Re: question about linker

On 11/12/2024 22:51, Waldek Hebisch wrote:

Bart <bc@freeuk.com> wrote:

On 07/12/2024 11:34, Janis Papanagnou wrote:

On 06.12.2024 02:20, Bart wrote:

>

I used to be puzzled by this too: 'while' can both start a while
statement, and it can delimit a do-while statement. How is that possible?

>
A keyword (like 'while') is just a lexical token that can be used in
different syntactical contexts.

>
Is it common to both start and end a statememt with the same keyword?

 Technically, 'do' loop ends in a semicolon.

A do-loop ends with:
        while (cond);
A while loop starts with:
        while (cond)
Sometimes a while loop starts and ends with:
        while (cond);

Concerning "start and end a statememt with the same keyword",
this is clearly common when statement consist of a single
symbol, like empty statement in many langages, including C and
Algol68

I think it's uncommon. In Algol68 you both start and end a comment with 'comment', or 'co' or '#', which is a spectacularly bad idea. It's easy to get them out of step. Or if looking at an isolated delimiter, you can't tell, if in the middle of code, whether it is the beginning or end of the comment.

I think it was already noted that this is not a valid C statement.
      do while(cond) ; while (cond);
 is valid, and indeed may be confusing to humans.  But people
would normally write it as
      do {
         while(cond) ;
     } while (cond);

That's still pretty confusing! In languages where 'while' doesn't do two jobs, you see 'while' and know instantly what it denotes. With C, you need to scan outwards from a 'while' to be able to infer from the surrounding context whether it is normal while-loop or is part of do-while.
I consider that undesirable.

or even
      do {
         while(cond) {
            ;
         }
     } while (cond);
 to stress that inner loop is "do nothing" loop.  In this form
structure is clear.

If you want human-oriented rule for parsing badly written code,
you may notice that body of a C loop can not be syntactually
empty, you need at least a semicolon.  So first 'while'
after 'do' must start an inner loop.

The inner loop might be this:
   do {
      s1;
      while (cond) {}
   } while (cond);
Here, it is now the first { that is fragile. But this is a general problem in C with braces, when braces are optional. A missing/extra brace might be offset by an extra/missing brace later on. That can result in a still-valid structure, but the wrong one.

Concerning "subtle", note that changing say '*' to '+' can
signifcantly change parse tree and semantics in almost any
programming language.  Most people are not bothered by this
and take care when writing/reading operators.  Similarly
C programmers are careful when writing/reading semicolons.

; and to a lesser extent {} are regarded as punctuation. The problem is more in leaving them out or adding extra ones.
Being generally able to silently make typos is a separate problem that occurs everywhere including my language.
But it is not as sensitive as C is to messing up '; { }' because of stricter rules on layout, even though it looks more lax:
   repeat
     s1
     while cond do od
   until cond
There are no semicolons or braces to get wrong! Extra semicolons are either ignored or give a syntax. I can't even get while/repeat confused as they use different keywords.

You can design a laguage with more redundancy.  For example
one of languages that I use has
     if cond then expr1 else expr2 endif
     while cond do expr endwhile
     repeat expr endrepeat

OK, I should have read this first. Of course, that's already been done. C decided to keep its delicate grammar rather than tighten it up.