4 # :title: Markov plugin
6 # Author:: Tom Gilbert <tom@linuxbrit.co.uk>
7 # Copyright:: (C) 2005 Tom Gilbert
9 # Contribute to chat with random phrases built from word sequences learned
10 # by listening to chat
12 class MarkovPlugin < Plugin
13 Config.register Config::BooleanValue.new('markov.enabled',
15 :desc => "Enable and disable the plugin")
16 Config.register Config::IntegerValue.new('markov.probability',
18 :validate => Proc.new { |v| (0..100).include? v },
19 :desc => "Percentage chance of markov plugin chipping in")
20 Config.register Config::ArrayValue.new('markov.ignore',
22 :desc => "Hostmasks and channel names markov should NOT learn from (e.g. idiot*!*@*, #privchan).")
23 Config.register Config::ArrayValue.new('markov.readonly',
25 :desc => "Hostmasks and channel names markov should NOT talk to (e.g. idiot*!*@*, #privchan).")
26 Config.register Config::IntegerValue.new('markov.max_words',
28 :validate => Proc.new { |v| (0..100).include? v },
29 :desc => "Maximum number of words the bot should put in a sentence")
30 Config.register Config::FloatValue.new('markov.learn_delay',
32 :validate => Proc.new { |v| v >= 0 },
33 :desc => "Time the learning thread spends sleeping after learning a line. If set to zero, learning from files can be very CPU intensive, but also faster.")
34 Config.register Config::IntegerValue.new('markov.delay',
36 :validate => Proc.new { |v| v >= 0 },
37 :desc => "Wait short time before contributing to conversation.")
38 Config.register Config::IntegerValue.new('markov.answer_addressed',
40 :validate => Proc.new { |v| (0..100).include? v },
41 :desc => "Probability of answer when addressed by nick")
42 Config.register Config::ArrayValue.new('markov.ignore_patterns',
44 :desc => "Ignore these word patterns")
48 # upgrade a registry entry from 0.9.14 and earlier, converting the Arrays
49 # into Hashes of weights
50 def upgrade_entry(k, logfile)
51 logfile.puts "\t#{k.inspect}"
57 # wipe the current key
62 logfile.puts "\tEMPTY"
67 logfile.puts "\t#{ar.inspect}"
69 # re-encode key to UTF-8 and cleanup as needed
70 words = k.split.map do |w|
71 BasicUserMessage.strip_formatting(
72 @bot.socket.filter.in(w)
76 # old import that failed to split properly?
77 if words.length == 1 and words.first.include? '/'
80 at = unsplit.rindex('/')
81 words = [unsplit[0,at], unsplit[at+1..-1]]
84 # if any of the re-split/re-encoded words have spaces,
85 # or are empty, we would get a chain we can't convert,
87 if words.first.empty? or words.first.include?(' ') or
88 words.last.empty? or words.last.include?(' ')
89 logfile.puts "\tSKIPPED"
93 # former unclean CTCP, we can't convert this
94 if words.first[0] == 1
95 logfile.puts "\tSKIPPED"
99 # nonword CTCP => SKIP
100 # someword CTCP => nonword someword
101 if words.last[0] == 1
102 if words.first == "nonword"
103 logfile.puts "\tSKIPPED"
110 # intern the old keys
112 ['nonword', MARKER].include?(w) ? MARKER : w.chomp("\001")
115 newkey = words.join(' ')
116 logfile.puts "\t#{newkey.inspect}"
118 # the new key exists already, so we want to merge
119 if k != newkey and @registry.key? newkey
120 ar2 = @registry[newkey]
121 logfile.puts "\tMERGE"
122 logfile.puts "\t\t#{ar2.inspect}"
124 # and get rid of the key
125 @registry.delete(newkey)
131 @chains_mutex.synchronize do
132 if @chains.key? newkey
133 ar2 = @chains[newkey]
144 # we convert old words into UTF-8, cleanup, resplit if needed,
145 # and only get the first word. we may lose some data for old
146 # missplits, but this is the best we can do
147 w = BasicUserMessage.strip_formatting(
148 @bot.socket.filter.in(word).split.first
151 when /^\001\S+$/, "\001", ""
152 # former unclean CTCP or end of CTCP
155 # intern after clearing leftover end-of-actions if present
156 sym = w.chomp("\001")
163 logfile.puts "\tSKIPPED"
166 logfile.puts "\t#{[total, hash].inspect}"
167 @chains[newkey] = [total, hash]
172 # we load all the keys and then iterate over this array because
173 # running each() on the registry and updating it at the same time
175 keys = @registry.keys
176 # no registry, nothing to do
177 return if keys.empty?
