GUDHI / gudhi-devel

The GUDHI library is a generic open source C++ library, with a Python interface, for Topological Data Analysis (TDA) and Higher Dimensional Geometry Understanding.
https://gudhi.inria.fr/
MIT License
259 stars 66 forks source link

AlphaComplex default_filtration_value can use 'fast' #430

Open mglisse opened 3 years ago

mglisse commented 3 years ago

When computing a Delaunay triangulation (aka AlphaComplex with default_filtration_value=True), there is no point using Epeck. If possible, we should silently use Epick. If not, we should warn the user that using precision='fast' would safely give the same result as exact.

VincentRouvreau commented 3 years ago

https://github.com/GUDHI/gudhi-devel/blob/7e05e915adc1be285e04eb00d3ab7ba1b797f38d/src/python/include/Alpha_complex_interface.h#L59-L64

 if (fast_version_ || default_filtration_value) { 
   alpha_ptr_ = std::make_unique<Inexact_Alphacomplex_dD>(points_, exact_version_); 
 } else { 
   alpha_ptr_ = std::make_unique<Exact_Alphacomplex_dD>(points_, exact_version_); 
 } 
 alpha_ptr_->create_simplex_tree(simplex_tree, max_alpha_square, default_filtration_value); 

would solve this issue.