add QL specification section on module instantiations

[ginsbach]

• fixes https://github.com/github/codeql-core/issues/2452

[alexet]

LGTM but I think there are still some missing parts

For all the bits in the evaluation sections that kind-of happen "after substitution" we probably want something to just state that we handle predicates by using the normal algorithm but replacing all instantiation-relative predicates and types with the actual version. You probably want to say this should happen after we resolve calls and type references but I think we can get away with being a bit vague about precisely what that means.

The section on layer evaluation may want a note that it applies to "fully instantiated" types and predicates.

Otherwise the missing parts are really covered by the general section on components.

[ginsbach]

The section on layer evaluation may want a note that it applies to "fully instantiated" types and predicates.

I have added this in the fifth commit.

For all the bits in the evaluation sections that kind-of happen "after substitution" we probably want something to just state that we handle predicates by using the normal algorithm but replacing all instantiation-relative predicates and types with the actual version. You probably want to say this should happen after we resolve calls and type references but I think we can get away with being a bit vague about precisely what that means.

Can you elaborate a bit which section you mean and what changes you'd like to see?

[alexet]

Changes LGTM

The point I was trying to make is that the sections on Stratification and Layer evaluation only really apply to declared predicates. We don't need to go into great detail but a line saying it works by substituting after resolving is fine. Alternatively we could probably get away with by substituting after resolving types and predicates.

[ginsbach]

Changes LGTM

The point I was trying to make is that the sections on Stratification and Layer evaluation only really apply to declared predicates. We don't need to go into great detail but a line saying it works by substituting after resolving is fine. Alternatively we could probably get away with by substituting after resolving types and predicates.

I will consider a follow-up PR to polish the evaluation section (see #13513).