1 //===--- YAMLParser.cpp - Simple YAML parser ------------------------------===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // This file implements a YAML parser.
12 //===----------------------------------------------------------------------===//
14 #include "llvm/Support/YAMLParser.h"
16 #include "llvm/ADT/ilist.h"
17 #include "llvm/ADT/ilist_node.h"
18 #include "llvm/ADT/SmallVector.h"
19 #include "llvm/ADT/StringExtras.h"
20 #include "llvm/ADT/Twine.h"
21 #include "llvm/Support/ErrorHandling.h"
22 #include "llvm/Support/MemoryBuffer.h"
23 #include "llvm/Support/raw_ostream.h"
24 #include "llvm/Support/SourceMgr.h"
29 enum UnicodeEncodingForm {
30 UEF_UTF32_LE, ///< UTF-32 Little Endian
31 UEF_UTF32_BE, ///< UTF-32 Big Endian
32 UEF_UTF16_LE, ///< UTF-16 Little Endian
33 UEF_UTF16_BE, ///< UTF-16 Big Endian
34 UEF_UTF8, ///< UTF-8 or ascii.
35 UEF_Unknown ///< Not a valid Unicode encoding.
38 /// EncodingInfo - Holds the encoding type and length of the byte order mark if
39 /// it exists. Length is in {0, 2, 3, 4}.
40 typedef std::pair<UnicodeEncodingForm, unsigned> EncodingInfo;
42 /// getUnicodeEncoding - Reads up to the first 4 bytes to determine the Unicode
43 /// encoding form of \a Input.
45 /// @param Input A string of length 0 or more.
46 /// @returns An EncodingInfo indicating the Unicode encoding form of the input
47 /// and how long the byte order mark is if one exists.
48 static EncodingInfo getUnicodeEncoding(StringRef Input) {
49 if (Input.size() == 0)
50 return std::make_pair(UEF_Unknown, 0);
52 switch (uint8_t(Input[0])) {
54 if (Input.size() >= 4) {
56 && uint8_t(Input[2]) == 0xFE
57 && uint8_t(Input[3]) == 0xFF)
58 return std::make_pair(UEF_UTF32_BE, 4);
59 if (Input[1] == 0 && Input[2] == 0 && Input[3] != 0)
60 return std::make_pair(UEF_UTF32_BE, 0);
63 if (Input.size() >= 2 && Input[1] != 0)
64 return std::make_pair(UEF_UTF16_BE, 0);
65 return std::make_pair(UEF_Unknown, 0);
67 if ( Input.size() >= 4
68 && uint8_t(Input[1]) == 0xFE
71 return std::make_pair(UEF_UTF32_LE, 4);
73 if (Input.size() >= 2 && uint8_t(Input[1]) == 0xFE)
74 return std::make_pair(UEF_UTF16_LE, 2);
75 return std::make_pair(UEF_Unknown, 0);
77 if (Input.size() >= 2 && uint8_t(Input[1]) == 0xFF)
78 return std::make_pair(UEF_UTF16_BE, 2);
79 return std::make_pair(UEF_Unknown, 0);
81 if ( Input.size() >= 3
82 && uint8_t(Input[1]) == 0xBB
83 && uint8_t(Input[2]) == 0xBF)
84 return std::make_pair(UEF_UTF8, 3);
85 return std::make_pair(UEF_Unknown, 0);
88 // It could still be utf-32 or utf-16.
89 if (Input.size() >= 4 && Input[1] == 0 && Input[2] == 0 && Input[3] == 0)
90 return std::make_pair(UEF_UTF32_LE, 0);
92 if (Input.size() >= 2 && Input[1] == 0)
93 return std::make_pair(UEF_UTF16_LE, 0);
95 return std::make_pair(UEF_UTF8, 0);
100 /// Token - A single YAML token.
101 struct Token : ilist_node<Token> {
103 TK_Error, // Uninitialized token.
112 TK_BlockSequenceStart,
113 TK_BlockMappingStart,
115 TK_FlowSequenceStart,
127 /// A string of length 0 or more whose begin() points to the logical location
128 /// of the token in the input.
131 Token() : Kind(TK_Error) {}
138 struct ilist_sentinel_traits<Token> {
139 Token *createSentinel() const {
142 static void destroySentinel(Token*) {}
144 Token *provideInitialHead() const { return createSentinel(); }
145 Token *ensureHead(Token*) const { return createSentinel(); }
146 static void noteHead(Token*, Token*) {}
149 mutable Token Sentinel;
153 struct ilist_node_traits<Token> {
154 Token *createNode(const Token &V) {
155 return new (Alloc.Allocate<Token>()) Token(V);
157 static void deleteNode(Token *V) {}
159 void addNodeToList(Token *) {}
160 void removeNodeFromList(Token *) {}
161 void transferNodesFromList(ilist_node_traits & /*SrcTraits*/,
162 ilist_iterator<Token> /*first*/,
163 ilist_iterator<Token> /*last*/) {}
165 BumpPtrAllocator Alloc;
169 typedef ilist<Token> TokenQueueT;
172 /// @brief This struct is used to track simple keys.
174 /// Simple keys are handled by creating an entry in SimpleKeys for each Token
175 /// which could legally be the start of a simple key. When peekNext is called,
176 /// if the Token To be returned is referenced by a SimpleKey, we continue
177 /// tokenizing until that potential simple key has either been found to not be
178 /// a simple key (we moved on to the next line or went further than 1024 chars).
179 /// Or when we run into a Value, and then insert a Key token (and possibly
180 /// others) before the SimpleKey's Tok.
182 TokenQueueT::iterator Tok;
188 bool operator ==(const SimpleKey &Other) {
189 return Tok == Other.Tok;
194 /// @brief The Unicode scalar value of a UTF-8 minimal well-formed code unit
195 /// subsequence and the subsequence's length in code units (uint8_t).
196 /// A length of 0 represents an error.
197 typedef std::pair<uint32_t, unsigned> UTF8Decoded;
199 static UTF8Decoded decodeUTF8(StringRef Range) {
200 StringRef::iterator Position= Range.begin();
201 StringRef::iterator End = Range.end();
202 // 1 byte: [0x00, 0x7f]
203 // Bit pattern: 0xxxxxxx
204 if ((*Position & 0x80) == 0) {
205 return std::make_pair(*Position, 1);
207 // 2 bytes: [0x80, 0x7ff]
208 // Bit pattern: 110xxxxx 10xxxxxx
209 if (Position + 1 != End &&
210 ((*Position & 0xE0) == 0xC0) &&
211 ((*(Position + 1) & 0xC0) == 0x80)) {
212 uint32_t codepoint = ((*Position & 0x1F) << 6) |
213 (*(Position + 1) & 0x3F);
214 if (codepoint >= 0x80)
215 return std::make_pair(codepoint, 2);
217 // 3 bytes: [0x8000, 0xffff]
218 // Bit pattern: 1110xxxx 10xxxxxx 10xxxxxx
219 if (Position + 2 != End &&
220 ((*Position & 0xF0) == 0xE0) &&
221 ((*(Position + 1) & 0xC0) == 0x80) &&
222 ((*(Position + 2) & 0xC0) == 0x80)) {
223 uint32_t codepoint = ((*Position & 0x0F) << 12) |
224 ((*(Position + 1) & 0x3F) << 6) |
225 (*(Position + 2) & 0x3F);
226 // Codepoints between 0xD800 and 0xDFFF are invalid, as
227 // they are high / low surrogate halves used by UTF-16.
