1 /* 2 * Copyright 2015 Google Inc. 3 * 4 * Use of this source code is governed by a BSD-style license that can be 5 * found in the LICENSE file. 6 */ 7 8 #include "SkPDFMetadata.h" 9 10 #include "SkMD5.h" 11 #include "SkMilestone.h" 12 #include "SkPDFTypes.h" 13 #include "SkTo.h" 14 #include "SkUtils.h" 15 16 #include <utility> 17 18 #define SKPDF_STRING(X) SKPDF_STRING_IMPL(X) 19 #define SKPDF_STRING_IMPL(X) #X 20 #define SKPDF_PRODUCER "Skia/PDF m" SKPDF_STRING(SK_MILESTONE) 21 #define SKPDF_CUSTOM_PRODUCER_KEY "ProductionLibrary" 22 23 static constexpr SkTime::DateTime kZeroTime = {0, 0, 0, 0, 0, 0, 0, 0}; 24 25 static bool operator!=(const SkTime::DateTime& u, const SkTime::DateTime& v) { 26 return u.fTimeZoneMinutes != v.fTimeZoneMinutes || 27 u.fYear != v.fYear || 28 u.fMonth != v.fMonth || 29 u.fDayOfWeek != v.fDayOfWeek || 30 u.fDay != v.fDay || 31 u.fHour != v.fHour || 32 u.fMinute != v.fMinute || 33 u.fSecond != v.fSecond; 34 } 35 36 static SkString pdf_date(const SkTime::DateTime& dt) { 37 int timeZoneMinutes = SkToInt(dt.fTimeZoneMinutes); 38 char timezoneSign = timeZoneMinutes >= 0 ? '+' : '-'; 39 int timeZoneHours = SkTAbs(timeZoneMinutes) / 60; 40 timeZoneMinutes = SkTAbs(timeZoneMinutes) % 60; 41 return SkStringPrintf( 42 "D:%04u%02u%02u%02u%02u%02u%c%02d'%02d'", 43 static_cast<unsigned>(dt.fYear), static_cast<unsigned>(dt.fMonth), 44 static_cast<unsigned>(dt.fDay), static_cast<unsigned>(dt.fHour), 45 static_cast<unsigned>(dt.fMinute), 46 static_cast<unsigned>(dt.fSecond), timezoneSign, timeZoneHours, 47 timeZoneMinutes); 48 } 49 50 static bool utf8_is_pdfdocencoding(const char* src, size_t len) { 51 const uint8_t* end = (const uint8_t*)src + len; 52 for (const uint8_t* ptr = (const uint8_t*)src; ptr < end; ++ptr) { 53 uint8_t v = *ptr; 54 // See Table D.2 (PDFDocEncoding Character Set) in the PDF3200_2008 spec. 55 if ((v > 23 && v < 32) || v > 126) { 56 return false; 57 } 58 } 59 return true; 60 } 61 62 void write_utf16be(char** ptr, uint16_t value) { 63 *(*ptr)++ = (value >> 8); 64 *(*ptr)++ = (value & 0xFF); 65 } 66 67 // Please Note: This "abuses" the SkString, which "should" only hold UTF8. 68 // But the SkString is written as if it is really just a ref-counted array of 69 // chars, so this works, as long as we handle endiness and conversions ourselves. 70 // 71 // Input: UTF-8 72 // Output UTF-16-BE 73 static SkString to_utf16be(const char* src, size_t len) { 74 SkString ret; 75 const char* const end = src + len; 76 size_t n = 1; // BOM 77 for (const char* ptr = src; ptr < end;) { 78 SkUnichar u = SkUTF::NextUTF8(&ptr, end); 79 if (u < 0) { 80 break; 81 } 82 n += SkUTF::ToUTF16(u); 83 } 84 ret.resize(2 * n); 85 char* out = ret.writable_str(); 86 write_utf16be(&out, 0xFEFF); // BOM 87 for (const char* ptr = src; ptr < end;) { 88 SkUnichar u = SkUTF::NextUTF8(&ptr, end); 89 if (u < 0) { 90 break; 91 } 92 uint16_t utf16[2]; 93 size_t l = SkUTF::ToUTF16(u, utf16); 94 write_utf16be(&out, utf16[0]); 95 if (l == 2) { 96 write_utf16be(&out, utf16[1]); 97 } 98 } 99 SkASSERT(out == ret.writable_str() + 2 * n); 100 return ret; 101 } 102 103 // Input: UTF-8 104 // Output UTF-16-BE OR PDFDocEncoding (if that encoding is identical to ASCII encoding). 