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