Bug fixes and getting master up to date
Jørgen Dokken
Branch: dokken/ffc-interval-tables-and-quadrature-fixes:dokken/interval-tables-and-quadrature-fixes
Branch: fenics-project/ffc:master
Merged
Merged pull request
Merged next into master
Merged next into master
Update links for ufl and fiat to github. Replace constant in demos with coefficient of real space.
Only thing to get ci green is to update reference data, which i dont have access to do.
The change in elementtables.py updates the number of dofs in the same way as the unique_table_types are updated.
The second change in tools.py corrects the assumption that the quadrature points are uniquely defined through the number of quadrature points . That is not true, as then number of quadrature points for intervals are defined in FIAT as:
num_points_per_axis = (degree + 1 + 1) // 2
`. Similarly, third and fourth order triangle schemes have the same order (https://bitbucket.org/fenics-project/fiat/src/dd99c9e6b3d01d21f584215a2bfcde7e1bf3a3ad/FIAT/quadrature_schemes.py#lines-121:137). Therefore, I propose to only keep the highest order quadrature rule.Additionally, this test is never entered, as it would cause errors due to the boolean evaluation of a numpy array, and that the weights are compared to the points in the second assert statement
Corresponding pull request in dolfin with tests highlighting these issues will be added.