feat: DH-22030: dx event handlers plan by jnumainville · Pull Request #1373 · deephaven/deephaven-plugins

jnumainville · 2026-06-22T20:56:56Z

Adds plan for dx event handlers

mofojed · 2026-06-22T21:06:28Z

+
+`traceName` is the Plotly trace name, which for Deephaven charts is the `by` column value (or partition key) that identifies the trace. If the chart has no `by`/partitioning, `traceName` is the default Plotly trace name.
+
+### Layout Events (`on_relayout`)


Will these events fire for every data point that ticks in? Something to keep in mind.

This one should not fire for every data point. The events that would have been filtered out already.
We only update a couple things, such as title and legend title, in the layout after init. So besides that case, this mainly updates for user initiated adjustment such as zooming. This is why we freeze the layout too, so we can have the events such as zoom.

Might need to debounce when the user is zooming, will need to play with it.

Added debouncing note. The main spot is spamming the zoom button.

mofojed · 2026-06-22T21:09:32Z

+
+
+fig = dx.sunburst(
+    pop_by_continent, names="Continent", values="Pop", on_sunburst_click=handle_sunburst


It's kind of annoying that we need to do on_sunburst_click for the plot vs. just doing a on_click - we already know it's a sunburst chart, why not just do an on_click handler when calling from dx.sunburst?
I guess in dx.layer we could have multiple types of charts, so I imagine we need to have more information on the click event there; so do we have one on_click and add different metadata for the type of click, or do we force the on_click, on_sunburst_click when we're in a layer (we'll still need extra metadata to link it to which plot was clicked either way yeah?)

I have some generated plotly.js code I was playing with and after some further research I was missing something here. The plotly_click event fires already, even for the hierarchical plots, it just wasn't wired.
The hierarchical plots have different events because they actually change the visualized data, which is what I thought the key difference was, but the actual reason they exist is if you return False from the callback you can prevent the drill down.
I'm thinking that instead of that, we expose a drill_down param that can be used to disable drill down. Downside is it's not as flexible (no conditional disabling several layers down for example) but this is a helpful and simple switch for the basic use case where you just want the data out of the level you click on.
Also, we are filtering out most of the trace data. The idea is you can use the curve number to look up the specific trace if you need to. This has made me think it's worth piping through the trace type too.

mofojed · 2026-06-22T21:10:14Z

+
+
+def handle_legend_click(event):
+    print(f"Legend clicked: {event['traceName']} (curve {event['curveNumber']})")


Hmm do we want these events converted to snake_case instead of camelCase...

Yeah I'll do that

dsmmcken · 2026-06-22T21:15:37Z

+
+### `on_relayout` — Layout Change (Pan/Zoom)
+
+**Justification:** Synchronize the visible time range across multiple charts (e.g. pan one chart, update all others to the same window) or filter tables to the visible range.


I accept this justification, as it was an event I questioned. Concerned it might fire more than expected.

See comment to Bender. This specific one should be safe.

Copilot

Pull request overview

Adds a detailed implementation plan for DH-22030 to support server-side Python callbacks for Plotly Express (dx) chart interaction events, including payload schemas, API signatures, and a phased rollout across Python + JS.

Changes:

Specifies supported Plotly event → Python callback parameters (plus hierarchical drill_down behavior).
Documents event payload schemas and provides usage examples for each event type.
Outlines a phased implementation plan across DeephavenFigure, message handling, and frontend wiring (plus proposed test coverage).

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

mofojed

Not sold on the drill_down behaviour yet, and not sure the legend click disabling the default behaviour was mentioned at all.

mofojed · 2026-06-25T15:34:07Z

+### `drill_down`
+
+Hierarchical chart functions (`sunburst`, `treemap`, `icicle`) now accept an additional non-callback parameter, based on the functionality enabled by returning false from `plotly_sunburstclick` / `plotly_treemapclick` / `plotly_icicleclick` handlers.
+
+```python
+drill_down: bool = True
+```
+
+When `False`, clicking a segment does **not** trigger the default drill-down animation. The `on_click` callback still fires with the clicked point data. This is implemented on the JS side by returning `false` from the underlying `plotly_sunburstclick` / `plotly_treemapclick` / `plotly_icicleclick` handler and is a replacement for that functionality being directly exposed.


