Squashed 'src/univalue/' changes from 2740c4f..f32df99

f32df99 Merge branch '2016_04_unicode' into bitcoin
280b191 Merge remote-tracking branch 'jgarzik/master' into bitcoin
c9a716c Handle UTF-8
bed8dd9 Version 1.0.2.
5e7985a Merge pull request #14 from laanwj/2015_11_escape_plan

git-subtree-dir: src/univalue
git-subtree-split: f32df99e96d99ab49e5eeda16cac93747d388245

3 files changed, 135 insertions, 32 deletions

diff --git a/lib/univalue_read.cpp b/lib/univalue_read.cpp
index c7516b9628..95bac6958d 100644
--- a/lib/univalue_read.cpp
+++ b/lib/univalue_read.cpp
@@ -6,6 +6,7 @@
 #include <vector>
 #include <stdio.h>
 #include "univalue.h"
+#include "univalue_utffilter.h"
 
 using namespace std;
 
@@ -174,41 +175,31 @@ enum jtokentype getJsonToken(string& tokenVal, unsigned int& consumed,
         raw++;                                // skip "
 
         string valStr;
+        JSONUTF8StringFilter writer(valStr);
 
         while (*raw) {
-            if (*raw < 0x20)
+            if ((unsigned char)*raw < 0x20)
                 return JTOK_ERR;
 
             else if (*raw == '\\') {
                 raw++;                        // skip backslash
 
                 switch (*raw) {
-                case '"':  valStr += "\""; break;
-                case '\\': valStr += "\\"; break;
-                case '/':  valStr += "/"; break;
-                case 'b':  valStr += "\b"; break;
-                case 'f':  valStr += "\f"; break;
-                case 'n':  valStr += "\n"; break;
-                case 'r':  valStr += "\r"; break;
-                case 't':  valStr += "\t"; break;
+                case '"':  writer.push_back('\"'); break;
+                case '\\': writer.push_back('\\'); break;
+                case '/':  writer.push_back('/'); break;
+                case 'b':  writer.push_back('\b'); break;
+                case 'f':  writer.push_back('\f'); break;
+                case 'n':  writer.push_back('\n'); break;
+                case 'r':  writer.push_back('\r'); break;
+                case 't':  writer.push_back('\t'); break;
 
                 case 'u': {
                     unsigned int codepoint;
                     if (hatoui(raw + 1, raw + 1 + 4, codepoint) !=
                                raw + 1 + 4)
                         return JTOK_ERR;
-
-                    if (codepoint <= 0x7f)
-                        valStr.push_back((char)codepoint);
-                    else if (codepoint <= 0x7FF) {
-                        valStr.push_back((char)(0xC0 | (codepoint >> 6)));
-                        valStr.push_back((char)(0x80 | (codepoint & 0x3F)));
-                    } else if (codepoint <= 0xFFFF) {
-                        valStr.push_back((char)(0xE0 | (codepoint >> 12)));
-                        valStr.push_back((char)(0x80 | ((codepoint >> 6) & 0x3F)));
-                        valStr.push_back((char)(0x80 | (codepoint & 0x3F)));
-                    }
-
+                    writer.push_back_u(codepoint);
                     raw += 4;
                     break;
                     }
@@ -226,11 +217,13 @@ enum jtokentype getJsonToken(string& tokenVal, unsigned int& consumed,
             }
 
             else {
-                valStr += *raw;
+                writer.push_back(*raw);
                 raw++;
             }
         }
 
