</H3>
<p>
Although PHP does not support overloading functions natively, swig
will generate dispatch functions which will use <tt>%typecheck</tt>
typemaps to allow overloading. This dispatch function's operation and
precedence is described in <a
href="SWIGPlus.html#SWIGPlus_overloaded_methods">Wrapping
Overloaded Functions and Methods</a>.
</p>
<!-- This isn't correct for 1.3.30 and needs rewriting to reflect reality
<p>
Because PHP is a dynamically typed language, simple values can be
silently converted from one type to another. For example, integers,
doubles and strings silently convert to each other depending on
context. This situation make overloading slightly problematic because
given the following function:
</p>
<div class="code"><pre>
void doit( int i );
void doit( double i );
</pre></div>
<p>
it is questionable which to invoke when <tt>doit("2");</tt> is used in
PHP. The string <tt>"2"</tt> simultaneously represents the integer
<tt>2</tt> and the double <tt>2.0</tt>.
</p>
<p>
In order to provide the most natural experience to PHP programmers,
the default <tt>%typecheck</tt> implemented in <tt>php.swg</tt>
allows any simple type (integer, double, string) in PHP to be used for
any simple C type (int, double, char *). The function selected then
depends only on the argument type precedence defined by SWIG.
</p>
<p>
It should be noted that <tt>SWIGTYPE</tt> references and pointers will
not be silently converted. So these two functions:
</p>
<div class="code"><pre>
void doit( const Vector & );
void doit( int i );
</pre></div>
<p>
Cause less confusion and <tt>doit("2");</tt> will invoke the function
taking the integer argument.
</p>
-->
<H3><a name="Php_nn2_5"></a>32.2.5 Pointers and References</H3>
<p>
Pointers to C/C++ objects are represented
as PHP resources, rather like MySQL connection handles.
</p>
<p>
There are multiple ways to wrap pointers to simple types. Given the
following C method:
</p>
<div class="code"><pre>
void add( int *in1, int *in2, int *result);
</pre></div>
<p>
One can include <b>cpointer.i</b> to generate PHP wrappers to <tt>int
*</tt>.
</p>
<div class="code"><pre>
%module example
%include "cpointer.i"
%pointer_functions(int,intp)
void add( int *in1, int *in2, int *result);
</pre></div>
<p>
This will result in the following usage in PHP:
</p>
<div class="code"><pre>
<?php
include("example.php");
$in1=copy_intp(3);
$in2=copy_intp(5);
$result=new_intp();
add( $in1, $in2, $result );
echo "The sum " . intp_value($in1) . " + " . intp_value($in2) . " = " . intp_value( $result) . "\n";
?>
</pre></div>
<p>
An alternative would be to use the include <b>typemaps.i</b> which
defines named typemaps for INPUT, OUTPUT and INOUT variables. One
needs to either <tt>%apply</tt> the appropriate typemap or adjust the
parameter names as appropriate.
</p>
<div class="code"><pre>
%module example
%include "typemaps.i"
void add( int *INPUT, int *INPUT, int *OUTPUT);
</pre></div>
<p>
This will result in the following usage in PHP:
</p>
<div class="code"><pre>
<?php
include("example.php");
$in1 = 3;
$in2 = 5;
$result= add($in1,$in2); # Note using variables for the input is unnecessary.
echo "The sum $in1 + $in2 = $result\n";
?>
</pre></div>
<p>
Because PHP has a native concept of reference, it may seem more natural
to the PHP developer to use references to pass pointers. To enable
this, one needs to include <b>phppointers.i</b> which defines the
named typemap REFERENCE.
</p>
<div class="code"><pre>
%module example
%include "phppointers.i"
void add( int *REF, int *REF, int *REF);
</pre></div>
<p>
This will result in the following usage in PHP:
</p>
<div class="code"><pre>
<?php
include("example.php");
$in1 = 3;
$in2 = 5;
$result = 0;
add(&$in1,&$in2,&$result);
echo "The sum $in1 + $in2 = $result\n";
?>
</pre></div>
<p>
It is important to note that a php variable which is NULL when passed
by reference would end up passing a NULL pointer into the function.
In PHP, an unassigned variable (i.e. where the first reference to the
variable is not an assignment) is
NULL. In the above example, if any of the three variables had not
been assigned, a NULL pointer would have been passed into
<tt>add</tt>. Depending on the implementation of the function, this
may or may not be a good thing.
