Add JSON-specific escaping, which has different rules from JS escaping.BUG=http://crbug.com/11431TEST=base_unittests.exe --gtest_filter=StringEscapeTest.Json*

Review URL: http://codereview.chromium.org/113606

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@16485 0039d316-1c4b-4281-b951-d872f2087c98

6 files changed, 161 insertions, 146 deletions

diff --git a/base/json_writer.cc b/base/json_writer.cc
index aa66306..a95798e 100644
--- a/base/json_writer.cc
+++ b/base/json_writer.cc
@@ -92,16 +92,15 @@ void JSONWriter::BuildJSONString(const Value* const node,
 
     case Value::TYPE_STRING:
       {
+        std::string value;
+        bool result = node->GetAsString(&value);
+        DCHECK(result);
         if (escape) {
-          std::wstring value;
-          bool result = node->GetAsString(&value);
-          DCHECK(result);
-          AppendQuotedString(value);
+          string_escape::JsonDoubleQuote(UTF8ToUTF16(value),
+                                         true,
+                                         json_string_);
         } else {
-          std::string value;
-          bool result = node->GetAsString(&value);
-          DCHECK(result);
-          string_escape::JavascriptDoubleQuote(value, true, json_string_);
+          string_escape::JsonDoubleQuote(value, true, json_string_);
         }
         break;
       }
@@ -182,9 +181,9 @@ void JSONWriter::BuildJSONString(const Value* const node,
 }
 
 void JSONWriter::AppendQuotedString(const std::wstring& str) {
-  string_escape::JavascriptDoubleQuote(WideToUTF16Hack(str),
-                                       true,
-                                       json_string_);
+  string_escape::JsonDoubleQuote(WideToUTF16Hack(str),
+                                 true,
+                                 json_string_);
 }
 
 void JSONWriter::IndentLine(int depth) {
diff --git a/base/string_escape.cc b/base/string_escape.cc
index aafee11..0fecfa1 100644
--- a/base/string_escape.cc
+++ b/base/string_escape.cc
@@ -11,10 +11,13 @@
 namespace string_escape {
 
 // Try to escape |c| as a "SingleEscapeCharacter" (\n, etc).  If successful,
-// returns true and appends the escape sequence to |dst|.
+// returns true and appends the escape sequence to |dst|.  This isn't required
+// by the spec, but it's more readable by humans than the \uXXXX alternatives.
 template<typename CHAR>
-static bool JavascriptSingleEscapeChar(const CHAR c, std::string* dst) {
+static bool JsonSingleEscapeChar(const CHAR c, std::string* dst) {
   // WARNING: if you add a new case here, you need to update the reader as well.
+  // Note: \v is in the reader, but not here since the JSON spec doesn't
+  // allow it.
   switch (c) {
     case '\b':
       dst->append("\\b");
@@ -31,9 +34,6 @@ static bool JavascriptSingleEscapeChar(const CHAR c, std::string* dst) {
     case '\t':
       dst->append("\\t");
       break;
-    case '\v':
-      dst->append("\\v");
-      break;
     case '\\':
       dst->append("\\\\");
       break;
@@ -46,25 +46,24 @@ static bool JavascriptSingleEscapeChar(const CHAR c, std::string* dst) {
   return true;
 }
 
-void JavascriptDoubleQuote(const string16& str,
-                           bool put_in_quotes,
-                           std::string* dst) {
+template <class STR>
+void JsonDoubleQuoteT(const STR& str,
+                      bool put_in_quotes,
+                      std::string* dst) {
   if (put_in_quotes)
     dst->push_back('"');
 
-  for (string16::const_iterator it = str.begin(); it != str.end(); ++it) {
-    char16 c = *it;
-    if (!JavascriptSingleEscapeChar(c, dst)) {
-      if (c > 255) {
-        // Non-ascii values need to be unicode dst->
-        // TODO(tc): Some unicode values are handled specially. See
-        // spidermonkey code.
-        StringAppendF(dst, "\\u%04X", c);
-      } else if (c < 32 || c > 126) {
-        // Spidermonkey hex escapes these values.
-        StringAppendF(dst, "\\x%02X", c);
+  for (typename STR::const_iterator it = str.begin(); it != str.end(); ++it) {
+    typename ToUnsigned<typename STR::value_type>::Unsigned c = *it;
+    if (!JsonSingleEscapeChar(c, dst)) {
+      if (c < 32 || c > 126) {
+        // Technically, we could also pass through c > 126 as UTF8, but this is
+        // also optional.  It would also be a pain to implement here.
+        unsigned int as_uint = static_cast<unsigned int>(c);
+        StringAppendF(dst, "\\u%04X", as_uint);
       } else {
-        dst->push_back(static_cast<char>(c));
+        unsigned char ascii = static_cast<unsigned char>(*it);
+        dst->push_back(ascii);
       }
     }
   }
@@ -73,26 +72,16 @@ void JavascriptDoubleQuote(const string16& str,
     dst->push_back('"');
 }
 
