How do I find all the classes used in a PHP file?

Question

I am trying to use the tokenizer to scan a file to find all the defined classes, anything they extend, any created instances, and anytime they were statically invoked.

<?php

$tokens = token_get_all(file_get_contents($file));

$used_classes = array();
$defined_classes = array();
$variable_classes = array();

foreach($tokens as $i => $token) {

    if(is_array($token)) {

        if(isset($tokens[$i - 2][0], $tokens[$i - 1][0])) {

            // new [class]
            if ($tokens[$i - 2][0] == T_NEW AND $tokens[$i - 1][0] == T_WHITESPACE) {

                if($tokens[$i][0] == T_STRING) {
                    $used_classes[$token[1]] = TRUE;

                // new $variable()
                } elseif($tokens[$i][0] == T_VARIABLE) {    

                    // @todo, this is really broken. However, do best to look for the assignment
                    if(preg_match('~\$var\s*=\s*([\'"])((?:(?!\1).)*)\1~', $text, $match)) {
                        if(empty($extension_classes[$match[2]])) {
                            $used_classes[$match[2]] = TRUE;
                        }
                    } elseif($token[1] !== '$this') {
                        $variable_classes[$token[1]] = TRUE;
                    }
                }

            }

            // class [class]
            if ($tokens[$i - 2][0] == T_CLASS AND $tokens[$i - 1][0] == T_WHITESPACE) {

                if($tokens[$i][0] == T_STRING) {
                    $defined_classes[$token[1]] = TRUE;
                }
            }


            // @todo: find more classes \/

            // class [classname] extends [class] ???
            // [class]::method()???
        }
    }
}

How can I extend this code to find any additional instances of PHP classes like mentioned above?

• http://www.php.net/manual/en/tokens.php
• http://westhoffswelt.de/blog/class_dependency_graph_generation_update.html
• http://phpqatools.org/

Answer 1

Parsing and then interpreting PHP code is not something that can be solved well using a regex. You would need a something much more clever, like a state machine, that can actually understand things like scope, class names, inheritance etc to be able to do what you want.

It just so happens, that I happen to have written a PHP-to-Javascript converter based on a state-machine that will almost do most of what you want to do:

all the defined classes

Yes, all the classes create a ClassScope with all their variables listed and their methods are created as FunctionScope's, so you can tell which methods a class has.

anything they extend

Yes, every class has it's parent classes listed in ClassScope->$parentClasses

any created instances

Nope, but wouldn't be hard to add extra code to record these.

anytime they were statically invoked.

Nope - but that actually could be done with a regex.

Although it doesn't exactly solve your problem, the project as it stands would get you 95% of the way towards what you want to do, which would save a couple weeks work.