Discuss
  • WeakMaps functionality, within regular object semantics

WeakMaps functionality, within regular object semantics

# Hugh Jackson (13 years ago)

From a user perspective, I'm really happy to see WeakMaps in the

proposals. The syntax for it deviates from js' currently cohesive core object model, though; requiring getter and setter methods and having no literal format.

One alternative that may work neatly is a 'hold me weakly' operator. For example's sake, I'm going to use *, but there may be other eligible ones:

As a key:

var key = {} , obj = { *key: 'val' } // when key is eligible for GC, the key:value pair gets deleted

obj[key] // -> 'val'; retrieval by key doesn't require the operator

As a value:

var val = {} , obj = { key: *val } // likewise, when val is eligible for GC, the key:value pair gets deleted

As both a key and a value:

var val = {} , key = {} , obj = { *key: *val } // when either key or val is eligible for GC, the key:value pair gets deleted

Non-literal form

var key = {} , val = {} , obj = {}

obj[*key] = *val // behaves the same as obj = { *key: *val }

I'm almost 100% convinced that the reason type of thing isn't in the proposal is that there's a some glaring semantic or implementation issue; but I thought it'd be worth mentioning anyway. If not only because of the comparative neatness of the syntax, perhaps also for helping keep the language cohesive as it grows.

Best,

Hugh

# Allen Wirfs-Brock (13 years ago)

The "weak reference" semantics needs to apply to the entire collection in order to avoid key/value circularity induced leaks

# Hugh Jackson (13 years ago)

Well, at least my powers of prediction haven't left me!

Thanks for humouring me.

Best,

Hugh

# Axel Rauschmayer (13 years ago)

Isn’t that a general problem with weak references, though (one that they’d have to solve)? Does Java have that problem, too? But they do have finalizers.

I’d also expect the map management data to be tricky to manage with weak references.

# Allen Wirfs-Brock (13 years ago)

On May 21, 2012, at 8:03 AM, Axel Rauschmayer wrote:

Isn’t that a general problem with weak references, though (one that they’d have to solve)? Does Java have that problem, too? But they do have finalizers.

I’d also expect the map management data to be tricky to manage with weak references.

Take a look at some of the references at the bottom of the Weak Maps wiki page harmony:weak_maps#see_also

# Jason Orendorff (13 years ago)

On Mon, May 21, 2012 at 9:42 AM, Allen Wirfs-Brock <allen at wirfs-brock.com> wrote:

The "weak reference" semantics needs to apply to the entire collection in order to avoid key/value circularity induced leaks

I don't think this is quite right. The semantics need to refer to both (a) the object containing the entry and (b) the key of the entry. But nothing says that the object containing the entries has to be a WeakMap and not just any old object.