-void JavascriptDoubleQuote(const std::string& str,
-                           bool put_in_quotes,
-                           std::string* dst) {
-  if (put_in_quotes)
-    dst->push_back('"');
-
-  for (std::string::const_iterator it = str.begin(); it != str.end(); ++it) {
-    unsigned char c = *it;
-    if (!JavascriptSingleEscapeChar(c, dst)) {
-      // Hex encode if the character is non-printable 7bit ascii
-      if (c < 32 || c == 127) {
-        StringAppendF(dst, "\\x%02X", c);
-      } else {
-        dst->push_back(static_cast<char>(c));
-      }
-    }
-  }
+void JsonDoubleQuote(const std::string& str,
+                     bool put_in_quotes,
+                     std::string* dst) {
+  JsonDoubleQuoteT(str, put_in_quotes, dst);
+}
 
-  if (put_in_quotes)
-    dst->push_back('"');
+void JsonDoubleQuote(const string16& str,
+                     bool put_in_quotes,
+                     std::string* dst) {
+  JsonDoubleQuoteT(str, put_in_quotes, dst);
 }
 
 }  // namespace string_escape
diff --git a/base/string_escape.h b/base/string_escape.h
index 9c6233a..3f0bf3f 100644
--- a/base/string_escape.h
+++ b/base/string_escape.h
@@ -7,28 +7,25 @@
 #ifndef BASE_STRING_ESCAPE_H__
 #define BASE_STRING_ESCAPE_H__
 
+#include <string>
+
 #include "base/string16.h"
 
 namespace string_escape {
 
-// Escape |str| appropriately for a javascript string litereal, _appending_ the
-// result to |dst|. This will create standard escape sequences (\b, \n),
-// hex escape sequences (\x00), and unicode escape sequences (\uXXXX).
+// Escape |str| appropriately for a JSON string litereal, _appending_ the
+// result to |dst|. This will create unicode escape sequences (\uXXXX).
 // If |put_in_quotes| is true, the result will be surrounded in double quotes.
 // The outputted literal, when interpreted by the browser, should result in a
 // javascript string that is identical and the same length as the input |str|.
-void JavascriptDoubleQuote(const string16& str,
-                           bool put_in_quotes,
-                           std::string* dst);
-
-// Similar to the wide version, but for narrow strings.  It will not use
-// \uXXXX unicode escape sequences.  It will pass non-7bit characters directly
-// into the string unencoded, allowing the browser to interpret the encoding.
-// The outputted literal, when interpreted by the browser, could result in a
-// javascript string of a different length than the input |str|.
-void JavascriptDoubleQuote(const std::string& str,
-                           bool put_in_quotes,
-                           std::string* dst);
+void JsonDoubleQuote(const std::string& str,
+                     bool put_in_quotes,
+                     std::string* dst);
+
+void JsonDoubleQuote(const string16& str,
+                     bool put_in_quotes,
+                     std::string* dst);
+
 
 }  // namespace string_escape
 
diff --git a/base/string_escape_unittest.cc b/base/string_escape_unittest.cc
index e6a1806..d731dc1 100644
--- a/base/string_escape_unittest.cc
+++ b/base/string_escape_unittest.cc
@@ -6,59 +6,90 @@
 #include "base/string_escape.h"
 #include "base/string_util.h"
 
-TEST(StringEscapeTest, JavascriptDoubleQuote) {
-  static const char* kToEscape          = "\b\001aZ\"\\wee";
-  static const char* kEscaped           = "\\b\\x01aZ\\\"\\\\wee";
-  static const char* kEscapedQuoted     = "\"\\b\\x01aZ\\\"\\\\wee\"";
-  static const wchar_t* kUToEscape      = L"\b\x0001" L"a\x123fZ\"\\wee";
-  static const char* kUEscaped          = "\\b\\x01a\\u123FZ\\\"\\\\wee";
-  static const char* kUEscapedQuoted    = "\"\\b\\x01a\\u123FZ\\\"\\\\wee\"";
+namespace {
 
