Add new SplitString backend.

base/strings/string_split.cc

Index: base/strings/string_split.cc
diff --git a/base/strings/string_split.cc b/base/strings/string_split.cc
index 88a623664fcc1c256de60c110c3186ecc0034885..3b1311ea2cc14ea7ebe82c6825d26b88e5200187 100644
--- a/base/strings/string_split.cc
+++ b/base/strings/string_split.cc
@@ -12,26 +12,91 @@ namespace base {
 
 namespace {
 
-template <typename STR>
-void SplitStringT(const STR& str,
-                  const typename STR::value_type s,
-                  bool trim_whitespace,
-                  std::vector<STR>* r) {
-  r->clear();
-  size_t last = 0;
-  size_t c = str.size();
-  for (size_t i = 0; i <= c; ++i) {
-    if (i == c || str[i] == s) {
-      STR tmp(str, last, i - last);
-      if (trim_whitespace)
-        TrimWhitespace(tmp, TRIM_ALL, &tmp);
-      // Avoid converting an empty or all-whitespace source string into a vector
-      // of one empty string.
-      if (i != c || !r->empty() || !tmp.empty())
-        r->push_back(tmp);
-      last = i + 1;
+// PieceToOutputType converts a StringPiece as needed to a given output type,
+// which is either the same type of StringPiece (a NOP) or the corresponding
+// non-piece string type.
+//
+// The default converter is a NOP, it works when the OutputType is the
+// correct StringPiece.
+template<typename Str, typename OutputType>
+OutputType PieceToOutputType(BasicStringPiece<Str> piece) {
+  return piece;
+}
+template<>  // Convert StringPiece to std::string
+std::string PieceToOutputType<std::string, std::string>(StringPiece piece) {
+  return piece.as_string();
+}
+template<>  // Convert StringPiece16 to string16.
+string16 PieceToOutputType<string16, string16>(StringPiece16 piece) {
+  return piece.as_string();
+}
+
+// Returns either the ASCII or UTF-16 whitespace.
+template<typename Str> BasicStringPiece<Str> WhitespaceForType();
+template<> StringPiece16 WhitespaceForType<string16>() {
+  return kWhitespaceUTF16;
+}
+template<> StringPiece WhitespaceForType<std::string>() {
+  return kWhitespaceASCII;
+}
+
+// Optimize the single-character case to call find() on the string instead,
+// since this is the common case and can be made faster. This could have been
+// done with template specialization too, but would have been less clear.
+//
+// There is no corresponding FindFirstNotOf because StringPiece already
+// implements these different versions that do the optimized searching.
+size_t FindFirstOf(StringPiece piece, char c, size_t pos) {
+  return piece.find(c, pos);
+}
+size_t FindFirstOf(StringPiece16 piece, char16 c, size_t pos) {
+  return piece.find(c, pos);
+}
+size_t FindFirstOf(StringPiece piece, StringPiece one_of, size_t pos) {
+  return piece.find_first_of(one_of, pos);
+}
+size_t FindFirstOf(StringPiece16 piece, StringPiece16 one_of, size_t pos) {
+  return piece.find_first_of(one_of, pos);
+}
+
+// General string splitter template. Can take 8- or 16-bit input, can produce
+// the corresponding string or StringPiece output, and can take single- or
+// multiple-character delimiters.
+//
+// DelimiterType is either a character (Str::value_type> or a string piece of

