Change in panning behavior with bounds set

[lastbestmatt]

I've recently upgraded from 0.5 to the latest master, and noticed a change in panning inertia behavior when maxbounds are set. Previously, if you panned the map beyond the bounds, it would "bounce" back into bounds immediately when the pan was completed (mouse button released, for example). Now, if there was inertia when the pan was completed (map was flinged/flung? beyond the bounds), the animation to decelerate the inertia is allowed to finish, and then the map bounces back into bounds.

Is this the intended behavior? If so, is there an option somewhere to enable/disable it? I have not come across anything obvious in the documentation or a quick glance at the code, but I may not know what to search for.

Thanks.

[mourner]

I'll investigate this, it seems like a regression. I noticed this weird behavior but didn't realize it was better in 0.5.

[lastbestmatt]

Thanks. It looks like setMaxBounds now adds an event listener for moveend which calls _panInsideMaxBounds. I tried changing the listener to dragend, which still works if there's no inertia, but if there is then the inertia animation seems to supersede the drag end behavior.

[mourner]

@RungKutta git bisecting now...

[mourner]

First bad commit: db6d689

[lastbestmatt]

Thanks - it looks like this is tangentially related to a fix I recommended a while back (#1333). It looks like the concept of maxBounds was removed from the _onDragEnd handler in Map.Drag.js.

What's needed here is an if statement that ignores the inertia animation IF maxBounds is set AND the map is outside the bounds when dragging ends. I'm not sure the best way to accomplish this yet, but will keep looking. Any thoughts?

(Note that the way it used to work probably isn't ideal - it relied on the order in which two conflicting requestAnimFrame statements were called.)

[lastbestmatt]

Actually, it's probably a little more complicated than that, because the drag could finish in bounds, but inertia would still carry it beyond the bounds. Probably need to do the bounds check when computing offset in _onDragEnd. Perhaps this is what was meant by the following comment (in the old version of Map.Drag.js)?

// TODO predrag validation instead of animation

[mourner]

Fixed, please try it out!

[lastbestmatt]

Thanks much, nice work. I agree with your assessment that this fixes a lot of issues!

One very very minor note to think about related to dragging in the new version: if you drag the map beyond the bounds and let go, it "bounces" back into bounds as expected. If you fling it beyond the bounds, it "inertias" back into bounds. Same effect, but the inertia speed is much slower than the bounce speed.

I don't think this is even worth fixing, but I figured I'd mention it for the record.

[mourner]

@RungKutta yeah, I also noticed this quirk, but currently it's impossible to fix properly until we rewrite the whole panning animation code so that it doesn't use transitions and can be interrupted or corrected at any time. Then we will be able to flip the animation vector once it gets past max bounds.

[lastbestmatt]

Not to belabor a minor issue, but thinking about it a bit more, there's technically two cases here when flinging:

1. If the map is in bounds when you fling, the inertia could be enough to take it beyond the bounds. With your new code today the inertia is just modified so that the movement stops right when the map gets to the bounds, which works great as far as I'm concerned. You could instead let momentum carry it beyond bounds and then bounce back for visual effect, which I assume is what you mean by flipping the vector.
2. If you've already dragged it past bounds and then stop and let up on the mouse, it bounces back. If you've dragged it past and then let go with a fling instead, it will inertia back. This to me is the more noticeable issue (again, very minor). This problem could be fixed relatively easily by adding an inbounds check in _onDragEnd:

if (!offset.x || !offset.y || mapIsOutOfBounds()) {
    map.fire('moveend');
} etc etc

(the 'moveend' event should cause the normal bounceback to happen)

I don't know if there's already some mapIsOutOfBounds() method available - if not, it's again probably not worth adding just for this.

Thanks again!

[mourner]

@RungKutta I think I'll figure out the minor details after 0.7 release. Thanks for the comments!

[denpanosekai]

I have little to add other than this issue is affecting me as well (0.7.3).

Basically if _onDragEnd returns with 'moveend' I get the problem described by rungkutta. So as a workaround I have to enable inertia, have a very high threshold and right now I'm trying to figure out the part with (!offset.x || !offset.y)

Otherwise, if I start to drag, hold it, and drop the map (with setMaxBounds) out of bounds, it doesn't snap back to center.