1//===--- YAMLParser.cpp - Simple YAML parser ------------------------------===//
2//
3//                     The LLVM Compiler Infrastructure
4//
5// This file is distributed under the University of Illinois Open Source
6// License. See LICENSE.TXT for details.
7//
8//===----------------------------------------------------------------------===//
9//
10//  This file implements a YAML parser.
11//
12//===----------------------------------------------------------------------===//
13
14#include "llvm/Support/YAMLParser.h"
15
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"
25
26using namespace llvm;
27using namespace yaml;
28
29enum 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.
36};
37
38/// EncodingInfo - Holds the encoding type and length of the byte order mark if
39///                it exists. Length is in {0, 2, 3, 4}.
40typedef std::pair<UnicodeEncodingForm, unsigned> EncodingInfo;
41
42/// getUnicodeEncoding - Reads up to the first 4 bytes to determine the Unicode
43///                      encoding form of \a Input.
44///
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.
48static EncodingInfo getUnicodeEncoding(StringRef Input) {
49  if (Input.size() == 0)
50    return std::make_pair(UEF_Unknown, 0);
51
52  switch (uint8_t(Input[0])) {
53  case 0x00:
54    if (Input.size() >= 4) {
55      if (  Input[1] == 0
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);
61    }
62
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);
66  case 0xFF:
67    if (  Input.size() >= 4
68       && uint8_t(Input[1]) == 0xFE
69       && Input[2] == 0
70       && Input[3] == 0)
71      return std::make_pair(UEF_UTF32_LE, 4);
72
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);
76  case 0xFE:
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);
80  case 0xEF:
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);
86  }
87
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);
91
92  if (Input.size() >= 2 && Input[1] == 0)
93    return std::make_pair(UEF_UTF16_LE, 0);
94
95  return std::make_pair(UEF_UTF8, 0);
96}
97
98namespace llvm {
99namespace yaml {
100/// Token - A single YAML token.
101struct Token : ilist_node<Token> {
102  enum TokenKind {
103    TK_Error, // Uninitialized token.
104    TK_StreamStart,
105    TK_StreamEnd,
106    TK_VersionDirective,
107    TK_TagDirective,
108    TK_DocumentStart,
109    TK_DocumentEnd,
110    TK_BlockEntry,
111    TK_BlockEnd,
112    TK_BlockSequenceStart,
113    TK_BlockMappingStart,
114    TK_FlowEntry,
115    TK_FlowSequenceStart,
116    TK_FlowSequenceEnd,
117    TK_FlowMappingStart,
118    TK_FlowMappingEnd,
119    TK_Key,
120    TK_Value,
121    TK_Scalar,
122    TK_Alias,
123    TK_Anchor,
124    TK_Tag
125  } Kind;
126
127  /// A string of length 0 or more whose begin() points to the logical location
128  /// of the token in the input.
129  StringRef Range;
130
131  Token() : Kind(TK_Error) {}
132};
133}
134}
135
136namespace llvm {
137template<>
138struct ilist_sentinel_traits<Token> {
139  Token *createSentinel() const {
140    return &Sentinel;
141  }
142  static void destroySentinel(Token*) {}
143
144  Token *provideInitialHead() const { return createSentinel(); }
145  Token *ensureHead(Token*) const { return createSentinel(); }
146  static void noteHead(Token*, Token*) {}
147
148private:
149  mutable Token Sentinel;
150};
151
152template<>
153struct ilist_node_traits<Token> {
154  Token *createNode(const Token &V) {
155    return new (Alloc.Allocate<Token>()) Token(V);
156  }
157  static void deleteNode(Token *V) {}
158
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*/) {}
164
165  BumpPtrAllocator Alloc;
166};
167}
168
169typedef ilist<Token> TokenQueueT;
170
171namespace {
172/// @brief This struct is used to track simple keys.
173///
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.
181struct SimpleKey {
182  TokenQueueT::iterator Tok;
183  unsigned Column;
184  unsigned Line;
185  unsigned FlowLevel;
186  bool IsRequired;
187
188  bool operator ==(const SimpleKey &Other) {
189    return Tok == Other.Tok;
190  }
191};
192}
193
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.
