search

PMEAL / OpenPNM

A Python package for performing pore network modeling of porous media
http://openpnm.org
MIT License
442 stars 175 forks source link

check_network_health wrongly identifies disconnected clusters #879

Closed TomTranter closed 6 years ago

TomTranter commented 6 years ago

I've been trying to compare some tests between versions on a voronoi network. I use check_network_health to identify disconnected clusters and the like which can result from very small throats being closed off when a fiber radius is applied. It seems that in V1 the check_network_health uses a simpler function to identify clusters and the V2 uses topotools percolation models. I think topotools._bond_percolation is wrongly labeling some pores as uninvaded and assigning them to a different cluster. I checked the tmask parameter being passed to the cluster functions and they are the same length and contain all trues. I will check that throat conns is exactly the same. If it is there must be a bug in the percolation way of doing this function

ma-sadeghi commented 6 years ago

I'm experiencing the same issue.

jgostick commented 6 years ago

yes, I'm working on int! It's in the PR I'm working on, and now works just fine

TomTranter commented 6 years ago

We should probably keep better track of these issues throughout all the different branches rather than this issue list. It might also scare people off V1 as they mostly relate to V2 and having them in the same github repo is a pain. Will you be merging this in soon?

jgostick commented 6 years ago

That is the point of tests...once all the important things are tested, and pass, they the branch can be trusted. At this point the v2 branch is not full tested.

I'll merge my testing branch right away, as soon as it passes.

ma-sadeghi commented 6 years ago

Maybe we should add V2 label to the issues related to V2?