### # Copyright (c) 2002-2004, Jeremiah Fincher # All rights reserved. # # Redistribution and use in source and binary forms, with or without # modification, are permitted provided that the following conditions are met: # # * Redistributions of source code must retain the above copyright notice, # this list of conditions, and the following disclaimer. # * Redistributions in binary form must reproduce the above copyright notice, # this list of conditions, and the following disclaimer in the # documentation and/or other materials provided with the distribution. # * Neither the name of the author of this software nor the name of # contributors to this software may be used to endorse or promote products # derived from this software without specific prior written consent. # # THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" # AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE # IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE # ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE # LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR # CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF # SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS # INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN # CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) # ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE # POSSIBILITY OF SUCH DAMAGE. ### """ Accesses Google for various things. """ __revision__ = "$Id$" import supybot.plugins as plugins import re import cgi import sets import time import getopt import socket import urllib import xml.sax import SOAP import google import supybot.registry as registry import supybot.conf as conf import supybot.utils as utils import supybot.ircmsgs as ircmsgs import supybot.plugins as plugins import supybot.ircutils as ircutils import supybot.webutils as webutils import supybot.privmsgs as privmsgs import supybot.callbacks as callbacks import supybot.structures as structures def configure(advanced): from supybot.questions import output, expect, anything, something, yn output('To use Google\'t Web Services, you must have a license key.') if yn('Do you have a license key?'): key = something('What is it?') while len(key) != 32: output('That\'s not a valid Google license key.') if yn('Are you sure you have a valid Google license key?'): key = something('What is it?') else: key = '' break if key: conf.registerPlugin('Google', True) conf.supybot.plugins.Google.licenseKey.setValue(key) output("""The Google plugin has the functionality to watch for URLs that match a specific pattern. (We call this a snarfer) When supybot sees such a URL, it will parse the web page for information and reply with the results. Google has two available snarfers: Google Groups link snarfing and a google search snarfer.""") if yn('Do you want the Google Groups link snarfer enabled by ' 'default?'): conf.supybot.plugins.Google.groupsSnarfer.setValue(True) if yn('Do you want the Google search snarfer enabled by default?'): conf.supybot.plugins.Google.searchSnarfer.setValue(True) else: output("""You'll need to get a key before you can use this plugin. You can apply for a key at http://www.google.com/apis/""") def search(log, queries, **kwargs): # We have to keep stats here, rather than in formatData or elsewhere, # because not all searching functions use formatData -- fight, lucky, etc. assert not isinstance(queries, basestring), 'Old code: queries is a list.' try: for (i, query) in enumerate(queries): if len(query.split(None, 1)) > 1: queries[i] = repr(query) proxy = conf.supybot.protocols.http.proxy() if proxy: kwargs['http_proxy'] = proxy query = ' '.join(queries).decode('utf-8') data = google.doGoogleSearch(query, **kwargs) searches = conf.supybot.plugins.Google.state.searches() + 1 conf.supybot.plugins.Google.state.searches.setValue(searches) time = conf.supybot.plugins.Google.state.time() + data.meta.searchTime conf.supybot.plugins.Google.state.time.setValue(time) last24hours.enqueue(None) return