I was reading some TeX guides from where I have found this source:

Fjfi71PAVVFjbigskipRPWGAUU71727374 75,76Fjpar71727375Djifx
RrhC?yLRurtKFeLPFovPgaTLtReRomL;PABB71 72,73:Fjif.73.jelse
B73:jfiXF71PU71 72,73:PWs;AMM71F71diPAJJFRdriPAQQFRsreLPAI
I71Fo71dPA!!FRgiePBt'el@ lTLqdrYmu.Q.,Ke;vz vzLqpip.Q.,tz;
;Lql.IrsZ.eap,qn.i. i.eLlMaesLdRcna,;!;h htLqm.MRasZ.ilk,%
s$;z zLqs'.ansZ.Ymi,/sx ;LYegseZRyal,@i;@ TLRlogdLrDsW,@;G
LcYlaDLbJsW,SWXJW ree @rzchLhzsW,;WERcesInW qt.'oL.Rtrul;e
doTsW,Wk;Rri@stW aHAHHFndZPpqar.tridgeLinZpe.LtYer.W,:jbye

The text preceding the code is something like:

TeX is a macro language and the meaning of existing commands can be changed on the fly, and also new commands can be dened on the fly [Knu84]. As perhaps the most extreme example of this is David Carlisle's xii.tex TeX code, which is obtainable as http://mirrors.ctan.org/macros/plain/contrib/xii/xii.tex:

Now my question is, how this so-called extreme example of macro could be simplified to be readable to the beginner TeX/LaTeX users?

I have run the code through TeX and I was really surprised to see the result!

One more basic question I do have:

Could all TeX commands be used from any LaTeX file?


With the answer from hendrik-vogt and the related answer from joseph-wright, I was trying to understand the obfuscated code.

TeXifying the code after adding \traceall at the beginning gives perhaps useful information, but still quite stumbling blocks to me is something like:


That seems to me something like BF. Not very well, but still I am getting some hints what is going on behind the scene. Can someone help me what does the previous block of code means?

  • 3
    Hi Cylian, the whole point of that macro is to not be readable! The simplified version would simply be to type the resulting text.
    – Scott H.
    Commented Mar 29, 2013 at 6:52
  • 1
    I am pretty sure David would say that it is already simplified as much as can be. :-) Commented Mar 29, 2013 at 6:52
  • @ScottH.: But, I like to on which basis these entire string was written? Is this code somehow like Java Bytecodes that only could be interpreted by the compiler?
    – Cylian
    Commented Mar 29, 2013 at 7:03
  • 3
    We were asked about another such document very early on in the life of the site: tex.stackexchange.com/q/2272.
    – Joseph Wright
    Commented Mar 29, 2013 at 7:37
  • 13
    I'm very hurt that people don't find my code readable:( Joseph has already given one link, and this came up a day or two ago and there are some links to other sites where this is described tex.stackexchange.com/questions/104248/… Commented Mar 29, 2013 at 10:01

3 Answers 3


The short answer

There are actually three levels of obfuscation in that code: Firstly, some clever recursive macros are used to compress the text to just a few lines. Secondly, some kind of substitution and transposition ciphers are implemented by rather straightforward macros like \defR#1#2#3{#3#2#1} (where R is an active character). The third level is the use of category codes to avoid intellegible TeX-specific characters like {, }, # and \.

Some more explanations

I saw this TeX file more that 10 years ago and thought Eh? So I tried to take it apart step by step. On the way I learned a lot about TeX's macro language, and also about category codes. To begin with, I'll only post the very first steps I took. (I still have the files on my computer, and this one is named hae.1, "hae" being German and meaning "eh?")

\let~\catcode  ~`76  ~`A13  ~`F1  ~`j00  ~`P2
A''{w}      A;;{}    AZZ{LaL}    A//#1{#1i}   AHH{L}
Azz{en}     ASS{th}; A$${ev}     A@@{f}
ADD{Rgni}   AWW#1{}  ATT{ve}     A**{stRsam}  AGG{Rruo}
Aqq#1.#2.{#1#2#1}                AYY#1#2{#2#1#1}
A??{i*Lm}   A&&#1\fi{\fi#1}      AVV{\bigskipR}WG
AUU#1#2#3#4 #5,#6{\par#1#2#3#5D\ifx:#6\else&U#6\fi}L
AXX#1{VLnOSeL#1SLRyadR@oL RrhC?yLRurtK{eL}{ov}gaTLtReRomL;}
ABB#1 #2,#3:{\if.#3.\else B#3:\fiX{#1}U#1 #2,#3:}Ws;
AMM#1{#1di}  AJJ{Rdri}  AQQ{RsreL}  AII#1{o#1d}  A!!{Rgie}
Bt'el@ lTLqdrYmu.Q.,Ke;vz vzLqpip.Q.,tz;
;Lql.IrsZ.eap,qn.i. i.eLlMaesLdRcna,;!;h htLqm.MRasZ.ilk,%
s$;z zLqs'.ansZ.Ymi,/sx ;LYegseZRyal,@i;@ TLRlogdLrDsW,@;G
LcYlaDLbJsW,SWXJW ree @rzchLhzsW,;WERcesInW qt.'oL.Rtrul;e
doTsW,Wk;Rri@stW aHAHH{ndZ}pqar.tridgeLinZpe.LtYer.W,:\bye

You see, I replaced F with { and P with } throughout. Why? ~`F1 gives the character F category code 1 (since the active character ~ is \let to \catcode), and TeX understands such characters as opening braces. In the same way, P is given category code 2 by ~`P2, so it acts as a closing brace.

