Mitter/mitterlib/network/twitter.py

#!/usr/bin/python
# -*- coding: utf-8 -*-

# Mitter, a client for Twitter.
# Copyright (C) 2007, 2008 The Mitter Contributors
#
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program.  If not, see <http://www.gnu.org/licenses/>.

import urllib
import urllib2
import logging
import datetime
import base64
import htmlentitydefs
import re
import warnings
import htmlentitydefs

from httplib import BadStatusLine
from socket import error as socketError
#from mitterlib import htmlize

from networkbase import NetworkBase, NetworkData, auth_options, \
        NetworkDNSError, NetworkBadStatusLineError, NetworkLowLevelError, \
        NetworkInvalidResponseError, NetworkPermissionDeniedError, \
        MessageTooLongWarning

try:
    # Python 2.6/3.0 JSON parser
    import json
except ImportError:
    # Fallback to SimpleJSON
    import simplejson as json

# logging
_log = logging.getLogger('mitterlib.network.Twitter')

# the month names come directly from the site, so we are not affected by
# locale settings.
_month_names = [None, 'Jan', 'Feb', 'Mar', 'Apr', 'May', 'Jun', 'Jul', 'Aug',
        'Sep', 'Oct', 'Nov', 'Dec']


def htmlize(text):
    """Convert a normal text to the format required by Twitter (url-encoded
    and UTF-8 encoded."""
    # XXX: UTF-8 part is not working as it should, sadly.
    if not isinstance(text, unicode):
        text = unicode(text, 'utf-8')
#    hex_char = (lambda x: '%' + hex(x)[2:].rjust(2, '0').upper())
#
#    codes = []
#    for char in text:
#        char_code = ord(char)
#        bytes = []
#        mask = 0
#
#        ranges = [65535, 2047, 127]
#        for r in ranges:
#            if char_code > r:
#                mask >>= 1
#                mask |= 128
#
#                byte = char_code & 63
#                byte |= 128
#
#                bytes.insert(0, hex_char(byte))
#
#                char_code >>= 6
#
#        if mask:
#            # multibyte char
#            mask >>= 1
#            mask |= 128
#
#            byte = char_code | mask
#            bytes.insert(0, hex_char(byte))
#        else:
#            if char_code < 32:
#                # special char and it's not a multibyte char
#                bytes.insert(0, hex_char(char_code))
#            else:
#                # normal, printable char
#                bytes.insert(0, char)
#
#        codes.append(''.join(bytes))
#
#    text = ''.join(codes)
    new = []
    for char in text:
        if ord(char) in htmlentitydefs.codepoint2name:
            new.append('&%s;' % (htmlentitydefs.codepoint2name[ord(char)]))
        elif ord(char) == 37:   # '%'
            new.append('%25')
        else:
            new.append(char)

    return ''.join(new)

def _unhtml(text):
    """Convert text coming in HTML encoded to UTF-8 representations."""
    new_text = []
    copy_pos = 0
    _log.debug('Original text: %s', text)
    for code in re.finditer(r'&(\w+);', text):
        new_text.append(text[copy_pos:code.start()])
        entity = text[code.start()+1:code.end()-1]
        if entity in htmlentitydefs.name2codepoint:
            new_text.append(unichr(
                    htmlentitydefs.name2codepoint[entity]))
        else:
            new_text.append(code.group().decode('utf8'))
        copy_pos = code.end()

    new_text.append(text[copy_pos:])

    _log.debug('New text: %s', new_text)
    result = u''.join(new_text)
    _log.debug('Result: %s', result)
    return result


def _to_datetime(server_str):
    """Convert a date send by the server to a datetime object.
    Ex:
        from this:
            Tue Mar 13 00:12:41 +0000 2007
            to datetime.
    """
    date_info = server_str.split(' ')
    month = _month_names.index(date_info[1])
    day = int(date_info[2])
    year = int(date_info[5])

    time_info = date_info[3].split(':')
    hour = int(time_info[0])
    minute = int(time_info[1])
    second = int(time_info[2])

    return datetime.datetime(year, month, day, hour, minute, second)


def _make_datetime(response):
    """Converts dates on responses to datetime objects."""
    result = []
    for tweet in response:
        result.append(TwitterNetworkData(tweet))

    return result


class TwitterNetworkData(NetworkData):
    """A simple wrapper around NetworkData, to make things easier to convert
    twitter data into a NetworkData object."""

