fix(internal): use a plain Map as the backing cache for wrapped marshaller

[mhofman]

closes: #12111

Description

Avoid the virtual aware WeakMap overhead

I did not change the capacity and kept 50, a number I pulled out of my hat. Open to bikeshed it.

Security Considerations

None

Scaling Considerations

None

Documentation Considerations

None

Testing Considerations

I didn't end up re-testing the performance impact of this, but conceptually this has much less overhead.

Upgrade Considerations

Will be picked up by any future vat bundle being deployed

[gibson042]

We shouldn't be ignoring a parameter; either drop it from the signature or leave this function as-is and update the call sites to explicitly opt-out of weak keying.

[mhofman]

The call site needs to pass weakKey for non-default maker, and as I provide a defaut makeCacheMap implementation, it needs to accept a parameter (or it'd be a type error). I could prefix the param with _ to make the eslint error go away, but I kinda dislike that in general.

Would a comment explaining why I'm ignoring the param satisfy your concern?

Not awake, there is no makeCacheMap.

[mhofman]

Kept the ternary and added a comment

[gibson042]

This is still unsatisfying. Why do we need a parameter whose value is irrelevant? The function is not even exported from packages/internal/src/marshal/wrap-marshaller.js. Suggestion:

/**
 * Construct a cache map with default capacity, backed by a non-weak Map because
 * a) we assume execution under liveslots, in which WeakMap is virtualized, and
 * b) the mapping is bidirectional by default and thus would pin keys anyway.
 * @template K
 * @template V
 */
const makeDefaultCacheMap = () =>
  /** @type {WeakMapAPI<K, V>} */ (
    makeCacheMapKit(capacityOfDefaultCache, {
      makeMap: Map,
    }).cache
  );

[mhofman]

I would like the ability to simply revert to weak keys in the future without having to change the callsite.

[gibson042]

That seems like an anti-goal to me; clarity is far more important and there's no clear condition on when such a change would be effected. But if you can come up with good explanatory comment text, it might work out.