+const struct json_narrow_test_data {
+  const char* to_escape;
+  const char* escaped;
+} json_narrow_cases[] = {
+  {"\b\001aZ\"\\wee", "\\b\\u0001aZ\\\"\\\\wee"},
+  {"a\b\f\n\r\t\v\1\\.\"z",
+      "a\\b\\f\\n\\r\\t\\u000B\\u0001\\\\.\\\"z"},
+  {"b\x0f\x7f\xf0\xff!", "b\\u000F\\u007F\\u00F0\\u00FF!"},
+};
+
+}
+
+TEST(StringEscapeTest, JsonDoubleQuoteNarrow) {
+  for (size_t i = 0; i < arraysize(json_narrow_cases); ++i) {
+    std::string in = json_narrow_cases[i].to_escape;
+    std::string out;
+    string_escape::JsonDoubleQuote(in, false, &out);
+    EXPECT_EQ(std::string(json_narrow_cases[i].escaped), out);
+  }
+
+  std::string in = json_narrow_cases[0].to_escape;
+  std::string out;
+  string_escape::JsonDoubleQuote(in, false, &out);
+
+  // test quoting
+  std::string out_quoted;
+  string_escape::JsonDoubleQuote(in, true, &out_quoted);
+  EXPECT_EQ(out.length() + 2, out_quoted.length());
+  EXPECT_EQ(out_quoted.find(out), 1U);
+
+  // now try with a NULL in the string
+  std::string null_prepend = "test";
+  null_prepend.push_back(0);
+  in = null_prepend + in;
+  std::string expected = "test\\u0000";
+  expected += json_narrow_cases[0].escaped;
+  out.clear();
+  string_escape::JsonDoubleQuote(in, false, &out);
+  EXPECT_EQ(expected, out);
+}
+
+namespace {
+
+const struct json_wide_test_data {
+  const wchar_t* to_escape;
+  const char* escaped;
+} json_wide_cases[] = {
+  {L"b\uffb1\u00ff", "b\\uFFB1\\u00FF"},
+  {L"\b\001aZ\"\\wee", "\\b\\u0001aZ\\\"\\\\wee"},
+  {L"a\b\f\n\r\t\v\1\\.\"z",
+      "a\\b\\f\\n\\r\\t\\u000B\\u0001\\\\.\\\"z"},
+  {L"b\x0f\x7f\xf0\xff!", "b\\u000F\\u007F\\u00F0\\u00FF!"},
+};
+
+}
+
+TEST(StringEscapeTest, JsonDoubleQuoteWide) {
+
+  for (size_t i = 0; i < arraysize(json_wide_cases); ++i) {
+    std::string out;
+    string16 in = WideToUTF16(json_wide_cases[i].to_escape);
+    string_escape::JsonDoubleQuote(in, false, &out);
+    EXPECT_EQ(std::string(json_wide_cases[i].escaped), out);
+  }
+
+  string16 in = WideToUTF16(json_wide_cases[0].to_escape);
   std::string out;
+  string_escape::JsonDoubleQuote(in, false, &out);
+
+  // test quoting
+  std::string out_quoted;
+  string_escape::JsonDoubleQuote(in, true, &out_quoted);
+  EXPECT_EQ(out.length() + 2, out_quoted.length());
+  EXPECT_EQ(out_quoted.find(out), 1U);
 
