mirror of
https://github.com/Mikaela/Limnoria.git
synced 2024-12-28 21:52:37 +01:00
275 lines
11 KiB
Python
275 lines
11 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.
|
|
###
|
|
|
|
"""
|
|
Provides several commands that go out to websites and get things.
|
|
|
|
Commands include:
|
|
freshmeat
|
|
stockquote
|
|
foldoc
|
|
gkstats
|
|
zipcode
|
|
weather
|
|
slashdot
|
|
"""
|
|
|
|
from baseplugin import *
|
|
|
|
import re
|
|
import time
|
|
import urllib
|
|
import urllib2
|
|
import xml.dom.minidom
|
|
|
|
import utils
|
|
import debug
|
|
import privmsgs
|
|
import callbacks
|
|
|
|
class FreshmeatException(Exception):
|
|
pass
|
|
|
|
class Http(callbacks.Privmsg):
|
|
threaded = True
|
|
_fmProject = re.compile('<projectname_full>([^<]+)</projectname_full>')
|
|
_fmVersion = re.compile('<latest_version>([^<]+)</latest_version>')
|
|
_fmVitality = re.compile('<vitality_percent>([^<]+)</vitality_percent>')
|
|
_fmPopular=re.compile('<popularity_percent>([^<]+)</popularity_percent>')
|
|
_fmLastUpdated = re.compile('<date_updated>([^<]+)</date_updated>')
|
|
def freshmeat(self, irc, msg, args):
|
|
"""<project name>"""
|
|
project = privmsgs.getArgs(args)
|
|
url = 'http://www.freshmeat.net/projects-xml/%s' % project
|
|
try:
|
|
fd = urllib2.urlopen(url)
|
|
text = fd.read()
|
|
fd.close()
|
|
if text.startswith('Error'):
|
|
raise FreshmeatException, text
|
|
project = self._fmProject.search(text).group(1)
|
|
version = self._fmVersion.search(text).group(1)
|
|
vitality = self._fmVitality.search(text).group(1)
|
|
popularity = self._fmPopular.search(text).group(1)
|
|
lastupdated = self._fmLastUpdated.search(text).group(1)
|
|
irc.reply(msg,
|
|
'%s, last updated %s, with a vitality percent of %s '\
|
|
'and a popularity of %s, is in version %s.' % \
|
|
(project, lastupdated, vitality, popularity, version))
|
|
except FreshmeatException, e:
|
|
irc.reply(msg, debug.exnToString(e))
|
|
except Exception, e:
|
|
debug.recoverableException()
|
|
irc.reply(msg, debug.exnToString(e))
|
|
|
|
def stockquote(self, irc, msg, args):
|
|
"""<company symbol>"""
|
|
symbol = privmsgs.getArgs(args)
|
|
url = 'http://finance.yahoo.com/d/quotes.csv?s=%s'\
|
|
'&f=sl1d1t1c1ohgv&e=.csv' % symbol
|
|
try:
|
|
fd = urllib2.urlopen(url)
|
|
quote = fd.read()
|
|
fd.close()
|
|
except Exception, e:
|
|
irc.reply(msg, debug.exnToString(e))
|
|
return
|
|
data = quote.split(',')
|
|
#debug.printf(data) # debugging
|
|
if data[1] != '0.00':
|
|
irc.reply(msg,
|
|
'The current price of %s is %s, as of %s EST. '\
|
|
'A change of %s from the last business day.' %\
|
|
(data[0][1:-1], data[1], data[3][1:-1], data[4]))
|
|
return
|
|
else:
|
|
m = 'I couldn\'t find a listing for %s' % symbol