197typedef std::pair<uint32_t, unsigned> UTF8Decoded;
198
199static 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);
206  }
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);
216  }
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);
231  }
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);
245  }
246  return std::make_pair(0, 0);
247}
248
249namespace llvm {
250namespace yaml {
251/// @brief Scans YAML tokens from a MemoryBuffer.
252class Scanner {
253public:
254  Scanner(const StringRef Input, SourceMgr &SM);
255
256  /// @brief Parse the next token and return it without popping it.
257  Token &peekNext();
258
259  /// @brief Parse the next token and pop it from the queue.
260  Token getNext();
261
262  void printError(SMLoc Loc, SourceMgr::DiagKind Kind, const Twine &Message,
263                  ArrayRef<SMRange> Ranges = ArrayRef<SMRange>()) {
264    SM.PrintMessage(Loc, Kind, Message, Ranges);
265  }
266
267  void setError(const Twine &Message, StringRef::iterator Position) {
268    if (Current >= End)
269      Current = End - 1;
270
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.
273    if (!Failed)
274      printError(SMLoc::getFromPointer(Current), SourceMgr::DK_Error, Message);
275    Failed = true;
276  }
277
278  void setError(const Twine &Message) {
279    setError(Message, Current);
280  }
281
282  /// @brief Returns true if an error occurred while parsing.
283  bool failed() {
284    return Failed;
285  }
286
287private:
288  StringRef currentInput() {
289    return StringRef(Current, End - Current);
290  }
291
292  /// @brief Decode a UTF-8 minimal well-formed code unit subsequence starting
293  ///        at \a Position.
294  ///
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
297  /// is 0.
298  UTF8Decoded decodeUTF8(StringRef::iterator Position) {
299    return ::decodeUTF8(StringRef(Position, End - Position));
300  }
301
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
305  // the spec.
306  //
307  // See 4.2 [Production Naming Conventions] for the meaning of the prefixes.
308  //
309  // c-
310  //   A production starting and ending with a special character.
311  // b-
312  //   A production matching a single line break.
313  // nb-
314  //   A production starting and ending with a non-break character.
315  // s-
316  //   A production starting and ending with a white space character.
317  // ns-
318  //   A production starting and ending with a non-space character.
319  // l-
320  //   A production matching complete line(s).
321
322  /// @brief Skip a single nb-char[27] starting at Position.
323  ///
324  /// A nb-char is 0x9 | [0x20-0x7E] | 0x85 | [0xA0-0xD7FF] | [0xE000-0xFEFE]
325  ///                  | [0xFF00-0xFFFD] | [0x10000-0x10FFFF]
326  ///
327  /// @returns The code unit after the nb-char, or Position if it's not an
328  ///          nb-char.
329  StringRef::iterator skip_nb_char(StringRef::iterator Position);
330
331  /// @brief Skip a single b-break[28] starting at Position.
332  ///
333  /// A b-break is 0xD 0xA | 0xD | 0xA
334  ///
335  /// @returns The code unit after the b-break, or Position if it's not a
336  ///          b-break.
337  StringRef::iterator skip_b_break(StringRef::iterator Position);
338
339  /// @brief Skip a single s-white[33] starting at Position.
340  ///
341  /// A s-white is 0x20 | 0x9
342  ///
343  /// @returns The code unit after the s-white, or Position if it's not a
344  ///          s-white.
345  StringRef::iterator skip_s_white(StringRef::iterator Position);
346
347  /// @brief Skip a single ns-char[34] starting at Position.
348  ///
349  /// A ns-char is nb-char - s-white
350  ///
351  /// @returns The code unit after the ns-char, or Position if it's not a
352  ///          ns-char.
353  StringRef::iterator skip_ns_char(StringRef::iterator Position);
354
355  typedef StringRef::iterator (Scanner::*SkipWhileFunc)(StringRef::iterator);
356  /// @brief Skip minimal well-formed code unit subsequences until Func
357  ///        returns its input.
358  ///
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);
363
364  /// @brief Scan ns-uri-char[39]s starting at Cur.
365  ///
366  /// This updates Cur and Column while scanning.
367  ///
368  /// @returns A StringRef starting at Cur which covers the longest contiguous
369  ///          sequence of ns-uri-char.
370  StringRef scan_ns_uri_char();
371
372  /// @brief Scan ns-plain-one-line[133] starting at \a Cur.
