Bug in format of XML files: ambiguity between one-object and unnamed object properties

[chrisdembia]

We allow users to express single-value Object properties in two different ways in XML. For example, a property declared as OpenSim_DECLARE_PROPERTY(myfunc, GCVSpline, "comment"); can appear as:

<GCVSpline name="myfunc">
 ...
</GCVSpline>

or

<myfunc>
    <GCVSpline>
    ...
    </GCVSpline>
</myfunc>

This is explained on doxygen.

If I have an unnamed property OpenSim_DECLARE_UNNAMED_PROPERTY(GCVSpline, "comment");, it appears in XML as:

<GCVSpline>
 ...
</GCVSpline>

(the name attribute is ignored if it exists).

Now, if I have a class with the following two properties:

OpenSim_DECLARE_PROPERTY(myfunc, GCVSpline, "comment");
OpenSim_DECLARE_UNNAMED_PROPERTY(GCVSpline, "comment");

the XML fragment <GCVSpline name="myfunc"> </GCVSpline> is ambiguous but gets interpreted as the unnamed property.

I think the fix is simple: we remove the ability for unnamed properties to have a name attribute. Instead of ignoring the name attribute, we must make sure it does not exist.

@sherm1

[sherm1]

I think the fix is simple: we remove the ability for unnamed properties to have a name attribute.

That makes sense. A couple of thoughts:

• It really doesn't make sense for a component to have both an unnamed and named property of the same type. The unnamed properties are there because if there is only one property of that type then the type can serve as the name. If there are more than one, they ought all to be named. Maybe it would be better to diagnose this as an error rather than disambiguate with the name attribute?
• If there is only an unnamed property and a name is given it would probably be better to give an error message rather than ignoring the name as is done presently.

[chrisdembia]

Maybe it would be better to diagnose this as an error rather than disambiguate with the name attribute?

Sounds fine to me.