[Clang] Add a value_or helper to UnsignedOrNone

[Sirraide]

This adds two value_or overloads to UnsignedOrNone. I’ve had a use for these in something I’m currently working on and thought it might make sense to add them separately. The overload that takes an UnsignedOrNone might seem a bit strange at first, but I think it makes sense given how specific this class already is.

[llvmbot]

@llvm/pr-subscribers-clang

Author: None (Sirraide)

Changes

This adds two value_or overloads to UnsignedOrNone. I’ve had a use for these in something I’m currently working on and thought it might make sense to add them separately. The overload that takes an UnsignedOrNone might seem a bit strange at first, but I think it makes sense given how specific this class already is.

---

Full diff: https://github.com/llvm/llvm-project/pull/143516.diff

1 Files Affected:

• (modified) clang/include/clang/Basic/UnsignedOrNone.h (+8)

diff --git a/clang/include/clang/Basic/UnsignedOrNone.h b/clang/include/clang/Basic/UnsignedOrNone.h
index 659fd8c6487d2..5081e3c2a9e32 100644
--- a/clang/include/clang/Basic/UnsignedOrNone.h
+++ b/clang/include/clang/Basic/UnsignedOrNone.h
@@ -29,6 +29,14 @@ struct UnsignedOrNone {
   }
   constexpr unsigned toInternalRepresentation() const { return Rep; }
 
+  constexpr unsigned value_or(unsigned val) const {
+    return operator bool() ? **this : val;
+  }
+
+  constexpr UnsignedOrNone value_or(UnsignedOrNone val) const {
+    return operator bool() ? *this : val;
+  }
+
   explicit constexpr operator bool() const { return Rep != 0; }
   unsigned operator*() const {
     assert(operator bool());

[AaronBallman]

LGTM but if the in-tree use does not materialize, we should have this on our radar for removal. It might make sense to wait to land this until you're ready to use it in-tree in a PR.

[Sirraide]

LGTM but if the in-tree use does not materialize, we should have this on our radar for removal. It might make sense to wait to land this until you're ready to use it in-tree in a PR.

Yeah, that’s a good idea

[mizvekov]

No strong objections, but I'd prefer to wait as well.

This was one thing I looked into adding when I initially implemented it, but all potential uses turned out better without it.