373  StringRef scan_ns_plain_one_line();
374
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);
379
380  /// @brief Skip \a Distance UTF-8 code units. Updates \a Cur and \a Column.
381  void skip(uint32_t Distance);
382
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);
386
387  /// @brief If IsSimpleKeyAllowed, create and push_back a new SimpleKey.
388  void saveSimpleKeyCandidate( TokenQueueT::iterator Tok
389                             , unsigned AtColumn
390                             , bool IsRequired);
391
392  /// @brief Remove simple keys that can no longer be valid simple keys.
393  ///
394  /// Invalid simple keys are not on the current line or are further than 1024
395  /// columns back.
396  void removeStaleSimpleKeyCandidates();
397
398  /// @brief Remove all simple keys on FlowLevel \a Level.
399  void removeSimpleKeyCandidatesOnFlowLevel(unsigned Level);
400
401  /// @brief Unroll indentation in \a Indents back to \a Col. Creates BlockEnd
402  ///        tokens if needed.
403  bool unrollIndent(int ToColumn);
404
405  /// @brief Increase indent to \a Col. Creates \a Kind token at \a InsertPoint
406  ///        if needed.
407  bool rollIndent( int ToColumn
408                 , Token::TokenKind Kind
409                 , TokenQueueT::iterator InsertPoint);
410
411  /// @brief Skip whitespace and comments until the start of the next token.
412  void scanToNextToken();
413
414  /// @brief Must be the first token generated.
415  bool scanStreamStart();
416
417  /// @brief Generate tokens needed to close out the stream.
418  bool scanStreamEnd();
419
420  /// @brief Scan a %BLAH directive.
421  bool scanDirective();
422
423  /// @brief Scan a ... or ---.
424  bool scanDocumentIndicator(bool IsStart);
425
426  /// @brief Scan a [ or { and generate the proper flow collection start token.
427  bool scanFlowCollectionStart(bool IsSequence);
428
429  /// @brief Scan a ] or } and generate the proper flow collection end token.
430  bool scanFlowCollectionEnd(bool IsSequence);
431
432  /// @brief Scan the , that separates entries in a flow collection.
433  bool scanFlowEntry();
434
435  /// @brief Scan the - that starts block sequence entries.
436  bool scanBlockEntry();
437
438  /// @brief Scan an explicit ? indicating a key.
439  bool scanKey();
440
441  /// @brief Scan an explicit : indicating a value.
442  bool scanValue();
443
444  /// @brief Scan a quoted scalar.
445  bool scanFlowScalar(bool IsDoubleQuoted);
446
447  /// @brief Scan an unquoted scalar.
448  bool scanPlainScalar();
449
450  /// @brief Scan an Alias or Anchor starting with * or &.
451  bool scanAliasOrAnchor(bool IsAlias);
452
453  /// @brief Scan a block scalar starting with | or >.
454  bool scanBlockScalar(bool IsLiteral);
455
456  /// @brief Scan a tag of the form !stuff.
457  bool scanTag();
458
459  /// @brief Dispatch to the next scanning function based on \a *Cur.
460  bool fetchMoreTokens();
461
462  /// @brief The SourceMgr used for diagnostics and buffer management.
463  SourceMgr &SM;
464
465  /// @brief The original input.
466  MemoryBuffer *InputBuffer;
467
468  /// @brief The current position of the scanner.
469  StringRef::iterator Current;
470
471  /// @brief The end of the input (one past the last character).
472  StringRef::iterator End;
473
474  /// @brief Current YAML indentation level in spaces.
475  int Indent;
476
477  /// @brief Current column number in Unicode code points.
478  unsigned Column;
479
480  /// @brief Current line number.
481  unsigned Line;
482
483  /// @brief How deep we are in flow style containers. 0 Means at block level.
484  unsigned FlowLevel;
485
486  /// @brief Are we at the start of the stream?
487  bool IsStartOfStream;
488
489  /// @brief Can the next token be the start of a simple key?
490  bool IsSimpleKeyAllowed;
491
492  /// @brief True if an error has occurred.
493  bool Failed;
494
495  /// @brief Queue of tokens. This is required to queue up tokens while looking
496  ///        for the end of a simple key. And for cases where a single character
497  ///        can produce multiple tokens (e.g. BlockEnd).
