Add missing characters from PDF doc encoding (fixes #606)

Jay Berkenbilt · Jay Berkenbilt
1 parent 77c31305
Showing 10 changed files with 72 additions and 17 deletions
ChangeLog
libqpdf/QUtil.cc
libtests/qtest/qutil/qutil.out
libtests/qutil.cc
qpdf/qtest/qpdf/json-image-streams-all.out
qpdf/qtest/qpdf/json-image-streams-small.out
qpdf/qtest/qpdf/json-image-streams-specialized.out
qpdf/qtest/qpdf/json-image-streams.out
qpdf/qtest/qpdf/json-page-labels-num-tree.out
qpdf/qtest/qpdf/page_api_2-json.out
+2022-01-11  Jay Berkenbilt  <ejb@ql.org>
+
+	* Bug fix: add missing characters from PDF doc encoding.
+	Fixes #606.
+
 2021-12-29  Jay Berkenbilt  <ejb@ql.org>
 	* Add method QUtil::file_can_be_opened
@@ -37,8 +37,20 @@
 # include <sys/stat.h>
 #endif
-// First element is 128
+// First element is 24
+static unsigned short pdf_doc_low_to_unicode[] = {
+    0x02d8,    // 0x18    BREVE
+    0x02c7,    // 0x19    CARON
+    0x02c6,    // 0x1a    MODIFIER LETTER CIRCUMFLEX ACCENT
+    0x02d9,    // 0x1b    DOT ABOVE
+    0x02dd,    // 0x1c    DOUBLE ACUTE ACCENT
+    0x02db,    // 0x1d    OGONEK
+    0x02da,    // 0x1e    RING ABOVE
+    0x02dc,    // 0x1f    SMALL TILDE
+};
+// First element is 127
 static unsigned short pdf_doc_to_unicode[] = {
+    0xfffd,    // 0x7f    UNDEFINED
     0x2022,    // 0x80    BULLET
     0x2020,    // 0x81    DAGGER
     0x2021,    // 0x82    DOUBLE DAGGER
@@ -2032,6 +2044,30 @@ encode_pdfdoc(unsigned long codepoint)
     unsigned char ch = '\0';
     switch (codepoint)
     {
+      case 0x02d8:
+        ch = 0x18;
+        break;
+      case 0x02c7:
+        ch = 0x19;
+        break;
+      case 0x02c6:
+        ch = 0x1a;
+        break;
+      case 0x02d9:
+        ch = 0x1b;
+        break;
+      case 0x02dd:
+        ch = 0x1c;
+        break;
+      case 0x02db:
+        ch = 0x1d;
+        break;
+      case 0x02da:
+        ch = 0x1e;
+        break;
+      case 0x02dc:
+        ch = 0x1f;
+        break;
       case 0x2022:
         ch = 0x80;
         break;
@@ -2427,9 +2463,13 @@ QUtil::pdf_doc_to_utf8(std::string const&amp; val)
     {
         unsigned char ch = static_cast<unsigned char>(val.at(i));
         unsigned short ch_short = ch;
-        if ((ch >= 128) && (ch <= 160))
+        if ((ch >= 127) && (ch <= 160))
+        {
+            ch_short = pdf_doc_to_unicode[ch - 127];
+        }
+        else if ((ch >= 24) && (ch <= 31))
         {
-            ch_short = pdf_doc_to_unicode[ch - 128];
+            ch_short = pdf_doc_low_to_unicode[ch - 24];
         }
         result += QUtil::toUTF8(ch_short);
     }
@@ -69,6 +69,7 @@ HAGOOGAMAGOOGLE: 0
 <c0>Does * have fingers?