    def __init__(self, data):
        """Class initialization. Receives a dictionary with a single tweet."""
        NetworkData.__init__(self)

        self.id = data['id']
        self.name = data['user']['name']
        self.username = data['user']['screen_name']
        self.avatar = data['user']['profile_image_url']
        self.message_time = _to_datetime(data['created_at'])

        if 'favorited' in data:
            self.favourited = data['favorited']

        if 'in_reply_to_status_id' in data and data['in_reply_to_status_id']:
            self.parent = int(data['in_reply_to_status_id'])

        if 'retweeted_status' in data:
            self.reposted_by = self.username

            retweet_user = data['retweeted_status']['user']
            self.name = retweet_user['name']
            self.username = retweet_user['screen_name']
            self.avatar = retweet_user['profile_image_url']

            # also switch the text for the original text.
            data['text'] = data['retweeted_status']['text']

        # Twitter encodes a lot of HTML entities, which are not good when
        # you want to *display* then (e.g., "<" returns to us as "&lt;").
        # So we convert this here. Interfaces need to worry about converting
        # them if it becomes a problem.
        self.message = _unhtml(data['text'])

        return


class Connection(NetworkBase):
    """Base class to talk to twitter."""

    NAMESPACE = 'Twitter'
    SHORTCUT = 'tw'     # TODO: find a way to move this to the config file

    def is_setup(self):
        """Return True or False if the network is setup/enabled."""
        if (self._options[self.NAMESPACE]['username'] and
                self._options[self.NAMESPACE]['password']):
            # Consider the network enabled if there is an username and
            # password
            return True
        else:
            return False

    def __init__(self, options):
        self._options = options

    @property
    def server(self):
        if self._options[self.NAMESPACE]['https']:
            return self._options[self.NAMESPACE]['secure_server_url']
        else:
            return self._options[self.NAMESPACE]['server_url']

    def _common_headers(self):
        """Returns a string with the normal headers we should add on every
        request"""

        auth = base64.b64encode('%s:%s' % (
            self._options[self.NAMESPACE]['username'],
            self._options[self.NAMESPACE]['password']))

        headers = {
                'Authorization': 'Basic %s' % (auth),
                'User-Agent': self._user_agent}
        return headers

    def _request(self, resource, headers=None, body=None):
        """Send a request to the Twitter server. Once finished, call the
        function at callback."""

        url = '%s%s' % (self.server, resource)
        _log.debug('Request %s' % (url))

        request = urllib2.Request(url=url)
        request_headers = self._common_headers()
        if headers:
            request_headers.update(headers)

        for key in request_headers:
            _log.debug('Header: %s=%s' % (key, request_headers[key]))
            request.add_header(key, request_headers[key])

        if body:
            _log.debug('Body: %s' % (body))
            request.add_data(body)

        try:
            _log.debug('Starting request of %s' % (url))
            response = urllib2.urlopen(request)
            data = response.read()
        except urllib2.HTTPError, exc:
            _log.debug('HTTPError: %d' % (exc.code))
            _log.debug('HTTPError: response body:\n%s' % exc.read())
            # To me, I got a lot of 502 for "replies". It shows the
            # "Something is technically wrong" most of the time in the real
            # pages.
            if exc.code == 403:
                # Permission denied.
                raise NetworkPermissionDeniedError
            raise NetworkInvalidResponseError
        except urllib2.URLError, exc:
            _log.error('URL error: %s' % exc.reason)
            raise NetworkDNSError
        except BadStatusLine:
            _log.error('Bad status line (Twitter is going bananas)')
            raise NetworkBadStatusLineError
        except socketError:   # That's the worst exception ever.
            _log.error('Socket connection error')
            raise NetworkLowLevelError
        # TODO: Permission denied?