180 log "starting markov database conversion thread (v1 to v2, #{keys.length} keys)"
182 keys.each { |k| @upgrade_queue.push k }
183 @upgrade_queue.push nil
185 @upgrade_thread = Thread.new do
186 logfile = File.open(@bot.path('markov-conversion.log'), 'a')
187 logfile.puts "=== conversion thread started #{Time.now} ==="
188 while k = @upgrade_queue.pop
190 logfile.puts "Key #{ki} (#{@upgrade_queue.length} in queue):"
192 upgrade_entry(k, logfile)
193 rescue Exception => e
194 logfile.puts "=== ERROR ==="
195 logfile.puts e.pretty_inspect
196 logfile.puts "=== EREND ==="
198 sleep @bot.config['markov.learn_delay'] unless @bot.config['markov.learn_delay'].zero?
200 logfile.puts "=== conversion thread stopped #{Time.now} ==="
203 @upgrade_thread.priority = -1
206 attr_accessor :chains
210 @registry.set_default([])
211 if @registry.has_key?('enabled')
212 @bot.config['markov.enabled'] = @registry['enabled']
213 @registry.delete('enabled')
215 if @registry.has_key?('probability')
216 @bot.config['markov.probability'] = @registry['probability']
217 @registry.delete('probability')
219 if @bot.config['markov.ignore_users']
220 debug "moving markov.ignore_users to markov.ignore"
221 @bot.config['markov.ignore'] = @bot.config['markov.ignore_users'].dup
222 @bot.config.delete('markov.ignore_users'.to_sym)
225 @chains = @registry.sub_registry('v2')
226 @chains.set_default([])
227 @rchains = @registry.sub_registry('v2r')
228 @rchains.set_default([])
229 @chains_mutex = Mutex.new
230 @rchains_mutex = Mutex.new
232 @upgrade_queue = Queue.new
233 @upgrade_thread = nil
236 @learning_queue = Queue.new
237 @learning_thread = Thread.new do
238 while s = @learning_queue.pop
240 sleep @bot.config['markov.learn_delay'] unless @bot.config['markov.learn_delay'].zero?
243 @learning_thread.priority = -1
247 if @upgrade_thread and @upgrade_thread.alive?
248 debug 'closing conversion thread'
250 @upgrade_queue.push nil
252 debug 'conversion thread closed'
255 debug 'closing learning thread'
256 @learning_queue.clear
257 @learning_queue.push nil
258 @learning_thread.join
259 debug 'learning thread closed'
265 # pick a word from the registry using the pair as key.
266 def pick_word(word1, word2=MARKER, chainz=@chains)
267 k = "#{word1} #{word2}"
268 return MARKER unless chainz.key? k
270 pick_word_from_list wordlist
273 # pick a word from weighted hash
274 def pick_word_from_list(wordlist)
275 total = wordlist.first
277 return MARKER if total == 0
278 return hash.keys.first if hash.length == 1
291 def generate_string(word1, word2)
292 # limit to max of markov.max_words words
294 output = [word1, word2]
298 @chains.each_key(output) do |key|
299 if key.downcase.include? output
305 return nil if keys.empty?
306 output = keys[rand(keys.size)].split(/ /)
308 output = output.split(/ /) unless output.is_a? Array
309 input = [word1, word2]
310 while output.length < @bot.config['markov.max_words'] and (output.first != MARKER or output.last != MARKER) do
311 if output.last != MARKER
312 output << pick_word(output[-2], output[-1])
314 if output.first != MARKER
315 output.insert 0, pick_word(output[0], output[1], @rchains)
326 def help(plugin, topic="")
327 topic, subtopic = topic.split
331 "markov delay <value> => Set message delay"
335 "markov ignore add <hostmask|channel> => ignore a hostmask or a channel"
337 "markov ignore list => show ignored hostmasks and channels"
339 "markov ignore remove <hostmask|channel> => unignore a hostmask or channel"
341 "ignore hostmasks or channels -- topics: add, remove, list"
346 "markov readonly add <hostmask|channel> => read-only a hostmask or a channel"
348 "markov readonly list => show read-only hostmasks and channels"
350 "markov readonly remove <hostmask|channel> => unreadonly a hostmask or channel"
352 "restrict hostmasks or channels to read only -- topics: add, remove, list"
355 "markov status => show if markov is enabled, probability and amount of messages in queue for learning"
357 "markov probability [<percent>] => set the % chance of rbot responding to input, or display the current probability"
361 "markov chat about <word> [<another word>] => talk about <word> or riff on a word pair (if possible)"
363 "markov chat => try to say something intelligent"
366 ["markov learn from <file> [testing [<num> lines]] [using pattern <pattern>]:",
367 "learn from the text in the specified <file>, optionally using the given <pattern> to filter the text.",
368 "you can sample what would be learned by specifying 'testing <num> lines'"].join(' ')
370 "markov plugin: listens to chat to build a markov chain, with which it can (perhaps) attempt to (inanely) contribute to 'discussion'. Sort of.. Will get a *lot* better after listening to a lot of chat. Usage: 'chat' to attempt to say something relevant to the last line of chat, if it can -- help topics: ignore, readonly, delay, status, probability, chat, chat about, learn"
375 str = m.plainmessage.dup
376 str =~ /^(\S+)([:,;])/
377 if $1 and m.target.is_a? Irc::Channel and m.target.user_nicks.include? $1.downcase
378 str.gsub!(/^(\S+)([:,;])\s+/, "")
380 str.gsub!(/\s{2,}/, ' ') # fix for two or more spaces
385 return @bot.config['markov.probability']
389 if @bot.config['markov.enabled']
390 reply = _("markov is currently enabled, %{p}% chance of chipping in") % { :p => probability? }
391 l = @learning_queue.length
392 reply << (_(", %{l} messages in queue") % {:l => l}) if l > 0
393 l = @upgrade_queue.length
394 reply << (_(", %{l} chains to upgrade") % {:l => l}) if l > 0
396 reply = _("markov is currently disabled")
402 return false unless m
403 return true if m.private?
