drop tests for indexing problem with a parameter

[isaacsas]

Since indexing problems by a parameter has been dropped by SciMLBase we shouldn't test for this anymore.

[coveralls]

Pull Request Test Coverage Report for Build 7366660627

• 0 of 0 changed or added relevant lines in 0 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+1.3%) to 89.945%

---

Totals
Change from base Build 7364429899:	1.3%
Covered Lines:	2299
Relevant Lines:	2556

---

💛 - Coveralls

[TorkelE]

For reference, not sure whether this is implemented or not, but I think the plan is to make something like;

@parameters k
...
prob.ps[k]

work for parameter indexing. Basically, due to parameters having some property, allowing

prob[k]

directly (like what is done for initial conditions) becomes type unstable. hence the indexing has to be done on the .ps field instead. When it is clear that this is ready, we should move to that instead.

[ChrisRackauckas]

For reference, not sure whether this is implemented or not, but I think the plan is to make something like;

Set getp/setp.

But again this is separate from whether these tests should be here. The symbolic indexing interface has all of those things, but we are dropping specifically definition of names by defining f.syms

[isaacsas]

For some reason 1.10 CI seems bugged. Not sure why it is taking over an hour (I've already restarted it once).

[isaacsas]

OK, so we should just drop these tests completely then? Since otherwise there is no real way to create a problem with names here (i.e. we don't want to have to use MTK to run the tests...)

[ChrisRackauckas]

You can define a SymbolCache f.sys

[ChrisRackauckas]

setp will still allow for issues, along with manual mutation.

[isaacsas]

We probably need a custom dispatch on it then to update MassActionJumps?

[ChrisRackauckas]

Yes, @AayushSabharwal can probably help with that.

[isaacsas]

That would be great, thanks!

[AayushSabharwal]

We now have set_parameter! in SII which is used by setp internally, and can be implemented to do other bookkeeping actions when setp is used.

[isaacsas]

@AayushSabharwal, can you add the needed dispatches to that when parameters get modified, i.e. something like what we had before when modifying a problem:

if using_params(prob.massaction_jump)
    update_parameters!(prob.massaction_jump, prob.prob.p)
end

Alternatively, @ChrisRackauckas do we want to put this on users (since this could be expensive, so a faster workflow is to only call it once all parameter modifications are finished)? That does leave a gotcha someone could encounter unless they carefully read the docs (which is why the previous behavior here always updated directly after modifying a parameter).

[isaacsas]

But we do put calling the jump update function on users currently when modifying integrators within callbacks, so perhaps it isn't unreasonable to do that.

[isaacsas]

You can define a SymbolCache f.sys

Ok, that is very helpful to know. I was confused from looking at the example test file in SII and thought I’d have to completely build a custom struct satisfying the interface.

[ChrisRackauckas]

We'll make it a tutorial, https://docs.sciml.ai/SymbolicIndexingInterface/dev/simple_sii_sys/ it's just missing right now.

[AayushSabharwal]

SciMLBase master supports prob.ps[p] syntax now, so parameter indexing tests can use that instead

[AayushSabharwal]

SciMLBase#master isn't tagged yet, so this won't work

[isaacsas]

Where has the dispatch been added to ensure update_parameters!(jprob.massaction_jump, jprob.prob.p) is called when these parameter changes are made?

[AayushSabharwal]

bump

[isaacsas]

Handled in #408 now.