On 2024-01-18 14:34, Bart Schaefer wrote:
>    asc=( [one]=1 [two]=2 [nil]= )
>
> The "not clever" pairwise assignment
I don't want to sound to flippant.  Anything other than pairwise would 
obviously be much more difficult to implement and I myself have never 
seen any reason why simply maintaining the pairs should be considered a 
problem.  I did experience some crash and burns on first playing with A 
arrays, but it's not hard once you know how.
> Associative arrays are maintained as hash tables for fast access.
It's something way over my head.
> ... storing extra data about the order
> (such as a linked list through the hash elements).  Zsh opted to do
> neither of these on the premise that access to individual elements by
> key is far more frequent than access to the entire table.

Sure.  Perhaps Mark has some idea of the nuts and bolts of the thing in 
ksh -- whether or not it's a labor to implement the determined order or 
not -- but that's up to you devs.  As it is, for diagnostics I do end up 
printing entire arrays, and it would be nice to have a predicted order 
but not if it's going to be any trouble.  Funny, now that I try to break 
the order I can't:

typeset -A asc=()
asc=( [one]=1 [two]=2 [three]=3 [four]=4 [five]=5 )
print -l "$asc"
asc[six]=6
print -l "$asc"
asc[two]=22
print -l "$asc"
asc[three]=
print -l "$asc"
asc[three]=33
print -l "$asc"
asc[one]=
print -l "$asc"
asc[one]=11
print -l "$asc"

1 5 4 2 3
1 5 4 2 3 6
1 5 4 22 3 6
1 5 4 22  6
1 5 4 22 33 6
  5 4 22 33 6
11 5 4 22 33 6

... once it decides on 'one five four two three' ... it keeps it.  How 
hard would it be to get the first hash 'right'?





>