404 return true if m.prefixed?
405 @bot.config['markov.ignore'].each do |mask|
406 return true if m.channel.downcase == mask.downcase
407 return true if m.source.matches?(mask)
413 return false unless m
414 @bot.config['markov.readonly'].each do |mask|
415 return true if m.channel.downcase == mask.downcase
416 return true if m.source.matches?(mask)
421 def ignore(m, params)
422 action = params[:action]
423 user = params[:option]
426 if @bot.config['markov.ignore'].include? user
427 s = @bot.config['markov.ignore']
429 @bot.config['ignore'] = s
430 m.reply _("%{u} removed") % { :u => user }
432 m.reply _("not found in list")
436 if @bot.config['markov.ignore'].include?(user)
437 m.reply _("%{u} already in list") % { :u => user }
439 @bot.config['markov.ignore'] = @bot.config['markov.ignore'].push user
440 m.reply _("%{u} added to markov ignore list") % { :u => user }
443 m.reply _("give the name of a person or channel to ignore")
446 m.reply _("I'm ignoring %{ignored}") % { :ignored => @bot.config['markov.ignore'].join(", ") }
448 m.reply _("have markov ignore the input from a hostmask or a channel. usage: markov ignore add <mask or channel>; markov ignore remove <mask or channel>; markov ignore list")
452 def readonly(m, params)
453 action = params[:action]
454 user = params[:option]
457 if @bot.config['markov.readonly'].include? user
458 s = @bot.config['markov.readonly']
460 @bot.config['markov.readonly'] = s
461 m.reply _("%{u} removed") % { :u => user }
463 m.reply _("not found in list")
467 if @bot.config['markov.readonly'].include?(user)
468 m.reply _("%{u} already in list") % { :u => user }
470 @bot.config['markov.readonly'] = @bot.config['markov.readonly'].push user
471 m.reply _("%{u} added to markov readonly list") % { :u => user }
474 m.reply _("give the name of a person or channel to read only")
477 m.reply _("I'm only reading %{readonly}") % { :readonly => @bot.config['markov.readonly'].join(", ") }
479 m.reply _("have markov not answer to input from a hostmask or a channel. usage: markov readonly add <mask or channel>; markov readonly remove <mask or channel>; markov readonly list")
483 def enable(m, params)
484 @bot.config['markov.enabled'] = true
488 def probability(m, params)
489 if params[:probability]
490 @bot.config['markov.probability'] = params[:probability].to_i
493 m.reply _("markov has a %{prob}% chance of chipping in") % { :prob => probability? }
497 def disable(m, params)
498 @bot.config['markov.enabled'] = false
503 return false unless @bot.config['markov.enabled']
504 prob = m.address? ? @bot.config['markov.answer_addressed'] : probability?