498  TokenQueueT TokenQueue;
499
500  /// @brief Indentation levels.
501  SmallVector<int, 4> Indents;
502
503  /// @brief Potential simple keys.
504  SmallVector<SimpleKey, 4> SimpleKeys;
505};
506
507} // end namespace yaml
508} // end namespace llvm
509
510/// encodeUTF8 - Encode \a UnicodeScalarValue in UTF-8 and append it to result.
511static void encodeUTF8( uint32_t UnicodeScalarValue
512                      , SmallVectorImpl<char> &Result) {
513  if (UnicodeScalarValue <= 0x7F) {
514    Result.push_back(UnicodeScalarValue & 0x7F);
515  } else if (UnicodeScalarValue <= 0x7FF) {
516    uint8_t FirstByte = 0xC0 | ((UnicodeScalarValue & 0x7C0) >> 6);
517    uint8_t SecondByte = 0x80 | (UnicodeScalarValue & 0x3F);
518    Result.push_back(FirstByte);
519    Result.push_back(SecondByte);
520  } else if (UnicodeScalarValue <= 0xFFFF) {
521    uint8_t FirstByte = 0xE0 | ((UnicodeScalarValue & 0xF000) >> 12);
522    uint8_t SecondByte = 0x80 | ((UnicodeScalarValue & 0xFC0) >> 6);
523    uint8_t ThirdByte = 0x80 | (UnicodeScalarValue & 0x3F);
524    Result.push_back(FirstByte);
525    Result.push_back(SecondByte);
526    Result.push_back(ThirdByte);
527  } else if (UnicodeScalarValue <= 0x10FFFF) {
528    uint8_t FirstByte = 0xF0 | ((UnicodeScalarValue & 0x1F0000) >> 18);
529    uint8_t SecondByte = 0x80 | ((UnicodeScalarValue & 0x3F000) >> 12);
530    uint8_t ThirdByte = 0x80 | ((UnicodeScalarValue & 0xFC0) >> 6);
531    uint8_t FourthByte = 0x80 | (UnicodeScalarValue & 0x3F);
532    Result.push_back(FirstByte);
533    Result.push_back(SecondByte);
534    Result.push_back(ThirdByte);
535    Result.push_back(FourthByte);
536  }
537}
538
539bool yaml::dumpTokens(StringRef Input, raw_ostream &OS) {
540  SourceMgr SM;
541  Scanner scanner(Input, SM);
542  while (true) {
543    Token T = scanner.getNext();
544    switch (T.Kind) {
545    case Token::TK_StreamStart:
546      OS << "Stream-Start: ";
547      break;
548    case Token::TK_StreamEnd:
549      OS << "Stream-End: ";
550      break;
551    case Token::TK_VersionDirective:
552      OS << "Version-Directive: ";
553      break;
554    case Token::TK_TagDirective:
555      OS << "Tag-Directive: ";
556      break;
557    case Token::TK_DocumentStart:
558      OS << "Document-Start: ";
559      break;
560    case Token::TK_DocumentEnd:
561      OS << "Document-End: ";
562      break;
563    case Token::TK_BlockEntry:
564      OS << "Block-Entry: ";
565      break;
566    case Token::TK_BlockEnd:
567      OS << "Block-End: ";
568      break;
569    case Token::TK_BlockSequenceStart:
570      OS << "Block-Sequence-Start: ";
571      break;
572    case Token::TK_BlockMappingStart:
573      OS << "Block-Mapping-Start: ";
574      break;
575    case Token::TK_FlowEntry:
576      OS << "Flow-Entry: ";
577      break;
578    case Token::TK_FlowSequenceStart:
579      OS << "Flow-Sequence-Start: ";
580      break;
581    case Token::TK_FlowSequenceEnd:
582      OS << "Flow-Sequence-End: ";
583      break;
584    case Token::TK_FlowMappingStart:
585      OS << "Flow-Mapping-Start: ";
586      break;
587    case Token::TK_FlowMappingEnd:
588      OS << "Flow-Mapping-End: ";
589      break;
590    case Token::TK_Key:
591      OS << "Key: ";
592      break;
593    case Token::TK_Value:
594      OS << "Value: ";
595      break;
596    case Token::TK_Scalar:
597      OS << "Scalar: ";
598      break;
599    case Token::TK_Alias:
600      OS << "Alias: ";
601      break;
602    case Token::TK_Anchor:
603      OS << "Anchor: ";
604      break;
605    case Token::TK_Tag:
606      OS << "Tag: ";
