[css-typed-om] .equals methods for CSSStyleValue subclasses

[wilddamon]

Could be useful for some calculations

e.g. someLengthValue.equals(otherLengthValue)

CSSCalcLength({px: 5}).equals(CSSSimpleLength(5, 'px')) // returns true

@shans

[tabatkins]

Yes, they definitely all need one, or at least all the "numeric" ones.

[tabatkins]

So there are several cases here:

1. Both UnitValues, equal unit: just compare values
2. Both UnitValues, different but compatible units: convert values, compare?
3. Both CalcValues, all terms pair up with equal-unit terms: just compare values
4. Both CalcValues, all terms can group by canonical unit, then can all pair up: convert values, compare?
5. UnitValue & CalcValue, Calc has same unit: just compare values
6. UnitValue & CalcValue, Calc has compatible units: convert values, compare?

This all sound good? Any cases I'm missing?

[tabatkins]

Should we drop 0-valued units from the comparison, so "calc(1px + 0em)" is equal to "1px"?

What if the 0-valued units aren't compatible - is "calc(0px)" equal to "calc(0rad)"? (I assume no; we'll probably check that they have the same CSS type before checking the individual units.)

[shans]

I think .equal should be pretty strict equality. We can add a less strict version that does things like unit conversion and 0-valued unit elision if necessary.

So in other words:

1. yup
2. nope, they're different.
3. yep.
4. nope, they're different.
5. nope, they're different.
6. nope, they're different.

[tabatkins]

That certainly makes things easier to define.

I'm only really concerned about #5 - I don't see a good reason to distinguish between 5px and calc(5px).

[shans]

We decided to stick to strict equality for now (i.e. #5 distinguishes because they're different classes)

[tabatkins]

Unifying idea behind strict equality is "what would a Value Object do for ==, if/when that exists?" - it'll type-check then check for equality of the bytes in the underlying typed array. That gives us our current answer.

Loose equality can be achieved today by casting both to a known type and then calling .equals(), so it's almost just sugar. (Not quite - comparing non-trivial calcs can't be done that way.)

[nainar]

@tabatkins merged in a change from me (#412) and said that they would finish this up themselves. Just posting a status update since we are working on some of the Typed OM spec issues.

[nainar]

Need to add the text to make sure that we recurse over all our children here.

[tabatkins]

That is to say, currently the last steps assume the children of the functions are all unit-values; it just needs to do a proper recursion instead.

[nainar]

Closing this since the last bit of the text was added in PR #452