[clangd] Support type hints for `decltype(expr)`

[Trass3r]

clang-tools-extra/clangd/InlayHints.cpp

315

Should use dyn_cast instead of is+cast I think.

[v1nh1shungry]

Apply the comment's suggestion

[v1nh1shungry]

@Trass3r Thank you for the suggestion!

[v1nh1shungry]

Avoid type hints for dependent types

[v1nh1shungry]

• fix qualifiers
• fix recursive decltype (partly)
• support return type

[nridge]

Sorry for being a slow reviewer.

I think showing type hints for decltype(expr) would be a nice enhancement. @v1nh1shungry, are you interested in working further on this?

One high-level thought I had is: what if we attached the type hint to the closing ) of the decltype (and had it pertain only to the decltype(expr), not anything surrounding it like const or &)? It seems to me that this would both simplify the implementation, and allow us to show hints in places where decltype is used in a context unrelated to a variable or function declaration (for example, in something like using Foo = A<B<decltype(expr), C>, D>, if decltype(expr) was int, we would show using Foo = A<B<decltype(expr) : int, C>, D>. What do you think about this approach?

[v1nh1shungry]

Thanks for the feedback!

One high-level thought I had is: what if we attached the type hint to the closing ) of the decltype (and had it pertain only to the decltype(expr), not anything surrounding it like const or &)? It seems to me that this would both simplify the implementation, and allow us to show hints in places where decltype is used in a context unrelated to a variable or function declaration (for example, in something like using Foo = A<B<decltype(expr), C>, D>, if decltype(expr) was int, we would show using Foo = A<B<decltype(expr) : int, C>, D>. What do you think about this approach?

I think this approach is better.

I think showing type hints for decltype(expr) would be a nice enhancement. @v1nh1shungry, are you interested in working further on this?

Sure!

clang-tools-extra/clangd/unittests/InlayHintTests.cpp

1155

The only concern I have is how to deal with this situation.

When I wrote this patch I thought maybe I could remove the & and figure out the underlying type of decltype(0) recursively and then add the & back. But at that time I couldn't find a way to get the referenced type or the pointee type.

[nridge]

clang-tools-extra/clangd/unittests/InlayHintTests.cpp

1155

This is a problem we already have for auto. For example, if you write:

auto i = decltype(0)();

the type hint for the auto is decltype(0).

The reason is that we are printing the type with TypeHintPolicy, which has PrintCanonicalTypes=false (the default). This means the type is printed without "type sugar" (e.g. decltype, typedefs) being replaced with their "underlying type".

We could alternatively print it with PrintCanonicalTypes=true, which would make the hint given in this case better, but it would make other cases worse (e.g. replacing std::string with std::basic_string<...>).

For the purposes of this patch, I think it's fine to just live with whatever type is printed with the current printing policy.

If we want to make improvements in this area, we can do that in a separate patch, and in a way that affects both auto and decltype(expr).

[v1nh1shungry]

Got it! Thank you for the detailed explanation!

[v1nh1shungry]

reimplement and address review's suggestions

[nridge]

Thanks!

[v1nh1shungry]

@nridge Thank you for reviewing!

If this patch is ready to land, could you please help me commit it? Thanks again!

[nridge]

Done, thanks for the patch!

clang-tools-extra/clangd/InlayHints.cpp

bool VisitVarDecl(VarDecl *D) {
     if (auto *DD = dyn_cast<DecompositionDecl>(D)) {
       for (auto *Binding : DD->bindings()) {
         addTypeHint(Binding->getLocation(), Binding->getType(), /*Prefix=*/": ",
                     StructuredBindingPolicy);
       }
       return true;
     }
 
-    if (D->getType()->getContainedAutoType()) {
-      if (!D->getType()->isDependentType()) {
+    auto T = D->getType();
+    if (T->getContainedAutoType()) {
+      if (!T->isDependentType()) {
         // Our current approach is to place the hint on the variable
         // and accordingly print the full type
         // (e.g. for `const auto& x = 42`, print `const int&`).
         // Alternatively, we could place the hint on the `auto`
         // (and then just print the type deduced for the `auto`).
-        addTypeHint(D->getLocation(), D->getType(), /*Prefix=*/": ");
+        addTypeHint(D->getLocation(), T, /*Prefix=*/": ");
       }
     }
 
+    // Show type hint for `decltype(expr)`, e.g.
+    // for `decltype(0) i;`, show `decltype(0) i: int;`
+    if (const auto *DT = llvm::dyn_cast<DecltypeType>(T))
+      addTypeHint(D->getLocation(), DT->getUnderlyingType(),
+                  /*Prefix=*/": ");
+
     // Handle templates like `int foo(auto x)` with exactly one instantiation.
     if (auto *PVD = llvm::dyn_cast<ParmVarDecl>(D)) {
       if (D->getIdentifier() && PVD->getType()->isDependentType() &&
           !getContainedAutoParamType(D->getTypeSourceInfo()->getTypeLoc())
                .isNull()) {
         if (auto *IPVD = getOnlyParamInstantiation(PVD))
           addTypeHint(D->getLocation(), IPVD->getType(), /*Prefix=*/": ");
       }

clang-tools-extra/clangd/unittests/InlayHintTests.cpp

TEST(TypeHints, DecltypeAuto) {
   assertTypeHints(R"cpp(
     int x = 42;
     int& y = x;
     decltype(auto) $z[[z]] = y;
   )cpp",
                   ExpectedHint{": int &", "z"});
 }
 
+TEST(TypeHints, Decltype) {
+  assertTypeHints("decltype(0) $i[[i]];", ExpectedHint{": int", "i"});
+}
+
 TEST(TypeHints, NoQualifiers) {
   assertTypeHints(R"cpp(
     namespace A {
       namespace B {
         struct S1 {};
         S1 foo();
         auto $x[[x]] = foo();
 
         struct S2 {
           template <typename T>
           struct Inner {};
         };
         S2::Inner<int> bar();
         auto $y[[y]] = bar();
       }
     }