505 return true if prob > rand(100)
509 # Generates all sequence pairs from array
510 # seq_pairs [1,2,3,4] == [ [1,2], [2,3], [3,4]]
513 0.upto(arr.size-2) do |i|
514 res << [arr[i], arr[i+1]]
519 def set_delay(m, params)
520 if params[:delay] == "off"
521 @bot.config["markov.delay"] = 0
523 elsif !params[:delay]
524 m.reply _("Message delay is %{delay}" % { :delay => @bot.config["markov.delay"]})
526 @bot.config["markov.delay"] = params[:delay].to_i
531 def reply_delay(m, line)
533 if @bot.config['markov.delay'] > 0
534 @bot.timer.add_once(1 + rand(@bot.config['markov.delay'])) {
535 m.reply line, :nick => false, :to => :public
538 m.reply line, :nick => false, :to => :public
542 def random_markov(m, message)
543 return unless should_talk(m)
545 words = clean_message(m).split(/\s+/)
547 line = generate_string words.first, nil
549 if line and message.index(line) != 0
554 pairs = seq_pairs(words).sort_by { rand }
555 pairs.each do |word1, word2|
556 line = generate_string(word1, word2)
557 if line and message.index(line) != 0
562 words.sort_by { rand }.each do |word|
563 line = generate_string word.first, nil
564 if line and message.index(line) != 0
573 line = generate_string(params[:seed1], params[:seed2])
574 if line and line != [params[:seed1], params[:seed2]].compact.join(" ")
577 m.reply _("I can't :(")
581 def rand_chat(m, params)
582 # pick a random pair from the db and go from there
583 word1, word2 = MARKER, MARKER
585 @bot.config['markov.max_words'].times do
586 word3 = pick_word(word1, word2)
587 break if word3 == MARKER
589 word1, word2 = word2, word3
592 m.reply output.join(" ")
594 m.reply _("I can't :(")
599 lines.each { |l| @learning_queue.push l }
605 # in channel message, the kind we are interested in
606 message = m.plainmessage
609 message = "#{m.sourcenick} #{message}"
612 random_markov(m, message) unless readonly? m or m.replied?
613 learn clean_message(m)
617 def learn_triplet(word1, word2, word3)
618 k = "#{word1} #{word2}"
619 rk = "#{word2} #{word3}"
620 @chains_mutex.synchronize do
630 @chains[k] = [total, hash]
632 @rchains_mutex.synchronize do
637 t2, h2 = @rchains[rk]
643 @rchains[rk] = [total, hash]
648 def learn_line(message)
649 # debug "learning #{message.inspect}"
650 wordlist = message.strip.split(/\s+/).reject do |w|
651 @bot.config['markov.ignore_patterns'].map do |pat|
652 w =~ Regexp.new(pat.to_s)
653 end.select{|v| v}.size != 0
655 return unless wordlist.length >= 2
656 word1, word2 = MARKER, MARKER
658 wordlist.each do |word3|
659 learn_triplet(word1, word2, word3.to_sym)
660 word1, word2 = word2, word3
664 # TODO allow learning from URLs
665 def learn_from(m, params)
668 file = File.open(path, "r")
669 pattern = params[:pattern].empty? ? nil : Regexp.new(params[:pattern].to_s)
671 m.reply _("no such file")
676 m.reply _("the file is empty!")
682 range = case params[:lines]
684 Range.new(*params[:lines].split("..").map { |e| e.to_i })
686 Range.new(1, params[:lines].to_i)
688 Range.new(1, [@bot.config['send.max_lines'], 3].max)
692 next unless file.lineno >= range.begin
694 break if file.lineno == range.end
697 lines = lines.map do |l|
698 pattern ? l.scan(pattern).to_s : l
699 end.reject { |e| e.empty? }
703 m.reply _("example matches for that pattern at lines %{range} include: %{lines}") % {
704 :lines => lines.map { |e| Underline+e+Underline }.join(", "),
708 m.reply _("the pattern doesn't match anything at lines %{range}") % {
713 m.reply _("learning from the file without a pattern would learn, for example: ")
714 lines.each { |l| m.reply l }
721 file.each { |l| learn(l.scan(pattern).to_s) }
723 file.each { |l| learn(l.chomp) }
730 m.reply "Markov status: chains: #{@chains.length} forward, #{@rchains.length} reverse, queued phrases: #{@learning_queue.size}"
735 plugin = MarkovPlugin.new
736 plugin.map 'markov delay :delay', :action => "set_delay"
737 plugin.map 'markov delay', :action => "set_delay"
738 plugin.map 'markov ignore :action :option', :action => "ignore"
739 plugin.map 'markov ignore :action', :action => "ignore"
740 plugin.map 'markov ignore', :action => "ignore"
741 plugin.map 'markov readonly :action :option', :action => "readonly"
742 plugin.map 'markov readonly :action', :action => "readonly"
743 plugin.map 'markov readonly', :action => "readonly"
744 plugin.map 'markov enable', :action => "enable"
745 plugin.map 'markov disable', :action => "disable"
746 plugin.map 'markov status', :action => "status"
747 plugin.map 'markov stats', :action => "stats"
748 plugin.map 'chat about :seed1 [:seed2]', :action => "chat"
749 plugin.map 'chat', :action => "rand_chat"
750 plugin.map 'markov probability [:probability]', :action => "probability",
751 :requirements => {:probability => /^\d+%?$/}
752 plugin.map 'markov learn from :file [:testing [:lines lines]] [using pattern *pattern]', :action => "learn_from", :thread => true,
754 :testing => /^testing$/,
755 :lines => /^(?:\d+\.\.\d+|\d+)$/ }
757 plugin.default_auth('ignore', false)
758 plugin.default_auth('probability', false)
759 plugin.default_auth('learn', false)