105 // 106 // See sections 14.3.3 (Document Information Dictionary) and 7.9.2.2 (Text String Type) 107 // of the PDF32000_2008 spec. 108 static SkString convert(const SkString& s) { 109 return utf8_is_pdfdocencoding(s.c_str(), s.size()) ? s : to_utf16be(s.c_str(), s.size()); 110 } 111 static SkString convert(const char* src) { 112 size_t len = strlen(src); 113 return utf8_is_pdfdocencoding(src, len) ? SkString(src, len) : to_utf16be(src, len); 114 } 115 116 namespace { 117 static const struct { 118 const char* const key; 119 SkString SkPDF::Metadata::*const valuePtr; 120 } gMetadataKeys[] = { 121 {"Title", &SkPDF::Metadata::fTitle}, 122 {"Author", &SkPDF::Metadata::fAuthor}, 123 {"Subject", &SkPDF::Metadata::fSubject}, 124 {"Keywords", &SkPDF::Metadata::fKeywords}, 125 {"Creator", &SkPDF::Metadata::fCreator}, 126 }; 127 } // namespace 128 129 std::unique_ptr<SkPDFObject> SkPDFMetadata::MakeDocumentInformationDict( 130 const SkPDF::Metadata& metadata) { 131 auto dict = SkPDFMakeDict(); 132 for (const auto keyValuePtr : gMetadataKeys) { 133 const SkString& value = metadata.*(keyValuePtr.valuePtr); 134 if (value.size() > 0) { 135 dict->insertString(keyValuePtr.key, convert(value)); 136 } 137 } 138 if (metadata.fProducer.isEmpty()) { 139 dict->insertString("Producer", convert(SKPDF_PRODUCER)); 140 } else { 141 dict->insertString("Producer", convert(metadata.fProducer)); 142 dict->insertString(SKPDF_CUSTOM_PRODUCER_KEY, convert(SKPDF_PRODUCER)); 143 } 144 if (metadata.fCreation != kZeroTime) { 145 dict->insertString("CreationDate", pdf_date(metadata.fCreation)); 146 } 147 if (metadata.fModified != kZeroTime) { 148 dict->insertString("ModDate", pdf_date(metadata.fModified)); 149 } 150 return std::move(dict); 151 } 152 153 SkUUID SkPDFMetadata::CreateUUID(const SkPDF::Metadata& metadata) { 154 // The main requirement is for the UUID to be unique; the exact 155 // format of the data that will be hashed is not important. 156 SkMD5 md5; 157 const char uuidNamespace[] = "org.skia.pdf\n"; 158 md5.writeText(uuidNamespace); 159 double msec = SkTime::GetMSecs(); 160 md5.write(&msec, sizeof(msec)); 161 SkTime::DateTime dateTime; 162 SkTime::GetDateTime(&dateTime); 163 md5.write(&dateTime, sizeof(dateTime)); 164 md5.write(&metadata.fCreation, sizeof(metadata.fCreation)); 165 md5.write(&metadata.fModified, sizeof(metadata.fModified)); 166 167 for (const auto keyValuePtr : gMetadataKeys) { 168 md5.writeText(keyValuePtr.key); 169 md5.write("\037", 1); 170 const SkString& value = metadata.