Not a member of Pastebin yet?
Sign Up,
it unlocks many cool features!
- import sys, re
- css = open( sys.argv[1] , 'r' ).read()
- # remove comments - this will break a lot of hacks :-P
- css = re.sub( r's*/*s**/', "$$HACK1$$", css ) # preserve IE<6 comment hack
- css = re.sub( r'/*[sS]*?*/', "", css )
- css = css.replace( "$$HACK1$$", '/**/' ) # preserve IE<6 comment hack
- # url() doesn't need quotes
- css = re.sub( r'url((["'])([^)]*)1)', r'url(2)', css )
- # spaces may be safely collapsed as generated content will collapse them anyway
- css = re.sub( r's+', ' ', css )
- # shorten collapsable colors: #aabbcc to #abc
- css = re.sub( r'#([0-9a-f])1([0-9a-f])2([0-9a-f])3(s|;)', r'#1234', css )
- # fragment values can loose zeros
- css = re.sub( r':s*0(.d+([cm]m|e[mx]|in|p[ctx]))s*;', r':1;', css )
- for rule in re.findall( r'([^{]+){([^}]*)}', css ):
- # we don't need spaces around operators
- selectors = [re.sub( r'(?<=[[(>+=])s+|s+(?=[=~^$*|>+])])', r'', selector.strip() ) for selector in rule[0].split( ',' )]
- # order is important, but we still want to discard repetitions
- properties = {}
- porder = []
- for prop in re.findall( '(.*?):(.*?)(;|$)', rule[1] ):
- key = prop[0].strip().lower()
- if key not in porder: porder.append( key )
- properties[ key ] = prop[1].strip()
- # output rule if it contains any declarations
- if properties:
- print "%s{%s}" % ( ','.join( selectors ), ''.join(['%s:%s;' % (key, properties[key]) for key in porder])[:-1] )
Add Comment
Please, Sign In to add comment