META: Highlight & Selection write API - prototype

[lmeyerov]

Enable embedding users to trigger fast highlights & selections of items

Ex:

• User hovers over a name or group ID in outside tool
• Outside tool identifies corresponding node & edge IDs of interest
• Graphistry UI highlights those specific nodes/edges
• User clicks the name or group ID in outside tool
• Graphistry UI selection focuses on those (sticky)
• User clicks filter/exclude/... or other action for interacting with that selected subset
• If user didn't click, but instead mouse-out'd, up to UI to return Highlight state to empty

Out-of-scope: Prototype will -

• Work on internal entity Index, not ID, for PR. Index is the most recent index returned by the selection API.
• Only highlight/select the exact

Piggybacks on selection API work

• design
• client API - Update Graphistry client API to accept selection/highlight writes graphistry-js#111
• internal API - Update Graphistry internal API to accept selection/highlight writes graphistry-js#110
• perf check - https://github.com/graphistry/graphistry/issues/1860 (shared)
• docs & demo - Demo of new selection/highlight write JS API graphistry-js#112
• publish https://github.com/graphistry/graphistry/issues/1866
• (TBD Bugs)

[exrhizo]

Bugs are:

• When you inject a selection it causes a bunch of events, as many as 10.

• When you inject a selection, the onSelectionUpdate can have the wrong selection.labels

• hovering over the nodes to get thier label causes the above to fix it also causes the onSelection to flicker

These didn't reproduce for me with ./nc

[lmeyerov]

What does "hovering over the nodes to get thier label causes the above to fix it" mean? That the wrong label is shown, and hovering triggers an update that will show the correct one? I wonder if that means we're properly invalidating the label cache, but not subscribing to the invalidations properly in the labels container / triggering the wrong rendering effects for the rendering scheduler?

[lmeyerov]

I was actually more worried about filter states messing up local vs global indexing, so the above feels more doable :)

[lmeyerov]

@exrhizo I lost track, can you share status here?

[lmeyerov]

(I think this one is paused pending Gadi's feedback)