Purpose of stored-definition with several class definitions, and the within-clause

[henrikt-ma]

What is the purpose of allowing stored-definition to have zero or several class definitions?

If we are to keep it this way, the definition of the within-clause would need to be updated to better cover the case of not just a single class definition.

[HansOlsson]

What is the purpose of allowing stored-definition to have zero or several class definitions?

Historically it has been used (e.g., for "SaveTotal") - especially without within. Really annoying users can use it combination with within.

If we are to keep it this way, the definition of the within-clause would need to be updated to better cover the case of not just a single class definition.

I don't see that within causes a problem for it. Consider the following silly example of a package P containing a mo file as follows on Windows:

within P;
model MA
end MA;

model ma
end ma;

I don't know if there's a more relevant use case for this.

[HansOlsson]

Basically within allows tools to detect if files have just been moved around in the directory; thus it makes sense that there's only one within per file.

[henrikt-ma]

I don't see that within causes a problem for it. Consider the following silly example of a package P containing a mo file as follows on Windows:

within P;
model MA
end MA;

model ma
end ma;

There wouldn't be a clear way to name that file. So in addition to generalizing the within-clause to applying to every class in the stored-definition, we'd also have to say something about the filename-class mapping.

I don't know if there's a more relevant use case for this.

I'd prefer steering away from stored-definition with anything but a single class definition, rather than trying agree on exactly how it should work. As usual, tools would still keep support for legacy files with multiple classes for a long time. :)

[HansOlsson]

What is the purpose of allowing stored-definition to have zero or several class definitions?

If we are to keep it this way, the definition of the within-clause would need to be updated to better cover the case of not just a single class definition.

I checked https://specification.modelica.org/master/packages.html#mapping-a-package-class-hierarchy-into-a-single-file-nonstructured-entity and already in Modelica 3.0 (possibly earlier) it is split into three cases:

• A file stored inside a package: Here within is required and stored-definition may only contain one class.
• A top-level class loaded from ModelicaPath: within can be omitted (or include empty name), but stored-definition may only contain one class
• A top-level class loaded explicitly. within can be omitted (or include empty name) and stored-definition may contain multiple classes.

So, I don't see a major issue with the within-clause in combination with several class definitions.

[henrikt-ma]

• A file stored inside a package: Here within is required and stored-definition may only contain one class.

This is only when mapping a package hierarchy to the file system.

• A top-level class loaded from ModelicaPath: within can be omitted (or include empty name), but stored-definition may only contain one class

Right, so no need to worry about ModelicaPath here.

• A top-level class loaded explicitly. within can be omitted (or include empty name) and stored-definition may contain multiple classes.

What prevents MyClasses.mo from placing classes within some package?

Anyway, at least I have a difficulty extracting the rules and purpose of files with non-1 class definitions from the current text. Wouldn't it be easier to express clearly by describing such files in a separate paragraph somewhere?

[HansOlsson]

• A file stored inside a package: Here within is required and stored-definition may only contain one class.

This is only when mapping a package hierarchy to the file system.

Yes, that is the meaning of storing a file inside a package. Note that within is defined in this context.

• A top-level class loaded explicitly. within can be omitted (or include empty name) and stored-definition may contain multiple classes.

What prevents MyClasses.mo from placing classes within some package?

That within should correspond to the enclosing structure and there isn't any. Is there a need to clarify that within is used for consistency checking and not to move classes arbitrarily?

Anyway, at least I have a difficulty extracting the rules and purpose of files with non-1 class definitions from the current text. Wouldn't it be easier to express clearly by describing such files in a separate paragraph somewhere?

[henrikt-ma]

What prevents MyClasses.mo from placing classes within some package?

That within should correspond to the enclosing structure and there isn't any. Is there a need to clarify that within is used for consistency checking and not to move classes arbitrarily?

Yes, I think that would be helpful, especially if combined with something along the following lines in a separate section before File System Mapping of Package/Class:

A stored-definition not containing exactly one class definition can only be used to define top-level classes, cannot be used for mapping packages or class-hierarchies to files, and cannot be located using the MODELICAPATH. It follows that any within-clause (see below) in such a stored-definition shall not contain a name.

Any suggestions for what would be a good title of that section? How about Stored Definitions Containing Multiple Class Definitions?