2012-05-10




Currently, I'm reading ``The UNIX System'' by S.R.  Bourne.   The
book  is  similar to ``The UNIX Programming Environment'' by Ker-
nighan and Pike, but more targeted to users.  (Note that users of
Unix  are  tought sh, C and troff. This is part of the philosophy
of the system.)  While the book by Kernighan and Pike is a  clas-
sic,  the  book by Bourne is hardly known. I like Bourne's book a
lot because it gives very good and compact introductions  into  a
lot of topics.



Complex declarations in C are large obstacles for newcomers.  But
they  don't  require any guru knowledge, in fact, they don't even
require thinking. It's even better not to think at all  but  only
operate as state machine.  This  document  describes  how:   [0].
It's not at all difficult.



Have a look at this snippet of  command  line  argument  handling
code in C:

     while (argc > 0 && **argv == '-') {
             switch (argv[0][1]) {
             case 'e':
                     --argc; ++argv; if (argc == 0) usage();
                     include_event(*argv); --argc; ++argv;
                     break;
             case 'i':
                     --argc; ++argv; if (argc == 0) usage();
                     interval = atoi(*argv); if (interval == 0) usage();
                     --argc; ++argv;
                     break;
             default:
                     usage();
             }
     }

It appears as if this kind of compact code writing is typical for
lecturers.  They want to fill up their slides and provide the au-
dience as large a window into the code as possible (with readable
font size).

Take the one line:

     --argc; ++argv; if (argc == 0) usage();

It's four or five lines of code in a decent layouting style.

I believe that few code (i.e. few code entities and  short  iden-
tifiers)  is  good.   I believe that the brevity of C can improve
the readability of the code for programmers who are familiar with
this  style.   But  I think that the structure of the code should
follow a clear schema.  To me the rules are:

One statement, one line.  The body of each control statement is a
block  in  curly  braces,  having  the  closing brace on the same
column with the keyword.

These rules are violated here.  The code is too much  Perl-style.
That's when reading becomes difficult.

I can accept to have `argc--; argv++;' on one line together,  be-
cause it connects the two statements together, making them effec-
tively one operation.  A similar case  is  pushing  to  a  linked
list:

     node->next = head;
     head = node;

I'm indifferent whether to write them into one line or not.

I do like to have a short comment ``push'' here; I find  it  con-
venient.   When  the  statements  are  on two lines, I don't know
where to best put the comment. ;-) Still, I rather have  them  on
two lines.

[0]  http://ieng9.ucsd.edu/~cs30x/rt_lt.rule.html


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