Possibly returning null from getBox

[migueloller]

Hi Alex!

Thanks for making this library! Super helpful for all things DOM layout.

There's a pesky issue when measuring an element that has display: none. getComputedStyle will return the used values if the element has display: none and the computed values otherwise. What this means is that calling getBox on an element with display: none will result in an error as the used values could be a length (e.g., em, rem, etc.) or a keyword like auto.

Would you be open to making the signature of getBox (el: Element) => ?BoxModel? I thought of just wrapping the method in a try/catch in our codebase but we measure layout a large amount of times and a try/catch block would slow it down a lot.

There's more info on the quirks of getComputedStyle's resolved values here.

If this is something you're open to, I'd happily open a PR.

[alexreardon]

Interesting! Perhaps if it is display:none we return 0 for all values? rather than null?

🤔

[migueloller]

That's a possibility. I do wonder if that's the only case and is the "correct" source of truth for something like this. I tried finding an exhaustive list of what would cause this but had no luck finding one. Perhaps more research is required, or a dive into the spec.

If display: none is the only case for this, returning 0 for all values would definitely work, though! As people report other instances of similar issues, support for those could be added.

[alexreardon]

I think returning 0 when we detect that there is no px prefix is a reasonable way forward