Added RSS in the new plugin format.

This commit is contained in:
Jeremy Fincher 2005-02-02 06:03:09 +00:00
parent c409493d3d
commit 3f1b0335c2
7 changed files with 3178 additions and 0 deletions

1
plugins/RSS/README.txt Normal file
View File

@ -0,0 +1 @@
Insert a description of your plugin here, with any notes, etc. about using it.

60
plugins/RSS/__init__.py Normal file
View File

@ -0,0 +1,60 @@
###
# Copyright (c) 2005, 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 basic functionality for handling RSS/RDF feeds.
"""
import supybot
import supybot.world as world
# Use this for the version of this plugin. You may wish to put a CVS keyword
# in here if you're keeping the plugin in CVS or some similar system.
__version__ = "%%VERSION%%"
__author__ = supybot.authors.jemfinch
# This is a dictionary mapping supybot.Author instances to lists of
# contributions.
__contributors__ = {}
import config
import plugin
reload(plugin) # In case we're being reloaded.
# Add more reloads here if you add third-party modules and want them to be
# reloaded when this plugin is reloaded. Don't forget to import them as well!
if world.testing:
import test
Class = plugin.Class
configure = config.configure
# vim:set shiftwidth=4 tabstop=8 expandtab textwidth=78:

82
plugins/RSS/config.py Normal file
View File

@ -0,0 +1,82 @@
###
# Copyright (c) 2005, 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.
###
import supybot.conf as conf
import supybot.registry as registry
import supybot.callbacks as callbacks
def configure(advanced):
# This will be called by supybot to configure this module. advanced is
# a bool that specifies whether the user identified himself as an advanced
# user or not. You should effect your configuration by manipulating the
# registry as appropriate.
from supybot.questions import expect, anything, something, yn
conf.registerPlugin('RSS', True)
class AnnouncedFeeds(registry.SpaceSeparatedListOfStrings):
List = callbacks.CanonicalNameSet
RSS = conf.registerPlugin('RSS')
conf.registerChannelValue(RSS, 'bold', registry.Boolean(
True, """Determines whether the bot will bold the title of the feed when it
announces new news."""))
conf.registerChannelValue(RSS, 'headlineSeparator',
registry.StringSurroundedBySpaces(' || ', """Determines what string is used
to separate headlines in new feeds."""))
conf.registerChannelValue(RSS, 'announcementPrefix',
registry.StringWithSpaceOnRight('New news from ', """Determines what prefix
is prepended (if any) to the new news item announcements made in the
channel."""))
conf.registerChannelValue(RSS, 'announce',
AnnouncedFeeds([], """Determines which RSS feeds should be announced in the
channel; valid input is a list of strings (either registered RSS feeds or
RSS feed URLs) separated by spaces."""))
conf.registerGlobalValue(RSS, 'waitPeriod',
registry.PositiveInteger(1800, """Indicates how many seconds the bot will
wait between retrieving RSS feeds; requests made within this period will
return cached results."""))
conf.registerGlobalValue(RSS, 'feeds',
AnnouncedFeeds([], """Determines what feeds should be accessible as
commands."""))
conf.registerChannelValue(RSS, 'showLinks',
registry.Boolean(False, """Determines whether the bot will list the link
along with the title of the feed when the rss command is called.
supybot.plugins.RSS.announce.showLinks affects whether links will be
listed when a feed is automatically announced."""))
conf.registerGroup(RSS, 'announce')
conf.registerChannelValue(RSS.announce, 'showLinks',
registry.Boolean(False, """Determines whether the bot will list the link
along with the title of the feed when a feed is automatically
announced."""))
# vim:set shiftwidth=4 tabstop=8 expandtab textwidth=78