607      break;
608    case Token::TK_Error:
609      break;
610    }
611    OS << T.Range << "\n";
612    if (T.Kind == Token::TK_StreamEnd)
613      break;
614    else if (T.Kind == Token::TK_Error)
615      return false;
616  }
617  return true;
618}
619
620bool yaml::scanTokens(StringRef Input) {
621  llvm::SourceMgr SM;
622  llvm::yaml::Scanner scanner(Input, SM);
623  for (;;) {
624    llvm::yaml::Token T = scanner.getNext();
625    if (T.Kind == Token::TK_StreamEnd)
626      break;
627    else if (T.Kind == Token::TK_Error)
628      return false;
629  }
630  return true;
631}
632
633std::string yaml::escape(StringRef Input) {
634  std::string EscapedInput;
635  for (StringRef::iterator i = Input.begin(), e = Input.end(); i != e; ++i) {
636    if (*i == '\\')
637      EscapedInput += "\\\\";
638    else if (*i == '"')
639      EscapedInput += "\\\"";
640    else if (*i == 0)
641      EscapedInput += "\\0";
642    else if (*i == 0x07)
643      EscapedInput += "\\a";
644    else if (*i == 0x08)
645      EscapedInput += "\\b";
646    else if (*i == 0x09)
647      EscapedInput += "\\t";
648    else if (*i == 0x0A)
649      EscapedInput += "\\n";
650    else if (*i == 0x0B)
651      EscapedInput += "\\v";
652    else if (*i == 0x0C)
653      EscapedInput += "\\f";
654    else if (*i == 0x0D)
655      EscapedInput += "\\r";
656    else if (*i == 0x1B)
657      EscapedInput += "\\e";
658    else if ((unsigned char)*i < 0x20) { // Control characters not handled above.
659      std::string HexStr = utohexstr(*i);
660      EscapedInput += "\\x" + std::string(2 - HexStr.size(), '0') + HexStr;
661    } else if (*i & 0x80) { // UTF-8 multiple code unit subsequence.
662      UTF8Decoded UnicodeScalarValue
663        = decodeUTF8(StringRef(i, Input.end() - i));
664      if (UnicodeScalarValue.second == 0) {
665        // Found invalid char.
666        SmallString<4> Val;
667        encodeUTF8(0xFFFD, Val);
668        EscapedInput.insert(EscapedInput.end(), Val.begin(), Val.end());
669        // FIXME: Error reporting.
670        return EscapedInput;
671      }
672      if (UnicodeScalarValue.first == 0x85)
673        EscapedInput += "\\N";
674      else if (UnicodeScalarValue.first == 0xA0)
675        EscapedInput += "\\_";
676      else if (UnicodeScalarValue.first == 0x2028)
677        EscapedInput += "\\L";
678      else if (UnicodeScalarValue.first == 0x2029)
679        EscapedInput += "\\P";
680      else {
681        std::string HexStr = utohexstr(UnicodeScalarValue.first);
682        if (HexStr.size() <= 2)
683          EscapedInput += "\\x" + std::string(2 - HexStr.size(), '0') + HexStr;
684        else if (HexStr.size() <= 4)
685          EscapedInput += "\\u" + std::string(4 - HexStr.size(), '0') + HexStr;
686        else if (HexStr.size() <= 8)
687          EscapedInput += "\\U" + std::string(8 - HexStr.size(), '0') + HexStr;
688      }
689      i += UnicodeScalarValue.second - 1;
690    } else
691      EscapedInput.push_back(*i);
692  }
693  return EscapedInput;
694}
695
696Scanner::Scanner(StringRef Input, SourceMgr &sm)
697  : SM(sm)
698  , Indent(-1)
699  , Column(0)
700  , Line(0)
701  , FlowLevel(0)
702  , IsStartOfStream(true)
703  , IsSimpleKeyAllowed(true)
704  , Failed(false) {
705  InputBuffer = MemoryBuffer::getMemBuffer(Input, "YAML");
706  SM.AddNewSourceBuffer(InputBuffer, SMLoc());
707  Current = InputBuffer->getBufferStart();
708  End = InputBuffer->getBufferEnd();
709}
710
711Token &Scanner::peekNext() {
712  // If the current token is a possible simple key, keep parsing until we
713  // can confirm.