*(keyValuePtr.valuePtr); 171 md5.write(value.c_str(), value.size()); 172 md5.write("\036", 1); 173 } 174 SkMD5::Digest digest; 175 md5.finish(digest); 176 // See RFC 4122, page 6-7. 177 digest.data[6] = (digest.data[6] & 0x0F) | 0x30; 178 digest.data[8] = (digest.data[6] & 0x3F) | 0x80; 179 static_assert(sizeof(digest) == sizeof(SkUUID), "uuid_size"); 180 SkUUID uuid; 181 memcpy(&uuid, &digest, sizeof(digest)); 182 return uuid; 183 } 184 185 std::unique_ptr<SkPDFObject> SkPDFMetadata::MakePdfId(const SkUUID& doc, 186 const SkUUID& instance) { 187 // /ID [ <81b14aafa313db63dbd6f981e49f94f4> 188 // <81b14aafa313db63dbd6f981e49f94f4> ] 189 auto array = SkPDFMakeArray(); 190 static_assert(sizeof(SkUUID) == 16, "uuid_size"); 191 array->appendString( 192 SkString(reinterpret_cast<const char*>(&doc), sizeof(SkUUID))); 193 array->appendString( 194 SkString(reinterpret_cast<const char*>(&instance), sizeof(SkUUID))); 195 return std::move(array); 196 } 197 198 // Convert a block of memory to hexadecimal. Input and output pointers will be 199 // moved to end of the range. 200 static void hexify(const uint8_t** inputPtr, char** outputPtr, int count) { 201 SkASSERT(inputPtr && *inputPtr); 202 SkASSERT(outputPtr && *outputPtr); 203 while (count-- > 0) { 204 uint8_t value = *(*inputPtr)++; 205 *(*outputPtr)++ = SkHexadecimalDigits::gLower[value >> 4]; 206 *(*outputPtr)++ = SkHexadecimalDigits::gLower[value & 0xF]; 207 } 208 } 209 210 static SkString uuid_to_string(const SkUUID& uuid) { 211 // 8-4-4-4-12 212 char buffer[36]; // [32 + 4] 213 char* ptr = buffer; 214 const uint8_t* data = uuid.fData; 215 hexify(&data, &ptr, 4); 216 *ptr++ = '-'; 217 hexify(&data, &ptr, 2); 218 *ptr++ = '-'; 219 hexify(&data, &ptr, 2); 220 *ptr++ = '-'; 221 hexify(&data, &ptr, 2); 222 *ptr++ = '-'; 223 hexify(&data, &ptr, 6); 224 SkASSERT(ptr == buffer + 36); 225 SkASSERT(data == uuid.fData + 16); 226 return SkString(buffer, 36); 227 } 228 229 namespace { 230 class PDFXMLObject final : public SkPDFObject { 231 public: 232 PDFXMLObject(SkString xml) : fXML(std::move(xml)) {} 233 void emitObject(SkWStream* stream) const override { 234 SkPDFDict dict("Metadata"); 235 dict.insertName("Subtype", "XML"); 236 dict.insertInt("Length", fXML.size()); 237 dict.emitObject(stream); 238 static const char streamBegin[] = " stream\n"; 239 stream->writeText(streamBegin); 240 // Do not compress this. The standard requires that a 241 // program that does not understand PDF can grep for 242 // "<?xpacket" and extract the entire XML. 243 stream->write(fXML.c_str(), fXML.size()); 244 static const char streamEnd[] = "\nendstream"; 245 stream->writeText(streamEnd); 246 } 247 248 private: 249 const SkString fXML; 250 }; 251 } // namespace 252 253 static int count_xml_escape_size(const SkString& input) { 254 int extra = 0; 255 for (size_t i = 