[PATCH] D41537: Optionally add code completion results for arrow instead of dot

[Ivan Donchevskii via Phabricator via cfe-commits]

yvvan added a comment.

In https://reviews.llvm.org/D41537#1097763, @ilya-biryukov wrote:

> I should've suggested splitting the change into two earlier.


Next time I will do that from the beginning :)



================
Comment at: tools/libclang/CIndexCodeCompletion.cpp:309
+  /// before that result for the corresponding completion item.
+  std::vector<std::vector<FixItHint>> FixItsVector;
 };
----------------
ilya-biryukov wrote:
> Storing `vector<vector<>>` here makes looks like a hack. Even though it might seem more tricky, I suggest storing an opaque pointer to `vector<FixItHint>` in each `CXCompletionResult`.  Managing the lifetime of vectors in the `AllocatedCXCodeCompleteResults` seems totally fine, but there should be a way to get to the fixits in a similar way we can get to the completion string.
> More concretely, I suggest the following API:
> ```
> // === Index.h
> typedef void* CXCompletionFixIts;
> typedef struct {
>    // ...
>    CXCompletionFixIts FixIts;
> };
> 
> // Get the number of fix-its.
> unsigned clang_getCompletionNumFixIts(CXCompletionResult *Result);
> // ... Similar to getDiagnosticFixIt
> CXString clang_getCompletionFixIt((CXCompletionResult *Result, unsigned fixit_index, CXSourceRange *replacement_range);
> 
> 
> 
> // === Impl.cpp
> struct AllocatedCXCodeCompleteResults : public CXCodeCompleteResults {
> // .....
> // Pool for allocating non-empty fixit vectors in the CXCompletionResult.
> std::vector<std::vector<FixItHint>> FixItsVector
> };
> 
> unsigned clang_getCompletionNumFixIts(CXCompletionResult *Result) {
>   auto* FixIts = static_cast<std::vector<FixItHint>*>(Result->FixIts);
>   if (!FixIts)
>     return 0;
>   return FixIts->size();
> }
> ```
unsigned clang_getCompletionNumFixIts(CXCompletionResult *Result);

Do you mean appending CXCompletionResult struct with the "CXCompletionFixIts FixIts" member? Doesn't it break binary compatibility (changing the struct size)?


https://reviews.llvm.org/D41537