@bou For the moment, you MUST use the trick which @OSchaile shows (i.e. you MUST “randomize” X and Y and strongly hope that the current “Interpolate” method will return a “proper” value).
@couet No it does NOT do it properly. Everything still depends on how the bins are created / defined. You will get “blank pixels” in different places if you try one of:
tg2d->SetNpx(101); tg2d->SetNpy(101);
tg2d->SetNpx(102); tg2d->SetNpy(102);
tg2d->SetNpx(103); tg2d->SetNpy(103);
Yes, the trick will probably help but I have no guarantee that the returned value will be 100% OK. I admit that the probability should be (very) low.
Yes … as @OSchaile pointed it really depends if the point fall on a triangle border or not … and I have no clue how to fix that.
@couet How about you talk to those who maintain this “external code”? There have already been significant bug fixes implemented in this code between ROOT 5 and 6.
In ROOT 5 it was an other code (ROOT specific)
We move to something more robust and a lot faster (with the old implementation the macro we are talking about takes ages to execute).
The algorithm produces triangles … and produce them right … just plot the surface using TRI1 … you will see there is no holes. Then we extended the functionally to fill an underlaying histogram with the value of the triangle above a given bin… But that’s beyond the initial purpose of the triangles package we use.