Implement JSON v2 for String

Also refine the herustic for deciding whether to use hexadecimal notation for a string.

Implement JSON v2 for String
Also refine the herustic for deciding whether to use hexadecimal notation for a string.
Jay Berkenbilt
1 parent 16f4f94c
Showing 9 changed files with 68 additions and 52 deletions
libqpdf/QPDF_String.cc
libqpdf/qpdf/QPDF_String.hh
qpdf/qtest/qpdf/V4-clearmeta.pdf
qpdf/qtest/qpdf/direct-pages-json-objects.out
qpdf/qtest/qpdf/direct-pages-json-pages.out
qpdf/qtest/qpdf/good14.out
qpdf/qtest/qpdf/merge-dict.out
qpdf/qtest/qpdf/page_api_2-json-objects.out
qpdf/qtest/qpdf/page_api_2-json-pages.out
@@ -45,8 +45,32 @@ QPDF_String::unparse()
 JSON
 QPDF_String::getJSON(int json_version)
 {
-    // QXXXQ
-    return JSON::makeString(getUTF8Val());
+    if (json_version == 1) {
+        return JSON::makeString(getUTF8Val());
+    }
+    // See if we can unambiguously represent as Unicode.
+    bool is_unicode = false;
+    std::string result;
+    std::string candidate = getUTF8Val();
+    if (QUtil::is_utf16(this->val) || QUtil::is_explicit_utf8(this->val)) {
+        is_unicode = true;
+        result = candidate;
+    } else if (!useHexString()) {
+        std::string test;
+        if (QUtil::utf8_to_pdf_doc(candidate, test, '?') &&
+            (test == this->val)) {
+            // This is a PDF-doc string that can be losslessly encoded
+            // as Unicode.
+            is_unicode = true;
+            result = candidate;
+        }
+    }
+    if (is_unicode) {
+        result = "u:" + result;
+    } else {
+        result = "b:" + QUtil::hex_encode(this->val);
+    }
+    return JSON::makeString(result);
 }
  
 QPDFObject::object_type_e
@@ -61,41 +85,32 @@ QPDF_String::getTypeName() const
     return "string";
 }
  
-std::string
-QPDF_String::unparse(bool force_binary)
+bool
+QPDF_String::useHexString() const
 {
-    bool use_hexstring = force_binary;
-    if (!use_hexstring) {
-        unsigned int nonprintable = 0;
-        int consecutive_printable = 0;
-        for (unsigned int i = 0; i < this->val.length(); ++i) {
-            char ch = this->val.at(i);
-            // Note: do not use locale to determine printability.  The
-            // PDF specification accepts arbitrary binary data.  Some
-            // locales imply multibyte characters.  We'll consider
-            // something printable if it is printable in 7-bit ASCII.
-            // We'll code this manually rather than being rude and
-            // setting locale.
-            if ((ch == 0) ||
-                (!(is_ascii_printable(ch) || strchr("\n\r\t\b\f", ch)))) {
-                ++nonprintable;
-                consecutive_printable = 0;
-            } else {
-                if (++consecutive_printable > 5) {
-                    // If there are more than 5 consecutive printable
-                    // characters, I want to see them as such.
-                    nonprintable = 0;
-                    break;
-                }
+    // Heuristic: use the hexadecimal representation of a string if
+    // there are any non-printable (in PDF Doc encoding) characters or
+    // if too large of a proportion of the string consists of
+    // non-ASCII characters.
+    bool nonprintable = false;
+    unsigned int non_ascii = 0;
+    for (unsigned int i = 0; i < this->val.length(); ++i) {
+        char ch = this->val.at(i);
+        if ((ch == 0) ||
+            (!(is_ascii_printable(ch) || strchr("\n\r\t\b\f", ch)))) {
+            if ((ch >= 0) && (ch < 24)) {
+                nonprintable = true;
             }
-        }
-
-        // Use hex notation if more than 20% of the characters are not
-        // printable in plain ASCII.
-        if (5 * nonprintable > val.length()) {
-            use_hexstring = true;
+            ++non_ascii;
         }
     }
+    return (nonprintable || (5 * non_ascii > val.length()));
+}
+
+std::string
+QPDF_String::unparse(bool force_binary)
+{
+    bool use_hexstring = force_binary || useHexString();
     std::string result;
     if (use_hexstring) {
         result += "<" + QUtil::hex_encode(this->val) + ">";
@@ -20,6 +20,7 @@ class QPDF_String: public QPDFObject
     std::string getUTF8Val() const;
  
   private:
+    bool useHexString() const;
     std::string val;
 };
  