Hmmm so we'll have two props for a callback, e.g. dx.sunburst(..., on_click=handle_click, drill_down=False)?
Note there seems to be similar functionality when a legend is clicked: https://plotly.com/javascript/plotlyjs-events/#legend-click-events
Trying to think of how to hack this in a little more elegantly... since I'm not big on having extra props pollute the top level.
A couple options I think worth exploring:

If there's a callback, always return false for these, and then get the result from the callback - if the callback result is False, do nothing, otherwise trigger the default action (depends on if we can "trigger" those default actions)

Wrap the callback with something that has options, e.g. dx.click_handler(handle_click, default_behavior=False). I don't particularly like that but it would at least group the props with the callback handler rather than polluting the top scope.

It's more that on_click handles the data return and we replace on_sunburst_click (or at least the return value part of it) with the drill_down prop. They can be used independently and I don't like the idea of on_click having to possibly return a value because that would depend on chart type or alternatively using on_sunburst_click, which we agree is gross...
We could still do that with legend events too. Something legend_click_default: bool would determine the return value of the callback on the js side. If you just had legend_click_default = False but no callback the js side callback would just disable the clicking behavior.
Of course it's highly likely that most of the uses of drill_down=False would be coupled with the callback, but there could be the case of wanting to disable interactions. It looks like it's been asked for a few times on plotly's forum at least (example)
But, returning back to the point of not wanting extra props if we can avoid it, I'll have to ponder further. This also isn't flexible in terms of conditional anyways, a related concern.

Switched to option 1

mofojed · 2026-06-25T15:35:32Z

+
+`traceName` is the Plotly trace name, which for Deephaven charts is the `by` column value (or partition key) that identifies the trace. If the chart has no `by`/partitioning, `traceName` is the default Plotly trace name.
+
+### Layout Events (`on_relayout`)


Might need to debounce when the user is zooming, will need to play with it.

init plan

933ff63

jnumainville requested review from dsmmcken and mofojed June 22, 2026 20:56

mofojed reviewed Jun 22, 2026

View reviewed changes

dsmmcken reviewed Jun 22, 2026

View reviewed changes

init plan

020dd96

jnumainville requested review from dsmmcken and mofojed June 23, 2026 16:52

mofojed requested a review from Copilot June 23, 2026 20:45

Copilot started reviewing on behalf of mofojed June 23, 2026 20:45 View session

Copilot AI reviewed Jun 23, 2026

View reviewed changes

Comment thread plans/DH-22030-plotly-events.md

Comment thread plans/DH-22030-plotly-events.md

Comment thread plans/DH-22030-plotly-events.md

mofojed requested changes Jun 25, 2026

View reviewed changes

reworked drill down

6195092

jnumainville requested a review from mofojed June 30, 2026 19:28

mofojed approved these changes Jun 30, 2026

View reviewed changes


		`traceName` is the Plotly trace name, which for Deephaven charts is the `by` column value (or partition key) that identifies the trace. If the chart has no `by`/partitioning, `traceName` is the default Plotly trace name.

		### Layout Events (`on_relayout`)



		fig = dx.sunburst(
		pop_by_continent, names="Continent", values="Pop", on_sunburst_click=handle_sunburst



		def handle_legend_click(event):
		print(f"Legend clicked: {event['traceName']} (curve {event['curveNumber']})")


		### `on_relayout` — Layout Change (Pan/Zoom)

		Justification: Synchronize the visible time range across multiple charts (e.g. pan one chart, update all others to the same window) or filter tables to the visible range.

Uh oh!

Conversation

jnumainville commented Jun 22, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jnumainville Jun 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jnumainville Jun 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Uh oh!

mofojed left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

jnumainville Jun 23, 2026 •

edited

Loading

jnumainville Jun 23, 2026 •

edited

Loading