data except socket.error, e: if e.args[0] == 110: raise callbacks.Error, 'Connection timed out to Google.com.' else: raise callbacks.Error, 'Error connecting to Google.com.' except SOAP.HTTPError, e: log.info('HTTP Error accessing Google: %s', e) raise callbacks.Error, 'Error connecting to Google.com.' except SOAP.faultType, e: log.exception('Uncaught SOAP error:') raise callbacks.Error, 'Invalid Google license key.' except xml.sax.SAXException, e: log.exception('Uncaught SAX error:') raise callbacks.Error, 'Google returned an unparsable response. ' \ 'The full traceback has been logged.' except SOAP.Error, e: log.exception('Uncaught SOAP exception in Google.search:') raise callbacks.Error, 'Error connecting to Google.com.' class LicenseKey(registry.String): def setValue(self, s): if s and len(s) != 32: raise registry.InvalidRegistryValue, 'Invalid Google license key.' if s: registry.String.setValue(self, s) google.setLicense(self.value) if not s: registry.String.setValue(self, '') google.setLicense(self.value) class Language(registry.OnlySomeStrings): validStrings = ['lang_' + s for s in 'ar zh-CN zh-TW cs da nl en et fi fr ' 'de el iw hu is it ja ko lv lt no pt ' 'pl ro ru es sv tr'.split()] validStrings.append('') def normalize(self, s): if not s.startswith('lang_'): s = 'lang_' + s if not s.endswith('CN') or s.endswith('TW'): s = s.lower() else: s = s.lower()[:-2] + s[-2:] return s conf.registerPlugin('Google') conf.registerChannelValue(conf.supybot.plugins.Google, 'groupsSnarfer', registry.Boolean(False, """Determines whether the groups snarfer is enabled. If so, URLs at groups.google.com will be snarfed and their group/title messaged to the channel.""")) conf.registerChannelValue(conf.supybot.plugins.Google, 'searchSnarfer', registry.Boolean(False, """Determines whether the search snarfer is enabled. If so, messages (even unaddressed ones) beginning with the word 'google' will result in the first URL Google returns being sent to the channel.""")) conf.registerChannelValue(conf.supybot.plugins.Google, 'colorfulFilter', registry.Boolean(False, """Determines whether the word 'google' in the bot's output will be made colorful (like Google's logo).""")) conf.registerChannelValue(conf.supybot.plugins.Google, 'bold', registry.Boolean(True, """Determines whether results are bolded.""")) conf.registerChannelValue(conf.supybot.plugins.Google, 'maximumResults', registry.PositiveInteger(10, """Determines the maximum number of results returned from the google command.""")) conf.registerChannelValue(conf.supybot.plugins.Google, 'defaultLanguage', Language('lang_en', """Determines what default language is used in searches. If left empty, no specific language will be requested.""")) conf.registerChannelValue(conf.supybot.plugins.Google, 'safeSearch', registry.Boolean(True, "Determines whether safeSearch is on by default.")) conf.registerGlobalValue(conf.supybot.plugins.Google, 'licenseKey', LicenseKey('', """Sets the Google license key for using Google's Web Services API. This is necessary before you can do any searching with this module.""", private=True)) conf.registerGroup(conf.supybot.plugins.Google, 'state') conf.registerGlobalValue(conf.supybot.plugins.Google.state, 'searches', registry.Integer(0, """Used to keep the total number of searches Google has done for this bot. You shouldn't modify this.""")) conf.registerGlobalValue(conf.supybot.plugins.Google.state, 'time', registry.Float(0.0, """Used to keep the total amount of time Google has spent searching for this bot. You shouldn't modify this.""")) last24hours = structures.TimeoutQueue(86400) totalTime = conf.supybot.plugins.Google.state.time() searches = conf.supybot.plugins.Google.state.searches() class Google(callbacks.PrivmsgCommandAndRegexp): threaded = True callBefore = ['URL'] regexps = ['googleSnarfer', 'googleGroups'] def __init__(self): super(Google, self).