0; i < input.size(); ++i) { 256 if (input[i] == '&') { 257 extra += 4; // strlen("&") - strlen("&") 258 } else if (input[i] == '<') { 259 extra += 3; // strlen("<") - strlen("<") 260 } 261 } 262 return extra; 263 } 264 265 const SkString escape_xml(const SkString& input, 266 const char* before = nullptr, 267 const char* after = nullptr) { 268 if (input.size() == 0) { 269 return input; 270 } 271 // "&" --> "&" and "<" --> "<" 272 // text is assumed to be in UTF-8 273 // all strings are xml content, not attribute values. 274 size_t beforeLen = before ? strlen(before) : 0; 275 size_t afterLen = after ? strlen(after) : 0; 276 int extra = count_xml_escape_size(input); 277 SkString output(input.size() + extra + beforeLen + afterLen); 278 char* out = output.writable_str(); 279 if (before) { 280 strncpy(out, before, beforeLen); 281 out += beforeLen; 282 } 283 static const char kAmp[] = "&"; 284 static const char kLt[] = "<"; 285 for (size_t i = 0; i < input.size(); ++i) { 286 if (input[i] == '&') { 287 strncpy(out, kAmp, strlen(kAmp)); 288 out += strlen(kAmp); 289 } else if (input[i] == '<') { 290 strncpy(out, kLt, strlen(kLt)); 291 out += strlen(kLt); 292 } else { 293 *out++ = input[i]; 294 } 295 } 296 if (after) { 297 strncpy(out, after, afterLen); 298 out += afterLen; 299 } 300 // Validate that we haven't written outside of our string. 301 SkASSERT(out == &output.writable_str()[output.size()]); 302 *out = '\0'; 303 return output; 304 } 305 306 SkPDFIndirectReference SkPDFMetadata::MakeXMPObject( 307 const SkPDF::Metadata& metadata, 308 const SkUUID& doc, 309 const SkUUID& instance, 310 SkPDFDocument* docPtr) { 311 static const char templateString[] = 312 "<?xpacket begin=\"\" id=\"W5M0MpCehiHzreSzNTczkc9d\"?>\n" 313 "<x:xmpmeta xmlns:x=\"adobe:ns:meta/\"\n" 314 " x:xmptk=\"Adobe XMP Core 5.4-c005 78.147326, " 315 "2012/08/23-13:03:03\">\n" 316 "<rdf:RDF " 317 "xmlns:rdf=\"http://www.w3.org/1999/02/22-rdf-syntax-ns#\">\n" 318 "<rdf:Description rdf:about=\"\"\n" 319 " xmlns:xmp=\"http://ns.adobe.com/xap/1.0/\"\n" 320 " xmlns:dc=\"http://purl.org/dc/elements/1.1/\"\n" 321 " xmlns:xmpMM=\"http://ns.adobe.com/xap/1.0/mm/\"\n" 322 " xmlns:pdf=\"http://ns.adobe.com/pdf/1.3/\"\n" 323 " xmlns:pdfaid=\"http://www.aiim.org/pdfa/ns/id/\">\n" 324 "<pdfaid:part>2</pdfaid:part>\n" 325 "<pdfaid:conformance>B</pdfaid:conformance>\n" 326 "%s" // ModifyDate 327 "%s" // CreateDate 328 "%s" // xmp:CreatorTool 329 "<dc:format>application/pdf</dc:format>\n" 330 "%s" // dc:title 331 "%s" // dc:description 332 "%s" // author 333 "%s" // keywords 334 "<xmpMM:DocumentID>uuid:%s</xmpMM:DocumentID>\n" 335 "<xmpMM:InstanceID>uuid:%s</xmpMM:InstanceID>\n" 336 "%s" // pdf:Producer 337 "%s" // pdf:Keywords 338 "</rdf:Description>\n" 339 "</rdf:RDF>\n" 340 "</x:xmpmeta>\n" // Note: the standard suggests 4k of padding. 341 "<?xpacket end=\"w\"?