#!/usr/bin/env python
import BeautifulSoup
import HTMLParser
import os
import sgmllib
import sgmlop
import time

HTML_PATH = 'html'
XML_PATH = 'xml'

def get_data():
    html = xml = []
    html = [open(os.path.join(HTML_PATH, f)
        ).read() for f in os.listdir(HTML_PATH)]
    xml = [open(os.path.join(XML_PATH, f)
        ).read() for f in os.listdir(XML_PATH)]
        
    return html, xml
    
def parse(klass, data):
    errors = []
    for page in data:
        p = klass()
        try:
            p.feed(page)
        except Exception, e:
            errors.append(str(e))
            continue     
    return errors
    
def soup_parse(data):
    errors = []
    for page in data:
        try:
            BeautifulSoup.BeautifulSoup(page)
        except Exception, e:
            errors.append(str(e))
            continue
    return errors
    
def sgmlop_parse(klass, handler, data):
    errors = []
    for page in data:
        p = klass()
        p.register(handler)
        try:
            p.feed(page)
        except Exception, e:
            errors.append(str(e))
            continue
    return errors
    

def bench_with_time():
    html, xml = get_data()

    def time_htmlparser():
        # HTMLParser, only HTML
        html_start_time = time.time()
        errors = parse(HTMLParser.HTMLParser, html)
        html_time = time.time() - html_start_time
        print "HTMLParser, only HTML - time: %s, errors: %d" % (
            html_time, len(errors))
    
        # HTMLParser, only XML
        xml_start_time = time.time()
        errors = parse(HTMLParser.HTMLParser, xml)
        xml_time = time.time() - xml_start_time
        print "HTMLParser, only XML - time: %s, errors: %d" % (
            xml_time, len(errors))
        
        print "Total: %f" % (html_time + xml_time,)
        print

    
    def time_sgmllib():            
        # sgmllib.SGMLParser, only HTML
        html_start_time = time.time()
        errors = parse(sgmllib.SGMLParser, html)
        html_time = time.time() - html_start_time
        print "sgmllib.SGMLParser, only HTML - time: %s, errors: %d" % (
            html_time, len(errors))

        # sgmllib.SGMLParser, only XML
        xml_start_time = time.time()
        errors = parse(sgmllib.SGMLParser, xml)
        xml_time = time.time() - xml_start_time
        print "sgmllib.SGMLParser, only XML - time: %s, errors: %d" % (
            xml_time, len(errors))
            
        print "Total: %f" % (html_time + xml_time,)
        print
        

    def time_beautifulsoup():
        # BeautifulSoup, only HTML
        html_start_time = time.time()
        errors = soup_parse(html)
        html_time = time.time() - html_start_time
        print "BeautifulSoup, only HTML - time: %s, errors: %d" % (
            html_time, len(errors))

        # BeautifulSoup, only XML
        xml_start_time = time.time()
        errors = soup_parse(xml)
        xml_time = time.time() - xml_start_time
        print "BeautifulSoup, only XML - time: %s, errors: %d" % (
            xml_time, len(errors))
            
        print "Total: %f" % (html_time + xml_time,)
        print
        
    def time_sgmlop():
        class Handler(object):
            def handle_special(self, text):
                pass
            def handle_proc(self, target, value):
                pass
            def finish_starttag(self, tag, attrs):
                pass
            def finish_endtag(self, tag):
                pass
            def handle_data(self, data):
                pass
        
        # sgmlop.SGMLParser, only HTML
        html_start_time = time.time()
        errors = sgmlop_parse(sgmlop.SGMLParser, Handler(), html)
        html_time = time.time() - html_start_time
        print "sgmlop.SGMLParser, only HTML - time: %s, errors: %d" % (
            html_time, len(errors))

        # sgmlop.SGMLParser, only XML
        xml_start_time = time.time()
        errors = sgmlop_parse(sgmlop.SGMLParser, Handler(), xml)
        xml_time = time.time() - xml_start_time
        print "sgmlop.SGMLParser, only XML - time: %s, errors: %d" % (
            xml_time, len(errors))
            
        print "Total: %f" % (html_time + xml_time,)
        print
        

        # sgmlop.XMLParser, only HTML
        html_start_time = time.time()
        errors = sgmlop_parse(sgmlop.XMLParser, Handler(), html)
        html_time = time.time() - html_start_time
        print "sgmlop.XMLParser, only HTML - time: %s, errors: %d" % (
            html_time, len(errors))

        # sgmlop.XMLParser, only XML
        xml_start_time = time.time()
        errors = sgmlop_parse(sgmlop.XMLParser, Handler(), html)
        xml_time = time.time() - xml_start_time
        print "sgmlop.XMLParser, only XML - time: %s, errors: %d" % (
            xml_time, len(errors))
            
        print "Total: %f" % (html_time + xml_time,)
        print


    time_htmlparser()
    time_sgmllib()
    time_beautifulsoup()
    time_sgmlop()


if __name__ == '__main__':
    bench_with_time()