__init__() google.setLicense(self.registryValue('licenseKey')) _colorGoogles = {} def _getColorGoogle(self, m): s = m.group(1) ret = self._colorGoogles.get(s) if not ret: L = list(s) L[0] = ircutils.mircColor(L[0], 'blue') L[1] = ircutils.mircColor(L[1], 'red') L[2] = ircutils.mircColor(L[2], 'yellow') L[3] = ircutils.mircColor(L[3], 'blue') L[4] = ircutils.mircColor(L[4], 'green') L[5] = ircutils.mircColor(L[5], 'red') ret = ''.join(L) self._colorGoogles[s] = ret return ircutils.bold(ret) _googleRe = re.compile(r'\b(google)\b', re.I) def outFilter(self, irc, msg): if msg.command == 'PRIVMSG' and \ self.registryValue('colorfulFilter', msg.args[0]): s = msg.args[1] s = re.sub(self._googleRe, self._getColorGoogle, s) msg = ircmsgs.privmsg(msg.args[0], s, prefix=msg.prefix) return msg def formatData(self, data, bold=True, max=0): if isinstance(data, basestring): return data t = 'Search took %s seconds' % data.meta.searchTime results = [] if max: data.results = data.results[:max] for result in data.results: title = utils.htmlToText(result.title.encode('utf-8')) url = result.URL if title: if bold: title = ircutils.bold(title) results.append('%s: <%s>' % (title, url)) else: results.append(url) if not results: return 'No matches found (%s)' % t else: return '%s: %s' % (t, '; '.join(results)) def lucky(self, irc, msg, args): """ Does a google search, but only returns the first result. """ if not args: raise callbacks.ArgumentError data = search(self.log, args) if data.results: url = data.results[0].URL irc.reply(url) else: irc.reply('Google found nothing.') def google(self, irc, msg, args): """ [--{language,restrict}=] [--{notsafe,similar}] Searches google.com for the given string. As many results as can fit are included. --language accepts a language abbreviation; --restrict restricts the results to certain classes of things; --similar tells Google not to filter similar results. --notsafe allows possibly work-unsafe results. """ (optlist, rest) = getopt.getopt(args, '', ['language=', 'restrict=', 'notsafe', 'similar']) kwargs = {} if self.registryValue('safeSearch', channel=msg.args[0]): kwargs['safeSearch'] = 1 lang = self.registryValue('defaultLanguage', channel=msg.args[0]) if lang: kwargs['language'] = lang for (option, argument) in optlist: if option == '--notsafe': kwargs['safeSearch'] = False elif option == '--similar': kwargs['filter'] = False else: kwargs[option[2:]] = argument if not rest: raise callbacks.ArgumentError try: data = search(self.log, rest, **kwargs) except google.NoLicenseKey, e: irc.error('You must have a free Google web services license key ' 'in order to use this command. You can get one at ' '. Once you have one, you can ' 'set it with the command ' '"config supybot.plugins.Google.licenseKey ".') return bold = self.registryValue('bold', msg.args[0]) max = self.registryValue('maximumResults', msg.args[0]) irc.reply(self.formatData(data, bold=bold, max=max)) def metagoogle(self, irc, msg, args): """ [--(language,restrict)=] [--{similar,notsafe}] Searches google and gives all the interesting meta information about the search. See the help for the google command for a detailed description of the parameters. """ (optlist, rest) = getopt.getopt(args, '', ['language=', 'restrict=', 'notsafe', 'similar']) kwargs = {'language': 'lang_en', 'safeSearch': 1} for option, argument in optlist: if option == '--notsafe': kwargs['safeSearch'] = False elif option == '--similar': kwargs['filter'] = False else: kwargs[option[2:]] = argument data = search(self.log, rest, **kwargs) meta = data.meta categories = [d['fullViewableName'] for d in meta.directoryCategories] categories = [utils.dqrepr(s.replace('_', ' ')) for s in categories] if categories: categories = utils.commaAndify(categories) else: categories = '' s = 'Search for %r returned %s %s results in %s seconds.