fixes issue #96: registry refactor to introduce registry entry objects #97
Conversation
| @@ -171,7 +171,7 @@ When Reflect.Loader is called with no arguments, the following steps are taken: | |||
| 1. If NewTarget is *undefined*, then throw a *TypeError* exception. | |||
| 1. Let _O_ be OrdinaryCreateFromConstructor(NewTarget, "Reflect.Loader.prototype"). | |||
| 1. ReturnIfAbrupt(_O_). | |||
| 1. Let _registry_ to a new Registry(_O_). | |||
| 1. Let _registry_ to a new Registry(). | |||
There was a problem hiding this comment.
registry doesn't need a backpointer to loader anymore.
| 1. Let _entry_ be *this* value. | ||
| 1. If Type(_entry_) is not Object, throw a new *TypeError*. | ||
| 1. Let _stageValue_ be ToString(_stage_). | ||
| 1. If _stageValue_ is an abrupt completion, return a promise rejected with a new *TypeError* exception. |
There was a problem hiding this comment.
Actually stageValue.[[value]]. Alternately I think you can just replace this whole step with RejectIfAbrupt(stageValue).
There was a problem hiding this comment.
yeah, this is very confusing because we use stageValue as a returned value and as a record of completion. I will use RejectIfAbrupt(stageValue).
* editorial changes to use ? instead of ReturnIfAbrupt. * no more assertions to validate public methods, instead using If and throw. * using verbose form to validate instances of Loader, Registry, Module and ModuleStatus. * RejectIfAbrupt() when possible. * all arguments in abstract operations are now validated. * Removing unnecessary returns in promises that are not returned to the caller or stored.
| 1. Set _entry_.[[Dependencies]] to _deps_. | ||
| 1. Set _entry_.[[Module]] to _instance_. | ||
| </emu-alg> | ||
|
|
||
| <emu-note> | ||
| [[Key]] wil identifies the dependency module status in the entry's corresponding loader. |
* corrections on intrinsic object definitions * CreateRegistry() abstract operation is the only way to create registry objects.
|
@caridy still looks like Registry has a Map of ModuleStatus as the inspection mechanism, which is still using Promises like candy for [[Result]]. |
|
@bmeck that's correct, but the inspection mechanism is now sync, you can essentially push new modules (resolved module entries) into the registry by creating new |
|
fixes #98 |
|
@caridy the registry cannot be inspected synchronously at all if none of the results can be inspected synchronously. it is fine if the user API uses promises for this, but the registry is still a no-go. Other comments around the repo look like they want to make subclassing the loader work with synchronous loading, but this is not possible if all registration results require turning the event loop. |
|
@caridy perhaps you can go through the inspection workflow in less text than the whole spec to show how to synchronously get the results? |
|
@bmeck we can have a quick chat, and we can go over the details. just ping me! |
|
|
||
| The value of Reflect.Loader.prototype is an ordinary object. | ||
| The value of Loader.prototype is %LoaderPrototype%. |
There was a problem hiding this comment.
The initial value of Loader.prototype is the intrinsic object %LoaderPrototype%.
|
r+ with one fix for the issue where |
…ry-objects fixes issue #96: registry refactor to introduce registry entry objects
HTML Version
https://rawgit.com/caridy/6c921c618d3cc2b162ef/raw/2528a3224b0962a415fefc65565ac51bac2f2ec0/loader-refactor-registry-entry.html