Functions missing derivative functions, that are already present

[modelica-trac-importer]

Reported by veje.johannes on 16 Feb 2016 10:04 UTC
In Media.IdealGases.Common.Functions there are three functions where the corresponding derivative function is available but not implemented.

These three functions are; h_T, h_Tlow, and s0_Tlow with the corresponding functions: h_T_der, h_Tlow_der, and s0_Tlow_der.

function h_T 
  ...
  annotation (Inline=false,smoothOrder=2);
end h_T;

For a fourth function, cp_Tlow, the derivative function is implemented.

function cp_Tlow 
  ...
  annotation (Inline=false, derivative(zeroDerivative=data) = cp_Tlow_der);
end cp_Tlow;

---

Migrated-From: https://trac.modelica.org/Modelica/ticket/1909

[modelica-trac-importer]

Modified by beutlich on 16 Feb 2016 14:16 UTC

[modelica-trac-importer]

Comment by hubertus on 16 Feb 2016 14:35 UTC
There are two options for getting derivatives of functions:

1. the developer writes a derivative function and uses the derivative-annotation or
2. the tool computes the derivative automatically through AD (automatic differentiation) based on the smoothOrder annotation given by the developer.

I have chosen the latter, and I think most tools are fine with that. Note that one has to do either one or the other. But: this code was written when Dymola was the only tool that supported the Media library. So I'd like to use this ticket to get input from all tool developers:

• do you support computation of derivatives based on the SmoothOrder annotation?
• which of the two options do you prefer?

Note that the derivative functions are historical, Dymola supported that before supporting SmoothOrder.

[modelica-trac-importer]

Comment by beutlich on 16 Feb 2016 14:46 UTC
For the record: SimulationX supports both options.

[modelica-trac-importer]

Comment by hansolsson on 17 Feb 2016 16:37 UTC
Dymola also supports both - and I see several reasons to prefer smoothOrder:

• Reducing the risk of errors by having less code
• Declarative - by stating that the function is 2 times continuously differentiable instead of giving a derivative

The cases where a function has a smooth derivative, but it cannot be computed using automatic differentiation of the algorithm as it is written is pretty rare (but I am sure some have found those cases).

[modelica-trac-importer]

Comment by jmattsson on 18 Feb 2016 08:25 UTC
JModelica.org supports both.

[beutlich]

As again pointed here by @thorade the derivative or smoothOrder annotation, respectively, is missing for s0_Tlow.