Yaml writer bug after removing some originally duplicate reactions

[lavdwall]

In python: When doing mechanism reduction based on reaction rates of progress, it is possible that duplicate reactions in the original mechanism are no longer duplicate in the reduced mechanism. Those reactions can be identified and the "duplicate" property can be set to False. However, since this property change is not pushed to the underlying C++ object, when the reduced mechanism is written to a *yaml file using the YamlWriter, the originally duplicate reactions are still indicated as being duplicate, and reading the reduced mechanism gives an error.

Steps to reproduce Taking GRI-3.0 as example, in which reaction 87 and 88 are duplicate. I will remove reaction 88 and set reaction 87 to non-duplicate, then write that "reduced" model to yaml. Importing this new reduced model gives an error.

gas = ct.Solution('gri30.yaml')
reactions = [gas.reaction(i) for i in range(gas.n_reactions) if i != 88]
reactions[87].duplicate = False
gas_reduced = ct.Solution(name="gas", thermo="ideal-gas", kinetics="gas", species=gas.species(), reactions=reactions)
gas_reduced.write_yaml('gri30_test_red.yaml')

Now read the new mechanism: gas_reduced = ct.Solution('gri30_test_red.yaml')

Behavior

Gives the following error:

InputFileError thrown by Kinetics::checkDuplicates: Error on line 1186 of ./gri30_test_red.yaml: No duplicate found for declared duplicate reaction number 87 (H2O2 + OH <=> H2O + HO2)

[ischoegl]

Thanks, @lavdwall for reporting the issue. I can reproduce, where

*******************************************************************************
InputFileError thrown by Kinetics::checkDuplicates:
Error on line 1186 of ./gri30_test_red.yaml:
No duplicate found for declared duplicate reaction number 87 (H2O2 + OH <=> H2O + HO2)
|  Line |
|  1181 |   - equation: 2 OH <=> H2O + O
|  1182 |     rate-constant: {A: 35.7, b: 2.4, Ea: -8.82824e+06}
|  1183 |   - equation: HO2 + OH <=> H2O + O2
|  1184 |     rate-constant: {A: 1.45e+10, b: 0.0, Ea: -2.092e+06}
|  1185 |     duplicate: true
>  1186 >   - equation: H2O2 + OH <=> H2O + HO2
              ^
|  1187 |     rate-constant: {A: 2.0e+09, b: 0.0, Ea: 1.786568e+06}
|  1188 |     duplicate: true
|  1189 |   - equation: C + OH <=> CO + H
*******************************************************************************

[ischoegl]

On closer inspection, the duplicate flag is correctly passed to the C++ layer; however, buffered input data still hold the original value and overwrite things in Reaction::parameters. See https://github.com/Cantera/cantera/blob/90b4481554dd2014e09398fecba713275cd890bc/src/kinetics/Reaction.cpp#L196-L202 Reversing the order of operations (first write input, then overwrite with actual parameter values) breaks a surprising amount of logic.