Editorial: Model the bindings of a declarative Env Rec as a List of Records

[jmdyck]

Making this a draft PR because I'm not sure about module Env Recs.

[jmdyck]

Some alternatives for DeclarativeBindingRecord:

(1)
Replace [[IsInitialized]]: Boolean, [[BoundValue]]: ECMAScript language value
with [[BoundValue]]: ~uninitialized~ or ECMAScript language value
[Later: I did this.]

(2)
To avoid the weirdness of the [[BoundValue]] field in indirect bindings...

(2a)
Instead of defining the sub-type ImportDeclarativeBindingRecord, allow the [[BoundValue]] field of a DeclarativeBindingRecord to be a record with just the [[TargetModuleRec]] and [[TargetName]] fields.

(2b)
Or you could have an abstract base type that is specialized into either having a [[BoundValue]] field or the [[TargetModuleRec]] and [[TargetName]] fields. (E.g. DeclarativeBindingRecord specializes to DirectDeclarativeBindingRecord and IndirectDeclarativeBindingRecord.) This is probably nicest from a theoretical standpoint, but 2a is basically equivalent without having to set up another type hierarchy.

[Later: I went with 2b, basically.]

[jmdyck]

(forced-pushed after rebase to master)

[jmdyck]

(force-pushed to resolve merge conflicts)

[jmdyck]

Okay, this PR is ready for review.

Issue #2639 might affect SetMutableBinding, but probably not much.

[michaelficarra]

Suggested change

	<p>A <dfn>DeclarativeBinding</dfn> is either a SimpleDeclarativeBinding or an ImportDeclarativeBinding. SimpleDeclarativeBindings can appear in any Declarative Environment Record, but ImportDeclarativeBindings can only appear in a Module Environment Record.</p>
	<p>A <dfn variants="DeclarativeBindings">DeclarativeBinding</dfn> is either a SimpleDeclarativeBinding or an ImportDeclarativeBinding. SimpleDeclarativeBindings can appear in any Declarative Environment Record, but ImportDeclarativeBindings can only appear in a Module Environment Record.</p>

Same for SimpleDeclarativeBinding and ImportDeclarativeBinding.

[jmdyck]

Done.

[michaelficarra]

This field needs to be moved to the description of DeclarativeBinding since we refer to it on values that are known only to be a DeclarativeBinding.

[jmdyck]

Are you saying it needs to be that way in order for the spec to be "correct" in some sense? Because it sounds like you're talking about statically typing the pseudocode, and you have in mind a specific rule for how that should work.

As a counter-example, consider ComposeWriteEventBytes, which has _W_.[[ByteIndex]], where _W_ is either a WriteSharedMemory event or a ReadModifyWriteSharedMemory event, each of which separately declares a [[ByteIndex]] field. There are several other such examples.

[michaelficarra]

Can't we just say SimpleDeclarativeBinding in place of DeclarativeBinding in the above step then?

edit: Same below.

[jmdyck]

I don't think the spec does that.

That is, in a phrase like:

the [thing] in [some list] that satisfies [some constraint]

I don't think the [thing] phrase is ever restrictive. Rather, it's a descriptor that's true of every element of [some list]. All the restriction happens in [some constraint].