|
|
irc.error(msg, m)
|
|
return
|
|
|
|
def foldoc(self, irc, msg, args):
|
|
"""<something to lookup on foldoc>"""
|
|
search = '+'.join([urllib.quote(arg) for arg in args])
|
|
url = 'http://foldoc.doc.ic.ac.uk/foldoc/foldoc.cgi?query=%s' % search
|
|
try:
|
|
fd = urllib2.urlopen(url)
|
|
html = fd.read()
|
|
fd.close()
|
|
except Exception, e:
|
|
irc.error(msg, debug.exnToString(e))
|
|
text = html.split('<P>\n', 2)[1]
|
|
text = text.replace('.\n', '. ')
|
|
text = text.replace('\n', ' ')
|
|
text = utils.htmlToText(text)
|
|
irc.reply(msg, text.strip())
|
|
|
|
_gkrating = re.compile(r'<font color="#FFFF33">(\d+)</font>')
|
|
_gkgames = re.compile(r's: </td><td class=sml>(\d+)</td></tr>')
|
|
_gkrecord = re.compile(r'"#FFFF00">(\d+)[^"]+"#FFFF00">(\d+)[^"]+'\
|
|
'"#FFFF00">(\d+)')
|
|
_gkteam = re.compile('Team:\s+(.*?)\s+Last time')
|
|
_gkseen = re.compile('seen on GK:\s+([^[]+)')
|
|
def gkstats(self, irc, msg, args):
|
|
"""<name>"""
|
|
name = privmsgs.getArgs(args)
|
|
gkprofile = 'http://www.gameknot.com/stats.pl?%s' % name
|
|
try:
|
|
fd = urllib2.urlopen(gkprofile)
|
|
profile = fd.read()
|
|
fd.close()
|
|
rating = self._gkrating.search(profile).group(1)
|
|
games = self._gkgames.search(profile).group(1)
|
|
(w, l, d) = self._gkrecord.search(profile).groups()
|
|
profile = utils.htmlToText(profile)
|
|
seen = self._gkseen.search(profile).group(1)
|
|
if profile.find('Team:') >= 0:
|
|
team = self._gkteam.search(profile).group(1)
|
|
irc.reply(msg, '%s (team: %s) is rated %s and has %s active ' \
|
|
'games and a record of W-%s, L-%s, D-%s. ' \
|
|
'%s was last seen on Gameknot %s' % \
|
|
(name, team, rating, games, w, l, d, name, seen))
|
|
else:
|
|
irc.reply(msg, '%s is rated %s and has %s active games ' \
|
|
'and a record of W-%s, L-%s, D-%s. ' \
|
|
'%s was last seen on Gameknot %s' % \
|
|
(name, rating, games, w, l, d, name, seen))
|
|
except AttributeError:
|
|
if profile.find('User %s not found!' % name) != -1:
|
|
irc.error(msg, 'No user %s exists.')
|
|
else:
|
|
irc.error(msg, 'The format of the page was odd.')
|
|
except urllib2.URLError:
|
|
irc.error(msg, 'Couldn\'t connect to gameknot.com.')
|
|
|
|
_zipcode = re.compile(r'Local Forecast for (.*), (.*?) ')
|
|
def zipcode(self, irc, msg, args):
|
|
"""<US zip code>"""
|
|
zip = privmsgs.getArgs(args)
|
|
url = "http://www.weather.com/weather/local/%s?lswe=%s" % (zip, zip)
|
|
try:
|
|
html = urllib2.urlopen(url).read()
|
|
(city, state) = self._zipcode.search(html).groups()
|
|
irc.reply(msg, '%s, %s' % (city, state))
|
|
except AttributeError:
|
|
irc.error(msg, 'the format of the page was odd.')
|
|
except urllib2.URLError:
|
|
irc.error(msg, 'Couldn\'t open search page.')