228 if (codepoint >= 0x800 &&
229 (codepoint < 0xD800 || codepoint > 0xDFFF))
230 return std::make_pair(codepoint, 3);
232 // 4 bytes: [0x10000, 0x10FFFF]
233 // Bit pattern: 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx
234 if (Position + 3 != End &&
235 ((*Position & 0xF8) == 0xF0) &&
236 ((*(Position + 1) & 0xC0) == 0x80) &&
237 ((*(Position + 2) & 0xC0) == 0x80) &&
238 ((*(Position + 3) & 0xC0) == 0x80)) {
239 uint32_t codepoint = ((*Position & 0x07) << 18) |
240 ((*(Position + 1) & 0x3F) << 12) |
241 ((*(Position + 2) & 0x3F) << 6) |
242 (*(Position + 3) & 0x3F);
243 if (codepoint >= 0x10000 && codepoint <= 0x10FFFF)
244 return std::make_pair(codepoint, 4);
246 return std::make_pair(0, 0);
251 /// @brief Scans YAML tokens from a MemoryBuffer.
254 Scanner(const StringRef Input, SourceMgr &SM);
256 /// @brief Parse the next token and return it without popping it.
259 /// @brief Parse the next token and pop it from the queue.
262 void printError(SMLoc Loc, SourceMgr::DiagKind Kind, const Twine &Message,
263 ArrayRef<SMRange> Ranges = ArrayRef<SMRange>()) {
264 SM.PrintMessage(Loc, Kind, Message, Ranges);
267 void setError(const Twine &Message, StringRef::iterator Position) {
271 // Don't print out more errors after the first one we encounter. The rest
272 // are just the result of the first, and have no meaning.
274 printError(SMLoc::getFromPointer(Current), SourceMgr::DK_Error, Message);
278 void setError(const Twine &Message) {
279 setError(Message, Current);
282 /// @brief Returns true if an error occurred while parsing.
288 StringRef currentInput() {
289 return StringRef(Current, End - Current);
292 /// @brief Decode a UTF-8 minimal well-formed code unit subsequence starting
295 /// If the UTF-8 code units starting at Position do not form a well-formed
296 /// code unit subsequence, then the Unicode scalar value is 0, and the length
298 UTF8Decoded decodeUTF8(StringRef::iterator Position) {
299 return ::decodeUTF8(StringRef(Position, End - Position));
302 // The following functions are based on the gramar rules in the YAML spec. The
303 // style of the function names it meant to closely match how they are written
304 // in the spec. The number within the [] is the number of the grammar rule in
307 // See 4.2 [Production Naming Conventions] for the meaning of the prefixes.
310 // A production starting and ending with a special character.
312 // A production matching a single line break.
314 // A production starting and ending with a non-break character.
316 // A production starting and ending with a white space character.
318 // A production starting and ending with a non-space character.
320 // A production matching complete line(s).
322 /// @brief Skip a single nb-char[27] starting at Position.
324 /// A nb-char is 0x9 | [0x20-0x7E] | 0x85 | [0xA0-0xD7FF] | [0xE000-0xFEFE]
325 /// | [0xFF00-0xFFFD] | [0x10000-0x10FFFF]
327 /// @returns The code unit after the nb-char, or Position if it's not an
329 StringRef::iterator skip_nb_char(StringRef::iterator Position);
331 /// @brief Skip a single b-break[28] starting at Position.
333 /// A b-break is 0xD 0xA | 0xD | 0xA
335 /// @returns The code unit after the b-break, or Position if it's not a
337 StringRef::iterator skip_b_break(StringRef::iterator Position);
339 /// @brief Skip a single s-white[33] starting at Position.
341 /// A s-white is 0x20 | 0x9
343 /// @returns The code unit after the s-white, or Position if it's not a
345 StringRef::iterator skip_s_white(StringRef::iterator Position);
347 /// @brief Skip a single ns-char[34] starting at Position.
349 /// A ns-char is nb-char - s-white
351 /// @returns The code unit after the ns-char, or Position if it's not a
353 StringRef::iterator skip_ns_char(StringRef::iterator Position);
355 typedef StringRef::iterator (Scanner::*SkipWhileFunc)(StringRef::iterator);
356 /// @brief Skip minimal well-formed code unit subsequences until Func
357 /// returns its input.
359 /// @returns The code unit after the last minimal well-formed code unit
360 /// subsequence that Func accepted.
361 StringRef::iterator skip_while( SkipWhileFunc Func
362 , StringRef::iterator Position);
364 /// @brief Scan ns-uri-char[39]s starting at Cur.
366 /// This updates Cur and Column while scanning.
368 /// @returns A StringRef starting at Cur which covers the longest contiguous
369 /// sequence of ns-uri-char.
370 StringRef scan_ns_uri_char();
372 /// @brief Scan ns-plain-one-line[133] starting at \a Cur.
373 StringRef scan_ns_plain_one_line();
375 /// @brief Consume a minimal well-formed code unit subsequence starting at
376 /// \a Cur. Return false if it is not the same Unicode scalar value as
377 /// \a Expected. This updates \a Column.
378 bool consume(uint32_t Expected);
380 /// @brief Skip \a Distance UTF-8 code units. Updates \a Cur and \a Column.
381 void skip(uint32_t Distance);
383 /// @brief Return true if the minimal well-formed code unit subsequence at
384 /// Pos is whitespace or a new line
385 bool isBlankOrBreak(StringRef::iterator Position);
387 /// @brief If IsSimpleKeyAllowed, create and push_back a new SimpleKey.
388 void saveSimpleKeyCandidate( TokenQueueT::iterator Tok
392 /// @brief Remove simple keys that can no longer be valid simple keys.
394 /// Invalid simple keys are not on the current line or are further than 1024
396 void removeStaleSimpleKeyCandidates();
398 /// @brief Remove all simple keys on FlowLevel \a Level.
399 void removeSimpleKeyCandidatesOnFlowLevel(unsigned Level);
401 /// @brief Unroll indentation in \a Indents back to \a Col. Creates BlockEnd
402 /// tokens if needed.
403 bool unrollIndent(int ToColumn);
405 /// @brief Increase indent to \a Col. Creates \a Kind token at \a InsertPoint
407 bool rollIndent( int ToColumn
408 , Token::TokenKind Kind
409 , TokenQueueT::iterator InsertPoint);
411 /// @brief Skip whitespace and comments until the start of the next token.
412 void scanToNextToken();
414 /// @brief Must be the first token generated.
415 bool scanStreamStart();
417 /// @brief Generate tokens needed to close out the stream.
418 bool scanStreamEnd();
420 /// @brief Scan a %BLAH directive.
421 bool scanDirective();
423 /// @brief Scan a ... or ---.
424 bool scanDocumentIndicator(bool IsStart);
426 /// @brief Scan a [ or { and generate the proper flow collection start token.
427 bool scanFlowCollectionStart(bool IsSequence);
429 /// @brief Scan a ] or } and generate the proper flow collection end token.
430 bool scanFlowCollectionEnd(bool IsSequence);
432 /// @brief Scan the , that separates entries in a flow collection.
433 bool scanFlowEntry();
435 /// @brief Scan the - that starts block sequence entries.
436 bool scanBlockEntry();
438 /// @brief Scan an explicit ? indicating a key.
441 /// @brief Scan an explicit : indicating a value.
444 /// @brief Scan a quoted scalar.
