[ClangImporter] Handle underscore-punctuated enum names (used by CoreMedia).

<rdar://problem/17594425>

Swift SVN r19732

Author: jrose-apple

lib/Basic/StringExtras.cpp

@@ -48,24 +48,32 @@ bool swift::isLinkingVerb(StringRef word) {
 void WordIterator::computeNextPosition() const {
   assert(Position < String.size() && "Already at end of string");
 
-  // Skip over any uppercase letters at the beginning of the word.
   unsigned i = Position, n = String.size();
+
+  // Treat _ as a word on its own. Don't coalesce.
+  if (String[i] == '_') {
+    NextPosition = i + 1;
+    NextPositionValid = true;
+    return;
+  }
+
+  // Skip over any uppercase letters at the beginning of the word.
   while (i < n && clang::isUppercase(String[i]))
     ++i;
 
   // If there was more than one uppercase letter, this is an
   // acronym.
   if (i - Position > 1) {
     // If we hit the end of the string, that's it. Otherwise, this
-    // word ends at the last uppercase letter, so that the next word
-    // starts with the last uppercase letter.
+    // word ends before the last uppercase letter if the next word is alphabetic
+    // (URL_Loader) or after the last uppercase letter if it's not (UTF_8).
     NextPosition = (i == n || !clang::isLowercase(String[i])) ? i : i-1;
     NextPositionValid = true;
     return;
   }
 
   // Skip non-uppercase letters.
-  while (i < n && !clang::isUppercase(String[i]))
+  while (i < n && !clang::isUppercase(String[i]) && String[i] != '_')
     ++i;
 
   NextPosition = i;
@@ -75,18 +83,28 @@ void WordIterator::computeNextPosition() const {
 void WordIterator::computePrevPosition() const {
   assert(Position > 0 && "Already at beginning of string");
 
-  // While we see non-uppercase letters, keep moving back.
   unsigned i = Position;
-  while (i > 0 && !clang::isUppercase(String[i-1]))
+
+  // While we see non-uppercase letters, keep moving back.
+  while (i > 0 && !clang::isUppercase(String[i-1]) && String[i-1] != '_')
     --i;
 
   // If we found any lowercase letters, this was a normal camel case
   // word (not an acronym).
   if (i < Position) {
-    // If we hit the beginning of the string, that's it. Otherwise,
-    // this word starts at the uppercase letter that terminated the
-    // search above.
-    PrevPosition = (i == 0 || !clang::isLowercase(String[i])) ? i : i-1;
+    // If we hit the beginning of the string, that's it. Otherwise, this
+    // word starts with an uppercase letter if the next word is alphabetic
+    // (URL_Loader) or after the last uppercase letter if it's not (UTF_8).
+    PrevPosition = i;
+    if (i != 0 && clang::isLowercase(String[i]) && String[i-1] != '_')
+      --PrevPosition;
+    PrevPositionValid = true;
+    return;
+  }
+
+  // Treat _ as a word on its own. Don't coalesce.
+  if (String[i-1] == '_') {
+    PrevPosition = i - 1;
     PrevPositionValid = true;
     return;
   }

lib/ClangImporter/ImportDecl.cpp

@@ -1317,18 +1317,21 @@ namespace {
       if (!commonPrefix.empty()) {
         StringRef checkPrefix = commonPrefix;
 
+        // Account for the 'EnumName_Constant' convention on enumerators.
+        if (checkPrefix.back() == '_' && !followedByNonIdentifier)
+          checkPrefix = checkPrefix.drop_back();
+
         // Account for the 'kConstant' naming convention on enumerators.
-        bool dropKPrefix = false;
         if (checkPrefix[0] == 'k' &&
             ((checkPrefix.size() >= 2 && clang::isUppercase(checkPrefix[1])) ||
              !followedByNonIdentifier)) {
-          checkPrefix = checkPrefix.substr(1);
-          dropKPrefix = true;
+          checkPrefix = checkPrefix.drop_front();
         }
 