|
|
|
|
|
|
_tempregex = re.compile('CLASS=obsTempTextA>(\d+)°F</b></td>',\
|
|
re.IGNORECASE)
|
|
_cityregex = re.compile(r'Local Forecast for (.*), (.*?) ')
|
|
_condregex = re.compile('CLASS=obsInfo2><b CLASS=obsTextA>(.*)</b></td>',\
|
|
re.IGNORECASE)
|
|
def weather(self, irc, msg, args):
|
|
"""<US zip code>"""
|
|
zip = privmsgs.getArgs(args)
|
|
url = "http://www.weather.com/weather/local/%s?lswe=%s" % (zip, zip)
|
|
try:
|
|
html = urllib2.urlopen(url).read()
|
|
city, state = self._cityregex.search(html).groups()
|
|
temp = self._tempregex.search(html).group(1)
|
|
conds = self._condregex.search(html).group(1)
|
|
irc.reply(msg, 'The current temperature in %s, %s is %dF with %s'\
|
|
' conditions' % (city, state, int(temp), conds))
|
|
except AttributeError:
|
|
irc.error(msg, 'the format of the page was odd.')
|
|
except urllib2.URLError:
|
|
irc.error(msg, 'Couldn\'t open the search page.')
|
|
|
|
_slashdotTime = 0.0
|
|
def slashdot(self, irc, msg, args):
|
|
"""takes no arguments"""
|
|
if time.time() - self._slashdotTime > 1800:
|
|
try:
|
|
fd = urllib2.urlopen('http://slashdot.org/slashdot.xml')
|
|
slashdotXml = fd.read()
|
|
fd.close()
|
|
dom = xml.dom.minidom.parseString(slashdotXml)
|
|
headlines = []
|
|
for headline in dom.getElementsByTagName('title'):
|
|
headlines.append(str(headline.firstChild.data))
|
|
self._slashdotResponse = ' :: '.join(headlines)
|
|
self._slashdotTime = time.time()
|
|
except urllib2.URLError, e:
|
|
irc.error(msg, str(e))
|
|
return
|
|
irc.reply(msg, self._slashdotResponse)
|
|
|
|
_geekquotere = re.compile('<p class="qt">(.*?)</p>')
|
|
def geekquote(self, irc, msg, args):
|
|
"[<multiline>]"
|
|
multiline = privmsgs.getArgs(args, needed=0, optional=1)
|
|
try:
|
|
fd = urllib2.urlopen('http://bash.org/?random1')
|
|
except urllib2.URLError:
|
|
irc.error(msg, 'Error connecting to geekquote server.')
|
|
return
|
|
html = fd.read()
|
|
fd.close()
|
|
if multiline:
|
|
m = self._geekquotere.search(html, re.M)
|
|
else:
|
|
m = self._geekquotere.search(html)
|
|
if m is None:
|
|
irc.error(msg, 'No quote found.')
|
|
return
|
|
quote = utils.htmlToText(m.group(1))
|
|
quote = ' // '.join(quote.splitlines())
|
|
irc.reply(msg, quote)
|
|
|
|
_acronymre = re.compile(r'<td[^w]+width="70[^>]+>(?:<b>)?([^<]+)(?:</b>)?')
|
|
def acronym(self, irc, msg, args):
|
|
"""<acronym>
|
|
|
|
displays the first 5 acronym matches"""
|
|
acronym = privmsgs.getArgs(args)
|
|
try:
|
|
url = 'http://www.acronymfinder.com/' \
|
|
'af-query.asp?String=exact&Acronym=%s' % acronym
|
|
request = urllib2.Request(url, headers={'User-agent':
|
|
'Mozilla/4.0 (compatible; MSIE 5.5; Windows NT 4.0)'})
|
|
fd = urllib2.urlopen(request)
|
|
except urllib2.URLError:
|
|
irc.error(msg, 'Couldn\'t connect to acronymfinder.com')
|
|
return
|
|
html = fd.read()
|
|
fd.close()
|
|
defs = self._acronymre.findall(html)
|
|
if len(defs) == 0:
|
|
irc.reply(msg, 'No definitions found.')
|
|
else:
|
|
defs=[repr(x.strip()) for x in defs[1:-1]][:5]
|
|
irc.reply(msg, '%s could be %s' % (acronym, ', or '.join(defs)))
|
|
|
|
Class = Http
|
|
# vim:set shiftwidth=4 tabstop=8 expandtab textwidth=78:
|