445 bool scanFlowScalar(bool IsDoubleQuoted);
447 /// @brief Scan an unquoted scalar.
448 bool scanPlainScalar();
450 /// @brief Scan an Alias or Anchor starting with * or &.
451 bool scanAliasOrAnchor(bool IsAlias);
453 /// @brief Scan a block scalar starting with | or >.
454 bool scanBlockScalar(bool IsLiteral);
456 /// @brief Scan a tag of the form !stuff.
459 /// @brief Dispatch to the next scanning function based on \a *Cur.
460 bool fetchMoreTokens();
462 /// @brief The SourceMgr used for diagnostics and buffer management.
465 /// @brief The original input.
466 MemoryBuffer *InputBuffer;
468 /// @brief The current position of the scanner.
469 StringRef::iterator Current;
471 /// @brief The end of the input (one past the last character).
472 StringRef::iterator End;
474 /// @brief Current YAML indentation level in spaces.
477 /// @brief Current column number in Unicode code points.
480 /// @brief Current line number.
483 /// @brief How deep we are in flow style containers. 0 Means at block level.
486 /// @brief Are we at the start of the stream?
487 bool IsStartOfStream;
489 /// @brief Can the next token be the start of a simple key?
490 bool IsSimpleKeyAllowed;
492 /// @brief Is the next token required to start a simple key?
493 bool IsSimpleKeyRequired;
495 /// @brief True if an error has occurred.
498 /// @brief Queue of tokens. This is required to queue up tokens while looking
499 /// for the end of a simple key. And for cases where a single character
500 /// can produce multiple tokens (e.g. BlockEnd).
501 TokenQueueT TokenQueue;
503 /// @brief Indentation levels.
504 SmallVector<int, 4> Indents;
506 /// @brief Potential simple keys.
507 SmallVector<SimpleKey, 4> SimpleKeys;
510 } // end namespace yaml
511 } // end namespace llvm
513 /// encodeUTF8 - Encode \a UnicodeScalarValue in UTF-8 and append it to result.
514 static void encodeUTF8( uint32_t UnicodeScalarValue
515 , SmallVectorImpl<char> &Result) {
516 if (UnicodeScalarValue <= 0x7F) {
517 Result.push_back(UnicodeScalarValue & 0x7F);
518 } else if (UnicodeScalarValue <= 0x7FF) {
519 uint8_t FirstByte = 0xC0 | ((UnicodeScalarValue & 0x7C0) >> 6);
520 uint8_t SecondByte = 0x80 | (UnicodeScalarValue & 0x3F);
521 Result.push_back(FirstByte);
522 Result.push_back(SecondByte);
523 } else if (UnicodeScalarValue <= 0xFFFF) {
524 uint8_t FirstByte = 0xE0 | ((UnicodeScalarValue & 0xF000) >> 12);
525 uint8_t SecondByte = 0x80 | ((UnicodeScalarValue & 0xFC0) >> 6);
526 uint8_t ThirdByte = 0x80 | (UnicodeScalarValue & 0x3F);
527 Result.push_back(FirstByte);
528 Result.push_back(SecondByte);
529 Result.push_back(ThirdByte);
530 } else if (UnicodeScalarValue <= 0x10FFFF) {
531 uint8_t FirstByte = 0xF0 | ((UnicodeScalarValue & 0x1F0000) >> 18);
532 uint8_t SecondByte = 0x80 | ((UnicodeScalarValue & 0x3F000) >> 12);
533 uint8_t ThirdByte = 0x80 | ((UnicodeScalarValue & 0xFC0) >> 6);
534 uint8_t FourthByte = 0x80 | (UnicodeScalarValue & 0x3F);
535 Result.push_back(FirstByte);
536 Result.push_back(SecondByte);
537 Result.push_back(ThirdByte);
538 Result.push_back(FourthByte);
542 bool yaml::dumpTokens(StringRef Input, raw_ostream &OS) {
544 Scanner scanner(Input, SM);
546 Token T = scanner.getNext();
548 case Token::TK_StreamStart:
549 OS << "Stream-Start: ";
551 case Token::TK_StreamEnd:
552 OS << "Stream-End: ";
554 case Token::TK_VersionDirective:
555 OS << "Version-Directive: ";
557 case Token::TK_TagDirective:
558 OS << "Tag-Directive: ";
560 case Token::TK_DocumentStart:
561 OS << "Document-Start: ";
563 case Token::TK_DocumentEnd:
564 OS << "Document-End: ";
566 case Token::TK_BlockEntry:
567 OS << "Block-Entry: ";
569 case Token::TK_BlockEnd:
572 case Token::TK_BlockSequenceStart:
573 OS << "Block-Sequence-Start: ";
575 case Token::TK_BlockMappingStart:
576 OS << "Block-Mapping-Start: ";
578 case Token::TK_FlowEntry:
579 OS << "Flow-Entry: ";
581 case Token::TK_FlowSequenceStart:
582 OS << "Flow-Sequence-Start: ";
584 case Token::TK_FlowSequenceEnd:
585 OS << "Flow-Sequence-End: ";
587 case Token::TK_FlowMappingStart:
588 OS << "Flow-Mapping-Start: ";
590 case Token::TK_FlowMappingEnd:
591 OS << "Flow-Mapping-End: ";
596 case Token::TK_Value:
599 case Token::TK_Scalar:
602 case Token::TK_Alias:
605 case Token::TK_Anchor:
611 case Token::TK_Error:
614 OS << T.Range << "\n";
615 if (T.Kind == Token::TK_StreamEnd)
617 else if (T.Kind == Token::TK_Error)
623 bool yaml::scanTokens(StringRef Input) {
625 llvm::yaml::Scanner scanner(Input, SM);
627 llvm::yaml::Token T = scanner.getNext();
628 if (T.Kind == Token::TK_StreamEnd)
630 else if (T.Kind == Token::TK_Error)
636 std::string yaml::escape(StringRef Input) {
637 std::string EscapedInput;
638 for (StringRef::iterator i = Input.begin(), e = Input.end(); i != e; ++i) {
640 EscapedInput += "\\\\";
642 EscapedInput += "\\\"";
644 EscapedInput += "\\0";
646 EscapedInput += "\\a";
648 EscapedInput += "\\b";
650 EscapedInput += "\\t";
652 EscapedInput += "\\n";
654 EscapedInput += "\\v";
656 EscapedInput += "\\f";
658 EscapedInput += "\\r";
660 EscapedInput += "\\e";
661 else if ((unsigned char)*i < 0x20) { // Control characters not handled above.
662 std::string HexStr = utohexstr(*i);
663 EscapedInput += "\\x" + std::string(2 - HexStr.size(), '0') + HexStr;
664 } else if (*i & 0x80) { // UTF-8 multiple code unit subsequence.
665 UTF8Decoded UnicodeScalarValue
666 = decodeUTF8(StringRef(i, Input.end() - i));
667 if (UnicodeScalarValue.second == 0) {
668 // Found invalid char.
670 encodeUTF8(0xFFFD, Val);
671 EscapedInput.insert(EscapedInput.end(), Val.begin(), Val.end());
672 // FIXME: Error reporting.
