Bukkit

@polo3400

I'm getting around to it. Should have it up within a day or two.

Note: I'm pretty sure, with the changes to the chat system in 1.3.1, this plugin WILL need updating and will probably be hopelessly broken until I get an update out.

@Puttanesca621

...What's the point of a filter if you can opt out of it?

I actually have a permission node that can be assigned to a player to let the player bypass it, but I haven't actually made it a usable permission yet. I'll stick that in the next release.

v2.0.0 is up!

If you check the downloads section on bukkitdev before a Bukkit mod has a chance to approve the file, you may need to navigate to the latest release manually by choosing the "Files" tab and picking "WordFilter v2.0."

After a lot of debating, I decided on this implementation:

Blacklist: If any word on this list is founed to be contained within a word that is said on the server, the message is in violation.

Fuzzylist: If any word on this list is found to be "similar" to any word said on the server, the word is in violation.

Whitelist: If a word has been found to be in violation, but it is on the whitelist, the word is allowed and no longer in violation.

All methods are case-insensitive.

Does anyone care about how the old way allowed you to replace words with funny replacements? Because right now the plan is to allow you to, once a message is found to have a violating word, either suppress the message entirely and nothing appears in server chat, or you can replace the message with a funny phrase or something. I don't know if anyone was particularly attached to the old way where you could replace a single word.

So, while I was reviewing the old codebase, I remembered how completely dissatisfied with my old solutions I was. They were simple and naive and easy to work around. All the options and stuff that I offered to change the comparison algorithm were crappy and cluttered. The whole thing was really disappointing for me, both as a developer and user.

So, this time, I'm coming at it from a different direction.

I've been playing around with algorithms for finding the similarity between Strings, and I think that's the way I'm going to go. I did some simple tests and if you filter words that have over a 20 or 30% similarity to a blacklisted word, you can catch a lot of the simple variants people try to use to get around filters.

Let's say "pookie" is a bad word we don't want said in chat.

Here's how similar certain words are according to the algorithm:

"pokie": 89%
"p00kie": 40%
"pkie": 50%
"pk": 0%
"p00k|e": 0%

Now, you'll notice the more "leet speak"-y we get the worse the algorithm performs. Using LeetSpeak is probably also the first thing people go to when they're trying to swear around a filter. So, I made some modifications to wisen it up to the intricacies of Leet Speak.

Now...:

"pokie": 89%
"p00kie": 100%
"pkie": 50%
"pk": 0%
"p00k|e": 100%

The algorithm works almost perfectly now. For this word, you could easily set the threshold at 50% and you'd filter out everything you wanted.

So, that's the hard part done. I'll sleep on it to see if I can come up with anything better, and for good measure I'll see if I can run the past day or two of chat on my server through the plugin and see what it picks up on, both to see how good it is at catching words and to see how many false positives we get.

Enough of the codebase for this plugin is obsolete that I'm going to rewrite it from scratch. Might have something up by tomorrow or tomorrow night, depends on how much I can get done.

@realkangabean

Can you open a ticket about what's not working? Post console output.

EDIT: Actually, I know why it's not working, ignore that.