Implement WeakRef #148
Implement WeakRef #148
Conversation
saghul
force-pushed
the
weakref
branch
3 times, most recently
from
78f00b6 to
49f6783
Compare
saghul
force-pushed
the
weakref
branch
2 times, most recently
from
936f51b to
f2487ce
Compare
|
@bnoordhuis LMK what you think. I went this route since it seems the simplest, given the use of weak references is discouraged. I'll add FinalizationRegistry after, once we agree on the approach. If this one is taken, it would be a new "kind" of weak reference I reckon. |
| @@ -43787,10 +43809,18 @@ static JSMapRecord *map_add_record(JSContext *ctx, JSMapState *s, | |||
| mr->map = s; | |||
| mr->empty = FALSE; | |||
| if (s->is_weak) { | |||
| pmr = get_first_weak_ref(key); | |||
| JSWeakRefRecord *wr, **pwr; | |||
| wr = js_malloc(ctx, sizeof(*wr)); | |||
There was a problem hiding this comment.
Would it be possible to embed the JSWeakRefRecord inside JSMapRecord so you don't have to do this allocation? That's one fewer failure state.
If you're worried about embiggening JSMapRecord with three words, you could drop the kind field and use the LSB of the map_record/weak_ref_data pointer to distinguish between them.
I suppose the counterargument to my suggestion is that WeakMap and WeakRef aren't very common and therefore not worth optimizing to death. :)
There was a problem hiding this comment.
I'm taking the way out on this one :-P
|
Yes maybe a hash table on the context with the object as a key. But that might add an extra pointer too... I'm happy with how it turned out for now, happy to revisit if / when needed :-) |
Ref: #54