675 if (UnicodeScalarValue.first == 0x85)
676 EscapedInput += "\\N";
677 else if (UnicodeScalarValue.first == 0xA0)
678 EscapedInput += "\\_";
679 else if (UnicodeScalarValue.first == 0x2028)
680 EscapedInput += "\\L";
681 else if (UnicodeScalarValue.first == 0x2029)
682 EscapedInput += "\\P";
684 std::string HexStr = utohexstr(UnicodeScalarValue.first);
685 if (HexStr.size() <= 2)
686 EscapedInput += "\\x" + std::string(2 - HexStr.size(), '0') + HexStr;
687 else if (HexStr.size() <= 4)
688 EscapedInput += "\\u" + std::string(4 - HexStr.size(), '0') + HexStr;
689 else if (HexStr.size() <= 8)
690 EscapedInput += "\\U" + std::string(8 - HexStr.size(), '0') + HexStr;
692 i += UnicodeScalarValue.second - 1;
694 EscapedInput.push_back(*i);
699 Scanner::Scanner(StringRef Input, SourceMgr &sm)
705 , IsStartOfStream(true)
706 , IsSimpleKeyAllowed(true)
707 , IsSimpleKeyRequired(false)
709 InputBuffer = MemoryBuffer::getMemBuffer(Input, "YAML");
710 SM.AddNewSourceBuffer(InputBuffer, SMLoc());
711 Current = InputBuffer->getBufferStart();
712 End = InputBuffer->getBufferEnd();
715 Token &Scanner::peekNext() {
716 // If the current token is a possible simple key, keep parsing until we
718 bool NeedMore = false;
720 if (TokenQueue.empty() || NeedMore) {
721 if (!fetchMoreTokens()) {
723 TokenQueue.push_back(Token());
724 return TokenQueue.front();
727 assert(!TokenQueue.empty() &&
728 "fetchMoreTokens lied about getting tokens!");
730 removeStaleSimpleKeyCandidates();
732 SK.Tok = TokenQueue.front();
733 if (std::find(SimpleKeys.begin(), SimpleKeys.end(), SK)
739 return TokenQueue.front();
742 Token Scanner::getNext() {
743 Token Ret = peekNext();
744 // TokenQueue can be empty if there was an error getting the next token.
745 if (!TokenQueue.empty())
746 TokenQueue.pop_front();
748 // There cannot be any referenced Token's if the TokenQueue is empty. So do a
749 // quick deallocation of them all.
750 if (TokenQueue.empty()) {
751 TokenQueue.Alloc.Reset();
757 StringRef::iterator Scanner::skip_nb_char(StringRef::iterator Position) {
760 // Check 7 bit c-printable - b-char.
761 if ( *Position == 0x09
762 || (*Position >= 0x20 && *Position <= 0x7E))
765 // Check for valid UTF-8.
766 if (uint8_t(*Position) & 0x80) {
767 UTF8Decoded u8d = decodeUTF8(Position);
769 && u8d.first != 0xFEFF
770 && ( u8d.first == 0x85
771 || ( u8d.first >= 0xA0
772 && u8d.first <= 0xD7FF)
773 || ( u8d.first >= 0xE000
774 && u8d.first <= 0xFFFD)
775 || ( u8d.first >= 0x10000
776 && u8d.first <= 0x10FFFF)))
777 return Position + u8d.second;
782 StringRef::iterator Scanner::skip_b_break(StringRef::iterator Position) {
785 if (*Position == 0x0D) {
786 if (Position + 1 != End && *(Position + 1) == 0x0A)
791 if (*Position == 0x0A)
797 StringRef::iterator Scanner::skip_s_white(StringRef::iterator Position) {
800 if (*Position == ' ' || *Position == '\t')
805 StringRef::iterator Scanner::skip_ns_char(StringRef::iterator Position) {
808 if (*Position == ' ' || *Position == '\t')
810 return skip_nb_char(Position);
813 StringRef::iterator Scanner::skip_while( SkipWhileFunc Func
814 , StringRef::iterator Position) {
816 StringRef::iterator i = (this->*Func)(Position);
824 static bool is_ns_hex_digit(const char C) {
825 return (C >= '0' && C <= '9')
826 || (C >= 'a' && C <= 'z')
827 || (C >= 'A' && C <= 'Z');
830 static bool is_ns_word_char(const char C) {
832 || (C >= 'a' && C <= 'z')
833 || (C >= 'A' && C <= 'Z');
836 StringRef Scanner::scan_ns_uri_char() {
837 StringRef::iterator Start = Current;
841 if (( *Current == '%'
843 && is_ns_hex_digit(*(Current + 1))
844 && is_ns_hex_digit(*(Current + 2)))
845 || is_ns_word_char(*Current)
846 || StringRef(Current, 1).find_first_of("#;/?:@&=+$,_.!~*'()[]")
847 != StringRef::npos) {
853 return StringRef(Start, Current - Start);
856 StringRef Scanner::scan_ns_plain_one_line() {
857 StringRef::iterator start = Current;
858 // The first character must already be verified.
861 if (Current == End) {
863 } else if (*Current == ':') {
864 // Check if the next character is a ns-char.
865 if (Current + 1 == End)
867 StringRef::iterator i = skip_ns_char(Current + 1);
868 if (Current + 1 != i) {
870 Column += 2; // Consume both the ':' and ns-char.
873 } else if (*Current == '#') {
874 // Check if the previous character was a ns-char.
875 // The & 0x80 check is to check for the trailing byte of a utf-8
876 if (*(Current - 1) & 0x80 || skip_ns_char(Current - 1) == Current) {
882 StringRef::iterator i = skip_nb_char(Current);
889 return StringRef(start, Current - start);
892 bool Scanner::consume(uint32_t Expected) {
893 if (Expected >= 0x80)
894 report_fatal_error("Not dealing with this yet");
897 if (uint8_t(*Current) >= 0x80)
898 report_fatal_error("Not dealing with this yet");
899 if (uint8_t(*Current) == Expected) {
907 void Scanner::skip(uint32_t Distance) {
912 bool Scanner::isBlankOrBreak(StringRef::iterator Position) {
915 if ( *Position == ' ' || *Position == '\t'
916 || *Position == '\r' || *Position == '\n')
921 void Scanner::saveSimpleKeyCandidate( TokenQueueT::iterator Tok
924 if (IsSimpleKeyAllowed) {
928 SK.Column = AtColumn;
929 SK.IsRequired = IsRequired;
930 SK.FlowLevel = FlowLevel;
931 SimpleKeys.push_back(SK);
935 void Scanner::removeStaleSimpleKeyCandidates() {
936 for (SmallVectorImpl<SimpleKey>::iterator i = SimpleKeys.begin();
937 i != SimpleKeys.end();) {
938 if (i->Line != Line || i->Column + 1024 < Column) {
940 setError( "Could not find expected : for simple key"
941 , i->Tok->Range.begin());
942 i = SimpleKeys.erase(i);
948 void Scanner::removeSimpleKeyCandidatesOnFlowLevel(unsigned Level) {
949 if (!SimpleKeys.empty() && (SimpleKeys.end() - 1)->FlowLevel == Level)
950 SimpleKeys.pop_back();
953 bool Scanner::unrollIndent(int ToColumn) {
955 // Indentation is ignored in flow.
959 while (Indent > ToColumn) {
960 T.Kind = Token::TK_BlockEnd;
961 T.Range = StringRef(Current, 1);
962 TokenQueue.push_back(T);
963 Indent = Indents.pop_back_val();
969 bool Scanner::rollIndent( int ToColumn
970 , Token::TokenKind Kind
971 , TokenQueueT::iterator InsertPoint) {
974 if (Indent < ToColumn) {
975 Indents.push_back(Indent);
980 T.Range = StringRef(Current, 0);
981 TokenQueue.insert(InsertPoint, T);
986 void Scanner::scanToNextToken() {
988 while (*Current == ' ' || *Current == '\t') {
993 if (*Current == '#') {
995 // This may skip more than one byte, thus Column is only incremented
997 StringRef::iterator i = skip_nb_char(Current);
1006 StringRef::iterator i = skip_b_break(Current);
1012 // New lines may start a simple key.