        # Introduced in Twitter in 2009.03.27
        response_headers = response.info()
        if 'X-RateLimit-Remaining' in response_headers:
            self._rate_limit = int(response_headers['X-RateLimit-Remaining'])
            _log.debug('Remaning hits: %d', self._rate_limit)
        elif 'x-ratelimit-remaining' in response_headers:
            self._rate_limit = int(response_headers['x-ratelimit-remaining'])
            _log.debug('Remaning hits: %d', self._rate_limit)
        else:
            self._rate_limit = None

        _log.debug('Request completed')
        _log.debug('info(%s): %s', type(response.info()), response.info())

        return json.loads(data)

    #
    # New network style methods
    #

    AUTH = [
                {'name': 'username',
                 'flags': ['-u', '--username'],
                 'prompt': 'Username',
                 'help': 'Your twitter username',
                 'type': 'str'},
                {'name': 'password',
                 'flags': ['-p', '--password'],
                 'prompt': 'Password',
                 'help': 'Your twitter password',
                 'type': 'passwd'}]

    @classmethod
    def options(self, options):
        """Add options related to Twitter."""
        options.add_group(self.NAMESPACE, 'Twitter network')
        options.add_option('-s', '--no-https',
                group=self.NAMESPACE,
                option='https',
                default=True,   # Secure connections by default
                help='Disable HTTPS (secure) connection with Twitter.',
                action='store_false')
        options.add_option(
                group=self.NAMESPACE,
                option='last_tweet',
                default=0,
                is_cmd_option=False)
        options.add_option(
                group=self.NAMESPACE,
                option='last_reply',
                default=0,
                is_cmd_option=False)
        options.add_option(
                group=self.NAMESPACE,
                option='server_url',
                default='http://api.twitter.com/1',
                is_cmd_option=False)
        options.add_option(
                group=self.NAMESPACE,
                option='secure_server_url',
                default='https://api.twitter.com/1',
                is_cmd_option=False)
        options.add_option(
                group=self.NAMESPACE,
                option='page_size',
                default=20,
                is_cmd_option=False)
        options.add_option(
                group=self.NAMESPACE,
                option='message_threshold',
                default=4,
                is_cmd_option=False)
        auth_options(self.NAMESPACE, options, self.AUTH)
        return

    def _timeline(self, config_var, url):
        """Request one of the lists of tweets."""
        last_id = int(self._options[self.NAMESPACE][config_var])
        _log.debug('%s: %d', config_var, last_id)

        params = {'count': self._options[self.NAMESPACE]['page_size']}

        if last_id > 0:
            params['since_id'] = last_id

        page = 1
        result = []
        response = [0]  # So we stay in the loop.
        high_id = 0

        while response: # Not the cleanest code
            # TODO: How the interfaces can interrupt this?
            if page > 1:
                params['page'] = page

            final_url = '?'.join([url, urllib.urlencode(params)])
            response = self._request(final_url)

            _log.debug('Page %d, %d results', page, len(response))

            if response:
                # extract the highest id in the respone and save it so we can
                # use it when requesting data again (using the since_id
                # parameter)

                top_tweet_id = response[0]['id']
                _log.debug('Top tweet: %d; Highest seen tweet: %d',
                    top_tweet_id, high_id)

                if top_tweet_id > high_id:
                    high_id = top_tweet_id

                response_data = _make_datetime(response)
                result.extend(response_data)
                page += 1   # Request the next page

                _log.debug('%d messages, %d threshold' % (len(response_data),
                    self._options[self.NAMESPACE]['message_threshold']))

                if (len(response_data) <=
                        self._options[self.NAMESPACE]['message_threshold']):
                    break

            if last_id == 0:
                # do not try to download everything if we don't have a
                # previous list (or we'll blow the available requests in one
                # short)
                break

