[js] Update js.html externs to Aug 2018 WebIDLs #7354
Conversation
…elds and types, remove deprecated fields, interface->typedef for non-implemented
…e unlikely to be used)
|
Regarding deprecations: at the moment I've retained deprecated files but I'm happy to delete these files for the 4.0.0 release as I don't expect them to have much real-world use |
…Aug 2018 WebIDLs)
|
Hold off merging just yet; I'm working through popular projects and checking for issues |
|
@Simn @ncannasse this looks like Haxe 4 material to me |
|
There’s a few more tweaks on the way but I expect it to be complete soon :) |
|
I'd say the new externs are ready to go I've tested sample projects for popular libraries:
|
|
@nadako: I'll let you pull the trigger on this. |
|
This is an awesome contribution, thanks a lot! Get ready for some backward-compatibility issues though :) |
|
Indeed, thanks a lot for this! I have added you to the Haxe core team so you can deal with the fallout. |
|
Hmm, I wonder why OpenFL compiled for you, I'm getting an error because of the missing |
some personal one with |
| Scrolls the page until the element gets into the view. | ||
| **/ | ||
| @:overload( function( ?arg : ScrollIntoViewOptions) : Void {} ) | ||
| function scrollIntoView( ?arg : Bool ) : Void; |
There was a problem hiding this comment.
I'm late to the party, but this change is problematic.
Having ?arg : Bool instead of ?arg : ScrollIntoViewOptions as the primary signature for scrollIntoView breaks type inference and makes the use of ScrollIntoViewOptions really annoying.
Is there a way to prioritize in the generator?
Edit: there seem to be other APIs with the same problem, e.g. animate below.
There was a problem hiding this comment.
Relying on type-inference on overloaded functions is asking for trouble. But if this is a case where we flipped the default then it might make sense to stick with the previous order.
There was a problem hiding this comment.
Are we supposed to ECheckType every "object" argument in overloaded functions?
This seems really annoying to use =/
el.scrollIntoView(({behavior: SMOOTH, block: NEAREST} :js.html.ScrollIntoViewOptions));There was a problem hiding this comment.
The order comes from the WebIDLs but I could try to come up with a heuristic to pick the 'best' primary signature
(relevant part of the generator: https://github.com/HaxeFoundation/html-externs/blob/master/bin/Haxe.py#L990)
There was a problem hiding this comment.
But that is already typed. How is type-inference even involved here? A small example would probably help.
There was a problem hiding this comment.
Sorry, it's probably a vocabulary issue from me. And also some wrong guesses (animate seems fine for example).
I cannot do this:
el.scrollIntoView({behavior: SMOOTH, block: NEAREST}); without adding the ECheckType. I get an Unknown identifier : SMOOTH error.
Edit: try-haxe example
There was a problem hiding this comment.
Ah I see. That's probably something we should try to fix in the compiler, but I'm not sure if that's really possible with our overload architecture. Please open an issue about this, preferable with an example that doesn't require any JS externs.
There was a problem hiding this comment.
In the most recent versions of the compiler, EitherType seems to be better at inference (resolves this specific case) so I could switch to that in the externs (try-haxe)
However I guess the best solution would be to somehow have the same level of inference for overloads
(I actually initially switched to overloads for these cases because it had better type inference than EitherType)
Motivation
The existing js.html externs are from 2 years ago and missing a number of important and popular APIs. This PR brings them up to date with the latest Mozilla WebIDLs as well making a number of typing improvements and bugfixes
--
The externs were generated using a fork of HaxeFoundation/html-externs and fork of @markknol's document generator.
The changes include many missing fields, bug fixes, new APIs and new classes. Some of the headline changes are:
Issues resolved include:
The PR
deprecatesremoves the following files (which represent types no longer available in browsers or were misplaced originally):js/html/AnimationEffectReadOnly.hxjs/html/AnimationEffectTiming.hxjs/html/AnimationEffectTimingProperties.hxjs/html/AnimationEffectTimingReadOnly.hxjs/html/AppletElement.hxjs/html/ApplicationCache.hxjs/html/AudioChannel.hxjs/html/AudioContextState.hxjs/html/CSSCharsetRule.hxjs/html/CSSPrimitiveValue.hxjs/html/CSSUnknownRule.hxjs/html/CSSValue.hxjs/html/CSSValueList.hxjs/html/ComputedTimingProperties.hxjs/html/ContentElement.hxjs/html/DOMCursor.hxjs/html/DOMTransaction.hxjs/html/DOMTransactionEvent.hxjs/html/DOMTransactionEventInit.hxjs/html/DesktopNotification.hxjs/html/DesktopNotificationCenter.hxjs/html/ElementRegistrationOptions.hxjs/html/KeyframeEffectReadOnly.hxjs/html/NotifyPaintEvent.hxjs/html/RGBColor.hxjs/html/RecordErrorEvent.hxjs/html/RecordErrorEventInit.hxjs/html/Rect.hxjs/html/ShadowElement.hxjs/html/audio/OfflineAudioCompletionEvent.hxjs/html/svg/AltGlyphElement.hxjs/html/svg/ZoomEvent.hxjs/html/AudioContextState.hx(moved tojs.html.audio)The following files have been removed without deprecation note because they represent internal types that should not have been generated (so no one will be using them in the wild)
js/html/ChromeFilePropertyBag.hxjs/html/CommandEvent.hxjs/html/DummyInterface.hxjs/html/DummyInterfaceWorkers.hxjs/html/LifecycleCallbacks.hxjs/html/MessagePortList.hxjs/html/ObjectURLOptions.hxjs/html/SimpleGestureEvent.hxjs/html/UncaughtRejectionObserver.hxjs/html/svg/Document.hxjs/html/XMLStylesheetProcessingInstruction.hxWebGL extensions have been renamed and moved to
js.html.webgl.extensions, this is to avoid naming conflicts with the previous scheme (WEBGL_color_buffer_floatandEXT_color_buffer_floatwere previously incorrectly mapped to the same file).