Allow non-reactant orders for non-electrochemical reactions

[imitrichev]

Good day! I need an ability to have non-reactant partial orders for reactions. It will make possible to define global reactions and use Cantera to calculate their kinetics.
So, I propose to add this ability, when new nonreactant_orders option in cti is true.
(or, maybe, I have something missed, and there is a specific class for global kinetics in Cantera?)

It was intended to have non-reactant reaction orders only for electrochemical reactions, as one can see.
Now, if one uses cti (and I always use it), cti->ctml converter ctml_writer.py blocks "non-reactant order" for any reaction (link).

I've also tried example sofc.py with sofc.cti adding non-reactant order for edge_reaction.
edge_reaction("O(m) + (ox) + 2 electron <=> (m) + O''(ox)",
[5.0e13, 0.0, 120.0], beta = 0.5, id="edge-f3", order= "H(m):0.01")
It expectedly fails.

order specified for non-reactant: H(m)
....
File "/usr/local/lib/python2.7/site-packages/cantera/ctml_writer.py", line 1143, in init
raise CTI_Error("order specified for non-reactant: "+o)
CTI_Error

However, the tests are passed as they do not make use of ctml_writer.
Is that check in ctml_writer.py redundant? We have one more check in Reaction::validate() here.

Many thanks for your attention!
Ivan.

[speth]

Indeed, as a result of the refactoring which introduced the C++ Reaction class, the handling of reaction orders was generalized in a way which means that non-reactant orders can be used without any trouble.

I agree that by default, Cantera should still treat these as errors without a specific override, and that adding a nonreactant_orders option to the CTI reaction class, which parallels the negative_orders option would be appropriate.

There are a number of checks in ctml_writer which are redundant with checks performed in the C++ code. These checks used to be more important when the process for importing a mechanism required manually converting the CTI file to XML, so that errors could be caught in the conversion step rather than only when finally trying to import the mechanism.

[imitrichev]

Well, I will check how it works and prepare a small pull request in the near future!