An older version of numpy is used in the scipy Travis CI python 2.6 testing suite, and this is causing issues in our code because numpy.linalg.det in calc_circumcenters is operating on a 3D array, while it appears that only 2D is supported in this older version of numpy.
We should perhaps consider checking the installed version of numpy in the code and writing a fall-back solution for python 2.6, primarily so the test passes. It may not be pretty, but I think we can say that if users are using python 2.6 and the older version of numpy, then they will perhaps suffer some degree of performance loss due to the fallback code (most people should be using more recent versions of python and numpy these days).
An older version of
numpy
is used in thescipy
Travis CI python 2.6 testing suite, and this is causing issues in our code becausenumpy.linalg.det
incalc_circumcenters
is operating on a 3D array, while it appears that only 2D is supported in this older version of numpy.We should perhaps consider checking the installed version of numpy in the code and writing a fall-back solution for python 2.6, primarily so the test passes. It may not be pretty, but I think we can say that if users are using python 2.6 and the older version of numpy, then they will perhaps suffer some degree of performance loss due to the fallback code (most people should be using more recent versions of python and numpy these days).