 ---- transcoding
 bidirectional pdf doc done
+bidirectional pdf doc low done
 bidirectional win ansi done
 bidirectional mac roman done
 analysis done
@@ -85,6 +86,8 @@ alternatives
 2: 83a9e99e
 0: 717561636b
 done alternatives
+w˘wˇwˆw˙w˝w˛w˚w˜w�w
+done low characters
 ---- whoami
 quack1
 quack2
@@ -308,12 +308,12 @@ void utf8_to_ascii_test()
 void transcoding_test(std::string (*to_utf8)(std::string const&),
                       std::string (*from_utf8)(std::string const&, char),
-                      int last, std::string unknown)
+                      int first, int last, std::string unknown)
 {
     std::string in(" ");
     std::string out;
     std::string back;
-    for (int i = 128; i <= last; ++i)
+    for (int i = first; i <= last; ++i)
     {
         in.at(0) = static_cast<char>(static_cast<unsigned char>(i));
         out = (*to_utf8)(in);
@@ -355,13 +355,16 @@ void print_alternatives(std::string const&amp; str)
 void transcoding_test()
 {
     transcoding_test(&QUtil::pdf_doc_to_utf8,
-                     &QUtil::utf8_to_pdf_doc, 160, "\x9f");
+                     &QUtil::utf8_to_pdf_doc, 127, 160, "\x9f");
     std::cout << "bidirectional pdf doc done" << std::endl;
+    transcoding_test(&QUtil::pdf_doc_to_utf8,
+                     &QUtil::utf8_to_pdf_doc, 24, 31, "?");
+    std::cout << "bidirectional pdf doc low done" << std::endl;
     transcoding_test(&QUtil::win_ansi_to_utf8,
-                     &QUtil::utf8_to_win_ansi, 160, "?");
+                     &QUtil::utf8_to_win_ansi, 128, 160, "?");
     std::cout << "bidirectional win ansi done" << std::endl;
     transcoding_test(&QUtil::mac_roman_to_utf8,
-                     &QUtil::utf8_to_mac_roman, 255, "?");
+                     &QUtil::utf8_to_mac_roman, 128, 255, "?");
     std::cout << "bidirectional mac roman done" << std::endl;
     check_analyze("pi = \317\200", true, true, false);
     check_analyze("pi != \317", true, false, false);
@@ -396,6 +399,10 @@ void transcoding_test()
     print_alternatives(utf8);
     print_alternatives("quack");
     std::cout << "done alternatives" << std::endl;
+    std::string low = QUtil::pdf_doc_to_utf8(
+        "w\030w\031w\032w\033w\034w\035w\036w\037w\177w");
+    std::cout << low << std::endl;
+    std::cout << "done low characters" << std::endl;
 }
 void print_whoami(char const* str)
@@ -604,7 +604,7 @@
     "trailer": {
       "/ID": [
         "S¶Ł”łîð\u000e¢¬\u0007}_)\u0012¶",
-        "'+“‰¤V2«PP ç`m\"\u001d"
+        "'+“‰¤V2«PP ç`m\"˛"
       ],
       "/Root": "1 0 R",
       "/Size": 31
@@ -615,8 +615,8 @@
     },
     "trailer": {
       "/ID": [
-        "Z§¯•Py»’~’46\u001dı\u0011¢",
-        "Z§¯•Py»’~’46\u001dı\u0011¢"
+        "Z§¯•Py»’~’46˛ı\u0011¢",
+        "Z§¯•Py»’~’46˛ı\u0011¢"
       ],
       "/Root": "1 0 R",
       "/Size": 31
@@ -604,7 +604,7 @@
     "trailer": {
       "/ID": [
         "S¶Ł”łîð\u000e¢¬\u0007}_)\u0012¶",
-        "'+“‰¤V2«PP ç`m\"\u001d"
+        "'+“‰¤V2«PP ç`m\"˛"
       ],
       "/Root": "1 0 R",
       "/Size": 31
@@ -604,7 +604,7 @@
     "trailer": {
       "/ID": [
         "S¶Ł”łîð\u000e¢¬\u0007}_)\u0012¶",
-        "'+“‰¤V2«PP ç`m\"\u001d"
+        "'+“‰¤V2«PP ç`m\"˛"
       ],
       "/Root": "1 0 R",
       "/Size": 31
@@ -1518,8 +1518,8 @@
     "99 0 R": 47,
     "trailer": {
       "/ID": [
-        "’ù\u0019Þxtó¼\\·¯½\u001eŁ7»",
-        "\rþ\u0018©LÞ\u000fKýÈl\u0003¯\u0019\u0001\u000e"
+        "’ùˇÞxtó¼\\·¯½˚Ł7»",
+        "\rþ˘©LÞ\u000fKýÈl\u0003¯ˇ\u0001\u000e"
       ],
       "/Root": "1 0 R",
       "/Size": 100
@@ -178,8 +178,8 @@
     },
     "trailer": {
       "/ID": [
-        "û\u0018·ƒÿ{5⁄\u0005Ú−S*º‘o",
-        "÷\u0017ž³QY¿ÔÀ\u000f\u0012−¼ý\u001f\u0002"
+        "û˘·ƒÿ{5⁄\u0005Ú−S*º‘o",
+        "÷\u0017ž³QY¿ÔÀ\u000f\u0012−¼ý˜\u0002"
       ],
       "/Info": "2 0 R",
       "/Root": "1 0 R",