Multilingualism

This commit adjusts the speech parsing system to allow for an infinite
number of languages to be present in the same message. You can
transition freely between any language you are able to speak simply by
putting it's language key in the middle of the sentence.

Honestly, this was a massive pain in the ass, and there's probably still
broken stuff, even though I've spent around 8 hours testing and refining
this.

code/modules/mob/say.dm

@@ -132,17 +132,77 @@
 
 	return null
 
-//parses the language code (e.g. :j) from text, such as that supplied to say.
-//returns the language object only if the code corresponds to a language that src can speak, otherwise null.
-/mob/proc/parse_language(var/message)
-	var/prefix = copytext(message, 1, 2)
-	if(length(message) >= 1 && prefix == "!")
-		return GLOB.all_languages["Noise"]
+/datum/multilingual_say_piece
+	var/datum/language/speaking = null
+	var/message = ""
 
-	if(length(message) >= 2)
-		var/language_prefix = trim_right(lowertext(copytext(message, 1 ,4)))
-		var/datum/language/L = GLOB.language_keys[language_prefix]
-		if(can_speak_language(L))
-			return L
+/datum/multilingual_say_piece/New(datum/language/new_speaking, new_message)
+	. = ..()
+	speaking = new_speaking
+	if(new_message)
+		message = new_message
 
-	return null
+/mob/proc/find_valid_prefixes(message)
+	var/list/prefixes = list() // [["Common", start, end], ["Gutter", start, end]]
+	for(var/i in 1 to length(message))
+		var/selection = trim_right(lowertext(copytext(message, i, i + 3)))
+		var/datum/language/L = GLOB.language_keys[selection]
+		if(L != null && can_speak_language(L)) // What the fuck... remove the L != null check if you ever find out what the fuck is adding `null` to the languages list on absolutely random mobs... seriously what the hell...
+			prefixes[++prefixes.len] = list(L, i, i + length(selection))
+		else if(!L && i == 1)
+			prefixes[++prefixes.len] = list(get_default_language(), i, i)
+		else
+	return prefixes
+
+/proc/strip_prefixes(message)
+	. = ""
+	var/last_index = 1
+	for(var/i in 1 to length(message))
+		var/selection = trim_right(lowertext(copytext(message, i, i + 3)))
+		var/datum/language/L = GLOB.language_keys[selection]
+		if(L)
+			. += copytext(message, last_index, i)
+			last_index = i + 3
+		if(i + 1 > length(message))
+			. += copytext(message, last_index)
+
+// this returns a structured message with language sections
+// list(/datum/multilingual_say_piece(common, "hi"), /datum/multilingual_say_piece(farwa, "squik"), /datum/multilingual_say_piece(common, "meow!"))
+/mob/proc/parse_languages(message)
+	. = list()
+
+	// Noise language is a snowflake
+	if(copytext(message, 1, 2) == "!" && length(message) > 1)
+		return list(new /datum/multilingual_say_piece(GLOB.all_languages["Noise"], trim(strip_prefixes(copytext(message, 2)))))
+
+	// Scan the message for prefixes
+	var/list/prefix_locations = find_valid_prefixes(message)
+	if(!LAZYLEN(prefix_locations)) // There are no prefixes... or at least, no _valid_ prefixes.
+		. += new /datum/multilingual_say_piece(get_default_language(), trim(strip_prefixes(message))) // So we'll just strip those pesky things and still make the message.
+
+	for(var/i in 1 to length(prefix_locations))
+		var/current = prefix_locations[i] // ["Common", keypos]
+
+		// There are a few things that will make us want to ignore all other languages in - namely, HIVEMIND languages.
+		var/datum/language/L = current[1] 
+		if(L && L.flags & HIVEMIND)
+			. = new /datum/multilingual_say_piece(L, trim(strip_prefixes(message)))
+			break
+
+		if(i + 1 > length(prefix_locations)) // We are out of lookaheads, that means the rest of the message is in cur lang
+			var/spoke_message = handle_autohiss(trim(copytext(message, current[3])), L)
+			. += new /datum/multilingual_say_piece(current[1], spoke_message)
+		else
+			var/next = prefix_locations[i + 1] // We look ahead at the next message to see where we need to stop.
+			var/spoke_message = handle_autohiss(trim(copytext(message, current[3], next[2])), L)
+			. += new /datum/multilingual_say_piece(current[1], spoke_message)
+
+/* These are here purely because it would be hell to try to convert everything over to using the multi-lingual system at once */
+/proc/message_to_multilingual(message, datum/language/speaking = null)
+	. = list(new /datum/multilingual_say_piece(speaking, message))
+
+/proc/multilingual_to_message(list/message_pieces)
+	. = ""
+	for(var/datum/multilingual_say_piece/S in message_pieces)
+		. += S.message + " "
+	. = trim_right(.)