Saturday, October 23, 2010

Turing completeness

sed is a very powerful tool. A simple sed statement may turn a cat into cement. Observe: echo cat | sed statement
I was asked by mariom@ircnet whether is it possible to implement the Euklidean algorithm (the one that computes the greatest common divisor) in awk. awk is a Turing complete language, so the answer is yes, it is possible. There is a snippet proposed by mariom:

{
    a = $1;
    b = $2;
    while (b != 0) {
        c = a % b;
        a = b;
        b = c;
    }
    print a
}

Anyway, my response was that it is possible even in sed. Is it? Of course! Sed is a Turing complete language. Though I had no idea how to write it in sed. I use sed for simple substitutions only, but I could not admit that!

I started googling for arithmetic operations in sed and I found a dc implementation in pure sed. So the next question is how to implement the Euklidean algorithm in dc. It turned out to be quite simple, see:
[lalb%sclbsalcsblb0<F]sF sasblFxlap
This code assumes that there are two integers on the stack. So you can test it with something like that:
echo '20 25 [lalb%sclbsalcsblb0<F]sF sasblFxlap' | dc
Let's analyze what is happening there. There is F macro that is equivalent of "while" loop in awk code. It loads the values of a and b registers on the stack. Then replaces them with their reminder and saves it in the c registers.
lbsalcsb
statement just copies the value of b to a and c to b. Finaly it compares the value of b with 0. If former is greater it executes F (note: recurrence. It is the only way to implement a loop in dc). Otherwise it quits.
sasblFxlap
statement is an entry point. It saves values from stack in the a and b registers, then executes F and prints the content of the a register. So this sed script is an equivalent of awk script that executes Euklidean algothm.

Now it is enough to embed the dc script into dc.sed, et voila! Enjoy: Euclidean algorithm implemented in pure sed.