[cfe-commits] [PATCH] Allow RefactoringTool to write to memory instead of always to disk

[Edwin Vane]

================
Comment at: lib/Tooling/Refactoring.cpp:142
@@ +141,3 @@
+// saveRewrittenFiles() and getResults(), without cluttering Refactoring.h's
+// header with includes that are really only needed by the implementation.
+class RewriteHelper {
----------------
Edwin Vane wrote:
> Manuel Klimek wrote:
> > I think not putting something into the header is not a good reason to pull out an abstraction. I also generally don't like abstractions called "Helper" (I understand that you don't intend this to be an abstraction, but my point is that if there's not abstraction, I don't think it makes sense to pull one out just to reduce stuff in the header).
> Alright. I'm still new to this and take what I read in the LLVM coding standards at face value. I'm referring to the "include as little as possible". But even from a design standpoint, is it really necessary to expose the types that are really only needed by the implementation to every user of Refactoring.h?
With the three-method approach this point is moot but I'm still interested in the answer in general.


http://llvm-reviews.chandlerc.com/D273