stdcm: map shakes when adjusting time

[Castavo]

What happened?

When clicking on the time adjusting buttons of the form (among other inputs), the map "shakes" I guess it's reloading.

Enregistrement.de.l.ecran.2024-12-19.172215.mp4

What did you expect to happen?

I expect the map to either stay stable or to not reload at all when editing the time component

How can we reproduce it (as minimally and precisely as possible)?

1. Start a STDCM request
2. Fill in the origin
3. Fill in a waypoint / the destination (in order to have two pins placed on the map)
4. Now adjust the time component as shown on video
5. Witness

On which environments the bug occurs?

Recette (SNCF)

On which browser the bug occurs?

Firefox

OSRD version (top right corner Account button > Informations)

6580cfe

[theocrsb]

I can't reproduce the bug. Is it still there?

[Castavo]

This might be obscured by #10240

[sim51]

The map is recenter automatically when a user changes :

• traction engine
• time
• date
• tolerance
• A CH code
• a CI
• ...

Question : does the map should be automatically recenter on the "path" when the form's modifications don't impact it ?

[sim51]

I respond to my question : yes we need to recenter it (except for the time, date & tolerance).
When an input is changed, we call the API to find a path between the origin/destination, and the found path can be completely different than the previous one, and so the BBOX can changed.

I can reproduce it with :

• engine : 26000US - BB 26000
• Origin: valence
• Destination: Miramas
• via : orange

Bug explanation :

• when we change a field, a call is made to find a path (see useStaticPathfinding).
• during this process the pathfinding is set to undefined before to be set it with the API return
• For each new instance of the pathfinding (even undefined), the map is redraw and the BBOX computed. And if the path is undefined, the BBOX is computed with the markersInfo

So when we change the time field, we have one map render with a bbox for the marker, and a second later with pathfinding bbox. The blink effect comes from those 2 renders which can be different.

[thibautsailly]

This blinking doesn't scream "quality". Ideally, there should be a map state of "computing new path", and a transition between the two displays rather than a sudden move when the new path becomes available.
As a first step we can prevent it from happening when time related data is edited, and improve upon it later.

[SharglutDev]

I don't think we should recompute the pathfinding when we change an input not related to it (tolerance, date, time)