</p>
<p>
We chose to allow passing NULL pointers into functions because that is
sometimes required in C libraries. A NULL pointer can be created in
PHP in a number of ways: by using <tt>unset</tt> on an existing
variable, or assigning <tt>NULL</tt> to a variable.
</p>
<H3><a name="Php_nn2_6"></a>32.2.6 Structures and C++ classes</H3>
<p>
SWIG defaults to wrapping C++ structs and classes with PHP classes - this
is done by generating a PHP wrapper script which defines proxy classes
which calls a set of flat functions which actually wrap the C++ class.
You can disable this wrapper layer by passing the command-line option
"-noproxy" in which case you'll just get the flat functions.
</p>
<p>
This interface file
</p>
<div class="code"><pre>
%module vector
class Vector {
public:
double x,y,z;
Vector();
~Vector();
double magnitude();
};
struct Complex {
double re, im;
};
</pre></div>
<p>
Would be used in the following way from PHP5:
</p>
<div class="code"><pre>
<?php
require "vector.php";
$v = new Vector();
$v->x = 3;
$v->y = 4;
$v->z = 5;
echo "Magnitude of ($v->x,$v->y,$v->z) = " . $v->magnitude() . "\n";
$v = NULL; # destructor called.
$c = new Complex();
$c->re = 0;
$c->im = 0;
# $c destructor called when $c goes out of scope.
?>
</pre></div>
<p>
Member variables and methods are accessed using the <tt>-></tt> operator.
</p>
<H4><a name="Php_nn2_6_1"></a>32.2.6.1 Using <tt>-noproxy</tt></H4>
<p>
The <tt>-noproxy</tt> option flattens the object structure and
generates collections of named functions (these are the functions
which the PHP5 class wrappers call). The above example results
in the following PHP functions:
</p>
<div class="code"><pre>
new_Vector();
Vector_x_set($obj,$d);
Vector_x_get($obj);
Vector_y_set($obj,$d);
Vector_y_get($obj);
Vector_z_set($obj,$d);
Vector_z_get($obj);
Vector_magnitude($obj);
new_Complex();
Complex_re_set($obj,$d);
Complex_re_get($obj);
Complex_im_set($obj,$d);
Complex_im_get($obj);
</pre></div>
<H4><a name="Php_nn2_6_2"></a>32.2.6.2 Constructors and Destructors</H4>
<p>
The constructor is called when <tt>new Object()</tt> (or
<tt>new_Object()</tt> if using <tt>-noproxy</tt>) is used to create an
instance of the object. If multiple constructors are defined for an
object, function overloading will be used to determine which
constructor to execute.
</p>
<p>
Because PHP uses reference counting to manage resources, simple
assignment of one variable to another such as:
</p>
<div class="code"><pre>
$ref = $v;
</pre></div>
<p>
causes the symbol <tt>$ref</tt> to refer to the same underlying object
as <tt>$v</tt>. This does not result in a call to the C++ copy
constructor or copy assignment operator.
</p>
<p>
One can force execution of the copy constructor by using:
</p>
<div class="code"><pre>
$o_copy = new Object($o);
</pre></div>
<p>
Destructors are automatically called when all variables referencing
the instance are reassigned or go out of scope. The destructor is not
available to be called manually. To force a destructor to be called
the programmer can either reassign the variable or call
<tt>unset($v)</tt>
</p>
<H4><a name="Php_nn2_6_3"></a>32.2.6.3 Static Member Variables</H4>
<p>
Static member variables in C++ are not wrapped as such in PHP
as it does not appear to be possible to intercept accesses to such variables.
Therefore, static member variables are
wrapped using a class function with the same name, which
returns the current value of the class variable. For example
</p>
<div class="code"><pre>
%module example
class Ko {
static int threats;
};
</pre></div>
<p>
would be accessed in PHP as,
</p>
<div class="code"><pre>
include("example.php");
echo "There has now been " . Ko::threats() . " threats\n";
</pre></div>
<p>
To set the static member variable, pass the value as the argument to the class
function, e.g.
</p>
<div class="code"><pre>
Ko::threats(10);
echo "There has now been " . Ko::threats() . " threats\n";
</pre></div>
<H4><a name="Php_nn2_6_4"></a>32.2.6.4 Static Member Functions</H4>
<p>
Static member functions are supported in PHP using the
<tt>class::function()</tt> syntax. For example
</p>
<div class="code"><pre>
%module example
class Ko {
static void threats();
};
</pre></div>
would be executed in PHP as,
<div class="code"><pre>
include("example.php");
Ko::threats();
</pre></div>
<H3><a name="Php_nn2_7"></a>32.2.7 PHP Pragmas, Startup and Shutdown code</H3>
<p>
To place PHP code in the generated "example.php" file one can use the
<b>code</b> pragma. The code is inserted after loading the shared
object.
