Array filter: Positional access

[ChristianGruen]

Array filter: Positional access

[michaelhkay]

The reason I left out numeric subscripting is that it's not clear whether it should return a single member or an array of members.

You would probably expect

let $array := [1, 2, 3, 4, 5] return $array?[3]

to return 3 rather than [3].

Overloading filtering and subscripting works for sequences because an item is the same thing as a sequence of length 1. That equivalence doesn't exist for arrays, so if you overload the operator there's immense scope for confusion.

As the spec says:

Unlike filter expressions for sequences, there is no special treatment of numeric predicate values. This is because it would be unclear whether such an expression should return the member at the specified position, or a singleton array containing that member. Numeric values for the predicate are disallowed to avoid any possible confusion.

[ChristianGruen]

Unlike filter expressions for sequences, there is no special treatment of numeric predicate values. This is because it would be unclear whether such an expression should return the member at the specified position, or a singleton array containing that member. Numeric values for the predicate are disallowed to avoid any possible confusion.

I’ve seen the comment, but I don’t find it persuasive. [$n] and [position() = $n] have always been equivalent; if we believe that this syntax could be immensely confusing for arrays and possibly maps, we should definitely drop positional access in array/map predicates in general. Next, now that we allow EXPR[$n1, $n2, ...], the single integer is just a special case.

My preference, however, would be to make predicates completely symmetric, no matter if they are part of XPath steps, Filter Expressions or Array/Map Filters.