1014 IsSimpleKeyAllowed = true;
1018 bool Scanner::scanStreamStart() {
1019 IsStartOfStream = false;
1021 EncodingInfo EI = getUnicodeEncoding(currentInput());
1024 T.Kind = Token::TK_StreamStart;
1025 T.Range = StringRef(Current, EI.second);
1026 TokenQueue.push_back(T);
1027 Current += EI.second;
1031 bool Scanner::scanStreamEnd() {
1032 // Force an ending new line if one isn't present.
1040 IsSimpleKeyAllowed = false;
1043 T.Kind = Token::TK_StreamEnd;
1044 T.Range = StringRef(Current, 0);
1045 TokenQueue.push_back(T);
1049 bool Scanner::scanDirective() {
1050 // Reset the indentation level.
1053 IsSimpleKeyAllowed = false;
1055 StringRef::iterator Start = Current;
1057 StringRef::iterator NameStart = Current;
1058 Current = skip_while(&Scanner::skip_ns_char, Current);
1059 StringRef Name(NameStart, Current - NameStart);
1060 Current = skip_while(&Scanner::skip_s_white, Current);
1062 if (Name == "YAML") {
1063 Current = skip_while(&Scanner::skip_ns_char, Current);
1065 T.Kind = Token::TK_VersionDirective;
1066 T.Range = StringRef(Start, Current - Start);
1067 TokenQueue.push_back(T);
1073 bool Scanner::scanDocumentIndicator(bool IsStart) {
1076 IsSimpleKeyAllowed = false;
1079 T.Kind = IsStart ? Token::TK_DocumentStart : Token::TK_DocumentEnd;
1080 T.Range = StringRef(Current, 3);
1082 TokenQueue.push_back(T);
1086 bool Scanner::scanFlowCollectionStart(bool IsSequence) {
1088 T.Kind = IsSequence ? Token::TK_FlowSequenceStart
1089 : Token::TK_FlowMappingStart;
1090 T.Range = StringRef(Current, 1);
1092 TokenQueue.push_back(T);
1094 // [ and { may begin a simple key.
1095 saveSimpleKeyCandidate(TokenQueue.back(), Column - 1, false);
1097 // And may also be followed by a simple key.
1098 IsSimpleKeyAllowed = true;
1103 bool Scanner::scanFlowCollectionEnd(bool IsSequence) {
1104 removeSimpleKeyCandidatesOnFlowLevel(FlowLevel);
1105 IsSimpleKeyAllowed = false;
1107 T.Kind = IsSequence ? Token::TK_FlowSequenceEnd
1108 : Token::TK_FlowMappingEnd;
1109 T.Range = StringRef(Current, 1);
1111 TokenQueue.push_back(T);
1117 bool Scanner::scanFlowEntry() {
1118 removeSimpleKeyCandidatesOnFlowLevel(FlowLevel);
1119 IsSimpleKeyAllowed = true;
1121 T.Kind = Token::TK_FlowEntry;
1122 T.Range = StringRef(Current, 1);
1124 TokenQueue.push_back(T);
1128 bool Scanner::scanBlockEntry() {
1129 rollIndent(Column, Token::TK_BlockSequenceStart, TokenQueue.end());
1130 removeSimpleKeyCandidatesOnFlowLevel(FlowLevel);
1131 IsSimpleKeyAllowed = true;
1133 T.Kind = Token::TK_BlockEntry;
1134 T.Range = StringRef(Current, 1);
1136 TokenQueue.push_back(T);
1140 bool Scanner::scanKey() {
1142 rollIndent(Column, Token::TK_BlockMappingStart, TokenQueue.end());
1144 removeSimpleKeyCandidatesOnFlowLevel(FlowLevel);
1145 IsSimpleKeyAllowed = !FlowLevel;
1148 T.Kind = Token::TK_Key;
1149 T.Range = StringRef(Current, 1);
1151 TokenQueue.push_back(T);
1155 bool Scanner::scanValue() {
1156 // If the previous token could have been a simple key, insert the key token
1157 // into the token queue.
1158 if (!SimpleKeys.empty()) {
1159 SimpleKey SK = SimpleKeys.pop_back_val();
1161 T.Kind = Token::TK_Key;
1162 T.Range = SK.Tok->Range;
1163 TokenQueueT::iterator i, e;
1164 for (i = TokenQueue.begin(), e = TokenQueue.end(); i != e; ++i) {
1168 assert(i != e && "SimpleKey not in token queue!");
1169 i = TokenQueue.insert(i, T);
1171 // We may also need to add a Block-Mapping-Start token.
1172 rollIndent(SK.Column, Token::TK_BlockMappingStart, i);
1174 IsSimpleKeyAllowed = false;
1177 rollIndent(Column, Token::TK_BlockMappingStart, TokenQueue.end());
1178 IsSimpleKeyAllowed = !FlowLevel;
1182 T.Kind = Token::TK_Value;
1183 T.Range = StringRef(Current, 1);
1185 TokenQueue.push_back(T);
1189 // Forbidding inlining improves performance by roughly 20%.
1190 // FIXME: Remove once llvm optimizes this to the faster version without hints.
1191 LLVM_ATTRIBUTE_NOINLINE static bool
1192 wasEscaped(StringRef::iterator First, StringRef::iterator Position);
1194 // Returns whether a character at 'Position' was escaped with a leading '\'.
1195 // 'First' specifies the position of the first character in the string.
1196 static bool wasEscaped(StringRef::iterator First,
1197 StringRef::iterator Position) {
1198 assert(Position - 1 >= First);
1199 StringRef::iterator I = Position - 1;
1200 // We calculate the number of consecutive '\'s before the current position
1201 // by iterating backwards through our string.
1202 while (I >= First && *I == '\\') --I;
1203 // (Position - 1 - I) now contains the number of '\'s before the current
1204 // position. If it is odd, the character at 'Position' was escaped.
1205 return (Position - 1 - I) % 2 == 1;
1208 bool Scanner::scanFlowScalar(bool IsDoubleQuoted) {
1209 StringRef::iterator Start = Current;
1210 unsigned ColStart = Column;
1211 if (IsDoubleQuoted) {
1214 while (Current != End && *Current != '"')
1216 // Repeat until the previous character was not a '\' or was an escaped
1218 } while ( Current != End
1219 && *(Current - 1) == '\\'
1220 && wasEscaped(Start + 1, Current));
1224 // Skip a ' followed by another '.
1225 if (Current + 1 < End && *Current == '\'' && *(Current + 1) == '\'') {
1228 } else if (*Current == '\'')
1230 StringRef::iterator i = skip_nb_char(Current);
1232 i = skip_b_break(Current);
1246 skip(1); // Skip ending quote.