389
plugins/RSS/plugin.py Normal file
View File

@ -0,0 +1,389 @@
###
# 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.
###
import time
import socket
import sgmllib
import threading
import rssparser
import supybot.conf as conf
import supybot.utils as utils
import supybot.world as world
from supybot.commands import *
import supybot.ircutils as ircutils
import supybot.registry as registry
import supybot.callbacks as callbacks
def getFeedName(irc, msg, args, state):
if not registry.isValidRegistryName(args[0]):
irc.errorInvalid('feed name', name,
'Feed names must not include spaces.')
state.args.append(callbacks.canonicalName(args.pop(0)))
addConverter('feedName', getFeedName)
class RSS(callbacks.Privmsg):
"""This plugin is useful both for announcing updates to RSS feeds in a
channel, and for retrieving the headlines of RSS feeds via command. Use
the "add" command to add feeds to this plugin, and use the "announce"
command to determine what feeds should be announced in a given channel."""
threaded = True
def __init__(self, irc):
self.__parent = super(RSS, self)
self.__parent.__init__(irc)
self.feedNames = callbacks.CanonicalNameSet()
self.locks = {}
self.lastRequest = {}
self.cachedFeeds = {}
self.gettingLockLock = threading.Lock()
for name in self.registryValue('feeds'):
self._registerFeed(name)
try:
url = self.registryValue('feeds.%s' % name)
except registry.NonExistentRegistryEntry:
self.log.warning('%s is not a registered feed, removing.',name)
continue
self.makeFeedCommand(name, url)
self.getFeed(url) # So announced feeds don't announce on startup.
def _registerFeed(self, name, url=''):
self.registryValue('feeds').add(name)
group = self.registryValue('feeds', value=False)
group.register(name, registry.String(url, ''))
def __call__(self, irc, msg):
self.__parent.__call__(irc, msg)
irc = callbacks.SimpleProxy(irc, msg)
newFeeds = {}
for channel in irc.state.channels:
feeds = self.registryValue('announce', channel)
for name in feeds:
commandName = callbacks.canonicalName(name)
if self.isCommand(commandName):
name = commandName
url = self.getCommand(name).url
else:
url = name
if self.willGetNewFeed(url):
newFeeds.setdefault((url, name), []).append(channel)
for ((url, name), channels) in newFeeds.iteritems():
# We check if we can acquire the lock right here because if we
# don't, we'll possibly end up spawning a lot of threads to get
# the feed, because this thread may run for a number of bytecodes
# before it switches to a thread that'll get the lock in
# _newHeadlines.
if self.acquireLock(url, blocking=False):
try:
t = threading.Thread(target=self._newHeadlines,
name=format('Fetching %u', url),
args=(irc, channels, name, url))
self.log.info('Checking for announcements at %u', url)
world.threadsSpawned += 1
t.setDaemon(True)
t.start()
finally:
self.releaseLock(url)
time.sleep(0.1) # So other threads can run.
def buildHeadlines(self, headlines, channel, config='announce.showLinks'):
newheadlines = []
if self.registryValue(config, channel):
for headline in headlines:
if headline[1]:
newheadlines.append(format('%s %u', *headline))
else:
newheadlines.append(format('%s', headline[0]))
else:
for headline in headlines:
newheadlines = [format('%s', h[0]) for h in headlines]
return newheadlines
def _newHeadlines(self, irc, channels, name, url):
try:
# We acquire the lock here so there's only one announcement thread
# in this code at any given time. Otherwise, several announcement
# threads will getFeed (all blocking, in turn); then they'll all
# want to sent their news messages to the appropriate channels.
# Note that we're allowed to acquire this lock twice within the
# same thread because it's an RLock and not just a normal Lock.
self.acquireLock(url)
try:
oldresults = self.cachedFeeds[url]
oldheadlines = self.getHeadlines(oldresults)
except KeyError:
oldheadlines = []
newresults = self.getFeed(url)
newheadlines = self.getHeadlines(newresults)
def canonicalize(headline):
return (tuple(headline[0].lower().split()), headline[1])
oldheadlines = set(map(canonicalize, oldheadlines))
for (i, headline) in enumerate(newheadlines):
if canonicalize(headline) in oldheadlines:
newheadlines[i] = None
newheadlines = filter(None, newheadlines) # Removes Nones.
if newheadlines:
for channel in channels:
bold = self.registryValue('bold', channel)
sep = self.registryValue('headlineSeparator', channel)
prefix = self.registryValue('announcementPrefix', channel)
pre = format('%s%s: ', prefix, name)
if bold:
pre = ircutils.bold(pre)
sep = ircutils.bold(sep)
headlines = self.buildHeadlines(newheadlines, channel)
irc.replies(headlines, prefixer=pre, joiner=sep,
to=channel, prefixName=False, private=True)
finally:
self.releaseLock(url)
def willGetNewFeed(self, url):
now = time.time()
wait = self.registryValue('waitPeriod')
if url not in self.lastRequest or now - self.lastRequest[url] > wait:
return True
else:
return False
def acquireLock(self, url, blocking=True):
try:
self.gettingLockLock.acquire()
try:
lock = self.locks[url]
except KeyError:
lock = threading.RLock()
self.locks[url] = lock
return lock.acquire(blocking=blocking)
finally:
self.gettingLockLock.release()
def releaseLock(self, url):
self.locks[url].release()
def getFeed(self, url):
def error(s):
return {'items': [{'title': s}]}
try:
# This is the most obvious place to acquire the lock, because a
# malicious user could conceivably flood the bot with rss commands
# and DoS the website in question.
self.acquireLock(url)
if self.willGetNewFeed(url):
try:
self.log.debug('Downloading new feed from %u', url)
