iwla/plugins/pre_analysis/H001_robot.py

import re
from iwla import IWLA

from awstats_robots_data import awstats_robots

PLUGIN_CLASS = 'HTTP'
API_VERSION = 1

def get_plugins_infos():
    infos = {'class' : PLUGIN_CLASS,
             'min_version' : API_VERSION,
             'max_version' : -1}
    return infos

def load():
    global awstats_robots
    print '==> Generating robot dictionary'

    awstats_robots = map(lambda (x) : re.compile(x, re.IGNORECASE), awstats_robots)
    
    return True

# Basic rule to detect robots

def hook(iwla):
    hits = iwla.getCurrentVisists()
    for k in hits.keys():
        super_hit = hits[k]

        if super_hit['robot']: continue

        isRobot = False
        referers = 0

        first_page = super_hit['requests'][0]
        if first_page['time_decoded'].tm_mday == super_hit['last_access'].tm_mday:
            for r in awstats_robots:
                if r.match(first_page['http_user_agent']):
                    super_hit['robot'] = 1
                    continue

# 1) no pages view --> robot
        if not super_hit['viewed_pages']:
            super_hit['robot'] = 1
            continue

# 2) pages without hit --> robot
        if not super_hit['viewed_hits']:
            super_hit['robot'] = 1
            continue

        for hit in super_hit['requests']:
# 3) /robots.txt read
            if hit['extract_request']['http_uri'] == '/robots.txt':
                isRobot = True
                break

# 4) Any referer for hits
            if not hit['is_page'] and hit['http_referer']:
                referers += 1

        if isRobot:
            super_hit['robot'] = 1
            continue

        if super_hit['viewed_hits'] and not referers:
            super_hit['robot'] = 1
            continue