Implement ReferenceCell::point_is_inside.

[bangerth]

This is the generalized equivalent of GeometryInfo::is_inside_unit_cell().

/rebuild

[peterrum]

FYI @nfehn

[nfehn]

Maybe formulate as question is_point_inside().

[bangerth]

I don't like is_point_inside as a name. It sounds grammatically wrong.

Would contains_point be better? As in

  cell->contains_point(p)

[nfehn]

Maybe just is_inside(), in analogy to std::is_same, std::is_enum, etc.

(point.is_inside(cell) would be easier to read, but does not make sense from a software design perspective. I assume this is the reason why you repeated „point“ in the function name?)

[nfehn]

Maybe such geometry functions with tolerances should not use default parameters.

I think the documentation does not exactly specify what eps = 0 means. I would understand the docu as follows: A point lying on the boundary together with eps=0 will result in a return value of true. However, the formulation in the first sentence "if the given point is inside the reference cell ..." goes in a different direction.

I would prefer tolerance (often used in GridTools) over eps. Is it really numerical roundoff in the sense of eps?

[bangerth]

I made those changes.

[bangerth]

What do others think about the naming of the function?

[peterrum]

What do others think about the naming of the function?

I am fine with contains_point(). One would read cell.contains_point(a) as cell containts point a. If one wants to be shorter, one could also write contains(). But nut sure if that becomes ambiguous.

[nfehn]

I understand "contain" in the sense of information/data, not geometrically.

[nfehn]

Assume you want to know whether a cell (of one triangulation) is enclosed by another triangulation tria. The example tria->contains_cell(cell) in my opinion clearly demonstrates that the "contains" nomenclature is misleading. Put differently, we should find a wording that is intuitive not only for the cell/point example, but also for tria/cell and other examples.

[Rombur]

I don't have a strong opinion on the name. If we want something more general, we could use the same naming convention as boost. Use within if a box is totally inside another box and intersect if a boxes is totally or partially inside an other one. So for this function, something like point_within_cell.

[nfehn]

cell.point_within_cell(point) might be quite hard to read (then, we would need a free function like point_within_cell(point,cell) in my opinion).

What about cell.encloses(point) as one more suggestion ;)

[bangerth]

On 3/8/22 05:01, Niklas Fehn wrote: I understand "contain" in the sense of information/data, not geometrically.

That's not how native speakers would interpret it, though. "contains" is used in sets, for example. I think "contains" is pretty clearly geometric. (There's also the (shipping) container, which quite literally "contains" a bunch of objects in the same way as a dam contains the water behind it.)

[masterleinad]

I like contains_point.

[bangerth]

contains_point() it is. I like that name better than point_is_inside() anyway :-)

So updated!