ffc team mailing list archive

[Martin Sandve Alnæs <martinal@xxxxxxxxx>]

I've implemented a regression test output format that includes the
hierarchial structure of the forms, written as .json files (with a limited
feature set, not a lot of code), and gives us a more structured and easier
to read output. The lack of a counter also means the diffs will be robust
w.r.t. additions of new fields. It's already in ffc trunk (you can see a
minor formatting failing on the precise buildbot), with some improvements
in the ffc-ufc-geometry branch where I had some opportunity to test it in
practice. Here's the old vs new output from the same data (both formats are
printed from the same values during an execution of the regression test
code):


martinal@martinal-vbox:~/dev/fenics/ffc/ffc-ufc-geometry/test/regression$
diff ./output/r_auto/ProjectionManifold.out
./references/output/ProjectionManifold.out
351c351
< 308_tabulate_tensor = 0.1885333553623032 0.05519970816170585
0.02613510507389164 -0.4585219919280353 0.05519970816170588
0.3450726103877174 0.1304041493610624 0.4585219919280353
0.02613510507389163 0.1304041493610624 0.2637377966641111
-0.4585219919280353 -0.4585219919280353 0.4585219919280353
-0.4585219919280353 0
---
> 308_tabulate_tensor = 0.1885333553623032 0.05519970816170585
0.02613510507389164 -0.9999999999999998 0.05519970816170588
0.3450726103877174 0.1304041493610624 0.9999999999999998
0.02613510507389163 0.1304041493610624 0.2637377966641111
-0.9999999999999998 -0.9999999999999998 0.9999999999999998
-0.9999999999999998 0


martinal@martinal-vbox:~/dev/fenics/ffc/ffc-ufc-geometry/test/regression$
./recdiff.py ./output/r_auto/ProjectionManifold.json
./references/output/ProjectionManifold.json
Running diff on files ./output/r_auto/ProjectionManifold.json and
./references/output/ProjectionManifold.json
form_0:
  form_0.cell_integral:
    form_0.cell_integral.tabulate_tensor:
      3: -0.458521991928 != -1.0
      7: 0.458521991928 != 1.0
      11: -0.458521991928 != -1.0
      12: -0.458521991928 != -1.0
      13: 0.458521991928 != 1.0
      14: -0.458521991928 != -1.0