yangzhang75 opened a new issue, #8582:
URL: https://github.com/apache/texera/issues/8582
### Task Summary
`WorkflowEditorComponent` and `MiniMapComponent` each construct a
`joint.dia.Paper` bound to the joint graph owned by `WorkflowActionService`,
which is `providedIn: "root"` and lives for the whole tab. Neither
`ngOnDestroy` disposes its paper, and a repo-wide search finds no
`paper.remove()` anywhere in `frontend/src/app`.
* `WorkflowEditorComponent.ngAfterViewInit` -> `attachMainJointPaper(...)`
-> `new joint.dia.Paper({ model: this.jointGraph, ... })`; its `ngOnDestroy`
disconnects a resize observer, removes a keydown listener and resets the
heat-map view, and leaves the paper.
* `MiniMapComponent.ngAfterViewInit` constructs its own paper on the same
graph; its `ngOnDestroy` only writes a `localStorage` flag.
So every remount of either component adds a paper that keeps listening to
the one graph and rendering into a DOM node that is no longer on the page. The
Hub's read-only workflow preview already remounts the editor, and the Form View
mounts a second editor for its workflow preview strip.
The visible risk is pointer handling: two live papers on one model both
receive the model's events, and paper-wide interactivity (`setInteractivity`,
which is what makes operators draggable) is per paper. A stale paper carrying
the Form View's `structureLocked` settings is a plausible cause of the
"undraggable operators" symptom recorded in the comments that made the canvas
<-> Form View switch a full page load (see #8580).
Proposed: call `paper.remove()` in both `ngOnDestroy`s, with a test in each
spec that the paper is disposed when the fixture is destroyed.
Found while working on #8580; kept out of that PR because it is a
pre-existing leak on any remount, not specific to the switch, and wants its own
tests.
### Task Type
- [x] Refactor / Cleanup
- [x] Performance
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
