#276 Now using name only to identify signals

[restenb]

Solves #276

[markaren]

In scenario_parser.cpp you should remove find_variable_type and find_causality functions.
In defaults struct and parse_defaults you should remove causality and type
parse_scenario needs to be updated to get the required info from the variable_description

[restenb]

In scenario_parser.cpp you should remove find_variable_type and find_causality functions.

Don't think so, they are still in use for creating a scenario::variable_action.

[markaren]

In scenario_parser.cpp you should remove find_variable_type and find_causality functions.

Don't think so, they are still in use for creating a scenario::variable_action.

That is why you should update parse_scenario to get the required info from the variable_description. I did the same change in #275, but you can be more effective. I loop through the variable list three times. Once is enough.

Edit: And you can remove type and causality form the scenario files.

[restenb]

Okay, now I see. I wasn't familiar with the implementation of scenarios, but on closer inspection this sounds reasonable - type and causality should indeed no longer be needed in the scenario data.

Edit: I'll wait for @eidekrist to comment as he's the original author of this code. Feel free to implement your suggested change on this PR instead.

[markaren]

A compile error was just fixed, but file_observer.cpp and stuff related to it LogConfig (?) still needs to be updated for this PR to be ok.

[markaren]

Any hope of getting this PR moving? I think @eidekrist needs to verify and finish what remains.

[eidekrist]

I'll try to get on this as soon as I can. Looks like we'll need a companion PR for cse-client as well.

[markaren]

Finalizing the java wrapper and noticed that the C interface also needs to be updated. It should not be necessary to pass the variable_type when using the time_series_observer

[markaren]

Finalizing the java wrapper and noticed that the C interface also needs to be updated. It should not be necessary to pass the variable_type when using the time_series_observer

I looked a little bit into this, and I reckon it has to be "fixed" in a separate PR. I still think it is uneccessary for clients of the library to pass this information when the library can find it on his own. But the changes required for making it requires some substantial refactoring.

[eidekrist]

Finalizing the java wrapper and noticed that the C interface also needs to be updated. It should not be necessary to pass the variable_type when using the time_series_observer

I looked a little bit into this, and I reckon it has to be "fixed" in a separate PR. I still think it is uneccessary for clients of the library to pass this information when the library can find it on his own. But the changes required for making it requires some substantial refactoring.

I reckon you're talking about the cse_observer_start_observing() method? In that case, passing the variable_type is necessary, since the value reference is only unique for a given type. See src/test/data/fmi1/identity.fmu, where there are variables for four types, all with the value reference = 0.

[markaren]

Finalizing the java wrapper and noticed that the C interface also needs to be updated. It should not be necessary to pass the variable_type when using the time_series_observer

I looked a little bit into this, and I reckon it has to be "fixed" in a separate PR. I still think it is uneccessary for clients of the library to pass this information when the library can find it on his own. But the changes required for making it requires some substantial refactoring.

I reckon you're talking about the cse_observer_start_observing() method? In that case, passing the variable_type is necessary, since the value reference is only unique for a given type. See src/test/data/fmi1/identity.fmu, where there are variables for four types, all with the value reference = 0.

I had to double-check the spec to see if this was true. It apparently is for both 1.0 and 2.0. Why would they do something like that? Well, well.

[eidekrist]

I am now a happy man, and believe this is good to go. Companion PR for cse-client here. While looking at file_observer.cpp I couldn't help but do some changes here and there, resulting in an improvement (IMO) in the log file output.