mirror of
https://github.com/Mikaela/Limnoria.git
synced 2024-12-30 06:32:34 +01:00
349 lines
13 KiB
Python
349 lines
13 KiB
Python
#!/usr/bin/env python
|
|
|
|
###
|
|
# 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.
|
|
###
|
|
|
|
"""
|
|
Keeps track of URLs posted to a channel, along with relevant context. Allows
|
|
searching for URLs and returning random URLs. Also provides statistics on the
|
|
URLs in the database.
|
|
"""
|
|
|
|
__revision__ = "$Id$"
|
|
|
|
import supybot.plugins as plugins
|
|
|
|
import os
|
|
import re
|
|
import sets
|
|
import time
|
|
import getopt
|
|
import urlparse
|
|
import itertools
|
|
|
|
import supybot.conf as conf
|
|
import supybot.utils as utils
|
|
import supybot.ircmsgs as ircmsgs
|
|
import supybot.webutils as webutils
|
|
import supybot.ircutils as ircutils
|
|
import supybot.privmsgs as privmsgs
|
|
import supybot.registry as registry
|
|
import supybot.callbacks as callbacks
|
|
|
|
def configure(advanced):
|
|
from supybot.questions import output, expect, anything, something, yn
|
|
conf.registerPlugin('URL', True)
|
|
if yn("""This plugin offers a snarfer that will go to tinyurl.com and get
|
|
a shorter version of long URLs that are sent to the channel.
|
|
Would you like this snarfer to be enabled?""", default=False):
|
|
conf.supybot.plugins.URL.tinyurlSnarfer.setValue(True)
|
|
if yn("""This plugin also offers a snarfer that will try to fetch the
|
|
title of URLs that it sees in the channel. Would like you this
|
|
snarfer to be enabled?""", default=False):
|
|
conf.supybot.plugins.URL.titleSnarfer.setValue(True)
|
|
|
|
conf.registerPlugin('URL')
|
|
conf.registerChannelValue(conf.supybot.plugins.URL, 'tinyurlSnarfer',
|
|
registry.Boolean(False, """Determines whether the
|
|
tinyurl snarfer is enabled. This snarfer will watch for URLs in the
|
|
channel, and if they're sufficiently long (as determined by
|
|
supybot.plugins.URL.tinyurlSnarfer.minimumLength) it will post a smaller
|
|
from tinyurl.com."""))
|
|
conf.registerChannelValue(conf.supybot.plugins.URL.tinyurlSnarfer,
|
|
'minimumLength',
|
|
registry.PositiveInteger(48, """The minimum length a URL must be before the
|
|
tinyurl snarfer will snarf it."""))
|
|
conf.registerChannelValue(conf.supybot.plugins.URL, 'titleSnarfer',
|
|
registry.Boolean(False, """Determines whether the bot will output the HTML
|
|
title of URLs it sees in the channel."""))
|
|
conf.registerChannelValue(conf.supybot.plugins.URL, 'nonSnarfingRegexp',
|
|
registry.Regexp(None, """Determines what URLs are to be snarfed and stored
|
|
in the database in the channel; URLs matching the regexp given will not be
|
|
snarfed. Give the empty string if you have no URLs that you'd like to
|
|
exclude from being snarfed."""))
|
|
|
|
class URLDB(object):
|
|
def __init__(self, channel, log):
|
|
self.log = log
|
|
self.filename = plugins.makeChannelFilename(channel, 'URL.db')
|
|
|
|
def _getFile(self):
|
|
try:
|
|
fd = file(self.filename)
|
|
return fd
|
|
except EnvironmentError, e:
|
|
self.log.warning('Couldn\'t open %s: %s',
|
|
self.filename, utils.exnToString(e))
|
|
return None
|
|
|
|
def _formatRecord(self, url, nick):
|
|
return '%s %s\n' % (url, nick)
|
|
|
|
def addUrl(self, url, nick):
|
|
fd = file(self.filename, 'a')
|
|
fd.write(self._formatRecord(url, nick))
|
|
fd.close()
|
|
|
|
def numUrls(self):
|
|
fd = self._getFile()
|
|
if fd is None:
|
|
return 0
|
|
try:
|
|
return itertools.ilen(fd)
|
|
finally:
|
|
fd.close()
|
|
|
|
def getUrlsAndNicks(self, p=None):
|
|
L = []
|
|
fd = self._getFile()
|
|
if fd is None:
|
|
return []
|
|
try:
|
|
for line in fd:
|
|
line = line.strip()
|
|
(url, nick) = line.split()
|
|
if p(url, nick):
|
|
L.append((url, nick))
|
|
seen = sets.Set()
|
|
L.reverse()
|
|
for (i, (url, nick)) in enumerate(L):
|
|
if url in seen:
|
|
L[i] = None
|
|
else:
|
|
seen.add(url)
|
|
L = filter(None, L)
|
|
return L
|
|
finally:
|
|
fd.close()
|
|
|
|
def getUrls(self, p):
|
|
return [url for (url, nick) in self.getUrlsAndNicks(p)]
|
|
|
|
def vacuum(self):
|
|
out = utils.transactionalFile(self.filename)
|
|
notAdded = 0
|
|
urls = self.getUrlsAndNicks(lambda *args: True)
|
|
seen = sets.Set()
|
|
for (i, (url, nick)) in enumerate(urls):
|
|
if url not in seen:
|
|
seen.add(url)
|
|
else:
|
|
urls[i] = None
|
|
notAdded += 1
|
|
urls.reverse()
|
|
for urlNick in urls:
|
|
if urlNick is not None:
|
|
out.write(self._formatRecord(*urlNick))
|
|
out.close()
|
|
self.log.info('Vacuumed %s, removed %s records.',
|
|
self.filename, notAdded)
|
|
|
|
class URL(callbacks.PrivmsgCommandAndRegexp):
|
|
priority = 100 # lower than 99, the normal priority.
