#813 get t_eval from data #819
Conversation
Codecov Report
@@ Coverage Diff @@
## master #819 +/- ##
==========================================
+ Coverage 98.21% 98.22% +0.01%
==========================================
Files 181 180 -1
Lines 10241 10260 +19
==========================================
+ Hits 10058 10078 +20
+ Misses 183 182 -1
Continue to review full report at Codecov.
|
There was a problem hiding this comment.
Thanks @rtimms , really glad this is fixed! I think this is the right approach too (automatic time setting in the simulation) - just see comment about when the warning is triggered
pybamm/simulation.py
Outdated
| # of the data points. We only raise a warning here, as users | ||
| # may genuinely only want the solution returned at some | ||
| # specified points. | ||
| if set(time_data).issubset(set(t_eval)) is False: |
There was a problem hiding this comment.
I'm wondering if this will have problems with round-off errors? And should be also allow t_eval with a sampling time smaller than the smallest gap in the data, but that doesn't hit exactly the time points?
There was a problem hiding this comment.
yeah good point re round-off errors. yeah that would probably be fine. not sure of the best way of testing this for all cases and give an informative message (e.g. if t_eval or the data non-uniform, would you then just check the biggest gap in t_eval and smallest in the data and just warn that the user should check to ensure they have enough resolution in t_eval?). perhaps doing that is best, as doing something naive like just checking the number of values might pass if e.g. t_eval had loads of points clustered near the start and not many later
There was a problem hiding this comment.
ah yeah I hadn't thought of the non-uniform t_eval case. I think we can allow both options: checking that either the biggest gap in t_eval is smaller than the smallest in the data, or that each point in the data has a corresponding point in t_eval within some small error tolerance, should cover most cases
There was a problem hiding this comment.
I've updated to just check max dt_eval vs min dt_data as this also catches the case where t_eval = t_data (since dt is equal here). let me know your thoughts and i'll change or merge as is
There was a problem hiding this comment.
I don't think this would always catch the case t_eval = t_data, for example if t_eval=t_data = [0,1,5,6] then max(t_eval) = 4 > 1 = min(t_data)
There was a problem hiding this comment.
oh yeah good point haha
Description
When using the
Simulationclass with the current provided as data (i..e drive cycles)solvenow setst_eval(if it isn't provided) to be the times in the data (non-dimensionlised). Ift_evalis provided,solvechecks that it contains all of the points in the data, and raises a warning if it doesn't. Maybe this should raise an actual exception to be totally fool proof, but there may be cases where people only want the solution returned at a subset of points. Happy for input on this. I think if people are solving in the long-winded way (i.e. not usingSimulation) then we should just trust them to pickt_evalappropriately.Fixes #813
Type of change
Please add a line in the relevant section of CHANGELOG.md to document the change (include PR #) - note reverse order of PR #s. If necessary, also add to the list of breaking changes.
Key checklist:
$ flake8$ python run-tests.py --unit$ cd docsand then$ make clean; make htmlYou can run all three at once, using
$ python run-tests.py --quick.Further checks: