What should the value of water tightness tolerance be set to?

[makeclean]

A question remains as to what tolerance do we test the model for, it is my assumption thus far that we at should test at the very minimum the faceting tolerance, the image below shows 2 models faceted to the same accuracy, one model (red) was modified such that one vertex was moved vertically by 1.0e-2 cm (increased for plotting to exaggerate the differences)

Checking for faceting at a tolerance equal to or more than the faceted tolerance would not pick up this gap, however testing at 10*f_tol would have found the gap.

On this basis,

• I suggest we define the checking for water_tightness tolerance at a level greater than the faceting tolerance.

[makeclean]

Added tools into testing to allow the arbitrary modification of the vertices in the model 81654d9, the makes the generation of non water_tight models trivial, more accurately allows the arbitrary "loosening" of the model.

[pshriwise]

Unclear on where we're at with this. Does this still need resolution?

[makeclean]

I think you resolved this in #5, you either test to the faceting tolerance or you test using skin. I think for robust models where complete watertightness is important the skin check is the way to go, if only it were faster.

[pshriwise]

Agreed. There were updates made to the moab skinner in the bitbucket repo. I'm going to have a look tomorrow and see if the changes will speed it up. If they do I'll create a branch to update to the repository version.

[gonuke]

Don't we test to ensure that the vertices of one surface are identical to the vertices of it's neighbor, rather than just their proximity? Am I misunderstanding the algorithm?

[pshriwise]

If check_watertight isn't passed some checking tolerance, then yes, it does exactly that. If it is passed some water_tightness tolerance by the user, the check will be done by proximity.

I think that this is ok for now so long as it is well documented, as Andy was checking on.