Re: Loops (was Re: do { quit; } else { })

On 15/04/2025 14:19, Kaz Kylheku wrote:

On 2025-04-15, bart <bc@freeuk.com> wrote:

On 15/04/2025 08:17, Janis Papanagnou wrote:

and someone already posted another iteration type on linked lists
>
      for (node = list; node; node = node->next)

>
These two I get, and I can tell you that they're all WHILE loops. To
make make them tidier, I'd have fewer complaints about them if they
instead looked like this:
>
    while (node = list; node; node = node->next)

 That's mostly bike shedding. It could also have been called:
      do (node = list; node; node = node->next)
      loop (node = list; node; node = node->next)

That's sort of the point. It is not 'for' as is it generally understood. It's more of a souped-up while.
Thats's fine. But it means a real 'for' loop doesn't exist in C; you have to emulate it using that 3-way construct, which is naff, and also error prone.

Lisp called this kind of general variable stepping loop "do"
    (do ((var1 init1 step1)
        (var2 init1 step2)
        ...)
       (until-this-is-true also-this ... yield-this-result)
    do-this
    and-this
    ...)
 Naming things can be hard. The reason that the loop is called for
is that most instance of for have a focus on the specific variable
that is initialized, and the guard condition has to do with it
hitting some specific terminating value.

In what I recognise as 'for' loops, any variable is specified only once in the header. Any initialisation and increment is handled automatically.
The same applies if it loops over a set of values; that might mean managing some implicit index. But hey, this is a HLL, and it's the compiler's job to that.

So, you can optionally add extra elements to a while (x).
>
It would be necessary to either define whether while(x; y) means while
(; x; y) or while (x; y), or to always require either zero or two
semicolons inside (...).
>
That could have kept 'for' for how it works in other languages (the ones
that haven't just blindly copied C's version!).

 Common Lisp uses "for" to indicate several kinds of variable
stepping clauses inside the loop syntax, some of which are:
    [4]> (loop for x in '(a b c)
              for y = 1 then (1+ y)
              for z = (* y 10)
              with w = (* y 10)
              collect (list x y z w))
   ((A 1 10 10) (B 2 20 10) (C 3 30 10))

I've looked in the past at CLisp and its loops. My view is that the number of possibilities it provides is over the job. It's like somebody was tasked with thinking up as many as possible, and making them all available.
I don't know what your example does. Are these nested loops? Parallel ones?
My own reduce to this:
    for [i,]x in A [when cond] do       # i exposes internal index
and the older:
    for i [:= A] to/downto B [by C] [when cond] do
Languages like CPython go to town a bit more, and do deconstructing etc. However the only thing I've copied from that is an optional 'else' part.
Unusual loops are rare and the requirements diverse enough that I can't see the point of trying to provide dedicated features beyond the basics.
For custom loops, people can start off with:
    do
       .....
    end
then add what they like out of ordinary features.

When loop is used to simulate the classic Lisp do loop,
it's done with the "for var = init then step" clauses.
 There is a "for var from N to M" clause, as well as "for var below N"
which goes from 0 to N-1.

That's a poor way to do it. I suggested elsewhere using only inclusive ranges, and using '-1' as a postfix 'operator' to treat it as exclusive.
(Inclusive ranges are more natural:
     Monday .. Friday
     'A' .. 'Z'
     Iterate over the elements in X
People don't expect that first to stop at Thursday, or to only go up to 'Y', or to iterate over all elements except the last!)

Since it can always be trivially be written as:

 

     for (;cond;)

 That is simply untidy looking.

But this:
   for (index = A ; index <= B; index += 1)
is fine!