Regression test for the values of the 1D solution
samcunliffe opened this issue · comments
Expect a damped sine wave. Should do a stupid regression test of these values.
The expected solution ☝️ obtained from the initial commit of Janosch's code. Errors here are:
t = 1, L2-error = 0.07976660467116257
L-infty-L2 error u = 0.09457065797536773
L-infty-L2 error u_t = 0.7751446961112378
L2-L2 error u_t = 0.5385788727685025
And in the current HEAD
of main
:
The convergence reason code is 3
elapsed time 0.05649995803833008 seconds
t = 1, L2-error = 0.7071067811865475
L-infty-L2 error u = 0.7071067811865475
L-infty-L2 error u_t = 4.442882938158366
L2-L2 error u_t = 3.141592653589793
So a failing test looks "easy". Can just check the error dictionary to make sure they are small. Slow(ish) manual bisection ongoing....
Looks like the bug was introduced during #17. My test passes on the commit before, fails after #17, although #35 is also needed to get the 1d solver to run.
commit 4efff748d542c73674e17375aa7bb50b6a049bac (HEAD)
Author: Sam Cunliffe <samcunliffe@users.noreply.github.com>
Date: Thu Oct 5 17:15:32 2023 +0100
Smoke test and fix solve_1d.py (#35)
commit 69b68ed62add8af6ad8f27b92aa1b36ed0843f6a
Author: Sam Cunliffe <samcunliffe@users.noreply.github.com>
Date: Sun Oct 1 19:35:42 2023 +0100
Linting part two (#17)
Note that the above commit hashes are different to main
because we force-rewrote history in order to solve #34.