895728 - make Web IDL bindings support distinguishing primitive types and DOMString

Status: RESOLVED FIXED

(Core :: DOM: Core & HTML, defect)

Description

[Cameron McCormack (:heycam)]

Web IDL now supports distinguishing primitive types and DOMString for overload resolution and enum types.

Comment 1

[Boris Zbarsky [:bzbarsky]]

Attachment: Make primitive types be distinguishable from strings and enums.

Comment 2

[Boris Zbarsky [:bzbarsky]]

Attachment: With a bit more cleanup

Comment 3

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 778265 [details] [diff] [review]
With a bit more cleanup

Spec is changing more.  :(

Comment 4

[Cameron McCormack (:heycam)]

Spec is now updated to have boolean, numeric types and DOMString all be distinguishable.

Comment 5

[Boris Zbarsky [:bzbarsky]]

Attachment: part 1. Introduce "boolean" and "numeric" types, since those will be treated differently for purposes of distinguishability and overload resolution.

Comment 6

[Boris Zbarsky [:bzbarsky]]

Attachment: part 3. Fix overload resolution to work with the new boolean/numeric/string setup.

Comment 7

[Boris Zbarsky [:bzbarsky]]

Attachment: part 4. Fix unions to work with the new boolean/numeric/string setup.

Comment 8

[Boris Zbarsky [:bzbarsky]]

Attachment: part 2. Make booleans, numerics, and strings/enums all be distinguishable from each other.

Comment 9

[:Ms2ger (he/him; ⌚ UTC+1/+2)]

Comment on attachment 778798 [details] [diff] [review]
part 3.  Fix overload resolution to work with the new boolean/numeric/string setup.

Review of attachment 778798 [details] [diff] [review]:
-----------------------------------------------------------------

::: dom/bindings/Codegen.py
@@ +5070,5 @@
> +            if numericSignature:
> +                addCase(numericSignature, numericCondition)
> +            if stringSignature:
> +                addCase(stringSignature, None)
> +            

trailing ws

Comment 10

[:Ms2ger (he/him; ⌚ UTC+1/+2)]

Comment on attachment 778799 [details] [diff] [review]
part 4.  Fix unions to work with the new boolean/numeric/string setup.

Review of attachment 778799 [details] [diff] [review]:
-----------------------------------------------------------------

::: dom/bindings/Codegen.py
@@ +2930,5 @@
>          else:
>              templateBody = CGGeneric()
>  
> +        stringTypes = filter(lambda t: t.isString() or t.isEnum(),
> +                             memberTypes)

[t if t.isString() or t.isEnum() for t in memberTypes] would be more idiomatic, I think.

Comment 11

[Boris Zbarsky [:bzbarsky]]

Attachment: Part 4 with the filter() bits nixed.

> trailing ws

Fixed.

> [t if t.isString() or t.isEnum() for t in memberTypes] would be more idiomatic, I think.

Good catch.  I was just copying the existing code, but that was already suboptimal.  ;)

Comment 12

[Kyle Huey (Exited; not receiving bugmail, old account, do not use)]

Comment on attachment 778800 [details] [diff] [review]
part 2.  Make booleans, numerics, and strings/enums all be distinguishable from each other.

Review of attachment 778800 [details] [diff] [review]:
-----------------------------------------------------------------

::: dom/bindings/parser/WebIDL.py
@@ +2165,5 @@
> +            return (other.isNumeric() or other.isString() or other.isEnum() or
> +                    other.isInterface() or other.isObject() or
> +                    other.isCallback() or other.isDictionary() or
> +                    other.isSequence() or other.isArray() or
> +                    other.isDate())

Wouldn't this be easier as 'return not other.isBoolean()'?

@@ +2171,5 @@
> +            return (other.isBoolean() or other.isString() or other.isEnum() or
> +                    other.isInterface() or other.isObject() or
> +                    other.isCallback() or other.isDictionary() or
> +                    other.isSequence() or other.isArray() or
> +                    other.isDate())

Similarly here?

Comment 13

[Kyle Huey (Exited; not receiving bugmail, old account, do not use)]

Comment on attachment 778798 [details] [diff] [review]
part 3.  Fix overload resolution to work with the new boolean/numeric/string setup.

Review of attachment 778798 [details] [diff] [review]:
-----------------------------------------------------------------

Note Ms2ger's comment.

Comment 14

[Boris Zbarsky [:bzbarsky]]

> Wouldn't this be easier as 'return not other.isBoolean()'?

It'd have to be 'return not other.isBoolean() and not other.isAny()', I think.  Which is why I didn't do it: it's a less-direct translation of the distinguishability table to do it that way...

I can change it if you prefer; let me know.

Comment 15

[Boris Zbarsky [:bzbarsky]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/11874cb86560
https://hg.mozilla.org/integration/mozilla-inbound/rev/7cc767d92af6
https://hg.mozilla.org/integration/mozilla-inbound/rev/c4e3b826d3f3
https://hg.mozilla.org/integration/mozilla-inbound/rev/d879c9cff194

Comment 16

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/11874cb86560
https://hg.mozilla.org/mozilla-central/rev/7cc767d92af6
https://hg.mozilla.org/mozilla-central/rev/c4e3b826d3f3
https://hg.mozilla.org/mozilla-central/rev/d879c9cff194