@@ -65,8 +65,8 @@
     ],
     "trailer": {
       "/ID": [
-        "\u0013#¥ﬁ|WzfsU…©6ŸÎ<",
-        "7,¿DöÛ‹«`Ù&<\u000f\u000bÒj"
+        "b:1323a5937c577a66735583a93698ce3c",
+        "b:372cbf44f6db88ab60d9263c0f0bd26a"
       ],
       "/Root": "1 0 R",
       "/Size": 7
@@ -89,8 +89,8 @@
     },
     "trailer": {
       "/ID": [
-        "\u0013#¥ﬁ|WzfsU…©6ŸÎ<",
-        "7,¿DöÛ‹«`Ù&<\u000f\u000bÒj"
+        "b:1323a5937c577a66735583a93698ce3c",
+        "b:372cbf44f6db88ab60d9263c0f0bd26a"
       ],
       "/Root": "1 0 R",
       "/Size": 7
@@ -9,7 +9,7 @@ three lines
 (string with \nCRLF and\nCR and\nLF)
  and another
    indentation
-(\001B%DEF)<01>
+<014225444546><01>
 <8a8b>
 (ab)
 <8c><dd> ) >
 {
-  "/k1": "scalar1",
+  "/k1": "u:scalar1",
   "/k2": 16059,
   "/k3": {
-    "/a": "a",
-    "/b": "conflict: seen",
+    "/a": "u:a",
+    "/b": "u:conflict: seen",
     "/c": [
       2,
       3
@@ -12,7 +12,7 @@
       "/y": 25,
       "/z": 26
     },
-    "/e": "e"
+    "/e": "u:e"
   },
   "/k4": {
     "/A": 65,
@@ -24,11 +24,11 @@
   "/k5": [
     "/one",
     2,
-    "three",
+    "u:three",
     [
       "/four"
     ],
-    "two"
+    "u:two"
   ]
 }
 /A
@@ -9,8 +9,8 @@
       "/Type": "/Catalog"
     },
     "2 0 R": {
-      "/CreationDate": "D:20120621124041",
-      "/Producer": "Apex PDFWriter"
+      "/CreationDate": "u:D:20120621124041",
+      "/Producer": "u:Apex PDFWriter"
     },
     "3 0 R": {
       "/Count": 3,
@@ -77,8 +77,8 @@
     "10 0 R": 47,
     "trailer": {
       "/ID": [
-        "û˘·ƒÿ{5⁄\u0005Ú−S*º‘o",
-        "÷\u0017ž³QY¿ÔÀ\u000f\u0012−¼ý˜\u0002"
+        "b:fb18b786ff7b358705da8a532aba8f6f",
+        "b:f7179eb35159bfd4c00f128abcfd1f02"
       ],
       "/Info": "2 0 R",
       "/Root": "1 0 R",
@@ -41,8 +41,8 @@
       "/Type": "/Catalog"
     },
     "2 0 R": {
-      "/CreationDate": "D:20120621124041",
-      "/Producer": "Apex PDFWriter"
+      "/CreationDate": "u:D:20120621124041",
+      "/Producer": "u:Apex PDFWriter"
     },
     "3 0 R": {
       "/Count": 3,
@@ -129,8 +129,8 @@
     },
     "trailer": {
       "/ID": [
-        "û˘·ƒÿ{5⁄\u0005Ú−S*º‘o",
-        "÷\u0017ž³QY¿ÔÀ\u000f\u0012−¼ý˜\u0002"
+        "b:fb18b786ff7b358705da8a532aba8f6f",
+        "b:f7179eb35159bfd4c00f128abcfd1f02"
       ],
       "/Info": "2 0 R",
       "/Root": "1 0 R",