Codecov Report

Attention: Patch coverage is 40.00000% with 3 lines in your changes missing coverage. Please review.

Project coverage is 73.12%. Comparing base (9d0e54a) to head (f8dc0ab).

@@            Coverage Diff             @@
##             main    #1760      +/-   ##
==========================================
- Coverage   73.12%   73.12%   -0.01%     
==========================================
  Files         381      381              
  Lines       54175    54172       -3     
  Branches     9226     9223       -3     
==========================================
- Hits        39615    39612       -3     
+ Misses      11600    11599       -1     
- Partials     2960     2961       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

What level of compatibility are we trying to maintain for the clib API between Cantera 3.0 and 3.1? So far, I think most of the changes have been consistent with what we do in the Python / C++ interface where user code should continue to work as long as it wasn't using any functions that were deprecated in Cantera 3.0. However, I'd also be OK with adopting a less conservative approach for clib to enable more rapid improvements.

What level of compatibility are we trying to maintain for the clib API between Cantera 3.0 and 3.1? So far, I think most of the changes have been consistent with what we do in the Python / C++ interface where user code should continue to work as long as it wasn't using any functions that were deprecated in Cantera 3.0. However, I'd also be OK with adopting a less conservative approach for clib to enable more rapid improvements.

Good question. With the legacy MATLAB toolbox gone, I believe that all of the remaining API's are de-facto "experimental" (.NET is unfinished and Fortran only captures a small subset of features). We can certainly include placeholders so symbols are still in place, but at the same time breaking changes may be a feature. Specifically, I think it's dangerous to use the 3.1 version of the experimental MATLAB toolbox with a 3.0 clib (or, more realistically, a 3.0 experimental toolbox with 3.1 clib). Until we have non-experimental API's that run through clib, I believe we should be able to cycle more rapidly. Changes in this PR are one case in point: the old clib was inconsistent, so specialized calls were necessary for a handful of methods that didn't fit the norm.