mirror of
https://github.com/Mikaela/Limnoria.git
synced 2024-11-09 12:29:22 +01:00
267 lines
10 KiB
Python
267 lines
10 KiB
Python
#!/usr/bin/env python
|
|
|
|
###
|
|
# Copyright (c) 2002, 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.
|
|
###
|
|
|
|
"""
|
|
Acceses Google for various things.
|
|
"""
|
|
|
|
from baseplugin import *
|
|
|
|
import re
|
|
import sets
|
|
import time
|
|
import getopt
|
|
import urllib2
|
|
|
|
import google
|
|
|
|
import utils
|
|
import ircmsgs
|
|
import privmsgs
|
|
import callbacks
|
|
import structures
|
|
|
|
def configure(onStart, afterConnect, advanced):
|
|
from questions import expect, anything, something, yn
|
|
print 'To use Google\'t Web Services, you must have a license key.'
|
|
if yn('Do you have a license key?') == 'y':
|
|
key = anything('What is it?')
|
|
onStart.append('load Google')
|
|
onStart.append('googlelicensekey %s' % key)
|
|
if yn('Google depends on the Alias module for some commands. ' \
|
|
'Is the Alias module loaded?') == 'n':
|
|
if yn('Would you like to load the Alias module now?') == 'y':
|
|
onStart.append('load Alias')
|
|
else:
|
|
print 'You can still use the Google module, but you won\'t ' \
|
|
'be asked any further questions.'
|
|
return
|
|
onStart.append('alias googlelinux "google --restrict=linux $1"')
|
|
onStart.append('alias googlebsd "google --restrict=bsd $1"')
|
|
onStart.append('alias googlemac "google --restrict=mac $1"')
|
|
else:
|
|
print 'You\'ll need to get a key before you can use this plugin.'
|
|
print 'You can apply for a key at http://www.google.com/apis/'
|
|
|
|
totalSearches = 0
|
|
totalTime = 0
|
|
last24hours = structures.queue()
|
|
|
|
def search(*args, **kwargs):
|
|
global totalSearches, totalTime, last24hours
|
|
data = google.doGoogleSearch(*args, **kwargs)
|
|
now = time.time()
|
|
totalSearches += 1
|
|
totalTime += data.meta.searchTime
|
|
last24hours.enqueue(now)
|
|
while last24hours and now - last24hours.peek() > 86400:
|
|
last24hours.dequeue()
|
|
return data
|
|
|
|
class Google(callbacks.PrivmsgCommandAndRegexp):
|
|
threaded = True
|
|
regexps = sets.Set(['googleSnarfer', 'googleGroups'])
|
|
def __init__(self):
|
|
callbacks.PrivmsgCommandAndRegexp.__init__(self)
|
|
self.total = 0
|
|
self.totalTime = 0
|
|
self.last24hours = structures.queue()
|
|
|
|
def formatData(self, data):
|
|
time = '(search took %s seconds)' % data.meta.searchTime
|
|
results = []
|
|
for result in data.results:
|
|
title = utils.htmlToText(result.title.encode('utf-8'))
|
|
url = result.URL
|
|
if title:
|
|
results.append('\x02%s\x02: %s' % (title, url))
|
|
else:
|
|
results.append(url)
|
|
if not results:
|
|
return 'No matches found %s' % time
|
|
else:
|
|
s = ircutils.privmsgPayload(results, ' :: ', 375)
|
|
return '%s %s' % (s, time)
|
|
|
|
def googlelicensekey(self, irc, msg, args):
|
|
"""<key>
|
|
|
|
Sets the Google license key for using Google's Web Services API. This
|
|
is necessary before you can do any searching with this module.
|
|
"""
|
|
key = privmsgs.getArgs(args)
|
|
google.setLicense(key)
|
|
irc.reply(msg, conf.replySuccess)
|
|
|
|
googlelicensekey = privmsgs.checkCapability(googlelicensekey, 'admin')
|
|
|
|
def google(self, irc, msg, args):
|
|
"""<search string> [--{language,restrict,safe,filter}=<value>]
|
|
|
|
Searches google.com for the given string. As many results as can fit
|
|
are included. Use options to set different values for the options
|
|
Google accepts.
|
|
"""
|
|
(optlist, rest) = getopt.getopt(args, '', ['language=', 'restrict=',
|
|
'safe=', 'filter='])
|
|
kwargs = {'language': 'lang_en', 'safeSearch': 1}
|
|
for (option, argument) in optlist:
|
|
kwargs[option[2:]] = argument
|
|
searchString = privmsgs.getArgs(rest)
|
|
data = search(searchString, **kwargs)
|
|
irc.reply(msg, self.formatData(data))
|
|
|
|
def metagoogle(self, irc, msg, args):
|
|
"""<search string> [--(language,restrict,safe,filter)=<value>]
|
|
|
|
Searches google and gives all the interesting meta information about
|
|
the search.
|
|
"""
|
|
(optlist, rest) = getopt.getopt(args, '', ['language=', 'restrict=',
|
|
'safe=', 'filter='])
|
|
kwargs = {'language': 'lang_en', 'safeSearch': 1}
|
|
for option, argument in optlist:
|
|
kwargs[option[2:]] = argument
|
|
searchString = privmsgs.getArgs(rest)
|
|
data = search(searchString, **kwargs)
|
|
meta = data.meta
|
|
categories = [d['fullViewableName'] for d in meta.directoryCategories]
|
|
categories = [repr(s.replace('_', ' ')) for s in categories]
|
|
if len(categories) > 1:
|
|
categories = ', and '.join([', '.join(categories[:-1]),
|
|
categories[-1]])
|
|
elif not categories:
|
|
categories = ''
|
|
else:
|
|
categories = categories[0]
|
|
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(msg, s)
|
|
|
|
def googlefight(self, irc, msg, args):
|
|
"""<search string> <search string> [<search string> ...]
|
|
|
|
Returns the results of each search, in order, from greatest number
|
|
of results to least.
|
|
"""
|
|
|
|
results = []
|
|
for arg in args:
|
|
data = search(arg)
|
|
results.append((data.meta.estimatedTotalResultsCount, arg))
|
|
results.sort()
|
|
results.reverse()
|
|
s = ', '.join(['%r: %s' % (s, i) for (i, s) in results])
|
|
irc.reply(msg, s)
|
|
|
|
def googlesite(self, irc, msg, args):
|
|
"""<site> <search string>
|
|
|
|
Searches Google on a specific site.
|
|
"""
|
|
(site, s) = privmsgs.getArgs(args, needed=2)
|
|
searchString = 'site:%s %s' % (site, s)
|
|
data = search(searchString, language='lang_en', safeSearch=1)
|
|
irc.reply(msg, self.formatData(data))
|
|
|
|
def googlespell(self, irc, msg, args):
|
|
"<word>"
|
|
word = privmsgs.getArgs(args)
|
|
result = google.doSpellingSuggestion(word)
|
|
if result:
|
|
irc.reply(msg, result)
|
|
else:
|
|
irc.reply(msg, 'No spelling suggestion made.')
|
|
|
|
def googleinfo(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(last24hours)
|
|
irc.reply(msg, 'This google module has been called %s time%stotal; '\
|
|
'%s time%sin the past 24 hours. ' \
|
|
'Google has spent %s seconds searching for me.' % \
|
|
(totalSearches, totalSearches != 1 and 's ' or ' ',
|
|
recent, recent != 1 and 's ' or ' ',
|
|
totalTime))
|
|
|
|
def googleSnarfer(self, irc, msg, match):
|
|
r"^google\s+(.*)$"
|
|
searchString = match.group(1)
|
|
data = search(searchString, safeSearch=1)
|
|
if data.results:
|
|
url = data.results[0].URL
|
|
irc.queueMsg(ircmsgs.privmsg(ircutils.replyTo(msg), url))
|
|
else:
|
|
irc.queueMsg(ircmsgs.privmsg(ircutils.replyTo(msg),
|
|
'No results for "%s"' % searchString))
|
|
|
|
_ggThread = re.compile(r'<br>Subject: ([^<]+)<br>')
|
|
_ggGroup = re.compile(r'Newsgroups: <a[^>]+>([^<]+)</a>')
|
|
def googleGroups(self, irc, msg, match):
|
|
r"http://groups.google.com/[^\s]+"
|
|
request = urllib2.Request(match.group(0), headers=\
|
|
{'User-agent': 'Mozilla/4.0 (compatible; MSIE 5.5; Windows NT 4.0)'})
|
|
fd = urllib2.urlopen(request)
|
|
text = fd.read()
|
|
fd.close()
|
|
if match.group(0).find('&prev=/') >= 0:
|
|
path = re.search('view the <a href=([^>]+)>no',text)
|
|
if path is None:
|
|
return
|
|
url = 'http://groups.google.com'
|
|
request = urllib2.Request('%s%s' % (url,path.group(1)),
|
|
headers={'User-agent': 'Mozilla/4.0 (compatible; MSIE 5.5;'
|
|
'Windows NT 4.0)'})
|
|
fd = urllib2.urlopen(request)
|
|
text = fd.read()
|
|
fd.close()
|
|
mThread = self._ggThread.search(text)
|
|
mGroup = self._ggGroup.search(text)
|
|
if mThread and mGroup:
|
|
irc.queueMsg(ircmsgs.privmsg(ircutils.replyTo(msg),
|
|
'Google Groups: %s, %s' % (mGroup.group(1), mThread.group(1))))
|
|
else:
|
|
irc.queueMsg(ircmsgs.privmsg(msg.args[0],
|
|
'That doesn\'t appear to be a proper Google Groups page.'))
|
|
|
|
|
|
|
|
Class = Google
|
|
|
|
# vim:set shiftwidth=4 tabstop=8 expandtab textwidth=78:
|