2010-12-16



debianforum.de is a place I like much, but most  because  of  the
little  programming  questions  that can often be solved with the
standard tools (sed, wc, awk, tr, etc) in nice ways.  I  love  to
find cool and short and shorter solutions, just for fun.  For me,
these threads start to become really interesting when the  origi-
nal  question  is solved and the ``but I have a sorter solution''
game starts. :-)

Yesterday exactly such a thread came up.  [0] Although I have few
time  currently,  I  couldn't  resist  to go to find even shorter
solutions.

The question had been:

     Produce all numbers from ``00 00 00 00 00 00'' to  ``FF
     FF FF FF FF FF'', in this format, one per line.

The problem itself is not hard to solve, but with how few charac-
ters could one solve it?


I do like to use awk for such kinds of problems.

The first version I posted (75 chars):

     awk 'BEGIN{for(i=0;i<2^48;i++)printf("%012X\n",i)}'|sed 's,..,& ,g;s, $,,'


The second version (70 chars):

     awk 'BEGIN{while(i<2^48)printf("%012X\n",i++)}'|sed 's,.., &,g;s,.,,'

I did not need to initialize the variable because  all  get  ini-
tialized  to  0  (or  "") on first use.  Using for or while loops
does not matter, they need the same amout of chars.  In sed:  In-
stead of appending the space to the pairs, I prepended it, then I
could remove the anchor because the dot  would  match  the  first
char.


The third version (68 chars):

     id|awk '{while(i<2^48)printf("%012X\n",i++)}'|sed 's,.., &,g;s,.,,'

I posted the second version together with the third  one  because
the latter is very ugly, only to remove two more chars.


Then I realized that awk's integers  are  32  bit  unsigned.   No
number  larger  than 2^32-1 was generated.  I knew that bc(1) had
no such limits, hence I wrote a new version  that  would  produce
correct results.

The fourth version (94 chars):

     echo 'obase=16;while(i<2^48)i++'|bc|awk '{printf("%0*d",12-length,0)}1'|sed 's,.., &,g;s,.,,'

Bc has, AFAIR, no possibility to format the output, thus the  awk
post processor.


When someone came up with a C++ approach, I converted it into C:

     #include <stdio.h>

     int
     main(void)
     {
             unsigned long long i=1LL<<24;
             while (i<1LL<<48) {
                     printf("%012llX\n",i++);
             }
             return 0;
     }

This should be valid C99 code.

Of course this is quite wordy, hence I  stripped  everything  the
compiler does not need so much:

     long long i;main(){while(i<1LL<<48)printf("%012llX\n",i++);}

For the IOCCC I now would need to start using  the  preprocessor.
;-)

This program produces only the numbers, the sed script  needs  to
be appended.  Hence surely no better.


An approach in Python3 by dimi is better:

     python3 -c "for i in range(2**48): print('%012X'%i)"

(It does not run with Python 2.)

As in the C program, the sed script is still needed.  Then,  with
75 chars, it's less than the bc version:

     python3 -c "for i in range(2**48): print('%012X'%i)"|sed 's,.., &,g;s,.,,'

Maybe one could shorten the Python part  further  more;  I  don't
know Python well.


Later in the thread the original had been extended to the wish to
filter  out  all  numbers with three equal hex-bytes next to each
other.  This addition to the problem, unfortunately, hadn't  been
described exact enough, thus some corner-cases are unsure.

My solution therefore is:

     grep -v '\(..\) \1 \1'


Comments of any kind are welcome!

[0]  http://debianforum.de/forum/viewtopic.php?f=28&t=125571


http://marmaro.de/lue/        markus schnalke <meillo@marmaro.de>