The next thing is to understand ~`76 and ~`j00. The first makes 7 behave like the macro parameter character # (category code 6), and the second makes j behave like the control sequence character \ (category code 0), so I've replaced 7 with # and j with \ throughout. This enhances readability quite a bit already. Moreover, I added some white space and line breaks, which helps some more.

The key point now is to understand what A does. This is an "active character" (category code 13) due to ~`A13, so it behaves like a control sequence. So what does the definition \defA#1{~`#113\def} mean? A takes one argument #1. Then ~ acts as \catcode, so A gives it's argument category code 13, and then it issues an additional \def.

So how does this fit with the output of \tracingall you got? The first line


says the following: To the left of -> you see that the active character A takes one argument #1 (recall that 7 acts as #); to the right of -> you see the corresponding expansion. Now the second line


says that #1 should be substituted with L, so that the expansion is ~`L13\def. Now ~ was \let to \catcode, so the category code assignment \catcode`L13 is performed next (making L an active character); then the \def is executed. This is what you see in the next two lines:


(Unfortunately \tracingall doesn't say anything specific about the execution of \catcode and \def.)

Let's look at the usage A''{w}. This expands to \catcode`'13\def'{w}, so ' is made an active character, and it's given a definition, namely that ' should expand to w (one of the transposition ciphers)! Just one more example: A??{i*Lm} makes ? active and gives it the definition i*Lm, which in turn expands to istRsam m since * expands to stRsam and L to a space. The final result is istmas m since R acts as a transposition cipher, as mentioned in the very beginning. And now we're able to understand the little piece RrhC?y in the code – it expands to Christmas my!

A challenge

If you remove all the cipher and catcode business, you just have some recursive macros that represent the text rather efficiently. Can anyone do it with less than 479 characters?

\let~\def~\U#1,#2:{\par#1ing \if.#2.\else\U#2:\fi}~\,{\def\,{and
}}~\;#1~#2 #3,#4:{\if.#4.\else\;#4:\fi\bigskip On the #1#2th day of
Christmas my true love gave to me\U#1#3,#4:}~~#1 {}\;twel~f ve drummers
drumm,eleven~ \ pipers pip,ten~ \ lords a leap,nin~ e ladies danc,eigh~
t maids a milk,seven~ \ swans a swimm,six~ \ geese a lay,fi~f ve gold
rings~,four~ \ calling birds~,th~ird\ ~ ree french hens~,~second\ ~ two
turtle doves~,~first\ ~ \,a partridge in a pear tree.~,:\bye
  • I really appreciate the approach.
    – Cylian
    Commented Mar 29, 2013 at 7:28
  • 1
    Inspired by your helpful start, I started to decipher it myself, but somewhere around \def B it becomes too painful to follow. I salute you for your efforts!
    – morbusg
    Commented Mar 29, 2013 at 13:22
  • @morbusg: B just does the recursion :-) Commented Mar 29, 2013 at 17:33
  • Unputdownable! Wish I could vote it more than once! One question, could you tell me a book from where I could acquire some knowledge to learn this kind of tricks. Anyway, thank you very much for your help.
    – Cylian
    Commented Mar 30, 2013 at 6:52
  • In the code at the end, I think, in the definition of \U, one should transform \U#2:\fi to \fi\U#2:, so that we don't get too many levels of nested \ifs, as in the original xii.tex, or one doesn't have to do so?
    – awllower
    Commented Mar 27, 2018 at 6:24

I think actually it is as simple as it can be already. It has two \ which is, I admit, rather complicated. I did try to simplify it so that it only had one, or even better no, backslash but I don't think that is possible.

  • By two backslashes, do you mean the \catcode`j=00? I.e., why the second zero?
    – morbusg
    Commented Mar 29, 2013 at 10:31
  • @morbusg: No, the code starts with \let~\, so the density of \ is rather high at the beginning :-) And the double 0 is just to get full justification of the code. Commented Mar 29, 2013 at 10:37
  • 1
    Can you make the \let, \catcode, def and bye more obscure? Commented Mar 29, 2013 at 11:27
  • 1
    @Karl'sstudents let and catcode no as I say I tried to avoid having any backslash but 2 was the lowest I could get (without slipping a back door into the latex format which was tempting) bye I could get rid of but only by having csname endcsname somewhere in the code and I was trying to avoid too many control names even with j Commented Mar 29, 2013 at 11:29
  • 10
    @HendrikVogt when the code was published in TUGBoat Barbara Beeton claimed that it should be lords-a-leaping, I told her she was the editor, she was welcome to add the - where needed:-) Commented Mar 29, 2013 at 18:45

Here is one simplification I found from https://bitbucket.org/ambrevar/xii.tex/overview by Pierre Neidhardt:


\def\verse#1 #2,#3:{\if.#3.\par #2{}\else\par#2\verse#3:\fi}

\def\day#1{\bigskip On the #1 day of Christmas my true love gave to me}

\def\gen#1 #2,#3:{\if.#3.\else \gen#3:\fi\day{#1}\verse#1 #2,#3:}

twelfth twelve drummers drumming,%
eleven eleven pipers piping,%
tenth ten lords a leaping,%
ninth nine ladies dancing,%
eighth eight maids a milk,%
seventh seven swans a swimm,%
sixth six geese a lay,%
fifth five gold rings,%
fourth fourr calling birds,%
third three french hens,% 
second two turtle doves,%
first \endverse\def\endverse{and }a partridge in a pear tree.,:


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .