[css-grid] Sizing algorithm for span >= 2 should be a generalization of span 1

[Loirooriol]

I don't understand why there is a track sizing algorithm using items with a span of 1 (and intrinsic minimum and non-flexible maximum), and another one for items with a span >= 2 (not spanning tracks with flexible maximums).

I think both cases should use the same algorithm, just with the extra space distribution being more trivial to perform for items with a span of 1. However, they aren't exactly the same:

• With span >= 2, tracks with intrinsic minimums end up at least as big as the minimum contribution. For span 1, this is only ensured for auto minimums.
• With span >= 2, an auto minimum under a sizing min/max-content constraint behaves as min/max-content. With span 1 it's not exactly like this, because the value is later clamped between the minimum contribution and the max track sizing function (or fit-content argument).

Why the differences?

[fantasai]

Wrt the first point, the following inequality holds: minimum contribution ≤ min-content contribution ≤ max-content contribution. This allowed us to have a simpler structure for span=1. (It wasn't possible for span > 1 because of the way spanning items are handled... We could have written a super generalized algorithm, but this would have been harder for readers to follow. It's easier to think about the span=1 case using the simplified algorithm, and then work up to spanning items, where there is some extra complexity.)

Wrt the second point. see #2303 (comment) :)

[Loirooriol]

minimum contribution ≤ min-content contribution

OK I was strongly suspecting this, but it's not obvious, because the definition of the minimum contribution is complex, and requires a deep understanding of css-sizing. Maybe add a note?

We could have written a super generalized algorithm, but this would have been harder for readers to follow.

Then could you also add another note saying that applying the algorithm for spanning items to non-spanning ones would produce the same result, it would just be less simple?

Wrt the second point. see #2303 (comment) :)

That looks good, thanks.

[fantasai]

Then could you also add another note saying that applying the algorithm for spanning items to non-spanning ones would produce the same result, it would just be less simple?

At what point in the section do you think this is helpful to point out?

[Loirooriol]

Well for example in https://drafts.csswg.org/css-grid/#algo-single-span-items after "consider the items in it with a span of 1" it could say that it's just a simplification, but with the same result, of the algorithm for spanning items.

Or maybe in https://drafts.csswg.org/css-grid/#algo-spanning-items after "Repeat incrementally for items with greater spans until all items have been considered" it could say that performing an initial iteration for items with a span of 1 would have the same effect as running the non-spanning algorithm.

[fantasai]

@Loirooriol Done. Close out the issue if you are satisfied? :)

[Loirooriol]

Yes, thanks!