Normative: change default function length value to not include optional arguments

[ljharb]

Fixes #264.

• A few legacy methods that have deviating lengths are documented explicitly.
• Any unneeded explicit lengths are removed.
• Needed legacy explicit lengths are added.
• Spacing inside optional argument brackets is made consistent
• Fixed a bug with a DataView note having the wrong argument name
• replaced two instances of a spreaded argument using an ellipsis with three dots

The DataView prototype get/set methods have inconsistent lengths across all browsers, so rather than select one of them as legacy, I decided to let them fall into the default category, which hopefully implementations will align with.

[ljharb]

(I count 59 explicit length removals, 16 explicit length additions)

[domenic]

It would be good to list functions whose normative lengths have changed, even if we think their previous lengths were a bug.

[ljharb]

The only functions that applies to is the DataView ones:

• DataView.prototype.getFloat32 (sometimes 0 or 2, now 1)
• DataView.prototype.getFloat64 (sometimes 0 or 2, now 1)
• DataView.prototype.getInt16 (sometimes 0 or 2, now 1)
• DataView.prototype.getInt32 (sometimes 0 or 2, now 1)
• DataView.prototype.getUint16 (sometimes 0 or 2, now 1)
• DataView.prototype.getUint32 (sometimes 0 or 2, now 1)
• DataView.prototype.setFloat32 (sometimes 0 or 3, now 2)
• DataView.prototype.setFloat64 (sometimes 0 or 3, now 2)
• DataView.prototype.setInt16 (sometimes 0 or 3, now 2)
• DataView.prototype.setInt32 (sometimes 0 or 3, now 2)
• DataView.prototype.setUint16 (sometimes 0 or 3, now 2)
• DataView.prototype.setUint32 (sometimes 0 or 3, now 2)

Note that the "int 8" methods had no optional arguments, and thus were always 2 - this brings all the DataView get and set methods into alignment with 1 for get, and 2 for set - since they all have the same number of required arguments respectively.

[domenic]

I thought e.g. Map's length changed, from previously 1 (since optional arguments were counted) to 0 (since now they are not).

[ljharb]

Map has always had an explicit length of 0, per #264 (comment). There are very few instances of functions deviating from the proposed Chapter 17 language, and most are Date prototype methods.

[bterlson]

This note's language is needlessly pedantic but you can leave it :-P

[ljharb]

I don't mind simplifying it, I just wanted to be explicit :-) Suggestions welcome.

[anba]

Issues

• length for Object.assign removed
• length=1 for Number.prototype.toString missing
• length for Math.min and Max.max removed, but Math.hypot unchanged
• length for Date is defined as part of 20.3.3
• length=7 for Date.UTC missing
• %TypedArray%.prototype.set changed from 2 -> 1, why?
• %TypedArray%.prototype.set should be below 22.2.3.22, not 22.2.3.22.1
• length for DataView is defined as part of 24.2.3
• JSON.parse and JSON.stringify length missing

[bterlson]

...

[ljharb]

oops, will fix momentarily in a new PR

[anba]

I wanted to write proper review comments tomorrow, because I didn't expect this gets merged on a weekend. :-/ (The comments above are basically just my personal notes, that's why they're extra so brief. )

[ljharb]

@anba re "length for Date is defined as part of 20.3.3" - should I remove it there and put it below the constructor like the rest?

[bterlson]

Sorry @anba :( Hopefully everything can be fixed in a subsequent PR?

[anba]

@ljharb I've just checked the other built-in constructors, except for the Array constructor, all constructor functions define their length property as part of the "Properties of the Constructor" section ("Besides the length property (whose value is ), the constructor has the following properties:"). I think I'd prefer if all built-in constructors have an explicit length definition (like the Array constructor), and the subordinate clauses in "Properties of the Constructor" are removed.

@bterlson No worries. 😄

[ljharb]

@anba agreed and done in the imminent PR!