Update `Modelica_StateGraph2` library to be Modelica 3.6/3.7 compliant

[modelica-trac-importer]

Reported by dietmarw on 1 Sep 2010 07:51 UTC
It would be reasonable to update this library with respect to Modelica 3.2 conventions:

• move resource file into Resources, e.g.:
  • folder Images should be moved to Resources/Images
  • scripts should be moved to Resources/Scripts/Dymola
• use of improved icons provided by the MSL3.2

---

Migrated-From: https://trac.modelica.org/Modelica/ticket/398

[modelica-trac-importer]

Modified by dietmarw on 8 Feb 2013 23:25 UTC

[modelica-trac-importer]

Changelog removed by dietmarw on 8 Feb 2013 23:25 UTC

[dietmarw]

@MartinOtter I was wondering, are there any plans to work on this? Dymola keeps shipping newer versions but I can't find any commits to this library that reflect the changes done in the version(s) shipped with Dymola.

[tbeu]

@MartinOtter @tobolar Dymola ships v2.0.3 which resolved the Resource/Images issue. This probably should also be reflected by this repository.

[HansOlsson]

The image-files have been corrected.

However, there are additional issues (I believe I have listed the first two previously):

• Connections.uniqueRoot is non-standard. This not just an extra argument for error messages (as I previously thought), but also differs by not allowing edges between different roots (even non spanning-tree edges). That could maybe sort of be implemented in Modelica 3.4 by using Connections.root and only have assert(false, ""); in the corresponding equalityConstraint. (Assuming tools handle this idiom - otherwise the diagnostics for such models will be annoyingly bad.)
• Connections.uniqueRootIndices is non-standard and used e.g. for checking parallel components. I don't see any way of implementing the check for parallel components using Modelica 3.4. The check that every loop has a delayed transition could be implemented using Modelica 3.4 (the connectors could have an extra variable indicating number of immediate transitions for it; loops then causing integer equations that cannot be solved).
• There are some illegal uses of expandable connector variables in e.g. Modelica_StateGraph2.Examples.Applications.ControlledTanks.Internal.TankController2. That could be corrected.
• The libraryinfo.mos is still at top-level. Currently Dymola does not support that special script further down. That will be improved in Dymola 2020.

Lower importance (these vendor-annotations are optional):

• Several models in Modelica_StateGraph2.Examples.WrongStateGraphs use annotation(__Dymola_inhibitCheck=true). I don't know how we handle such models. We already have annotation (__ModelicaAssociation(TestCase(shouldPass = false))), which is somewhat similar but different. The library "works" even if the annotation is not supported; it just generates errors for the deliberately incorrect models. One alternative would be to not include these models.

The goal would still be to update Modelica.StateGraph based on this.

[HansOlsson]

Created modelica/ModelicaSpecification#3701 for the first two issues.

[HansOlsson]

Note: Replacing uniqueRoot by root, and uniqueRootIndices by 1:nEntry kind of works (well, the sane idea would be to remove more of the logic), and generates a standard-conformant library without any extensions beyond 3.6.

Correct models work the same, but the following incorrect models would be accepted:

• Modelica_StateGraph2.Examples.WrongStateGraphs.TwoInitialSteps
• Modelica_StateGraph2.Examples.WrongStateGraphs.ParallelWithTwoInitialSteps
• Modelica_StateGraph2.Examples.WrongStateGraphs.UnsafeParallel1
• Modelica_StateGraph2.Examples.WrongStateGraphs.UnsafeParallel4

Regarding that version:

• If we were considering replacing Modelica.StateGraph in a new major version including that version would have been an improvement (but including it alongside Modelica.StateGraph in MSL 4.2.0 would be too confusing).
• It would be possible to make it as a compatibility release of this library.
• It might be possible to update Modelica.StateGraph based on this, with e.g., InitialStep just a short class =Step(initialStep=true);