Regexp to match “foo/bar” and not “foobar” wanted

Question

I'm looking for a regex that i can use in my tokenizer to compile a config file. Here is a snippet out of a class in php:

private $token = array(
    "PATH" => "([a-zA-Z\_-]+\.|\*\.)+([a-zA-Z\_-]+|\*)",
    "MIXED" => "[a-zA-Z0-9-_\(\)\/]{2,}",
    "STRING" => "[a-zA-Z-_]{2,}"
);

private function getToken($string) {
    foreach($this->token as $name => $pattern) {
        preg_match("/^".$pattern."/", $string, $match);
        if(!empty($match))
            return array($name, $match[0]);
    }

    return false;
}

"MIXED" should match "foo/bar" and not "foobar" and "STRING" should match "foobar" and not "foo/bar". Currently "foobar" and "foo/bar" are "MIXED".

How do i write this "AND NOT" in a single pattern down?

Thank you.

Cougar is on the right path. What you want is to express the idea "contains at least one slash". — Kaz

Kaz Kaz · Accepted Answer · 2012-04-07T18:25:46

This pattern will match any sequence of letters, digits, underscores, hyphens and slashes which contains at least one slash:

[a-zA-Z0-9-_/]*\/[a-zA-Z0-9-_/]*

So this gives you a general idea how to reject tokens like abc while matching ab/c. This is very similar to distinguishing floating-point constants from integer constants.

You should probably be tokenizing inputs like foo/bar(255) as four tokens: foo/bar ( 255 and ).

Otherwise enforcing this slash requirement is complicated. The naive ways mean that MIXED can be something like these:

foo(255/255)
foo(/)

or even:

)/-

just because it contains a slash somewhere, not necessarily where you want.

Clarify your requirements.

Regexp to match “foo/bar” and not “foobar” wanted

1 Answers