[clang] Move SPELLING_CACHE into CodeCompletion (PR #177586)

[Jannick Kremer via cfe-commits]

================
@@ -3149,6 +3073,107 @@ def __str__(self) -> str:
     VERTICAL_SPACE = 20
 
 
+class CompletionChunk:
+    class SpellingCacheAlias:
+        """
+        A temporary utility that acts as an alias to CompletionChunk.SPELLING_CACHE.
+        This will be removed without deprecation warning in a future release.
+        Please do not use it directly!
+        """
+
+        deprecation_message = (
+            "'SPELLING_CACHE' has been moved into the scope of 'CompletionChunk' "
+            "and adapted to use 'CompletionChunkKind's as keys instead of their "
+            "enum values. Please adapt all uses of 'SPELLING_CACHE' to use "
+            "'CompletionChunk.SPELLING_CACHE' instead. The old 'SPELLING_CACHE' "
+            "will be removed in a future release."
+        )
+
+        def __getitem__(self, value: int):
+            warnings.warn(self.deprecation_message, DeprecationWarning)
+            return CompletionChunk.SPELLING_CACHE[CompletionChunkKind.from_id(value)]
+        
+        def __contains__(self, value: int):
+            warnings.warn(self.deprecation_message, DeprecationWarning)
+            return CompletionChunkKind.from_id(value) in CompletionChunk.SPELLING_CACHE
----------------
DeinAlptraum wrote:

I had to add an extra class so we can put some processing into the `__getitem__` operator, for conversion and to add a deprecation warning. I'm currently only replicating the `[]` and `in` operators, but technically there are a bunch more usages that would break here, such as if you call `SPELLING_CACHE.keys()` or similar. I could cover for this as well by adding additional methods, but trying to perfectly replicate `dict()` behavior in this temporary Alias class seems like a bit of a waste of effort... what do you think?

https://github.com/llvm/llvm-project/pull/177586