[danakj, 2015/06/11 23:58:34]

typo here with value_type>

+// multiple characters (BasicStringPiece<Str>). StringPiece has version of

[danakj, 2015/06/11 23:58:34]

has a version

+// find for both of these cases, and the single-character version is the most
+// common and can be implemented faster, which is why this is a template.
+template<typename Str, typename OutputStringType, typename DelimiterType>
+static std::vector<OutputStringType> SplitStringT(
+    BasicStringPiece<Str> str,
+    DelimiterType delimiter,
+    WhitespaceHandling whitespace,
+    SplitResult result_type) {
+  std::vector<OutputStringType> result;
+  if (str.empty())
+    return result;
+
+  size_t start = 0;
+  while (start != Str::npos) {
+    size_t end = FindFirstOf(str, delimiter, start);
+
+    BasicStringPiece<Str> piece;
+    if (end == Str::npos) {
+      piece = str.substr(start);
+      start = Str::npos;
+    } else {
+      piece = str.substr(start, end - start);
+      start = end + 1;
     }
+
+    if (whitespace == TRIM_WHITESPACE)
+      piece = TrimString(piece, WhitespaceForType<Str>(), TRIM_ALL);
+
+    if (result_type == SPLIT_WANT_ALL || !piece.empty())
+      result.push_back(PieceToOutputType<Str, OutputStringType>(piece));
   }
+  return result;
 }
 
 bool SplitStringIntoKeyValue(const std::string& line,
@@ -62,8 +127,8 @@ bool SplitStringIntoKeyValue(const std::string& line,
 
 template <typename STR>
 void SplitStringUsingSubstrT(const STR& str,
-                                    const STR& s,
-                                    std::vector<STR>* r) {
+                             const STR& s,
+                             std::vector<STR>* r) {
   r->clear();
   typename STR::size_type begin_index = 0;
   while (true) {
@@ -83,64 +148,89 @@ void SplitStringUsingSubstrT(const STR& str,
   }
 }
 
-template<typename STR>
-void SplitStringAlongWhitespaceT(const STR& str, std::vector<STR>* result) {
-  result->clear();
-  const size_t length = str.length();
-  if (!length)
-    return;
-
-  bool last_was_ws = false;
-  size_t last_non_ws_start = 0;
-  for (size_t i = 0; i < length; ++i) {
-    switch (str[i]) {
-      // HTML 5 defines whitespace as: space, tab, LF, line tab, FF, or CR.
-      case L' ':
-      case L'\t':
-      case L'\xA':
-      case L'\xB':
-      case L'\xC':
-      case L'\xD':
-        if (!last_was_ws) {
-          if (i > 0) {
-            result->push_back(
-                str.substr(last_non_ws_start, i - last_non_ws_start));
-          }
-          last_was_ws = true;
-        }
-        break;
-
-      default:  // Not a space character.
-        if (last_was_ws) {
-          last_was_ws = false;
-          last_non_ws_start = i;
-        }
-        break;
-    }
+}  // namespace
+
+std::vector<std::string> SplitString(StringPiece input,
+                                     StringPiece separators,
+                                     WhitespaceHandling whitespace,
+                                     SplitResult result_type) {
+  if (separators.size() == 1) {
+    return SplitStringT<std::string, std::string, char>(
+        input, separators[0], whitespace, result_type);
   }
-  if (!last_was_ws) {
-    result->push_back(
-        str.substr(last_non_ws_start, length - last_non_ws_start));
+  return SplitStringT<std::string, std::string, StringPiece>(
+      input, separators, whitespace, result_type);
+}
+
+std::vector<string16> SplitString(StringPiece16 input,
+                                  StringPiece16 separators,
+                                  WhitespaceHandling whitespace,
+                                  SplitResult result_type) {
+  if (separators.size() == 1) {
+    return SplitStringT<string16, string16, char16>(
+        input, separators[0], whitespace, result_type);
   }
+  return SplitStringT<string16, string16, StringPiece16>(
+      input, separators, whitespace, result_type);
 }
 