</p>
<div class="code"><pre>
%module example
%pragma(php) code="
# This code is inserted into example.php
echo \"example.php execution\\n\";
"
</pre></div>
<p>
Results in the following in "example.php"
</p>
<div class="code"><pre>
# This code is inserted into example.php
echo "example.php execution\n";
</pre></div>
<p>
The <b>include</b> pragma is a short cut to add include statements to
the example.php file.
</p>
<div class="code"><pre>
%module example
%pragma(php) code="
include \"include.php\";
"
%pragma(php) include="include.php" // equivalent.
</pre></div>
<p>
The <b>phpinfo</b> pragma inserts code in the
<tt>PHP_MINFO_FUNCTION</tt> which is called from PHP's
phpinfo() function.
</p>
<div class="code"><pre>
%module example;
%pragma(php) phpinfo="
zend_printf("An example of PHP support through SWIG\n");
php_info_print_table_start();
php_info_print_table_header(2, \"Directive\", \"Value\");
php_info_print_table_row(2, \"Example support\", \"enabled\");
php_info_print_table_end();
"
</pre></div>
<p>
To insert code into the <tt>PHP_MINIT_FUNCTION</tt>, one can use
either <tt>%init</tt> or <tt>%minit</tt>.
</p>
<div class="code"><pre>
%module example;
%init {
zend_printf("Inserted into PHP_MINIT_FUNCTION\n");
}
%minit {
zend_printf("Inserted into PHP_MINIT_FUNCTION\n");
}
</pre></div>
<p>
To insert code into the <tt>PHP_MSHUTDOWN_FUNCTION</tt>, one can use
either <tt>%shutdown</tt> or <tt>%mshutdown</tt>.
</p>
<div class="code"><pre>
%module example;
%mshutdown {
zend_printf("Inserted into PHP_MSHUTDOWN_FUNCTION\n");
}
</pre></div>
<p>
The <tt>%rinit</tt> and <tt>%rshutdown</tt> statements are very similar but insert code
into the request init (PHP_RINIT_FUNCTION) and request shutdown (PHP_RSHUTDOWN_FUNCTION) code respectively.
</p>
<H2><a name="Php_nn3"></a>32.3 Cross language polymorphism</H2>
<p>
Proxy classes provide a more natural, object-oriented way to access
extension classes. As described above, each proxy instance has an
associated C++ instance, and method calls to the proxy are passed to the
C++ instance transparently via C wrapper functions.
</p>
<p>
This arrangement is asymmetric in the sense that no corresponding
mechanism exists to pass method calls down the inheritance chain from
C++ to PHP. In particular, if a C++ class has been extended in PHP
(by extending the proxy class), these extensions will not be visible
from C++ code. Virtual method calls from C++ are thus not able access
the lowest implementation in the inheritance chain.
</p>
<p>
Changes have been made to SWIG 1.3.18 to address this problem and make
the relationship between C++ classes and proxy classes more symmetric.
To achieve this goal, new classes called directors are introduced at the
bottom of the C++ inheritance chain. Support for generating PHP classes
has been added in SWIG 1.3.40. The job of the directors is to route
method calls correctly, either to C++ implementations higher in the
inheritance chain or to PHP implementations lower in the inheritance
chain. The upshot is that C++ classes can be extended in PHP and from
C++ these extensions look exactly like native C++ classes. Neither C++
code nor PHP code needs to know where a particular method is
implemented: the combination of proxy classes, director classes, and C
wrapper functions takes care of all the cross-language method routing
transparently.
</p>
<H3><a name="Php_nn3_1"></a>32.3.1 Enabling directors</H3>
<p>
The director feature is disabled by default. To use directors you
must make two changes to the interface file. First, add the "directors"
option to the %module directive, like this:
</p>
<div class="code">
<pre>
%module(directors="1") modulename
</pre>
</div>
<p>
Without this option no director code will be generated. Second, you
must use the %feature("director") directive to tell SWIG which classes
and methods should get directors. The %feature directive can be applied
globally, to specific classes, and to specific methods, like this:
</p>
<div class="code">
<pre>
// generate directors for all classes that have virtual methods
%feature("director");
// generate directors for all virtual methods in class Foo
%feature("director") Foo;
// generate a director for just Foo::bar()
%feature("director") Foo::bar;
</pre>
</div>
<p>
You can use the %feature("nodirector") directive to turn off
directors for specific classes or methods. So for example,
</p>
<div class="code">
<pre>
%feature("director") Foo;
%feature("nodirector") Foo::bar;
</pre>
</div>
<p>
will generate directors for all virtual methods of class Foo except
bar().
