Actually, the first 3 examples work exactly how the simple model I offered in my previous posts would expect them to.
In the first case the concatenation with () creates a new list to loop over so that the addition to @a has no consequences while the second case adds to @a keeping the end out of reach like a carrot on a stick dangling before the horse.
Same for the third case. After the second value was poped from the array, the loop naturally was finished early. But the fourth case shows the problem. Seems the values of the "anonymous array" are aliased to the real values and have a problem when that value isn't there anymore.
But I can see why "Don't do it" is the better answer. Otherwise too much of the internas would spill out into the language design.
In reply to Re^4: How foreach loops decide what to iterate through (push array inside foreach)
by jethro
in thread How foreach loops decide what to iterate through
by lefthanded
| For: | Use: | ||
| & | & | ||
| < | < | ||
| > | > | ||
| [ | [ | ||
| ] | ] |