1248 T.Kind = Token::TK_Scalar;
1249 T.Range = StringRef(Start, Current - Start);
1250 TokenQueue.push_back(T);
1252 saveSimpleKeyCandidate(TokenQueue.back(), ColStart, false);
1254 IsSimpleKeyAllowed = false;
1259 bool Scanner::scanPlainScalar() {
1260 StringRef::iterator Start = Current;
1261 unsigned ColStart = Column;
1262 unsigned LeadingBlanks = 0;
1263 assert(Indent >= -1 && "Indent must be >= -1 !");
1264 unsigned indent = static_cast<unsigned>(Indent + 1);
1266 if (*Current == '#')
1269 while (!isBlankOrBreak(Current)) {
1270 if ( FlowLevel && *Current == ':'
1271 && !(isBlankOrBreak(Current + 1) || *(Current + 1) == ',')) {
1272 setError("Found unexpected ':' while scanning a plain scalar", Current);
1276 // Check for the end of the plain scalar.
1277 if ( (*Current == ':' && isBlankOrBreak(Current + 1))
1279 && (StringRef(Current, 1).find_first_of(",:?[]{}")
1280 != StringRef::npos)))
1283 StringRef::iterator i = skip_nb_char(Current);
1290 // Are we at the end?
1291 if (!isBlankOrBreak(Current))
1295 StringRef::iterator Tmp = Current;
1296 while (isBlankOrBreak(Tmp)) {
1297 StringRef::iterator i = skip_s_white(Tmp);
1299 if (LeadingBlanks && (Column < indent) && *Tmp == '\t') {
1300 setError("Found invalid tab character in indentation", Tmp);
1306 i = skip_b_break(Tmp);
1315 if (!FlowLevel && Column < indent)
1320 if (Start == Current) {
1321 setError("Got empty plain scalar", Start);
1325 T.Kind = Token::TK_Scalar;
1326 T.Range = StringRef(Start, Current - Start);
1327 TokenQueue.push_back(T);
1329 // Plain scalars can be simple keys.
1330 saveSimpleKeyCandidate(TokenQueue.back(), ColStart, false);
1332 IsSimpleKeyAllowed = false;
1337 bool Scanner::scanAliasOrAnchor(bool IsAlias) {
1338 StringRef::iterator Start = Current;
1339 unsigned ColStart = Column;
1342 if ( *Current == '[' || *Current == ']'
1343 || *Current == '{' || *Current == '}'
1347 StringRef::iterator i = skip_ns_char(Current);
1354 if (Start == Current) {
1355 setError("Got empty alias or anchor", Start);
1360 T.Kind = IsAlias ? Token::TK_Alias : Token::TK_Anchor;
1361 T.Range = StringRef(Start, Current - Start);
1362 TokenQueue.push_back(T);
1364 // Alias and anchors can be simple keys.
1365 saveSimpleKeyCandidate(TokenQueue.back(), ColStart, false);
1367 IsSimpleKeyAllowed = false;
1372 bool Scanner::scanBlockScalar(bool IsLiteral) {
1373 StringRef::iterator Start = Current;
1374 skip(1); // Eat | or >
1376 StringRef::iterator i = skip_nb_char(Current);
1380 i = skip_b_break(Current);
1382 // We got a line break.
1388 // There was an error, which should already have been printed out.
1396 if (Start == Current) {
1397 setError("Got empty block scalar", Start);
1402 T.Kind = Token::TK_Scalar;
1403 T.Range = StringRef(Start, Current - Start);
1404 TokenQueue.push_back(T);
1408 bool Scanner::scanTag() {
1409 StringRef::iterator Start = Current;
1410 unsigned ColStart = Column;
1412 if (Current == End || isBlankOrBreak(Current)); // An empty tag.
1413 else if (*Current == '<') {
1419 // FIXME: Actually parse the c-ns-shorthand-tag rule.
1420 Current = skip_while(&Scanner::skip_ns_char, Current);
1424 T.Kind = Token::TK_Tag;
1425 T.Range = StringRef(Start, Current - Start);
1426 TokenQueue.push_back(T);
1428 // Tags can be simple keys.
1429 saveSimpleKeyCandidate(TokenQueue.back(), ColStart, false);
1431 IsSimpleKeyAllowed = false;
1436 bool Scanner::fetchMoreTokens() {
1437 if (IsStartOfStream)
1438 return scanStreamStart();
1443 return scanStreamEnd();
1445 removeStaleSimpleKeyCandidates();
1447 unrollIndent(Column);
1449 if (Column == 0 && *Current == '%')
1450 return scanDirective();
1452 if (Column == 0 && Current + 4 <= End
1454 && *(Current + 1) == '-'
1455 && *(Current + 2) == '-'
1456 && (Current + 3 == End || isBlankOrBreak(Current + 3)))
1457 return scanDocumentIndicator(true);
1459 if (Column == 0 && Current + 4 <= End
1461 && *(Current + 1) == '.'
1462 && *(Current + 2) == '.'
1463 && (Current + 3 == End || isBlankOrBreak(Current + 3)))
1464 return scanDocumentIndicator(false);
1466 if (*Current == '[')
1467 return scanFlowCollectionStart(true);
1469 if (*Current == '{')
1470 return scanFlowCollectionStart(false);
1472 if (*Current == ']')
1473 return scanFlowCollectionEnd(true);
1475 if (*Current == '}')
1476 return scanFlowCollectionEnd(false);
1478 if (*Current == ',')
1479 return scanFlowEntry();
1481 if (*Current == '-' && isBlankOrBreak(Current + 1))
1482 return scanBlockEntry();
1484 if (*Current == '?' && (FlowLevel || isBlankOrBreak(Current + 1)))
1487 if (*Current == ':' && (FlowLevel || isBlankOrBreak(Current + 1)))
1490 if (*Current == '*')
1491 return scanAliasOrAnchor(true);
1493 if (*Current == '&')
1494 return scanAliasOrAnchor(false);
1496 if (*Current == '!')
1499 if (*Current == '|' && !FlowLevel)
1500 return scanBlockScalar(true);
1502 if (*Current == '>' && !FlowLevel)
1503 return scanBlockScalar(false);
1505 if (*Current == '\'')
1506 return scanFlowScalar(false);
1508 if (*Current == '"')
1509 return scanFlowScalar(true);
1511 // Get a plain scalar.
1512 StringRef FirstChar(Current, 1);
1513 if (!(isBlankOrBreak(Current)
1514 || FirstChar.find_first_of("-?:,[]{}#&*!|>'\"%@`") != StringRef::npos)
1515 || (*Current == '-' && !isBlankOrBreak(Current + 1))
1516 || (!FlowLevel && (*Current == '?' || *Current == ':')
1517 && isBlankOrBreak(Current + 1))
1518 || (!FlowLevel && *Current == ':'
1519 && Current + 2 < End
1520 && *(Current + 1) == ':'
1521 && !isBlankOrBreak(Current + 2)))
1522 return scanPlainScalar();
1524 setError("Unrecognized character while tokenizing.");
1528 Stream::Stream(StringRef Input, SourceMgr &SM)
1529 : scanner(new Scanner(Input, SM))
1532 Stream::~Stream() {}
1534 bool Stream::failed() { return scanner->failed(); }
1536 void Stream::printError(Node *N, const Twine &Msg) {
1537 SmallVector<SMRange, 1> Ranges;
1538 Ranges.push_back(N->getSourceRange());
1539 scanner->printError( N->getSourceRange().Start
1540 , SourceMgr::DK_Error
1545 void Stream::handleYAMLDirective(const Token &t) {
1546 // TODO: Ensure version is 1.x.
1549 document_iterator Stream::begin() {
1551 report_fatal_error("Can only iterate over the stream once");
1553 // Skip Stream-Start.