results = rssparser.parse(url)
if 'bozo_exception' in results:
## for (k, v) in results.items():
## s = '%r: %r' % (k, v)
## if len(s) <= 80:
## print s
raise results['bozo_exception']
except sgmllib.SGMLParseError:
self.log.exception('Uncaught exception from rssparser:')
raise callbacks.Error, 'Invalid (unparsable) RSS feed.'
except socket.timeout:
return error('Timeout downloading feed.')
except Exception, e:
# These seem mostly harmless. We'll need reports of a
# kind that isn't.
self.log.debug('Allowing bozo_exception %r through.', e)
self.cachedFeeds[url] = results
self.lastRequest[url] = time.time()
try:
return self.cachedFeeds[url]
except KeyError:
self.lastRequest[url] = 0
return error('Unable to download feed.')
finally:
self.releaseLock(url)
def getHeadlines(self, feed):
headlines = []
for d in feed['items']:
if 'title' in d:
title = utils.web.htmlToText(d['title']).strip()
link = d.get('link')
if link:
headlines.append((title, link))
else:
headlines.append((title, None))
return headlines
def makeFeedCommand(self, name, url):
docstring = format("""[<number of headlines>]
Reports the titles for %s at the RSS feed %u. If
<number of headlines> is given, returns only that many headlines.
RSS feeds are only looked up every supybot.plugins.RSS.waitPeriod
seconds, which defaults to 1800 (30 minutes) since that's what most
websites prefer.
""", name, url)
if url not in self.locks:
self.locks[url] = threading.RLock()
if hasattr(self.__class__, name) and \
not hasattr(getattr(self, name), 'url'):
s = format('I already have a command in this plugin named %s.',name)
raise callbacks.Error, s
def f(self, irc, msg, args):
args.insert(0, url)
self.rss(irc, msg, args)
f = utils.changeFunctionName(f, name, docstring)
f.url = url # Used by __call__.
self.feedNames.add(name)
setattr(self.__class__, name, f)
self._registerFeed(name, url)
def add(self, irc, msg, args, name, url):
"""<name> <url>
Adds a command to this plugin that will look up the RSS feed at the
given URL.
"""
self.makeFeedCommand(name, url)
irc.replySuccess()
add = wrap(add, ['feedName', 'url'])
def remove(self, irc, msg, args, name):
"""<name>
Removes the command for looking up RSS feeds at <name> from
this plugin.
"""
if name not in self.feedNames:
irc.error('That\'s not a valid RSS feed command name.')
return
self.feedNames.remove(name)
delattr(self.__class__, name)
conf.supybot.plugins.RSS.feeds.unregister(name)
irc.replySuccess()
remove = wrap(remove, ['feedName'])
def announce(self, irc, msg, args, channel, optlist, rest):
"""[<channel>] [--remove] [<name|url> ...]
Adds the list of <name|url> to the current list of announced feeds in
the channel given. Valid feeds include the names of registered feeds
as well as URLs for a RSS feeds. <channel> is only necessary if the
message isn't sent in the channel itself. If no arguments are
specified, replies with the current list of feeds to announce. If
--remove is given, the specified feeds will be removed from the list
of feeds to announce.
"""
remove = False
announce = conf.supybot.plugins.RSS.announce
for (option, _) in optlist:
if option == 'remove':
if not rest:
raise callbacks.ArgumentError
remove = True
def addFeed(feed):
if feed not in feeds:
feeds.add(feed)
def removeFeed(feed):
if feed in feeds:
feeds.remove(feed)
if rest:
if remove:
updater = removeFeed
else:
updater = addFeed
feeds = announce.get(channel)()
for feed in rest:
updater(feed)
announce.get(channel).setValue(feeds)
irc.replySuccess()
elif not rest:
feeds = format('%L', announce.get(channel)())
irc.reply(feeds or 'I am currently not announcing any feeds.')
return
announce = wrap(announce, [('checkChannelCapability', 'op'),
getopts({'remove':''}),
any(first('url', 'feedName'))])
def rss(self, irc, msg, args, url, n):
"""<url> [<number of headlines>]
Gets the title components of the given RSS feed.
If <number of headlines> is given, return only that many headlines.
"""
self.log.debug('Fetching %u', url)
feed = self.getFeed(url)
if irc.isChannel(msg.args[0]):
channel = msg.args[0]
else:
channel = None
headlines = self.getHeadlines(feed)
if not headlines:
irc.error('Couldn\'t get RSS feed.')
return
headlines = self.buildHeadlines(headlines, channel, 'showLinks')
if n:
headlines = headlines[:n]
sep = self.registryValue('headlineSeparator', channel)
if self.registryValue('bold', channel):
sep = ircutils.bold(sep)
irc.replies(headlines, joiner=sep)
rss = wrap(rss, ['url', additional('int')])
def info(self, irc, msg, args, url):
"""<url|feed>
Returns information from the given RSS feed, namely the title,
URL, description, and last update date, if available.
"""
try:
url = self.registryValue('feeds.%s' % url)
except registry.NonExistentRegistryEntry:
pass
feed = self.getFeed(url)
info = feed['channel']
if not info:
irc.error('I couldn\'t retrieve that RSS feed.')
return
# check the 'modified' key, if it's there, convert it here first
if 'modified' in feed:
seconds = time.mktime(feed['modified'])
now = time.mktime(time.gmtime())
when = utils.timeElapsed(now - seconds) + ' ago'
else:
when = 'time unavailable'
# The rest of the entries are all available in the channel key
response = format('Title: %s; URL: %u; '
'Description: %s; Last updated %s.',
info.get('title', 'unavailable').strip(),
info.get('link', 'unavailable').strip(),
info.get('description', 'unavailable').strip(),
when)
irc.reply(utils.str.normalizeWhitespace(response))
info = wrap(info, [first('url', 'feedName')])
Class = RSS
# vim:set shiftwidth=4 tabstop=4 expandtab textwidth=78:

