Jumpy ghosts and pants

[tmickel]

@rachel-fenichel

Now that things are really coming together, some subtle bugs are becoming more obvious.

Here is one where everything is ending up in the right place, but there's a quick "jump" of a couple of frames while things are being dragged. Any ideas of how to address this? For me in this case it only seems to happen when dragging from the left, which is interesting...

https://www.dropbox.com/s/y5qptwzv4co1n6x/jumpy-ghosts.mov?dl=0

[tmickel]

This seems similar and possibly related, possibly not:
https://www.dropbox.com/s/j4ty1fcf1vmqzpw/pants-jiggle.mov?dl=0

[rachel-fenichel]

I'm pretty sure I know what's going on here, I just haven't addressed it yet because I think the fix is going to be somewhat fragile.

Essentially, the connection that it's ghosting to is moving by the width of the ghost block, which is enough to cause it to go out of the snap range. For the pants, it should be the width of the first part of the pants block.

Changing the snap radius may help obscure the problem but wouldn't fix it. I think the fix would involve doing the search for closest connection across a wider radius when the ghost exists, but I don't think that fix would be sufficient.

We previously discussed just lying about where the connections are, to make them stay in the same place regardless of the offset caused by the drag. I don't like that approach because it's going to be very error-prone (it's almost impossible to debug connections that aren't located where their visual markers are) and because I think it would lead to some strange behaviour that would be noticeable if moving a block slowly across a row of blocks.

[rachel-fenichel]

One clarification on the steps here:

1. the connection goes out of snap range
2. the ghost block disappears
3. the stack heals, rerendering without the ghost block and updating connection locations
4. the next connection is now within range and the ghost block reappears

After discussion with Erik, not updating the connection locations in response to the ghost does actually make sense. Unfortunately it also might involve rewriting large chunks of the rendering code.

[tmickel]

That's a helpful breakdown. Not updating connection locations in response to the ghost makes sense to me as well. Yeah I can't immediately think about how to approach this with the rendering code.....

[rachel-fenichel]

@tmickel is this fixed by changing the snap radius?

[tmickel]

It does seem largely fixed - although we might want to fix it for smaller snap radii. What do you think?

[rachel-fenichel]

What's the downside to the larger snap radius at this poitn?

On Thu, Apr 14, 2016 at 11:57 AM Tim Mickel notifications@github.com
wrote:

It does seem largely fixed - although we might want to fix it for smaller
snap radii. What do you think?

—
You are receiving this because you were assigned.

Reply to this email directly or view it on GitHub
#139 (comment)

[tmickel]

No real downside. I'm fine with leaving it, although I thought you wanted to solve the problem in general.

The only problem left I see right now is this:
https://www.dropbox.com/s/apkziy7tjdnv6qn/pants-jump.mov?dl=0
I think it's the same issue?

[tmickel]

The bulk of this is fixed, so I'm going to close it now. May still be one more situation to look at, will report separately.