PM_model$new cannot handle constant values in the #PRI block

[Siel]

Describe the bug
When using PM_model$new with a model.txt that contains constant values in the #PRI block it fails:

> mod1 <- PM_model$new("model.txt")
Error in `FUN()`:
✖ Constants should be defined in the appropriate block, not #PRI.
Run `rlang::last_trace()` to see where the error occurred.
> 

Expected behavior
It should load the model correctly

Version information

• OS: Tested on MacOs
• R: 4.5.1

[mhovd]

Is this not expected behaviour?
Constant parameters should be defined in the secondary block, no?

[Siel]

It should be in the R code for the model, reading from model.txt it should generate the right code. or at least give the user a clear error message

[mnneely]

I would argue it does give a clear message: ✖ Constants should be defined in the appropriate block, not #PRI.

I agree with Markus. This is expected behavior with an informative error trap.

[Siel]

I just received a message from an user stuck in that error, we all understand the message directly points to the error, but it gives not path forward. Pmetrics could very easily fix the model.txt/generate the right R code.

[mnneely]

I think it does give a path forward, but we can consider automatically moving constants to a SEC block when reading model.txt files and informing the user. Can anyone think of a reason not to do that?

[mhovd]

Can anyone think of a reason not to do that?

The only, albeit not very user-friendly, reason I can come up with is the eventual deprecation of model.txt, and we should be careful with expanding features associated with it to reduce its use.

Given that a user was stuck at that message, it means that either the message is not informative enough, or Pmetrics was not flexible enough. So we should do something to improve the experience. I am leaning towards an expanded error message - Pmetrics no longer suppors fixed parameters, and injecting them from the primary block to others is 1) not transparent, and 2) not how it is handled in the list-based model definition in R.