tweet.py

#!/usr/bin/env python

# I tweet for swag

# ./tweet.py --learn
# ./tweet.py --markov
# ./tweet.py --markov --tweet

import argparse
from collections import Counter
from ConfigParser import ConfigParser
import markovify
import twitter


class TweetCorpus(object):
    def __init__(self, args):
        self.args = args
        config = ConfigParser()
        config.read('config.ini')
        self.api = twitter.Api(
            consumer_key=config.get('secrets', 'consumer_key'),
            consumer_secret=config.get('secrets', 'consumer_secret'),
            access_token_key=config.get('secrets', 'access_token_key'),
            access_token_secret=config.get('secrets', 'access_token_secret'),
            )
        self.corpus = []
        self.searches = [
            'PDSummit16',
            'pagerduty',
            'pagerduty devops',
            'automated infrastructure',
            'infrastructure automation',
            'continuous delivery',
            'pagerduty site reliability',
            'site reliability',
            'pagerduty oncall',
            'pagerduty slack',
            'devops',
            ]
        self.wordcount = Counter()
        self.mentions = Counter()
        self.corpus_file = 'corpus.txt'

    def parse_results(self, results):
        for status in results:
            with open(self.corpus_file, 'a') as out:
                self.corpus.append(status.text)
                out.write(status.text.encode("UTF-8"))
                out.write('\n'.encode("UTF-8"))

            for word in status.text.split():
                self.wordcount[word] += 1

            for mention in status.user_mentions:
                self.mentions[mention.screen_name] += 1

    def learn(self):
        # statuses = api.GetUserTimeline(screen_name='pagerduty')

        for search in self.searches:
            print '----%s----' % search

            results = self.api.GetSearch(search, count=100, result_type='mixed')
            print len(results)
            self.parse_results(results)

        for (mention, rank) in self.mentions.most_common(50):
            print(mention)
            results = self.api.GetUserTimeline(screen_name=mention)
            self.parse_results(results)

        print(len(self.corpus))
        print(len(self.wordcount), self.wordcount.most_common(50))
        print(len(self.mentions), self.mentions.most_common(50))

    def markov(self):
        with open(self.corpus_file) as f:
            text = f.read()
        text_model = markovify.Text(text)

        tweets = {}
        for i in range(50):
            tweet = text_model.make_short_sentence(130)
            tweet = tweet.replace('@', '')
            tweet = tweet.replace('#', '')
            tweet += ' #PDSummit16'
            print tweet
            if self.args.tweet:
                if not tweets.get(tweet):
                    try:
                        self.api.PostUpdate(tweet)
                    except twitter.error.TwitterError:
                        continue

if __name__ == '__main__':
    parser = argparse.ArgumentParser()
    parser.add_argument('--learn', action='store_true')
    parser.add_argument('--markov', action='store_true')
    parser.add_argument('--tweet', action='store_true')
    args = parser.parse_args()

    corpus = TweetCorpus(args)

    if args.learn:
        corpus.learn()
    if args.markov:
        corpus.markov()