1556 CurrentDoc.reset(new Document(*this));
1557 return document_iterator(CurrentDoc);
1560 document_iterator Stream::end() {
1561 return document_iterator();
1564 void Stream::skip() {
1565 for (document_iterator i = begin(), e = end(); i != e; ++i)
1569 Node::Node(unsigned int Type, OwningPtr<Document> &D, StringRef A)
1573 SMLoc Start = SMLoc::getFromPointer(peekNext().Range.begin());
1574 SourceRange = SMRange(Start, Start);
1577 Token &Node::peekNext() {
1578 return Doc->peekNext();
1581 Token Node::getNext() {
1582 return Doc->getNext();
1585 Node *Node::parseBlockNode() {
1586 return Doc->parseBlockNode();
1589 BumpPtrAllocator &Node::getAllocator() {
1590 return Doc->NodeAllocator;
1593 void Node::setError(const Twine &Msg, Token &Tok) const {
1594 Doc->setError(Msg, Tok);
1597 bool Node::failed() const {
1598 return Doc->failed();
1603 StringRef ScalarNode::getValue(SmallVectorImpl<char> &Storage) const {
1604 // TODO: Handle newlines properly. We need to remove leading whitespace.
1605 if (Value[0] == '"') { // Double quoted.
1606 // Pull off the leading and trailing "s.
1607 StringRef UnquotedValue = Value.substr(1, Value.size() - 2);
1608 // Search for characters that would require unescaping the value.
1609 StringRef::size_type i = UnquotedValue.find_first_of("\\\r\n");
1610 if (i != StringRef::npos)
1611 return unescapeDoubleQuoted(UnquotedValue, i, Storage);
1612 return UnquotedValue;
1613 } else if (Value[0] == '\'') { // Single quoted.
1614 // Pull off the leading and trailing 's.
1615 StringRef UnquotedValue = Value.substr(1, Value.size() - 2);
1616 StringRef::size_type i = UnquotedValue.find('\'');
1617 if (i != StringRef::npos) {
1618 // We're going to need Storage.
1620 Storage.reserve(UnquotedValue.size());
1621 for (; i != StringRef::npos; i = UnquotedValue.find('\'')) {
1622 StringRef Valid(UnquotedValue.begin(), i);
1623 Storage.insert(Storage.end(), Valid.begin(), Valid.end());
1624 Storage.push_back('\'');
1625 UnquotedValue = UnquotedValue.substr(i + 2);
1627 Storage.insert(Storage.end(), UnquotedValue.begin(), UnquotedValue.end());
1628 return StringRef(Storage.begin(), Storage.size());
1630 return UnquotedValue;
1633 return Value.rtrim(" ");
1636 StringRef ScalarNode::unescapeDoubleQuoted( StringRef UnquotedValue
1637 , StringRef::size_type i
1638 , SmallVectorImpl<char> &Storage)
1640 // Use Storage to build proper value.
1642 Storage.reserve(UnquotedValue.size());
1643 for (; i != StringRef::npos; i = UnquotedValue.find_first_of("\\\r\n")) {
1644 // Insert all previous chars into Storage.
1645 StringRef Valid(UnquotedValue.begin(), i);
1646 Storage.insert(Storage.end(), Valid.begin(), Valid.end());
1647 // Chop off inserted chars.
1648 UnquotedValue = UnquotedValue.substr(i);
1650 assert(!UnquotedValue.empty() && "Can't be empty!");
1652 // Parse escape or line break.
1653 switch (UnquotedValue[0]) {
1656 Storage.push_back('\n');
1657 if ( UnquotedValue.size() > 1
1658 && (UnquotedValue[1] == '\r' || UnquotedValue[1] == '\n'))
1659 UnquotedValue = UnquotedValue.substr(1);
1660 UnquotedValue = UnquotedValue.substr(1);
1663 if (UnquotedValue.size() == 1)
1664 // TODO: Report error.
1666 UnquotedValue = UnquotedValue.substr(1);
1667 switch (UnquotedValue[0]) {
1670 T.Range = StringRef(UnquotedValue.begin(), 1);
1671 setError("Unrecognized escape code!", T);
1676 // Remove the new line.
1677 if ( UnquotedValue.size() > 1
1678 && (UnquotedValue[1] == '\r' || UnquotedValue[1] == '\n'))
1679 UnquotedValue = UnquotedValue.substr(1);
1680 // If this was just a single byte newline, it will get skipped
1684 Storage.push_back(0x00);
1687 Storage.push_back(0x07);
1690 Storage.push_back(0x08);
1694 Storage.push_back(0x09);
1697 Storage.push_back(0x0A);
1700 Storage.push_back(0x0B);
1703 Storage.push_back(0x0C);
1706 Storage.push_back(0x0D);
1709 Storage.push_back(0x1B);
1712 Storage.push_back(0x20);
1715 Storage.push_back(0x22);
1718 Storage.push_back(0x2F);
1721 Storage.push_back(0x5C);
1724 encodeUTF8(0x85, Storage);
1727 encodeUTF8(0xA0, Storage);
1730 encodeUTF8(0x2028, Storage);
1733 encodeUTF8(0x2029, Storage);
1736 if (UnquotedValue.size() < 3)
1737 // TODO: Report error.
1739 unsigned int UnicodeScalarValue;
1740 if (UnquotedValue.substr(1, 2).getAsInteger(16, UnicodeScalarValue))
1741 // TODO: Report error.
1742 UnicodeScalarValue = 0xFFFD;
1743 encodeUTF8(UnicodeScalarValue, Storage);
1744 UnquotedValue = UnquotedValue.substr(2);
1748 if (UnquotedValue.size() < 5)
1749 // TODO: Report error.
1751 unsigned int UnicodeScalarValue;
1752 if (UnquotedValue.substr(1, 4).getAsInteger(16, UnicodeScalarValue))
1753 // TODO: Report error.
1754 UnicodeScalarValue = 0xFFFD;
1755 encodeUTF8(UnicodeScalarValue, Storage);
1756 UnquotedValue = UnquotedValue.substr(4);
1760 if (UnquotedValue.size() < 9)
1761 // TODO: Report error.
1763 unsigned int UnicodeScalarValue;
1764 if (UnquotedValue.substr(1, 8).getAsInteger(16, UnicodeScalarValue))
1765 // TODO: Report error.
1766 UnicodeScalarValue = 0xFFFD;
1767 encodeUTF8(UnicodeScalarValue, Storage);
1768 UnquotedValue = UnquotedValue.substr(8);
1772 UnquotedValue = UnquotedValue.substr(1);
1775 Storage.insert(Storage.end(), UnquotedValue.begin(), UnquotedValue.end());
1776 return StringRef(Storage.begin(), Storage.size());
1779 Node *KeyValueNode::getKey() {
1782 // Handle implicit null keys.
1784 Token &t = peekNext();
1785 if ( t.Kind == Token::TK_BlockEnd
1786 || t.Kind == Token::TK_Value
1787 || t.Kind == Token::TK_Error) {
1788 return Key = new (getAllocator()) NullNode(Doc);
1790 if (t.Kind == Token::TK_Key)
1791 getNext(); // skip TK_Key.
1794 // Handle explicit null keys.
1795 Token &t = peekNext();
1796 if (t.Kind == Token::TK_BlockEnd || t.Kind == Token::TK_Value) {
1797 return Key = new (getAllocator()) NullNode(Doc);
1800 // We've got a normal key.
1801 return Key = parseBlockNode();
1804 Node *KeyValueNode::getValue() {
1809 return Value = new (getAllocator()) NullNode(Doc);
1811 // Handle implicit null values.