        # only update the "last seen id" if everything goes alright
        if high_id > int(self._options[self.NAMESPACE][config_var]):
            _log.debug('Last tweet updated: %d', high_id)
            self._options[self.NAMESPACE][config_var] = high_id

        return result

    def messages(self):
        """Return a list of NetworkData objects for the main "timeline"."""
        return self._timeline('last_tweet', '/statuses/home_timeline.json')

    def message(self, message_id):
        """Retrieves the information of one message."""
        response = self._request('/statuses/show/%d.json' % (message_id))
        return TwitterNetworkData(response)

    def link(self, message):
        """Return a link directly to the message."""
        return 'http://twitter.com/%s/status/%s' % (message.username,
                message.id)

    def reply_prefix(self, message):
        """Returns the prefix needed for a reply."""
        return '@' + message.username + ' '

    def replies(self):
        """Return a list of NetworkData objects for the replies for the user
        messages."""
        return self._timeline('last_reply', '/statuses/replies.json')

    def available_requests(self):
        """Return the current user rate limit."""
        if self._rate_limit:
            return self._rate_limit

        data = self._request('/account/rate_limit_status.json')
        _log.debug('Requests: %s', data)
        return int(data['remaining_hits'])

    def update(self, status, reply_to=None):
        """Update the user status."""
        if len(status) > 140:
            warnings.warn('Message too long', MessageTooLongWarning)

        # In Python 2.5, urllib.urlencode calls str(), which removes the
        # unicodeness of the "status". So we need to convert those peski
        # accents to HTML entities, so everything falls into ASCII.

        body = {
            'status': htmlize(status),
            'source': 'mitter'}

        if reply_to:
            if isinstance(reply_to, NetworkData):
                body['in_reply_to_status_id'] = reply_to.id

                # This is to protect the user from himself. You don't *need*
                # to start a reply with a @<username>, but it looks really
                # confusing in the Twiter website. So if the line doesn't
                # start with the username of the original user, we add it
                # for the user.

                if not status.startswith('@' + reply_to.username):
                    body['status'] = '@' + reply_to.username + ' ' + \
                            status
            else:
                body['in_reply_to_status_id'] = reply_to

        _log.debug('Body: %s', body)
        body = urllib.urlencode(body)
        _log.debug('Message to twitter: %s' % (body))

        data = self._request('/statuses/update.json', body=body)
        # TODO: Check if twitter sends an error message when the message is
        #       too large.
        return TwitterNetworkData(data)

    def repost(self, message):
        """Repost a message."""
        body = urllib.urlencode({'id': message.id})
        resource = '/statuses/retweet/%d.json' % (message.id)
        data = self._request(resource, body=body)
        return TwitterNetworkData(data)

    def favourite(self, message):
        """Mark a message as favourite."""
        body = urllib.urlencode({'id': message.id})
        if not message.favourite:
            resource = '/favorites/create/%d.json' % (message.id)
        else:
            resource = '/favorites/destroy/%d.json' % (message.id)
        data = self._request(resource, body=body)
        return TwitterNetworkData(data)

    def delete_message(self, message):
        """Delete a message."""
        if isinstance(message, NetworkData):
            message = message.id    # We don't need anything else for Twitter

        # make a body, so _request makes it a post.
        body = urllib.urlencode({'id': message})
        resource = '/statuses/destroy/%s.json' % (message)
        response = self._request(resource, body=body)
        _log.debug('Delete response: %s', response)
        return True # Either we get a response or an exception before we reach
                    # this.
    def can_delete(self, message):
        """Check if the message belongs to the user. If so, returns True;
        False otherwise."""
        return (message.username == self._options[self.NAMESPACE]['username'])

    def can_reply(self, message):
        """Always return True; Twitter allows replying to any messages,
        including the ones from the user."""
        return True

    def can_repost(self, message):
        """Twitter ignores retweets from the user."""
        return not (message.username ==
                self._options[self.NAMESPACE]['username'])

    def can_favourite(self, message):
        """Always return True; Twitter allows favouriting/unfavouriting any
        messages."""
        return True