Immutable mode
See original GitHub issueAlong with Plotly.react
, it would be nice if we would never mutate gd.data
and gd.layout
, just make new changed objects and alert the user to the change. We can’t make this the default behavior in v1 since some users depend on gd.data
and gd.layout
keeping the identity of the objects they initially passed in - though there are some cases already where they are not. But that rule holds well enough that some folks depend on it, so immutability would have to be a config
argument.
In general all of the plot modification methods except Plotly.react
(Plotly.restyle
, relayout
, update
, extendTraces
etc…) modify gd.data
and gd.layout
in place to make the requested updates. Most (but not all) then emit an event you can use to see that the plot state has changed. But also there are modifications that happen automatically during drawing the plot, principally (if not entirely) pushing auto values (axis ranges, colorscale ranges) back to gd.data
and gd.layout
so that future minor updates don’t need to recalculate these. This too is something we might be able to avoid doing, but probably not until v2.
So there are two things we should do:
- Make all updates to
gd.data
andgd.layout
immutably, conditional on a config argument. - Streamline the way we alert the user that the state has changed, so that they can feed this new state back into their app state. It seems like once we have immutable changes, it might be possible without any changes to the events from plotly.js… the react wrapper would have to bind to all the events that might carry state changes initiated from the UI, and also check after every update initiated by the wrapper itself, whether
gd.data === prevData
andgd.layout === prevLayout
… but seems like it would be much easier for the react wrapper if there were a single “new state” event that collects all of these.
Issue Analytics
- State:
- Created 6 years ago
- Reactions:12
- Comments:7 (5 by maintainers)
Top GitHub Comments
Is there a PR for this yet or is this being developed internally?
I’m a big fan of this idea and approach, as it seems to really close the gap between React’s view of the world and the way plotly.js works 😃