-  // Test wide unicode escaping
-  out = "testy: ";
-  string_escape::JavascriptDoubleQuote(WideToUTF16(kUToEscape), false, &out);
-  ASSERT_EQ(std::string("testy: ") + kUEscaped, out);
-
-  out = "testy: ";
-  string_escape::JavascriptDoubleQuote(WideToUTF16(kUToEscape), true, &out);
-  ASSERT_EQ(std::string("testy: ") + kUEscapedQuoted, out);
-
-  // Test null and high bit / negative unicode values
-  string16 str16 = UTF8ToUTF16("TeSt");
-  str16.push_back(0);
-  str16.push_back(0xffb1);
-  str16.push_back(0x00ff);
-
-  out = "testy: ";
-  string_escape::JavascriptDoubleQuote(str16, false, &out);
-  ASSERT_EQ("testy: TeSt\\x00\\uFFB1\\xFF", out);
-
-  // Test escaping of 7bit ascii
-  out = "testy: ";
-  string_escape::JavascriptDoubleQuote(std::string(kToEscape), false, &out);
-  ASSERT_EQ(std::string("testy: ") + kEscaped, out);
-
-  out = "testy: ";
-  string_escape::JavascriptDoubleQuote(std::string(kToEscape), true, &out);
-  ASSERT_EQ(std::string("testy: ") + kEscapedQuoted, out);
-
-  // Test null, non-printable, and non-7bit
-  std::string str("TeSt");
-  str.push_back(0);
-  str.push_back(15);
-  str.push_back(127);
-  str.push_back(-16);
-  str.push_back(-128);
-  str.push_back('!');
-
-  out = "testy: ";
-  string_escape::JavascriptDoubleQuote(str, false, &out);
-  ASSERT_EQ("testy: TeSt\\x00\\x0F\\x7F\xf0\x80!", out);
-
-  // Test escape sequences
-  out = "testy: ";
-  string_escape::JavascriptDoubleQuote("a\b\f\n\r\t\v\1\\.\"z", false, &out);
-  ASSERT_EQ("testy: a\\b\\f\\n\\r\\t\\v\\x01\\\\.\\\"z", out);
+  // now try with a NULL in the string
+  string16 null_prepend = WideToUTF16(L"test");
+  null_prepend.push_back(0);
+  in = null_prepend + in;
+  std::string expected = "test\\u0000";
+  expected += json_wide_cases[0].escaped;
+  out.clear();
+  string_escape::JsonDoubleQuote(in, false, &out);
+  EXPECT_EQ(expected, out);
 }
diff --git a/base/string_util.cc b/base/string_util.cc
index 5923c10..c666960 100644
--- a/base/string_util.cc
+++ b/base/string_util.cc
@@ -37,35 +37,6 @@ struct EmptyStrings {
   const string16 s16;
 };
 
-// Hack to convert any char-like type to its unsigned counterpart.
-// For example, it will convert char, signed char and unsigned char to unsigned
-// char.
-template<typename T>
-struct ToUnsigned {
-  typedef T Unsigned;
-};
-
-template<>
-struct ToUnsigned<char> {
-  typedef unsigned char Unsigned;
-};
-template<>
-struct ToUnsigned<signed char> {
-  typedef unsigned char Unsigned;
-};
-template<>
-struct ToUnsigned<wchar_t> {
-#if defined(WCHAR_T_IS_UTF16)
-  typedef unsigned short Unsigned;
-#elif defined(WCHAR_T_IS_UTF32)
-  typedef uint32 Unsigned;
-#endif
-};
-template<>
-struct ToUnsigned<short> {
-  typedef unsigned short Unsigned;
-};
-
 // Used by ReplaceStringPlaceholders to track the position in the string of
 // replaced parameters.
 struct ReplacementOffset {
diff --git a/base/string_util.h b/base/string_util.h
index cb71553..f40cc8d 100644
--- a/base/string_util.h
+++ b/base/string_util.h
@@ -595,5 +595,33 @@ bool MatchPattern(const std::string& string, const std::string& pattern);
 //   std::numeric_limits<size_t>::max() / 2
 std::string HexEncode(const void* bytes, size_t size);
 
+// Hack to convert any char-like type to its unsigned counterpart.
+// For example, it will convert char, signed char and unsigned char to unsigned
+// char.
+template<typename T>
+struct ToUnsigned {
+  typedef T Unsigned;
+};
+
+template<>
+struct ToUnsigned<char> {
+  typedef unsigned char Unsigned;
+};
+template<>
+struct ToUnsigned<signed char> {
+  typedef unsigned char Unsigned;
+};
+template<>
+struct ToUnsigned<wchar_t> {
+#if defined(WCHAR_T_IS_UTF16)
+  typedef unsigned short Unsigned;
+#elif defined(WCHAR_T_IS_UTF32)
+  typedef uint32 Unsigned;
+#endif
+};
+template<>
+struct ToUnsigned<short> {
+  typedef unsigned short Unsigned;
+};
 
 #endif  // BASE_STRING_UTIL_H_