</p>
<p>
Directors can also be generated implicitly through inheritance.
In the following, class Bar will get a director class that handles
the methods one() and two() (but not three()):
</p>
<div class="code">
<pre>
%feature("director") Foo;
class Foo {
public:
Foo(int foo);
virtual void one();
virtual void two();
};
class Bar: public Foo {
public:
virtual void three();
};
</pre>
</div>
<p>
then at the PHP side you can define
</p>
<div class="targetlang">
<pre>
require("mymodule.php");
class MyFoo extends Foo {
function one() {
print "one from php\n";
}
}
</pre>
</div>
<H3><a name="Php_nn3_2"></a>32.3.2 Director classes</H3>
<p>
For each class that has directors enabled, SWIG generates a new class
that derives from both the class in question and a special
<tt>Swig::Director</tt> class. These new classes, referred to as director
classes, can be loosely thought of as the C++ equivalent of the PHP
proxy classes. The director classes store a pointer to their underlying
PHP object. Indeed, this is quite similar to the "_cPtr" and "thisown"
members of the PHP proxy classes.
</p>
<p>
For simplicity let's ignore the <tt>Swig::Director</tt> class and refer to the
original C++ class as the director's base class. By default, a director
class extends all virtual methods in the inheritance chain of its base
class (see the preceding section for how to modify this behavior).
Thus all virtual method calls, whether they originate in C++ or in
PHP via proxy classes, eventually end up in at the implementation in the
director class. The job of the director methods is to route these method
calls to the appropriate place in the inheritance chain. By "appropriate
place" we mean the method that would have been called if the C++ base
class and its extensions in PHP were seamlessly integrated. That
seamless integration is exactly what the director classes provide,
transparently skipping over all the messy extension API glue that binds
the two languages together.
</p>
<p>
In reality, the "appropriate place" is one of only two possibilities:
C++ or PHP. Once this decision is made, the rest is fairly easy. If the
correct implementation is in C++, then the lowest implementation of the
method in the C++ inheritance chain is called explicitly. If the correct
implementation is in PHP, the Zend API is used to call the method of the
underlying PHP object (after which the usual virtual method resolution
in PHP automatically finds the right implementation).
</p>
<p>
Now how does the director decide which language should handle the method call?
The basic rule is to handle the method in PHP, unless there's a good
reason not to. The reason for this is simple: PHP has the most
"extended" implementation of the method. This assertion is guaranteed,
since at a minimum the PHP proxy class implements the method. If the
method in question has been extended by a class derived from the proxy
class, that extended implementation will execute exactly as it should.
If not, the proxy class will route the method call into a C wrapper
function, expecting that the method will be resolved in C++. The wrapper
will call the virtual method of the C++ instance, and since the director
extends this the call will end up right back in the director method. Now
comes the "good reason not to" part. If the director method were to blindly
call the PHP method again, it would get stuck in an infinite loop. We avoid this
situation by adding special code to the C wrapper function that tells
the director method to not do this. The C wrapper function compares the
called and the declaring class name of the given method. If these are
not the same, then the C wrapper function tells the director to resolve
the method by calling up the C++ inheritance chain, preventing an
infinite loop.
</p>
<p>
One more point needs to be made about the relationship between director
classes and proxy classes. When a proxy class instance is created in
PHP, SWIG creates an instance of the original C++ class and assigns it
to <tt>->_cPtr</tt>. This is exactly what happens without directors
and is true even if directors are enabled for the particular class in
question. When a class <i>derived</i> from a proxy class is created,
however, SWIG then creates an instance of the corresponding C++ director
class. The reason for this difference is that user-defined subclasses
may override or extend methods of the original class, so the director
class is needed to route calls to these methods correctly. For
unmodified proxy classes, all methods are ultimately implemented in C++
so there is no need for the extra overhead involved with routing the
calls through PHP.
</p>
<H3><a name="Php_nn3_3"></a>32.3.3 Ownership and object destruction</H3>
<p>
Memory management issues are slightly more complicated with directors
than for proxy classes alone. PHP instances hold a pointer to the
associated C++ director object, and the director in turn holds a pointer
back to the PHP object. By default, proxy classes own their C++ director
object and take care of deleting it when they are garbage collected.