-}  // namespace
+std::vector<StringPiece> SplitStringPiece(StringPiece input,
+                                          StringPiece separators,
+                                          WhitespaceHandling whitespace,
+                                          SplitResult result_type) {
+  if (separators.size() == 1) {
+    return SplitStringT<std::string, StringPiece, char>(
+        input, separators[0], whitespace, result_type);
+  }
+  return SplitStringT<std::string, StringPiece, StringPiece>(
+      input, separators, whitespace, result_type);
+}
+
+std::vector<StringPiece16> SplitStringPiece(StringPiece16 input,
+                                            StringPiece16 separators,
+                                            WhitespaceHandling whitespace,
+                                            SplitResult result_type) {
+  if (separators.size() == 1) {
+    return SplitStringT<string16, StringPiece16, char16>(
+        input, separators[0], whitespace, result_type);
+  }
+  return SplitStringT<string16, StringPiece16, StringPiece16>(
+      input, separators, whitespace, result_type);
+}
 
 void SplitString(const string16& str,
                  char16 c,
-                 std::vector<string16>* r) {
+                 std::vector<string16>* result) {
   DCHECK(CBU16_IS_SINGLE(c));
-  SplitStringT(str, c, true, r);
+  *result = SplitStringT<string16, string16, char16>(
+      str, c, TRIM_WHITESPACE, SPLIT_WANT_ALL);
+
+  // Backward-compat hack: The old SplitString implementation would keep
+  // empty substrings, for example:
+  //    "a,,b" -> ["a", "", "b"]
+  //    "a, ,b" -> ["a", "", "b"]
+  // which the current code also does. But the old one would discard them when
+  // the only result was that empty string:
+  //    "  " -> []
+  // In the latter case, our new code will give [""]
+  if (result->size() == 1 && (*result)[0].empty())
+    result->clear();
 }
 
 void SplitString(const std::string& str,
                  char c,
-                 std::vector<std::string>* r) {
+                 std::vector<std::string>* result) {
 #if CHAR_MIN < 0
   DCHECK_GE(c, 0);
 #endif
   DCHECK_LT(c, 0x7F);
-  SplitStringT(str, c, true, r);
+  *result = SplitStringT<std::string, std::string, char>(
+      str, c, TRIM_WHITESPACE, SPLIT_WANT_ALL);
+
+  // Backward-compat hack, see above.
+  if (result->size() == 1 && (*result)[0].empty())
+    result->clear();
+
 }
 
 bool SplitStringIntoKeyValuePairs(const std::string& line,
@@ -182,31 +272,37 @@ void SplitStringUsingSubstr(const std::string& str,
   SplitStringUsingSubstrT(str, s, r);
 }
 
-void SplitStringDontTrim(const string16& str,
+void SplitStringDontTrim(StringPiece16 str,
                          char16 c,
-                         std::vector<string16>* r) {
+                         std::vector<string16>* result) {
   DCHECK(CBU16_IS_SINGLE(c));
-  SplitStringT(str, c, false, r);
+  *result = SplitStringT<string16, string16, char16>(
+      str, c, KEEP_WHITESPACE, SPLIT_WANT_ALL);
 }
 
-void SplitStringDontTrim(const std::string& str,
+void SplitStringDontTrim(StringPiece str,
                          char c,
-                         std::vector<std::string>* r) {
+                         std::vector<std::string>* result) {
 #if CHAR_MIN < 0
   DCHECK_GE(c, 0);
 #endif
   DCHECK_LT(c, 0x7F);
-  SplitStringT(str, c, false, r);
+  *result = SplitStringT<std::string, std::string, char>(
+      str, c, KEEP_WHITESPACE, SPLIT_WANT_ALL);
 }
 
 void SplitStringAlongWhitespace(const string16& str,
                                 std::vector<string16>* result) {
-  SplitStringAlongWhitespaceT(str, result);
+  *result = SplitStringT<string16, string16, StringPiece16>(
+      str, StringPiece16(kWhitespaceASCIIAs16),
+      TRIM_WHITESPACE, SPLIT_WANT_NONEMPTY);
 }
 
 void SplitStringAlongWhitespace(const std::string& str,
                                 std::vector<std::string>* result) {
-  SplitStringAlongWhitespaceT(str, result);
+  *result = SplitStringT<std::string, std::string, StringPiece>(
+      str, StringPiece(kWhitespaceASCII),
+      TRIM_WHITESPACE, SPLIT_WANT_NONEMPTY);
 }
 
 }  // namespace base