CSS选择器的正则表达式 - PHP

Does anyone know what the most robust regex would be to select all selectors from a css declaration? Below is CSS code for non-standard CSS that would be a good test case.

.somenormalstyledeclaration 
{
    width:100%;
    height:100%;
    background-image: url(images/fallback-gradient.png);
}
.somenormalstyle1, .somenormalstyle2 
{
    width:100%;
    height:100%;
    background-image: url(images/fallback-gradient.png);
}
.gradient-bg 
{
   background-color: #1a82f7;
   background-image: url(images/fallback-gradient.png);
   background-image: -webkit-gradient(linear, 0% 0%, 0% 100%, from(#2F2727), to(#1a82f7));
   background-image: -webkit-linear-gradient(top, #2F2727, #1a82f7);
   background-image:    -moz-linear-gradient(top, #2F2727, #1a82f7);
   background-image:     -ms-linear-gradient(top, #2F2727, #1a82f7);
   background-image:      -o-linear-gradient(top, #2F2727, #1a82f7);
}

@-webkit-keyframes bounce_circle 
{
    0% 
    {
        opacity:0.3;
    }
    50% 
    {
        opacity:1;
        background-color:#111
    }
    100% 
    {
        opacity:0.3;
    }
}

This should work mostly. It's a few steps, but not as good as a full parser.

1. Replace all newlines with a space character
2. Assuming you never use strings including { or }, you can replace /\{[^\}]\}/ with a ,
3. Split all the text into an array on each , and ; character
4. Trim all whitespace from the beginning and end of each selector

Every element of the array should be a selector or an "at-rule" (i.e. @charset utf-8;).

I have probably missed some other edge cases, but this should work as a quick-and-dirty solution for most cases.

While this doesn't deal with the @-webkit... it does yeild just ID and Class selectors.

I'm sure a regex wizard can be more efficient...

This link helped point the way. Even though it's javascript, I was able to get the gist.

// Snag file
$text = file_get_contents("file.css");

// Strip comments
$text = preg_replace('!/\*.*?\*/!s', '', $text);

// Clear out everything between brackets
$text = preg_replace('/{(.*?)}/si', '', $text);

// Get ID and class selectors
$pattern = '/([#|\.])([_a-z]+[_a-z0-9-]*)/mi';

// Array to store results.
$result = array();

// No "g' modifier in PHP, use preg_match_all for this kind of behaviour.
preg_match_all($pattern, $text, $result);

print "<pre>";
print_r ( $result );
print "</pre>";