iwla/plugins/pre_analysis/page_to_hit.py

import re
import time

from iwla import IWLA
from iplugin import IPlugin

# Basic rule to detect robots

class IWLAPreAnalysisPageToHit(IPlugin):

    def __init__(self, iwla):
        super(IWLAPreAnalysisPageToHit, self).__init__(iwla)
        self.API_VERSION = 1

    def load(self):
# Remove logo from indefero
        self.regexps = self.iwla.getConfValue('page_to_hit_conf', [])
        if not self.regexps: return False
        self.regexps = map(lambda(r): re.compile(r), self.regexps)

        return True

    def hook(self):
        start_time = self.iwla.getStartAnalysisTime()
        start_time = time.mktime(start_time)

        hits = self.iwla.getCurrentVisists()
        viewed_http_codes = self.iwla.getConfValue('viewed_http_codes', [200, 304])
        for (k, super_hit) in hits.items():
            if super_hit['robot']: continue

            for p in super_hit['requests']:
                if not p['is_page']: continue
                if time.mktime(p['time_decoded']) < start_time: continue
                uri = p['extract_request']['extract_uri']
                for r in self.regexps:
                    if r.match(uri):
                        p['is_page'] = False
                        super_hit['viewed_pages'] -= 1
                        super_hit['viewed_hits'] += 1
                        break