%s' % \ (meta.searchQuery, meta.estimateIsExact and 'exactly' or 'approximately', meta.estimatedTotalResultsCount, meta.searchTime, categories and ' Categories include %s.' % categories) irc.reply(s) _cacheUrlRe = re.compile('([^<]+)') def cache(self, irc, msg, args): """ Returns a link to the cached version of if it is available. """ url = privmsgs.getArgs(args) html = google.doGetCachedPage(url) m = self._cacheUrlRe.search(html) if m is not None: url = m.group(1) url = utils.htmlToText(url) irc.reply(url) else: irc.error('Google seems to have no cache for that site.') def fight(self, irc, msg, args): """ [ ...] Returns the results of each search, in order, from greatest number of results to least. """ results = [] for arg in args: data = search(self.log, [arg]) results.append((data.meta.estimatedTotalResultsCount, arg)) results.sort() results.reverse() if self.registryValue('bold', msg.args[0]): format = ircutils.bold else: format = repr s = ', '.join(['%s: %s' % (format(s), i) for (i, s) in results]) irc.reply(s) def spell(self, irc, msg, args): """ Returns Google's spelling recommendation for . """ word = privmsgs.getArgs(args) result = google.doSpellingSuggestion(word) if result: irc.reply(result) else: irc.reply('No spelling suggestion made. This could mean that ' 'the word you gave is spelled right; it could also ' 'mean that its spelling was too whacked out even for ' 'Google to figure out.') def stats(self, irc, msg, args): """takes no arguments Returns interesting information about this Google module. Mostly useful for making sure you don't go over your 1000 requests/day limit. """ recent = len(self.last24hours) time = self.registryValue('state.time') searches = self.registryValue('state.searches') irc.reply('This google module has made %s total; ' '%s in the past 24 hours. ' 'Google has spent %s seconds searching for me.' % (utils.nItems('search', searches), recent, time)) def googleSnarfer(self, irc, msg, match): r"^google\s+(.*)$" if not self.registryValue('searchSnarfer', msg.args[0]): return searchString = match.group(1) try: data = search(self.log, [searchString], safeSearch=1) except google.NoLicenseKey: return if data.results: url = data.results[0].URL irc.reply(url, prefixName=False) googleSnarfer = privmsgs.urlSnarfer(googleSnarfer) _ggThread = re.compile(r'
Subject: ([^<]+)
', re.I) _ggGroup = re.compile(r'Newsgroups: (?:]+>)?([^<]+)(?:)?', re.I) _ggThreadm = re.compile(r'view the ]+)>no', re.I) _ggSelm = re.compile(r'selm=[^&]+', re.I) def googleGroups(self, irc, msg, match): r"http://groups.google.[\w.]+/\S+\?(\S+)" if not self.registryValue('groupsSnarfer', msg.args[0]): return queries = cgi.parse_qsl(match.group(1)) queries = [q for q in queries if q[0] in ('threadm', 'selm')] if not queries: return queries.append(('hl', 'en')) url = 'http://groups.google.com/groups?%s' % urllib.urlencode(queries) text = webutils.getUrl(url) mThread = None mGroup = None if 'threadm=' in url: path = self._ggThreadm.search(text) if path is None: return url = 'http://groups.google.com%s' % path.group(1) text = webutils.getUrl(url) mThread = self._ggThread.search(text) mGroup = self._ggGroup.search(text) if mThread and mGroup: irc.reply('Google Groups: %s, %s' % (mGroup.group(1), mThread.group(1)), prefixName=False) else: irc.errorPossibleBug('That doesn\'t appear to be a proper ' 'Google Groups page.') googleGroups = privmsgs.urlSnarfer(googleGroups) _calcRe = re.compile(r'(.*?)', re.I) _calcSupRe = re.compile(r'(.*?)', re.I) _calcFontRe = re.compile(r'(.*?)') _calcTimesRe = re.compile(r'×') def calc(self, irc, msg, args): """ Uses Google's calculator to calculate the value of . """ expr = privmsgs.getArgs(args) expr = expr.replace('+', '%2B') expr = expr.replace(' ', '+') url = r'http://google.com/search?q=%s' % expr html = webutils.getUrl(url) match = self._calcRe.search(html) if match is not None: s = match.group(1) s = self._calcSupRe.sub(r'^(\1)', s) s = self._calcFontRe.sub(r',', s) s = self._calcTimesRe.sub(r'*', s) irc.reply(s) else: irc.reply('Google\'s calculator didn\'t come up with anything.') Class = Google # vim:set shiftwidth=4 tabstop=8 expandtab textwidth=78: