Would it be feasible to rely on the dataIndex to avoid using floating point number as a key?

This part of the code is quite hard to get. Especially because storedOffsets is not very explicit. Whay those offset are stored?

Maybe something like offsetsToTickIndexes/dataIndexToTickIndes with a comment

// Map the axis dataIndexes to the tick indexes related.
const dataIndexToTicks = new Map<number, Set<number>>();

We can't use dataIndex in this case, we need to know exactly where the tick is, as this is used to remove duplicate ticks.

Floats should be fairly consistent overall, they are just not precise. Can you see an issue that might be created by this?

Can you see an issue that might be created by this?

I don't get why we need the position and not the dataIndex.
The tickOffset is computed from the scale, the tickValue=tickValues[dataIndex], and tickPlacement

Either tickPlacement is the same for all groups. In that case, same tickOffset is equivalement to same dataIndex. Or tickPlacement can change from one group to another. But in that case it seems risky to rely on the fact that tickPlacement='start' of an item and tickPlacement='end' of the previous item get exactly the same floating number

Ah true, because the ticks are always in the same place 🤦
I'll change it

Or tickPlacement can change from one group to another. But in that case it seems risky to rely on the fact that tickPlacement='start' of an item and tickPlacement='end' of the previous item get exactly the same floating number

As long as the original numbers are the same the results should be the same.

Should we update the scale type to be D3DiscreteScale (and create that type) instead?