2571
plugins/RSS/rssparser.py Normal file

File diff suppressed because it is too large Load Diff

74
plugins/RSS/test.py Normal file
View File

@ -0,0 +1,74 @@
###
# 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.
###
from supybot.test import *
url = 'http://www.advogato.org/rss/articles.xml'
class RSSTestCase(ChannelPluginTestCase):
plugins = ('RSS',)
def testRssAddBadName(self):
self.assertError('rss add "foo bar" %s' % url)
def testCantAddFeedNamedRss(self):
self.assertError('rss add rss %s' % url)
def testCantRemoveMethodThatIsntFeed(self):
self.assertError('rss remove rss')
if network:
def testRssinfo(self):
self.assertNotError('rss info %s' % url)
self.assertNotError('rss add advogato %s' % url)
# For some reason conf.supybot.plugins.RSS.feeds.advogato is
# empty, which is why this test is failing. Works in practice,
# but not here. :(
self.assertNotError('rss info advogato')
self.assertNotError('rss info AdVogATo')
self.assertNotError('rss remove advogato')
def testRssinfoDoesTimeProperly(self):
self.assertNotRegexp('rss info http://slashdot.org/slashdot.rss',
'-1 years')
def testRss(self):
self.assertNotError('rss %s' % url)
m = self.assertNotError('rss %s 2' % url)
self.failUnless(m.args[1].count('||') == 1)
def testRssAdd(self):
self.assertNotError('rss add advogato %s' % url)
self.assertNotError('advogato')
self.assertNotError('rss advogato')
self.assertNotError('rss remove advogato')
self.assertError('advogato')
self.assertError('rss advogato')
# vim:set shiftwidth=4 tabstop=8 expandtab textwidth=78:

View File

@ -50,6 +50,7 @@ plugins = [
'Owner', 'Owner',
'QuoteGrabs', 'QuoteGrabs',
'Reply', 'Reply',
'RSS',
'Scheduler', 'Scheduler',
'ShrinkUrl', 'ShrinkUrl',
'Status', 'Status',