1813 Token &t = peekNext();
1814 if ( t.Kind == Token::TK_BlockEnd
1815 || t.Kind == Token::TK_FlowMappingEnd
1816 || t.Kind == Token::TK_Key
1817 || t.Kind == Token::TK_FlowEntry
1818 || t.Kind == Token::TK_Error) {
1819 return Value = new (getAllocator()) NullNode(Doc);
1822 if (t.Kind != Token::TK_Value) {
1823 setError("Unexpected token in Key Value.", t);
1824 return Value = new (getAllocator()) NullNode(Doc);
1826 getNext(); // skip TK_Value.
1829 // Handle explicit null values.
1830 Token &t = peekNext();
1831 if (t.Kind == Token::TK_BlockEnd || t.Kind == Token::TK_Key) {
1832 return Value = new (getAllocator()) NullNode(Doc);
1835 // We got a normal value.
1836 return Value = parseBlockNode();
1839 void MappingNode::increment() {
1846 CurrentEntry->skip();
1847 if (Type == MT_Inline) {
1853 Token T = peekNext();
1854 if (T.Kind == Token::TK_Key || T.Kind == Token::TK_Scalar) {
1855 // KeyValueNode eats the TK_Key. That way it can detect null keys.
1856 CurrentEntry = new (getAllocator()) KeyValueNode(Doc);
1857 } else if (Type == MT_Block) {
1859 case Token::TK_BlockEnd:
1865 setError("Unexpected token. Expected Key or Block End", T);
1866 case Token::TK_Error:
1872 case Token::TK_FlowEntry:
1873 // Eat the flow entry and recurse.
1876 case Token::TK_FlowMappingEnd:
1878 case Token::TK_Error:
1879 // Set this to end iterator.
1884 setError( "Unexpected token. Expected Key, Flow Entry, or Flow "
1893 void SequenceNode::increment() {
1900 CurrentEntry->skip();
1901 Token T = peekNext();
1902 if (SeqType == ST_Block) {
1904 case Token::TK_BlockEntry:
1906 CurrentEntry = parseBlockNode();
1907 if (CurrentEntry == 0) { // An error occurred.
1912 case Token::TK_BlockEnd:
1918 setError( "Unexpected token. Expected Block Entry or Block End."
1920 case Token::TK_Error:
1924 } else if (SeqType == ST_Indentless) {
1926 case Token::TK_BlockEntry:
1928 CurrentEntry = parseBlockNode();
1929 if (CurrentEntry == 0) { // An error occurred.
1935 case Token::TK_Error:
1939 } else if (SeqType == ST_Flow) {
1941 case Token::TK_FlowEntry:
1942 // Eat the flow entry and recurse.
1944 WasPreviousTokenFlowEntry = true;
1946 case Token::TK_FlowSequenceEnd:
1948 case Token::TK_Error:
1949 // Set this to end iterator.
1953 case Token::TK_StreamEnd:
1954 case Token::TK_DocumentEnd:
1955 case Token::TK_DocumentStart:
1956 setError("Could not find closing ]!", T);
1957 // Set this to end iterator.
1962 if (!WasPreviousTokenFlowEntry) {
1963 setError("Expected , between entries!", T);
1968 // Otherwise it must be a flow entry.
1969 CurrentEntry = parseBlockNode();
1970 if (!CurrentEntry) {
1973 WasPreviousTokenFlowEntry = false;
1979 Document::Document(Stream &S) : stream(S), Root(0) {
1980 if (parseDirectives())
1981 expectToken(Token::TK_DocumentStart);
1982 Token &T = peekNext();
1983 if (T.Kind == Token::TK_DocumentStart)
1987 bool Document::skip() {
1988 if (stream.scanner->failed())
1993 Token &T = peekNext();
1994 if (T.Kind == Token::TK_StreamEnd)
1996 if (T.Kind == Token::TK_DocumentEnd) {
2003 Token &Document::peekNext() {
2004 return stream.scanner->peekNext();
2007 Token Document::getNext() {
2008 return stream.scanner->getNext();
2011 void Document::setError(const Twine &Message, Token &Location) const {
2012 stream.scanner->setError(Message, Location.Range.begin());
2015 bool Document::failed() const {
2016 return stream.scanner->failed();
2019 Node *Document::parseBlockNode() {
2020 Token T = peekNext();
2021 // Handle properties.
2025 case Token::TK_Alias:
2027 return new (NodeAllocator) AliasNode(stream.CurrentDoc, T.Range.substr(1));
2028 case Token::TK_Anchor:
2029 if (AnchorInfo.Kind == Token::TK_Anchor) {
2030 setError("Already encountered an anchor for this node!", T);
2033 AnchorInfo = getNext(); // Consume TK_Anchor.
2035 goto parse_property;
2037 getNext(); // Skip TK_Tag.
2039 goto parse_property;
2045 case Token::TK_BlockEntry:
2046 // We got an unindented BlockEntry sequence. This is not terminated with
2048 // Don't eat the TK_BlockEntry, SequenceNode needs it.
2049 return new (NodeAllocator) SequenceNode( stream.CurrentDoc
2050 , AnchorInfo.Range.substr(1)
2051 , SequenceNode::ST_Indentless);
2052 case Token::TK_BlockSequenceStart:
2054 return new (NodeAllocator)
2055 SequenceNode( stream.CurrentDoc
2056 , AnchorInfo.Range.substr(1)
2057 , SequenceNode::ST_Block);
2058 case Token::TK_BlockMappingStart:
2060 return new (NodeAllocator)
2061 MappingNode( stream.CurrentDoc
2062 , AnchorInfo.Range.substr(1)
2063 , MappingNode::MT_Block);
2064 case Token::TK_FlowSequenceStart:
2066 return new (NodeAllocator)
2067 SequenceNode( stream.CurrentDoc
2068 , AnchorInfo.Range.substr(1)
2069 , SequenceNode::ST_Flow);
2070 case Token::TK_FlowMappingStart:
2072 return new (NodeAllocator)
2073 MappingNode( stream.CurrentDoc
2074 , AnchorInfo.Range.substr(1)
2075 , MappingNode::MT_Flow);
2076 case Token::TK_Scalar:
2078 return new (NodeAllocator)
2079 ScalarNode( stream.CurrentDoc
2080 , AnchorInfo.Range.substr(1)
2083 // Don't eat the TK_Key, KeyValueNode expects it.
2084 return new (NodeAllocator)
2085 MappingNode( stream.CurrentDoc
2086 , AnchorInfo.Range.substr(1)
2087 , MappingNode::MT_Inline);
2088 case Token::TK_DocumentStart:
2089 case Token::TK_DocumentEnd:
2090 case Token::TK_StreamEnd:
2092 // TODO: Properly handle tags. "[!!str ]" should resolve to !!str "", not
2094 return new (NodeAllocator) NullNode(stream.CurrentDoc);
2095 case Token::TK_Error:
2098 llvm_unreachable("Control flow shouldn't reach here.");
2102 bool Document::parseDirectives() {
2103 bool isDirective = false;
2105 Token T = peekNext();
2106 if (T.Kind == Token::TK_TagDirective) {
2107 handleTagDirective(getNext());
2109 } else if (T.Kind == Token::TK_VersionDirective) {
2110 stream.handleYAMLDirective(getNext());
2118 bool Document::expectToken(int TK) {
2119 Token T = getNext();
2121 setError("Unexpected token", T);