714  bool NeedMore = false;
715  while (true) {
716    if (TokenQueue.empty() || NeedMore) {
717      if (!fetchMoreTokens()) {
718        TokenQueue.clear();
719        TokenQueue.push_back(Token());
720        return TokenQueue.front();
721      }
722    }
723    assert(!TokenQueue.empty() &&
724            "fetchMoreTokens lied about getting tokens!");
725
726    removeStaleSimpleKeyCandidates();
727    SimpleKey SK;
728    SK.Tok = TokenQueue.front();
729    if (std::find(SimpleKeys.begin(), SimpleKeys.end(), SK)
730        == SimpleKeys.end())
731      break;
732    else
733      NeedMore = true;
734  }
735  return TokenQueue.front();
736}
737
738Token Scanner::getNext() {
739  Token Ret = peekNext();
740  // TokenQueue can be empty if there was an error getting the next token.
741  if (!TokenQueue.empty())
742    TokenQueue.pop_front();
743
744  // There cannot be any referenced Token's if the TokenQueue is empty. So do a
745  // quick deallocation of them all.
746  if (TokenQueue.empty()) {
747    TokenQueue.Alloc.Reset();
748  }
749
750  return Ret;
751}
752
753StringRef::iterator Scanner::skip_nb_char(StringRef::iterator Position) {
754  if (Position == End)
755    return Position;
756  // Check 7 bit c-printable - b-char.
757  if (   *Position == 0x09
758      || (*Position >= 0x20 && *Position <= 0x7E))
759    return Position + 1;
760
761  // Check for valid UTF-8.
762  if (uint8_t(*Position) & 0x80) {
763    UTF8Decoded u8d = decodeUTF8(Position);
764    if (   u8d.second != 0
765        && u8d.first != 0xFEFF
766        && ( u8d.first == 0x85
767          || ( u8d.first >= 0xA0
768            && u8d.first <= 0xD7FF)
769          || ( u8d.first >= 0xE000
770            && u8d.first <= 0xFFFD)
771          || ( u8d.first >= 0x10000
772            && u8d.first <= 0x10FFFF)))
773      return Position + u8d.second;
774  }
775  return Position;
776}
777
778StringRef::iterator Scanner::skip_b_break(StringRef::iterator Position) {
779  if (Position == End)
780    return Position;
781  if (*Position == 0x0D) {
782    if (Position + 1 != End && *(Position + 1) == 0x0A)
783      return Position + 2;
784    return Position + 1;
785  }
786
787  if (*Position == 0x0A)
788    return Position + 1;
789  return Position;
790}
791
792
793StringRef::iterator Scanner::skip_s_white(StringRef::iterator Position) {
794  if (Position == End)
795    return Position;
796  if (*Position == ' ' || *Position == '\t')
797    return Position + 1;
798  return Position;
799}
800
801StringRef::iterator Scanner::skip_ns_char(StringRef::iterator Position) {
802  if (Position == End)
803    return Position;
804  if (*Position == ' ' || *Position == '\t')
805    return Position;
806  return skip_nb_char(Position);
807}
808
809StringRef::iterator Scanner::skip_while( SkipWhileFunc Func
810                                       , StringRef::iterator Position) {
811  while (true) {
812    StringRef::iterator i = (this->*Func)(Position);
813    if (i == Position)
814      break;
815    Position = i;
816  }
817  return Position;
818}
819
820static bool is_ns_hex_digit(const char C) {
821  return    (C >= '0' && C <= '9')
822         || (C >= 'a' && C <= 'z')
823         || (C >= 'A' && C <= 'Z');
824}
825
826static bool is_ns_word_char(const char C) {
827  return    C == '-'
828         || (C >= 'a' && C <= 'z')
829         || (C >= 'A' && C <= 'Z');
830}
831
832StringRef Scanner::scan_ns_uri_char() {
833  StringRef::iterator Start = Current;
834  while (true) {
835    if (Current == End)
836      break;
837    if ((   *Current == '%'
838          && Current + 2 < End
839          && is_ns_hex_digit(*(Current + 1))
840          && is_ns_hex_digit(*(Current + 2)))
841        || is_ns_word_char(*Current)
842        || StringRef(Current, 1).find_first_of("#;/?:@&=+$,_.!~*'()[]")
843          != StringRef::npos) {
844      ++Current;
845      ++Column;
846    } else
847      break;
848  }
849  return StringRef(Start, Current - Start);
850}
851
852StringRef Scanner::scan_ns_plain_one_line() {
853  StringRef::iterator start = Current;
854  // The first character must already be verified.