>\n"; 342 343 SkString creationDate; 344 SkString modificationDate; 345 if (metadata.fCreation != kZeroTime) { 346 SkString tmp; 347 metadata.fCreation.toISO8601(&tmp); 348 SkASSERT(0 == count_xml_escape_size(tmp)); 349 // YYYY-mm-ddTHH:MM:SS[+|-]ZZ:ZZ; no need to escape 350 creationDate = SkStringPrintf("<xmp:CreateDate>%s</xmp:CreateDate>\n", 351 tmp.c_str()); 352 } 353 if (metadata.fModified != kZeroTime) { 354 SkString tmp; 355 metadata.fModified.toISO8601(&tmp); 356 SkASSERT(0 == count_xml_escape_size(tmp)); 357 modificationDate = SkStringPrintf( 358 "<xmp:ModifyDate>%s</xmp:ModifyDate>\n", tmp.c_str()); 359 } 360 SkString title = 361 escape_xml(metadata.fTitle, 362 "<dc:title><rdf:Alt><rdf:li xml:lang=\"x-default\">", 363 "</rdf:li></rdf:Alt></dc:title>\n"); 364 SkString author = 365 escape_xml(metadata.fAuthor, "<dc:creator><rdf:Bag><rdf:li>", 366 "</rdf:li></rdf:Bag></dc:creator>\n"); 367 // TODO: in theory, XMP can support multiple authors. Split on a delimiter? 368 SkString subject = escape_xml( 369 metadata.fSubject, 370 "<dc:description><rdf:Alt><rdf:li xml:lang=\"x-default\">", 371 "</rdf:li></rdf:Alt></dc:description>\n"); 372 SkString keywords1 = 373 escape_xml(metadata.fKeywords, "<dc:subject><rdf:Bag><rdf:li>", 374 "</rdf:li></rdf:Bag></dc:subject>\n"); 375 SkString keywords2 = escape_xml(metadata.fKeywords, "<pdf:Keywords>", 376 "</pdf:Keywords>\n"); 377 // TODO: in theory, keywords can be a list too. 378 379 SkString producer("<pdf:Producer>" SKPDF_PRODUCER "</pdf:Producer>\n"); 380 if (!metadata.fProducer.isEmpty()) { 381 // TODO: register a developer prefix to make 382 // <skia:SKPDF_CUSTOM_PRODUCER_KEY> a real XML tag. 383 producer = escape_xml( 384 metadata.fProducer, "<pdf:Producer>", 385 "</pdf:Producer>\n<!-- <skia:" SKPDF_CUSTOM_PRODUCER_KEY ">" 386 SKPDF_PRODUCER "</skia:" SKPDF_CUSTOM_PRODUCER_KEY "> -->\n"); 387 } 388 389 SkString creator = escape_xml(metadata.fCreator, "<xmp:CreatorTool>", 390 "</xmp:CreatorTool>\n"); 391 SkString documentID = uuid_to_string(doc); // no need to escape 392 SkASSERT(0 == count_xml_escape_size(documentID)); 393 SkString instanceID = uuid_to_string(instance); 394 SkASSERT(0 == count_xml_escape_size(instanceID)); 395 396 397 auto value = SkStringPrintf( 398 templateString, modificationDate.c_str(), creationDate.c_str(), 399 creator.c_str(), title.c_str(), subject.c_str(), author.c_str(), 400 keywords1.c_str(), documentID.c_str(), instanceID.c_str(), 401 producer.c_str(), keywords2.c_str()); 402 403 std::unique_ptr<SkPDFDict> dict = SkPDFMakeDict("Metadata"); 404 dict->insertName("Subtype", "XML"); 405 return SkPDFStreamOut(std::move(dict), 406 SkMemoryStream::MakeCopy(value.c_str(), value.size()), 407 docPtr, false); 408 } 409 410 #undef SKPDF_CUSTOM_PRODUCER_KEY 411 #undef SKPDF_PRODUCER 412 #undef SKPDF_STRING 413 #undef SKPDF_STRING_IMPL 414