Add point that reset variables must be initialised?

[kerimoyle]

What do people feel about adding a point to either:
- 2.9 The reset element, or
- 3.11 Interpretation of resets
that says that the reset variable must be initialised? I think it would make a huge amount of the operation of the resets clearer, it's easy to validate, and the resets won't actually work without it ...?

[MichaelClerx]

How do you mean "it must be initialised" ?

[kerimoyle]

It must have an initial_value attribute somewhere in the set? Or have I got that wrong? ie: you can't define its value through the maths A = 1 if you want to reset A anywhere ... ?

[MichaelClerx]

Yeah I think that that's the only way the reset rules make sense. We can't have both "x is always 3" and "set x to 7". Probably wherever we explain the difference between eternal truths and initial values would be the place to bring this up?

[kerimoyle]

Another confusing thing: It's not that initial values are only true at t=0, it's that they're true until changed (whether by integration, maths, or resets makes no difference). The "initial" throws me a little, as I associate it only with integration. Maybe we can use the attribute initial_value still, but refer to them in the text by something more generic, as we do with map_variables etc ...

[MichaelClerx]

That's cool!
We're only just working out what these things mean. It started with people saying "what constructs do we need to encode all the models we care about", I think, but now we're approaching it from the other end and saying "now that we've said X, what does that imply"? I like it

[MichaelClerx]

Initially, x=0, until changed by a reset or ODE ?

[kerimoyle]

... or a change in another variable in an equation? Anything that has a value that can change for any reason needs to be initialised directly through either initial_value or indirectly by other variables that it depends on being initialised? I've started trying to figure this out ... see: https://libcellml-tutorials.readthedocs.io/en/pr344_documentation/tutorials/asides/aside_constant_vs_variable.html (work in progress!!)

[MichaelClerx]

... or a change in another variable in an equation?

how so?

[kerimoyle]

A = B + C
B is changed by a reset or ODE? I guess A doesn't need to be initialised then ...

Initially, x=0, until changed by a reset or ODE ?

Can we not just say: x=0 until changed by something? Ditch the initial part?

[MichaelClerx]

Is "initial" not a shorthand for "when the vector t is some unspecified value t0" ?

[kerimoyle]

It implies that it's a state variable (to me anyway), rather than just a regular variable that you need to give a value to. But I guess this could be cleared up with some clarity in whatever section in chapter 3 of the normative, and some good examples in the informative.