|
|
regexps = ['titleSnarfer', 'tinyurlSnarfer']
|
|
_titleRe = re.compile('<title>(.*?)</title>', re.I | re.S)
|
|
def getDb(self, channel):
|
|
return URLDB(channel, self.log)
|
|
|
|
def doPrivmsg(self, irc, msg):
|
|
channel = msg.args[0]
|
|
if ircutils.isChannel(channel):
|
|
db = self.getDb(channel)
|
|
if ircmsgs.isAction(msg):
|
|
text = ircmsgs.unAction(msg)
|
|
else:
|
|
text = msg.args[1]
|
|
for url in webutils.urlRe.findall(text):
|
|
r = self.registryValue('nonSnarfingRegexp', channel)
|
|
if r and r.search(url):
|
|
self.log.debug('Skipping adding %r to db.', url)
|
|
continue
|
|
self.log.debug('Adding %r to db.', url)
|
|
db.addUrl(url, msg.nick)
|
|
callbacks.PrivmsgCommandAndRegexp.doPrivmsg(self, irc, msg)
|
|
|
|
def tinyurlSnarfer(self, irc, msg, match):
|
|
r"https?://[^\])>\s]{13,}"
|
|
channel = msg.args[0]
|
|
if not ircutils.isChannel(channel):
|
|
return
|
|
r = self.registryValue('nonSnarfingRegexp', channel)
|
|
if self.registryValue('tinyurlSnarfer', channel):
|
|
url = match.group(0)
|
|
if r and r.search(url) is not None:
|
|
self.log.debug('Not tinyUrlSnarfing %r', url)
|
|
return
|
|
minlen = self.registryValue('tinyurlSnarfer.minimumLength',channel)
|
|
if len(url) >= minlen:
|
|
tinyurl = self._getTinyUrl(url, channel)
|
|
if tinyurl is None:
|
|
self.log.warning('Couldn\'t get tinyurl for %r', url)
|
|
return
|
|
domain = webutils.getDomain(url)
|
|
s = '%s (at %s)' % (ircutils.bold(tinyurl), domain)
|
|
irc.reply(s, prefixName=False)
|
|
tinyurlSnarfer = privmsgs.urlSnarfer(tinyurlSnarfer)
|
|
|
|
def titleSnarfer(self, irc, msg, match):
|
|
r"https?://[^\])>\s]+"
|
|
channel = msg.args[0]
|
|
if not ircutils.isChannel(channel):
|
|
return
|
|
if callbacks.addressed(irc.nick, msg):
|
|
return
|
|
if self.registryValue('titleSnarfer', channel):
|
|
url = match.group(0)
|
|
r = self.registryValue('nonSnarfingRegexp', channel)
|
|
if r and r.search(url):
|
|
self.log.debug('Not titleSnarfing %r.', url)
|
|
return
|
|
try:
|
|
size = conf.supybot.protocols.http.peekSize()
|
|
text = webutils.getUrl(url, size=size)
|
|
except webutils.WebError, e:
|
|
self.log.info('Couldn\'t snarf title of %s, %s.', url, e)
|
|
return
|
|
m = self._titleRe.search(text)
|
|
if m is not None:
|
|
domain = webutils.getDomain(url)
|
|
title = utils.htmlToText(m.group(1).strip())
|
|
s = 'Title: %s (at %s)' % (title, domain)
|
|
irc.reply(s, prefixName=False)
|
|
titleSnarfer = privmsgs.urlSnarfer(titleSnarfer)
|
|
|
|
_tinyRe = re.compile(r'<blockquote><b>(http://tinyurl\.com/\w+)</b>')
|
|
def _getTinyUrl(self, url, channel, cmd=False):
|
|
try:
|
|
s = webutils.getUrl('http://tinyurl.com/create.php?url=%s' % url)
|
|
m = self._tinyRe.search(s)
|
|
if m is None:
|
|
tinyurl = None
|
|
else:
|
|
tinyurl = m.group(1)
|
|
return tinyurl
|
|
except webutils.WebError, e:
|
|
if cmd:
|
|
raise callbacks.Error, e
|
|
else:
|
|
self.log.warning(str(e))
|
|
|
|
def tiny(self, irc, msg, args):
|
|
"""<url>
|
|
|
|
Returns a TinyURL.com version of <url>
|
|
"""
|
|
url = privmsgs.getArgs(args)
|
|
if len(url) < 20:
|
|
irc.error('Stop being a lazy-biotch and type the URL yourself.')
