Fix psycopg2 syntax and PostgreSQL missing REPLACE

Things were so easy with sqlite3...  PostgreSQL does things differently.

Signed-off-by: Stefan Hajnoczi <stefanha@gmail.com>

generator/generator.py

@@ -67,14 +67,15 @@ def store_markov_chain(cursor, topic, mc):
     for current in mc.chain:
         bucket = mc.chain[current]
         for next_ in bucket:
-            cursor.execute('INSERT OR REPLACE INTO markov_chains (topic_id, current, next, count, created_at, updated_at) VALUES (%s, %s, %s, %s, "", "")', (topic_id, current, next_, bucket[next_]))
+            cursor.execute('DELETE FROM markov_chains WHERE topic_id=%s AND current=%s AND next=%s', (topic_id, current, next_))
+            cursor.execute('INSERT INTO markov_chains (topic_id, current, next, count, created_at, updated_at) VALUES (%s, %s, %s, %s, NOW(), NOW())', (topic_id, current, next_, bucket[next_]))
 
 def load_random_user(cursor, topic):
     cursor.execute('SELECT sender FROM tweets WHERE topic_id=%s ORDER BY RANDOM() LIMIT 1', (load_topic_id(cursor, topic),))
     return cursor.fetchone()[0]
 
 def store_tweet(cursor, topic, sender, text, fake):
-    cursor.execute('INSERT INTO tweets (text, fake, topic_id, sender, created_at, updated_at) VALUES (%s, %s, %s, %s, "", "")', (text, fake, load_topic_id(cursor, topic), sender))
+    cursor.execute('INSERT INTO tweets (text, fake, topic_id, sender, created_at, updated_at) VALUES (%s, %s, %s, %s, NOW(), NOW())', (text, fake, load_topic_id(cursor, topic), sender))
 
 class MarkovChain(object):
     def __init__(self, cardinality):
@@ -130,10 +131,9 @@ def generate(self):
             next_ = self._choose_next(self.chain[current])
             current = ' '.join(current.split()[1:] + [next_])
             if next_ == '@@start@@':
-                # Try again if we only generated one word.  This happens when
-                # people start and end tweets with a hash tag, and therefore
-                # makes the algorithm generate just the hash tag.
-                if len(sentence.split()) == 1:
+                # Try again if we only generated one word or the sentence
+                # became too long.
+                if len(sentence.split()) == 1 or len(sentence) > 140:
                     current = self._choose_first_tuple()
                     sentence = ' '.join(current.split()[1:])
                     continue