[resize-observer] The initial size of ResizeObservation

[cathiechen]

https://drafts.csswg.org/resize-observer/#dom-resizeobservation-resizeobservation-target-options

4. Set this internal lastReportedSize slot to (0,0)

Not very strong opinion about this.
We might miss this:

• Element's size isn't (0, 0)
• ro.observe(Element)
• Element's size change to (0, 0)

[cathiechen]

Add a case to reproduce this issue.
https://jsbin.com/qotevalowe/1/edit?html,console,output

[mrego]

Add a case to reproduce this issue.
https://jsbin.com/qotevalowe/1/edit?html,console,output

In that example if you remove the last line (the one that changes the size to 0x0), then you'll get notified for a size change of 100x100 (when nothing has actually changed). I guess it's because the initial lastReportedSize is 0x0 and it detects a change.

[cathiechen]

@mrego Yes, exactly like you said. :) We might also get notified when nothing has changed.

[atotic]

There was a discussion about what initial size should be at:
WICG/resize-observer#8

Options were:

• current size of the element
  • rejected because this might cause RO.observe() to trigger layout.
• (0,0)
  • causes RO to fire immediately after being observed, unless Element size is 0,0
  • 0,0 is also Element's size when hidden. RO will not fire initially until Element is shown.

• undefined
  • would cause RO to always fire, even when hidden.

I think (0,0) option is most useful, because it does not fire until Element is shown.

[Loirooriol]

This is inconsistent with IntersectionObserver, which always delivers a notification just after observing, since

1. When observing a target Element, previousThresholdIndex is set to -1.
2. When running the update intersection observations steps, thresholdIndex is set to a non-negative index:

   Set thresholdIndex to the index of the first entry in observer.thresholds whose value is greater than intersectionRatio, or the length of observer.thresholds if intersectionRatio is greater than or equal to the last entry in observer.thresholds.

3. This ensures a IntersectionObserverEntry will be queued:

   If thresholdIndex does not equal previousThresholdIndex or if isIntersecting does not equal previousIsIntersecting, queue an IntersectionObserverEntry

Similarly, I think that ResizeObservation should initialize lastReportedSizes to something invalid like [(-1,-1)], so that we consistently get a notification regardless of the current size.

contain-intrinsic-size: auto also needs something like this, it uses an internal ResizeObserver to store the last remembered size. And 0x0 is a valid size that should still be stored, so the callback should be invoked. See #7539

I think (0,0) option is most useful, because it does not fire until Element is shown.

I don't think this justifies not reporting an observation for elements that do have a box but it just happens to be 0x0. Also note IntersectionObserver notifies about disconnected elements:

new IntersectionObserver(console.log).observe(document.createElement("div"));

[emilio]

Agenda+ to confirm the resolution in #7539 (comment) applies to ResizeObserver too (or get a new one otherwise), since from discussion with @Loirooriol it wasn't quite clear from the minutes.

[css-meeting-bot]

The CSS Working Group just discussed [resize-observer] The initial size of ResizeObservation, and agreed to the following:

• RESOLVED: When first observing an element with ResizeObserver, lastReportedSize gets initialized with a -1 x -1 size

The full IRC log of that discussion

<emeyer> Topic: [resize-observer] The initial size of ResizeObservation
<emeyer> github: https://github.com//issues/3664
<emeyer> emilio: FTF resolved that zero was not a special size for contain intrinsic size
<emeyer> …The resolution seems to include resizeObserver, but in chats with Oriol it evolved that wasn’t clear
<TabAtkins> +
<TabAtkins> q+
<emeyer> …Worth clarifying
<Rossen_> ack TabAtkins
<emeyer> TabAtkins: The resolution was not intended to apply to resizeObserver but I’m fine with it applying
<emeyer> …Consistency with things like c-i-s has me leaning to fire immediately and have an undefined size in resizeObserver
<emeyer> Rossen_: Any objections to the proposed resolution?
<emilio> When observing an element with ResizeObserver, lastReportedSize gets initialized with a -1x-1 size
<emeyer> RESOLVED: When first observing an element with ResizeObserver, lastReportedSize gets initialized with a -1 x -1 size