misc

[filters.git] / README

This is a collection of filters to do all sorts of strange things to text.
B1ff, the Swedish Chef, and an eleet filter are included, along with many,
many more.

Since the original versions of some of these filters have copyright
problems, I have done a "clean room" re-implementation of the original lex
filters in perl. I did this without looking at the original code to the
filters at all, I simply observed their output and drew my own conclusions,
and wrote my own code. Then I compared the output of the original and new
filters when ran on large bodies of text, and fixed the things I had
missed.

Here are the results of my investigations of how the filters work:

eleet: 
	This is the simplest filter to figure out. The letters a-z of
	the alphabet are replaced with the following letters: 
		4 b c d 3 f g h 1 j |< l /\/\ /\/ 0 p q r 5 + u \/ \/\/ >< y z
        Note that the equivalent translation is done on upper-case letters.

fudd:

	This one is trivial too. A few transpositions happen differently if
	the letter is at the end of the word or not, and I think I found
	all the transpositions with only a few paragraphs of sample output.
	The only weird thing is that 'n.' is changed to "n, uh-hah-hah-hah."
	(Works well with sentances ending in "gun." ..)

chef:
	Personally my favorite filter. This took a bit of work to figure
	out, and I doubt I have everything correct. Note that due to a bug
	or an odd feature of the original program, it doesn't seem to think
	that the first letter of the first word is really the first letter 
	of a word and so some of the words below don't trigger for that first
	word. I did not emulate this behavior because I think it's probably a
	bug.

	The word "bork" is never changed, no matter what any of these 
	rules may say. Neither is "Bork".

	The following translations only happen to letters that are the first
	letter of a word of at least 2 letters in size:
		o -> oo
		O -> Oo

	These happen to letters that are not the last letter of a word:
		a -> e
		A -> e

	These translations only happen if the letter is not the first letter
	of a word:
		f -> ff
		i -> ee (but only the first `i' per word)
		o -> u
		u -> oo

	These translations always happen:
		v -> f
		V -> F
		w -> v
		W -> V
		an -> un
		An -> Un
		au -> oo
		Au -> Oo

	Any occurrence of "e" at the end of a word is changed to "e-a".
	
	Any occurrence of "e" and the beginning of a word is changed to "i".
	Same with capitals.

	Any occurrence of "the" is changed to "zee", any occurrence of
	"The", to "Zee".
	
	Any occurrence of "tion" in a word, to "shun".

	Any "th" at the end of a word is changed to "t".
	
	Any "en" at the end of a word is changed to "ee".

	Any "ow" at the end of a word is changed to "oo".

	And of course, after any `.', `?', or `!' that is at the end of a
	line, the Chef prints out a new line, and then the famous
	"Bork Bork Bork!", and then another newline.

	Known bugs in this implementation:
		- says "frunch", not "french".
		- still quite buggy in general.

b1ff:
	Unlike the other filters, b1ff does not try to be an exact duplicate
	of its predecessor.
	
	After upper-casing everything, b1ff does some word and sub-word
	substitutions, most of them misspellings.

	B1ff also changes punctuation - All commas become periods; all
	semicolons, commas. Question and exclamation marks are changed to
	things like "!!!!1!" and "?!?!?!". It looks like the original filter
	decided how long a sequence to print based on the text input before
	the end of sentence. I took a simpler route and just randomized it.

Written by Joey Hess <joey@kitenet.net>