         StringRef commonWithEnum = getCommonPluralPrefix(checkPrefix,
                                                          enumName.str());
-        commonPrefix = commonPrefix.slice(0, commonWithEnum.size()+dropKPrefix);
+        size_t delta = commonPrefix.size() - checkPrefix.size();
+        commonPrefix = commonPrefix.slice(0, commonWithEnum.size() + delta);
       }
       Impl.EnumConstantNamePrefixes.insert({decl, commonPrefix});
     }

test/ClangModules/objc_ns_options.swift

@@ -40,3 +40,7 @@ let bitmapFormat2: NSBitmapFormat2 = .NSU16a | .NSU32a;
 let bitmapFormat3: NSBitmapFormat3 = .NSU16b | .NSS32b;
 let bitmapFormat4: NSUBitmapFormat4 = .NSU16c | .NSU32c;
 let bitmapFormat5: NSABitmapFormat5 = .NSAA16d | .NSAB32d;
+
+// Drop trailing underscores when possible.
+let timeFlags: CMTimeFlags = .Valid | .HasBeenRounded
+let timeFlags2: CMTimeFlagsWithNumber = ._Valid | ._888

test/Inputs/clang-importer-sdk/usr/include/Foundation.h

@@ -269,6 +269,21 @@ typedef NS_OPTIONS(NSInteger, CBCharacteristicProperties) {
   CBCharacteristicPropertyIndicateEncryptionRequired /*NS_ENUM_AVAILABLE(10_9, 6_0)*/ = 0x200
 };
 
+// From CoreMedia
+typedef CF_OPTIONS(unsigned int, CMTimeFlags) {
+  kCMTimeFlags_Valid = 1UL<<0,
+  kCMTimeFlags_HasBeenRounded = 1UL<<1,
+  kCMTimeFlags_PositiveInfinity = 1UL<<2,
+  kCMTimeFlags_NegativeInfinity = 1UL<<3,
+  kCMTimeFlags_Indefinite = 1UL<<4,
+  kCMTimeFlags_ImpliedValueFlagsMask = kCMTimeFlags_PositiveInfinity | kCMTimeFlags_NegativeInfinity | kCMTimeFlags_Indefinite
+};
+typedef CF_OPTIONS(unsigned int, CMTimeFlagsWithNumber) {
+  kCMTimeFlagsWithNumber_Valid = 1UL<<0,
+  kCMTimeFlagsWithNumber_888 = 1UL<<1,
+};
+
+
 // Contrived name with a plural "-es"...normally these are "beeps".
 typedef NS_OPTIONS(NSInteger, AlertBuzzes) {
   AlertBuzzFunk,

unittests/Basic/StringExtrasTest.cpp

@@ -56,6 +56,36 @@ TEST(CamelCaseWordsTest, Iteration) {
   EXPECT_EQ(iter, words.end());
 }
 
+TEST(CamelCaseWordsTest, WordsWithUnderscores) {
+  auto words = camel_case::getWords("CF_Flags_789");
+  EXPECT_EQ(5, std::distance(words.begin(), words.end()));
+
+  auto iter = words.begin();
+  EXPECT_EQ("CF", *iter++);
+  EXPECT_EQ("_", *iter++);
+  EXPECT_EQ("Flags", *iter++);
+  EXPECT_EQ("_", *iter++);
+  EXPECT_EQ("789", *iter++);
+  EXPECT_EQ("789", *--iter);
+  EXPECT_EQ("_", *--iter);
+  EXPECT_EQ("Flags", *--iter);
+  EXPECT_EQ("_", *--iter);
+  EXPECT_EQ("CF", *--iter);
+
+  auto manyUnderscores = camel_case::getWords("___ABC");
+  EXPECT_EQ(4, std::distance(manyUnderscores.begin(), manyUnderscores.end()));
+
+  iter = manyUnderscores.begin();
+  EXPECT_EQ("_", *iter++);
+  EXPECT_EQ("_", *iter++);
+  EXPECT_EQ("_", *iter++);
+  EXPECT_EQ("ABC", *iter++);
+  EXPECT_EQ("ABC", *--iter);
+  EXPECT_EQ("_", *--iter);
+  EXPECT_EQ("_", *--iter);
+  EXPECT_EQ("_", *--iter);
+}
+
 TEST(ToLowercaseTest, Words) {
   llvm::SmallString<64> scratch;