Minor improvement to markov generator.

[plomlombot-irc.git] / plomlombot.py

diff --git a/plomlombot.py b/plomlombot.py
index 0dd9823612d83c84a54e3e935197589f88ea3d15..1be105cbdf51325b4817177cb34f29860c57ac1a 100755 (executable)
--- a/plomlombot.py
+++ b/plomlombot.py
@@ -162,7 +162,7 @@ def handle_command(command, argument, notice, target, session):
         notice("QUOTE #" + str(i + 1) + ": " + lines[i])
 
     def markov():
-        from random import choice
+        from random import choice, shuffle
         select_length = 2
         selections = []
 
@@ -190,16 +190,24 @@ def handle_command(command, argument, notice, target, session):
         if not os.access(markovfeed_name, os.F_OK):
             notice("NOT ENOUGH TEXT TO MARKOV.")
             return
+
+        # Lowercase incoming lines, ensure they end in a sentence end mark.
         file = open(markovfeed_name, "r")
         lines = file.readlines()
         file.close()
         tokens = []
+        sentence_end_markers = ".!?)("
         for line in lines:
-            line = line.replace("\n", "").lower()
+            line = line.lower().replace("\n", "")
+            if line[-1] not in sentence_end_markers:
+                line += "."
             tokens += line.split()
         if len(tokens) <= select_length:
             notice("NOT ENOUGH TEXT TO MARKOV.")
             return
+
+        # Replace URLs with escape string for now, so that the Markov selector
+        # won't see them as different strings. Stash replaced URLs in urls.
         urls = []
         url_escape = "\nURL"
         url_starts = ["http://", "https://", "<http://", "<https://"]
@@ -215,6 +223,10 @@ def handle_command(command, argument, notice, target, session):
                     urls += [tokens[i][:length]]
                     tokens[i] = url_escape + tokens[i][length:]
                     break
+
+        # For each snippet of select_length, use markov() to find continuation
+        # token from selections. Replace present users' names with malkovich.
+        # Start snippets with the beginning of a sentence, if possible.
         for i in range(len(tokens) - select_length):
             token_list = []
             for j in range(select_length + 1):
@@ -223,12 +235,19 @@ def handle_command(command, argument, notice, target, session):
         snippet = []
         for i in range(select_length):
             snippet += [""]
+        shuffle(selections)
+        for i in range(len(selections)):
+            if selections[i][0][-1] in sentence_end_markers:
+                for i in range(select_length):
+                    snippet[i] = selections[i][i + 1]
+                break
         msg = ""
+        malkovich = "malkovich"
         while 1:
             new_end = markov(snippet)
             for name in session.users_in_chan:
                 if new_end[:len(name)] == name.lower():
-                    new_end = "malkovich" + new_end[len(name):]
+                    new_end = malkovich + new_end[len(name):]
                     break
             if len(msg) + len(new_end) > 200:
                 break
@@ -236,12 +255,16 @@ def handle_command(command, argument, notice, target, session):
             for i in range(select_length - 1):
                 snippet[i] = snippet[i + 1]
             snippet[select_length - 1] = new_end
+
+        # Replace occurences of url escape string with random choice from urls.
         while True:
             index = msg.find(url_escape)
             if index < 0:
                 break
             msg = msg.replace(url_escape, choice(urls), 1)
-        notice(msg + "malkovich.")
+
+        # More meaningful ways to randomly end sentences.
+        notice(msg + malkovich + ".")
 
     def twt():
         def try_open(mode):
@@ -293,6 +316,7 @@ def handle_url(url, notice, show_url=False):
     except (requests.exceptions.TooManyRedirects,
             requests.exceptions.ConnectionError,
             requests.exceptions.InvalidURL,
+            UnicodeError,
             requests.exceptions.InvalidSchema) as error:
         notice("TROUBLE FOLLOWING URL: " + str(error))
         return