</p>
<p>
This relationship can be reversed by calling the special
<tt>->thisown</tt> property of the proxy class. After setting this
property to <tt>0</tt>, the director class no longer destroys the PHP
object. Assuming no outstanding references to the PHP object remain,
the PHP object will be destroyed at the same time. This is a good thing,
since directors and proxies refer to each other and so must be created
and destroyed together. Destroying one without destroying the other will
likely cause your program to segfault.
</p>
<p>
Here is an example:
</p>
<div class="code">
<pre>
class Foo {
public:
...
};
class FooContainer {
public:
void addFoo(Foo *);
...
};
</pre>
</div>
<br>
<div class="targetlang">
<pre>
$c = new FooContainer();
$a = new Foo();
$a->thisown = 0;
$c->addFoo($a);
</pre>
</div>
<p>
In this example, we are assuming that FooContainer will take care of
deleting all the Foo pointers it contains at some point.
</p>
<H3><a name="Php_nn3_4"></a>32.3.4 Exception unrolling</H3>
<p>
With directors routing method calls to PHP, and proxies routing them
to C++, the handling of exceptions is an important concern. By default, the
directors ignore exceptions that occur during method calls that are
resolved in PHP. To handle such exceptions correctly, it is necessary
to temporarily translate them into C++ exceptions. This can be done with
the %feature("director:except") directive. The following code should
suffice in most cases:
</p>
<div class="code">
<pre>
%feature("director:except") {
if ($error == FAILURE) {
throw Swig::DirectorMethodException();
}
}
</pre>
</div>
<p>
This code will check the PHP error state after each method call from a
director into PHP, and throw a C++ exception if an error occurred. This
exception can be caught in C++ to implement an error handler.
Currently no information about the PHP error is stored in the
Swig::DirectorMethodException object, but this will likely change in the
future.
</p>
<p>
It may be the case that a method call originates in PHP, travels up to
C++ through a proxy class, and then back into PHP via a director method.
If an exception occurs in PHP at this point, it would be nice for that
exception to find its way back to the original caller. This can be done
by combining a normal %exception directive with the
<tt>director:except</tt> handler shown above. Here is an example of a
suitable exception handler:
</p>
<div class="code">
<pre>
%exception {
try { $action }
catch (Swig::DirectorException &e) { SWIG_fail; }
}
</pre>
</div>
<p>
The class Swig::DirectorException used in this example is actually a
base class of Swig::DirectorMethodException, so it will trap this
exception. Because the PHP error state is still set when
Swig::DirectorMethodException is thrown, PHP will register the exception
as soon as the C wrapper function returns.
</p>
<H3><a name="Php_nn3_5"></a>32.3.5 Overhead and code bloat</H3>
<p>
Enabling directors for a class will generate a new director method for
every virtual method in the class' inheritance chain. This alone can
generate a lot of code bloat for large hierarchies. Method arguments
that require complex conversions to and from target language types can
result in large director methods. For this reason it is recommended that
you selectively enable directors only for specific classes that are
likely to be extended in PHP and used in C++.
</p>
<p>
Compared to classes that do not use directors, the call routing in the
director methods does add some overhead. In particular, at least one
dynamic cast and one extra function call occurs per method call from
PHP. Relative to the speed of PHP execution this is probably completely
negligible. For worst case routing, a method call that ultimately
resolves in C++ may take one extra detour through PHP in order to ensure
that the method does not have an extended PHP implementation. This could
result in a noticeable overhead in some cases.
</p>
<p>
Although directors make it natural to mix native C++ objects with PHP
objects (as director objects) via a common base class pointer, one
should be aware of the obvious fact that method calls to PHP objects
will be much slower than calls to C++ objects. This situation can be
optimized by selectively enabling director methods (using the %feature
directive) for only those methods that are likely to be extended in PHP.
</p>
<H3><a name="Php_nn3_6"></a>32.3.6 Typemaps</H3>
<p>
Typemaps for input and output of most of the basic types from director
classes have been written. These are roughly the reverse of the usual
input and output typemaps used by the wrapper code. The typemap
operation names are 'directorin', 'directorout', and 'directorargout'.
The director code does not currently use any of the other kinds of
typemaps. It is not clear at this point which kinds are appropriate and
need to be supported.
</p>
<H3><a name="Php_nn3_7"></a>32.3.7 Miscellaneous</H3>
<p> Director typemaps for STL classes are mostly in place, and hence you
should be able to use std::string, etc., as you would any other type.
</p>
</body>
</html>
