Decompose compound types

[sdkrystian]

Consider the following:

struct S { };

S x;
S* y;

which produces the following output:

<namespace name="">
  <struct name="S" id="pOYGF6pLJlICuiGO0Xj0daDb/to=">
    <file path="input.cpp" line="6" class="def"/>
  </struct>
  <variable name="y" id="cND4SpNe0NdSmFwUXrBAZmL6sLo=">
    <file path="input.cpp" line="9" class="def"/>
    <type name="S *"/>
  </variable>
  <variable name="x" id="ujOvtFFNPf6akih2fga4S9j6Y9M=">
    <file path="input.cpp" line="8" class="def"/>
    <type name="S" id="pOYGF6pLJlICuiGO0Xj0daDb/to="/>
  </variable>
</namespace>

It may be desirable to define types (i.e, TypeInfo) in a way that separates cv-qualifiers and (pointer/reference/array/function)ness from the unadorned types (so their IDs can be used to reference the declaration of the type).

[vinniefalco]

Well the rationale in the original clang-doc is that this would produce an explosion of small allocated TypeInfos. And how does separating the cv-qualifiers help the documentation?

[sdkrystian]

@vinniefalco look at https://clang.llvm.org/doxygen/classclang_1_1ASTContext.html#aa39dff434d30711957097bb5b9d04343 -- note how there is a link to ArrayType in the return type, even though it is a pointer to a cv-qualified type