[cfe-dev] Getting the FieldDecl of a template class from a MemberExpr

[Adrien Chauve]

Thanks! I'm going to work on that!

On Mon, Jun 6, 2011 at 19:01, Douglas Gregor <dgregor at apple.com> wrote:

>
> Sent from my iPhone
>
> On May 29, 2011, at 9:31 AM, Adrien Chauve <adrien.chauve at gmail.com>
> wrote:
>
> Thanks a lot!
> Here is what I came up with (I didn't make getInstantiatedFrom a member
> yet):
>
> /// getInstantiatedFrom - if the parent of this (FieldDecl* d) is
> an
> /// instanciation of a template class, return the corresponding FieldDecl*
> of this
> /// template
> class.
>
> FieldDecl* getInstantiatedFrom(FieldDecl const * const
> d)
>
> {
>
>     ClassTemplateSpecializationDecl const * const parent
> =
>
>
> dyn_cast<ClassTemplateSpecializationDecl>(d->getParent());
>
>
>
> CXXRecordDecls themselves can also be instantiations of member classes.
> There's a getInstantiatedFromMember() function (or something like it) to
> determine when this is the case.
>
>
>
>     // check the parent of this field is an instanciation of a class
> template
>     if (!parent || parent->getTemplateSpecializationKind() !=
> TSK_ImplicitInstantiation)
>         return
> 0;
>
>
>
> Explicit instantiations should also be considered here.
>
>
>
>     CXXRecordDecl const * const generic_parent
> =
>
>
> parent->getSpecializedTemplate()->getTemplatedDecl();
>
>
> assert(generic_parent);
>
>
>
>     for(CXXRecordDecl::field_iterator f=generic_parent->field_begin(),
> e=generic_parent->field_end(); f != e; ++f)
>
> {
>
>         if ((*f)->getNameAsString() ==
> d->getNameAsString())
>
>             return
> *f;
>
>
> }
>
>
>
> DeclContex::lookup() can perform this search faster.
>
>
>
>     // should never get
> here
>
>     return
> 0;
>
> }
>
>
> Is it what you had in mind? I would be glad to make it a patch and add some
> tests it it seems ok.
>
> Adrien
>
> On Fri, May 27, 2011 at 23:34, Douglas Gregor < <dgregor at apple.com>
> dgregor at apple.com> wrote:
>
>>
>> On May 27, 2011, at 2:32 PM, Johannes Schaub (litb) wrote:
>>
>> > Douglas Gregor wrote:
>> >
>> >>
>> >> On May 27, 2011, at 10:02 AM, Adrien Chauve wrote:
>> >>
>> >>> Hi,
>> >>>
>> >>> Using a RecursiveASTVisitor, I've been trying to rename field names
>> >>> against some conventions.
>> >>>
>> >>> Let's say I would like to rename Foo::bar into Foo::m_bar in the
>> >>> following code:
>> >>>
>> >>> template<typename T>
>> >>> struct Foo
>> >>> {
>> >>>    int bar;
>> >>> };
>> >>>
>> >>> int main()
>> >>> {
>> >>>    Foo<double> foo;
>> >>>    foo.bar = 2;
>> >>>
>> >>>    return 1;
>> >>> }
>> >>>
>> >>>
>> >>> In my custom visitor, I implement the following method:
>> >>>
>> >>> bool VisitMemberExpr(MemberExpr* member)
>> >>> {
>> >>>    ValueDecl* v = member->getMemberDecl(); // FieldDecl or
>> CXXMethodDecl
>> >>>    ...
>> >>> }
>> >>>
>> >>> But the FieldDecl I get when visiting the statement corresponding to
>> >>> "foo.bar=2;" is Foo<double>::bar, but not Foo::bar.
>> >>>
>> >>> So my question is: is it possible to get the FieldDecl of Foo<T>::bar
>> >>> from the FieldDecl of Foo<double>::bar ? at least when the struct is
>> not
>> >>> specialized? What happens when the struct is specialized?
>> >>
>> >> For fields, this is a pain. You'll actually have to look at the type
>> that
>> >> owns the field (a ClassTemplateSpecializationDecl), check whether it
>> was
>> >> instantiated vs. specialized (using getSpecializationKind()) and, if it
>> >> was instantiated, retrieve the class template from which is was
>> >> instantiated (getSpecializedTemplate()).
>> >>
>> >
>> > I was looking into the doxygen documentations, and I couldn't find how
>> to
>> > proceed from that.
>> >
>> > How to get the correct field decl of the class template that correspond
>> to
>> > the field decl of the instantiated specialization? Can one just take the
>> > std::distance of the field decl iterator for the specialization, and
>> with
>> > that, index into the field decl iterator of the class template?
>>
>> That works, or you can use name lookup.
>>
>>        - Doug
>>
>> _______________________________________________
>> cfe-dev mailing list
>>  <cfe-dev at cs.uiuc.edu>cfe-dev at cs.uiuc.edu
>>  <http://lists.cs.uiuc.edu/mailman/listinfo/cfe-dev>
>> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-dev
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20110606/af5eab8c/attachment.html>