search

modelica-tools / csv-compare

Tool to compare curves from one csv files with curves from other csv files using an adjustable tolerance
https://www.modelica.org
BSD 3-Clause "New" or "Revised" License
25 stars 15 forks source link

Issues with updated tube width logic #74

Closed casella closed 2 days ago

casella commented 3 days ago

Matthias Schaefer reports that as LTX started to use the updated CSV-compare tool (commit 6c456d6a) on their servers, they started to get worse results in many cases that were previously working fine.

One such case is ModelicaTest.Fluid.Dissipation.TestCases.PressureLoss.Bend. If you check the linked test report for variable result.IN_p.ports[1].h_outflow, there is a constant reference value of 293586.68524, with an upper and lower tube that are apparently giving a relative tolerance of 1e-12.

My original commit 18a20316 only tweaked the computation of baseY by replacing epsilon = 1e-12 with default_nominal = 0.001 in case both the minimum value of the reference signal and the difference between the minimum and the maximum were very small or zero. As far as I understand, that should not have caused such a problem. I'll have Matthias check with that commit, just in case.

@beutlich, after that commit, you applied several refactoring commits. Maybe something went wrong there? Could you please check with the above-mentioned test case?

Thanks!

casella commented 3 days ago

Keeping @GallLeo in the loop.

beutlich commented 2 days ago

Thanks for reporting. As stated here, such regressions are rather expected due to insufficient testing.

75 addresses two of the found regressions.