855  ++Current;
856  while (true) {
857    if (Current == End) {
858      break;
859    } else if (*Current == ':') {
860      // Check if the next character is a ns-char.
861      if (Current + 1 == End)
862        break;
863      StringRef::iterator i = skip_ns_char(Current + 1);
864      if (Current + 1 != i) {
865        Current = i;
866        Column += 2; // Consume both the ':' and ns-char.
867      } else
868        break;
869    } else if (*Current == '#') {
870      // Check if the previous character was a ns-char.
871      // The & 0x80 check is to check for the trailing byte of a utf-8
872      if (*(Current - 1) & 0x80 || skip_ns_char(Current - 1) == Current) {
873        ++Current;
874        ++Column;
875      } else
876        break;
877    } else {
878      StringRef::iterator i = skip_nb_char(Current);
879      if (i == Current)
880        break;
881      Current = i;
882      ++Column;
883    }
884  }
885  return StringRef(start, Current - start);
886}
887
888bool Scanner::consume(uint32_t Expected) {
889  if (Expected >= 0x80)
890    report_fatal_error("Not dealing with this yet");
891  if (Current == End)
892    return false;
893  if (uint8_t(*Current) >= 0x80)
894    report_fatal_error("Not dealing with this yet");
895  if (uint8_t(*Current) == Expected) {
896    ++Current;
897    ++Column;
898    return true;
899  }
900  return false;
901}
902
903void Scanner::skip(uint32_t Distance) {
904  Current += Distance;
905  Column += Distance;
906  assert(Current <= End && "Skipped past the end");
907}
908
909bool Scanner::isBlankOrBreak(StringRef::iterator Position) {
910  if (Position == End)
911    return false;
912  if (   *Position == ' ' || *Position == '\t'
913      || *Position == '\r' || *Position == '\n')
914    return true;
915  return false;
916}
917
918void Scanner::saveSimpleKeyCandidate( TokenQueueT::iterator Tok
919                                    , unsigned AtColumn
920                                    , bool IsRequired) {
921  if (IsSimpleKeyAllowed) {
922    SimpleKey SK;
923    SK.Tok = Tok;
924    SK.Line = Line;
925    SK.Column = AtColumn;
926    SK.IsRequired = IsRequired;
927    SK.FlowLevel = FlowLevel;
928    SimpleKeys.push_back(SK);
929  }
930}
931
932void Scanner::removeStaleSimpleKeyCandidates() {
933  for (SmallVectorImpl<SimpleKey>::iterator i = SimpleKeys.begin();
934                                            i != SimpleKeys.end();) {
935    if (i->Line != Line || i->Column + 1024 < Column) {
936      if (i->IsRequired)
937        setError( "Could not find expected : for simple key"
938                , i->Tok->Range.begin());
939      i = SimpleKeys.erase(i);
940    } else
941      ++i;
942  }
943}
944
945void Scanner::removeSimpleKeyCandidatesOnFlowLevel(unsigned Level) {
946  if (!SimpleKeys.empty() && (SimpleKeys.end() - 1)->FlowLevel == Level)
947    SimpleKeys.pop_back();
948}
949
950bool Scanner::unrollIndent(int ToColumn) {
951  Token T;
952  // Indentation is ignored in flow.
953  if (FlowLevel != 0)
954    return true;
955
956  while (Indent > ToColumn) {
957    T.Kind = Token::TK_BlockEnd;
958    T.Range = StringRef(Current, 1);
959    TokenQueue.push_back(T);
960    Indent = Indents.pop_back_val();
961  }
962
963  return true;
964}
965
966bool Scanner::rollIndent( int ToColumn
967                        , Token::TokenKind Kind
968                        , TokenQueueT::iterator InsertPoint) {
969  if (FlowLevel)
970    return true;
971  if (Indent < ToColumn) {
972    Indents.push_back(Indent);
973    Indent = ToColumn;
974
975    Token T;
976    T.Kind = Kind;
977    T.Range = StringRef(Current, 0);
978    TokenQueue.insert(InsertPoint, T);
979  }
980  return true;
981}
982
983void Scanner::scanToNextToken() {
984  while (true) {
985    while (*Current == ' ' || *Current == '\t') {
986      skip(1);
987    }
988
989    // Skip comment.