+        if (!writer.finalize())
+            return JTOK_ERR;
         tokenVal = valStr;
         consumed = (raw - rawStart);
         return JTOK_STRING;
diff --git a/lib/univalue_utffilter.h b/lib/univalue_utffilter.h
new file mode 100644
index 0000000000..0e330dce9c
--- /dev/null
+++ b/lib/univalue_utffilter.h
@@ -0,0 +1,119 @@
+// Copyright 2016 Wladimir J. van der Laan
+// Distributed under the MIT software license, see the accompanying
+// file COPYING or http://www.opensource.org/licenses/mit-license.php.
+#ifndef UNIVALUE_UTFFILTER_H
+#define UNIVALUE_UTFFILTER_H
+
+#include <string>
+
+/**
+ * Filter that generates and validates UTF-8, as well as collates UTF-16
+ * surrogate pairs as specified in RFC4627.
+ */
+class JSONUTF8StringFilter
+{
+public:
+    JSONUTF8StringFilter(std::string &s):
+        str(s), is_valid(true), codepoint(0), state(0), surpair(0)
+    {
+    }
+    // Write single 8-bit char (may be part of UTF-8 sequence)
+    void push_back(unsigned char ch)
+    {
+        if (state == 0) {
+            if (ch < 0x80) // 7-bit ASCII, fast direct pass-through
+                str.push_back(ch);
+            else if (ch < 0xc0) // Mid-sequence character, invalid in this state
+                is_valid = false;
+            else if (ch < 0xe0) { // Start of 2-byte sequence
+                codepoint = (ch & 0x1f) << 6;
+                state = 6;
+            } else if (ch < 0xf0) { // Start of 3-byte sequence
+                codepoint = (ch & 0x0f) << 12;
+                state = 12;
+            } else if (ch < 0xf8) { // Start of 4-byte sequence
+                codepoint = (ch & 0x07) << 18;
+                state = 18;
+            } else // Reserved, invalid
+                is_valid = false;
+        } else {
+            if ((ch & 0xc0) != 0x80) // Not a continuation, invalid
+                is_valid = false;
+            state -= 6;
+            codepoint |= (ch & 0x3f) << state;
+            if (state == 0)
+                push_back_u(codepoint);
+        }
+    }
+    // Write codepoint directly, possibly collating surrogate pairs
+    void push_back_u(unsigned int codepoint)
+    {
+        if (state) // Only accept full codepoints in open state
+            is_valid = false;
+        if (codepoint >= 0xD800 && codepoint < 0xDC00) { // First half of surrogate pair
+            if (surpair) // Two subsequent surrogate pair openers - fail
+                is_valid = false;
+            else
+                surpair = codepoint;
+        } else if (codepoint >= 0xDC00 && codepoint < 0xE000) { // Second half of surrogate pair
+            if (surpair) { // Open surrogate pair, expect second half
+                // Compute code point from UTF-16 surrogate pair
+                append_codepoint(0x10000 | ((surpair - 0xD800)<<10) | (codepoint - 0xDC00));
+                surpair = 0;
+            } else // Second half doesn't follow a first half - fail
+                is_valid = false;
+        } else {
+            if (surpair) // First half of surrogate pair not followed by second - fail
+                is_valid = false;
+            else
+                append_codepoint(codepoint);
+        }
+    }
+    // Check that we're in a state where the string can be ended
+    // No open sequences, no open surrogate pairs, etc
+    bool finalize()
+    {
+        if (state || surpair)
+            is_valid = false;
+        return is_valid;
+    }
+private:
+    std::string &str;
+    bool is_valid;
+    // Current UTF-8 decoding state
+    unsigned int codepoint;
+    int state; // Top bit to be filled in for next UTF-8 byte, or 0
+
+    // Keep track of the following state to handle the following section of
+    // RFC4627:
+    //
+    //    To escape an extended character that is not in the Basic Multilingual
+    //    Plane, the character is represented as a twelve-character sequence,
+    //    encoding the UTF-16 surrogate pair.  So, for example, a string
+    //    containing only the G clef character (U+1D11E) may be represented as
+    //    "\uD834\uDD1E".
+    //
+    //  Two subsequent \u.... may have to be replaced with one actual codepoint.
+    unsigned int surpair; // First half of open UTF-16 surrogate pair, or 0
+
+    void append_codepoint(unsigned int codepoint)
+    {
+        if (codepoint <= 0x7f)
+            str.push_back((char)codepoint);
+        else if (codepoint <= 0x7FF) {
+            str.push_back((char)(0xC0 | (codepoint >> 6)));
+            str.push_back((char)(0x80 | (codepoint & 0x3F)));
+        } else if (codepoint <= 0xFFFF) {
+            str.push_back((char)(0xE0 | (codepoint >> 12)));
+            str.push_back((char)(0x80 | ((codepoint >> 6) & 0x3F)));
+            str.push_back((char)(0x80 | (codepoint & 0x3F)));
+        } else if (codepoint <= 0x1FFFFF) {
+            str.push_back((char)(0xF0 | (codepoint >> 18)));
+            str.push_back((char)(0x80 | ((codepoint >> 12) & 0x3F)));
+            str.push_back((char)(0x80 | ((codepoint >> 6) & 0x3F)));
+            str.push_back((char)(0x80 | (codepoint & 0x3F)));
+        }
+    }
+};
+
+#endif
diff --git a/lib/univalue_write.cpp b/lib/univalue_write.cpp
index ceb4cc9166..cfbdad3284 100644
--- a/lib/univalue_write.cpp
+++ b/lib/univalue_write.cpp
@@ -8,8 +8,6 @@
 #include "univalue.h"
 #include "univalue_escapes.h"
 
-// TODO: Using UTF8
-
 using namespace std;
 
 static string json_escape(const string& inS)
@@ -23,15 +21,8 @@ static string json_escape(const string& inS)
 
         if (escStr)
             outS += escStr;
-
-        else if (ch < 0x80)
+        else
             outS += ch;
-
-        else { // TODO handle UTF-8 properly
-            char tmpesc[16];
-            sprintf(tmpesc, "\\u%04x", ch);
-            outS += tmpesc;
-        }
     }
 
     return outS;