proposal: x/exp/xiter: add {All,Any,None}{,2}

[leaxoy]

Proposal Details

Proposed to add below funcitons:

-- func Fold[A, E any](s Seq[E], init A, f func(A, E) A) A--
// Remove Fold because original proposal's `Reduce` takes an initial value.

func All[E any](s Seq[E], f func(E) bool) bool {
    return Fold(s, true, func(x bool, e E) bool { return x && f(e) })
}

func Any[E any](s Seq[E], f func(E) bool) bool {
    return ...
}

func None[E any](s Seq[E], f func(E) bool) bool {
    return ...
}

Same as seq2.

Different from reduce, fold needs to provide an initial value

[jimmyfrasche]

None is just !Any. Does not seem worth it.

Presumably these would short circuit since Any can stop when it gets a hit and All can stop when it gets a miss. Is that correct?

Should this be Any[B ~bool](iter.Seq[B]) B and AnyFunc[E any ](iter.Seq[E], func(E) bool? Any2 wouldn't make sense in this scheme but AnyFunc2 would.

(and likewise for All)

[adonovan]

Because none of these functions declare type parameters beyond those of Seq or Seq2, they could be expressed as methods of the named types Seq and Seq2:

package iter

// All reports whether all elements of the sequence satisfy the predicate f.
func (Seq[E]) All(f func(E) bool) bool
func (Seq[E]) Any(f func(E) bool) bool // shorthand for !All(Negate(f))
func (Seq[E]) None(f func(E) bool) bool // shorthand for !Any(f)

[leaxoy]

Because none of these functions declare type parameters beyond those of Seq or Seq2, they could be expressed as methods of the named types Seq and Seq2:

package iter

// All reports whether all elements of the sequence satisfy the predicate f.
func (Seq[E]) All(f func(E) bool) bool
func (Seq[E]) Any(f func(E) bool) bool // shorthand for !All(Negate(f))
func (Seq[E]) None(f func(E) bool) bool // shorthand for !Any(f)

If Seq is a type alias as mentioned in the original proposal, we cannot add methods on it.

[adonovan]

If Seq is a type alias as mentioned in the original proposal, we cannot add methods on it.

iter.Seq is not an alias; generic aliases are not even part of the go1.23 spec. The discussion about whether it should have been an alias is just words after the fact.

[adamroyjones]

I'm here from #68945. (Thanks for connecting the two @seankhliao—I didn't see this when searching.)

Given the existence of slices.Contains and slices.ContainsFunc, we could alternatively have (for consistency)

func Contains[E comparable](seq Seq[E], e E) bool
func ContainsFunc[E any](seq Seq[E], func(e E) bool) bool

• Any is equivalent to ContainsFunc.
• None is the negation of Any.
• All(seq, f) is the same as None(seq, !f) which is the same as !Any(seq, !f).

[seankhliao]

xiter was withdrawn #61898