|
|
return
|
|
channel = msg.args[0]
|
|
snarf = self.registryValue('tinyurlSnarfer', channel)
|
|
minlen = self.registryValue('tinyurlSnarfer.minimumLength', channel)
|
|
dontSnarf = False
|
|
r = self.registryValue('nonSnarfingRegexp', channel)
|
|
if r is not None:
|
|
dontSnarf = r.search(url)
|
|
dontSnarf = not dontSnarf
|
|
if snarf and len(url) >= minlen and dontSnarf:
|
|
self.log.debug('Not applying tiny command, snarfer is active.')
|
|
return
|
|
tinyurl = self._getTinyUrl(url, channel, cmd=True)
|
|
domain = webutils.getDomain(url)
|
|
s = '%s (at %s)' % (ircutils.bold(tinyurl), domain)
|
|
if tinyurl is not None:
|
|
irc.reply(s)
|
|
else:
|
|
s = 'Could not parse the TinyURL.com results page.'
|
|
irc.errorPossibleBug(s)
|
|
tiny = privmsgs.thread(tiny)
|
|
|
|
def stats(self, irc, msg, args):
|
|
"""[<channel>]
|
|
|
|
Returns the number of URLs in the URL database. <channel> is only
|
|
required if the message isn't sent in the channel itself.
|
|
"""
|
|
channel = privmsgs.getChannel(msg, args)
|
|
db = self.getDb(channel)
|
|
db.vacuum()
|
|
count = db.numUrls()
|
|
irc.reply('I have %s in my database.' % utils.nItems('URL', count))
|
|
|
|
def last(self, irc, msg, args):
|
|
"""[<channel>] [--{from,with,proto}=<value>] --{nolimit}
|
|
|
|
Gives the last URL matching the given criteria. --from is from whom
|
|
the URL came; --proto is the protocol the URL used; --with is something
|
|
inside the URL; If --nolimit is given, returns all the URLs that are
|
|
found. to just the URL. <channel> is only necessary if the message
|
|
isn't sent in the channel itself.
|
|
"""
|
|
channel = privmsgs.getChannel(msg, args)
|
|
(optlist, rest) = getopt.getopt(args, '', ['from=', 'with=',
|
|
'proto=', 'nolimit',])
|
|
predicates = []
|
|
nolimit = False
|
|
for (option, arg) in optlist:
|
|
if option == '--nolimit':
|
|
nolimit = True
|
|
elif option == '--from':
|
|
def from_(url, nick, arg=arg):
|
|
return nick.lower() == arg.lower()
|
|
predicates.append(from_)
|
|
elif option == '--with':
|
|
def with(url, nick, arg=arg):
|
|
return arg in url
|
|
predicates.append(with)
|
|
elif option == '--proto':
|
|
def proto(url, nick, arg=arg):
|
|
return url.startswith(arg)
|
|
predicates.append(proto)
|
|
db = self.getDb(channel)
|
|
def predicate(url, nick):
|
|
for predicate in predicates:
|
|
if not predicate(url, nick):
|
|
return False
|
|
return True
|
|
urls = db.getUrls(predicate)
|
|
if not urls:
|
|
irc.reply('No URLs matched that criteria.')
|
|
else:
|
|
if nolimit:
|
|
urls = ['<%s>' % url for url in urls]
|
|
s = ', '.join(urls)
|
|
else:
|
|
# We should optimize this with another URLDB method eventually.
|
|
s = urls[0]
|
|
irc.reply(s)
|
|
|
|
|
|
Class = URL
|
|
|
|
# vim:set shiftwidth=4 tabstop=8 expandtab textwidth=78:
|