990    if (*Current == '#') {
991      while (true) {
992        // This may skip more than one byte, thus Column is only incremented
993        // for code points.
994        StringRef::iterator i = skip_nb_char(Current);
995        if (i == Current)
996          break;
997        Current = i;
998        ++Column;
999      }
1000    }
1001
1002    // Skip EOL.
1003    StringRef::iterator i = skip_b_break(Current);
1004    if (i == Current)
1005      break;
1006    Current = i;
1007    ++Line;
1008    Column = 0;
1009    // New lines may start a simple key.
1010    if (!FlowLevel)
1011      IsSimpleKeyAllowed = true;
1012  }
1013}
1014
1015bool Scanner::scanStreamStart() {
1016  IsStartOfStream = false;
1017
1018  EncodingInfo EI = getUnicodeEncoding(currentInput());
1019
1020  Token T;
1021  T.Kind = Token::TK_StreamStart;
1022  T.Range = StringRef(Current, EI.second);
1023  TokenQueue.push_back(T);
1024  Current += EI.second;
1025  return true;
1026}
1027
1028bool Scanner::scanStreamEnd() {
1029  // Force an ending new line if one isn't present.
1030  if (Column != 0) {
1031    Column = 0;
1032    ++Line;
1033  }
1034
1035  unrollIndent(-1);
1036  SimpleKeys.clear();
1037  IsSimpleKeyAllowed = false;
1038
1039  Token T;
1040  T.Kind = Token::TK_StreamEnd;
1041  T.Range = StringRef(Current, 0);
1042  TokenQueue.push_back(T);
1043  return true;
1044}
1045
1046bool Scanner::scanDirective() {
1047  // Reset the indentation level.
1048  unrollIndent(-1);
1049  SimpleKeys.clear();
1050  IsSimpleKeyAllowed = false;
1051
1052  StringRef::iterator Start = Current;
1053  consume('%');
1054  StringRef::iterator NameStart = Current;
1055  Current = skip_while(&Scanner::skip_ns_char, Current);
1056  StringRef Name(NameStart, Current - NameStart);
1057  Current = skip_while(&Scanner::skip_s_white, Current);
1058
1059  if (Name == "YAML") {
1060    Current = skip_while(&Scanner::skip_ns_char, Current);
1061    Token T;
1062    T.Kind = Token::TK_VersionDirective;
1063    T.Range = StringRef(Start, Current - Start);
1064    TokenQueue.push_back(T);
1065    return true;
1066  }
1067  return false;
1068}
1069
1070bool Scanner::scanDocumentIndicator(bool IsStart) {
1071  unrollIndent(-1);
1072  SimpleKeys.clear();
1073  IsSimpleKeyAllowed = false;
1074
1075  Token T;
1076  T.Kind = IsStart ? Token::TK_DocumentStart : Token::TK_DocumentEnd;
1077  T.Range = StringRef(Current, 3);
1078  skip(3);
1079  TokenQueue.push_back(T);
1080  return true;
1081}
1082
1083bool Scanner::scanFlowCollectionStart(bool IsSequence) {
1084  Token T;
1085  T.Kind = IsSequence ? Token::TK_FlowSequenceStart
1086                      : Token::TK_FlowMappingStart;
1087  T.Range = StringRef(Current, 1);
1088  skip(1);
1089  TokenQueue.push_back(T);
1090
1091  // [ and { may begin a simple key.
1092  saveSimpleKeyCandidate(TokenQueue.back(), Column - 1, false);
1093
1094  // And may also be followed by a simple key.
1095  IsSimpleKeyAllowed = true;
1096  ++FlowLevel;
1097  return true;
1098}
1099
1100bool Scanner::scanFlowCollectionEnd(bool IsSequence) {
1101  removeSimpleKeyCandidatesOnFlowLevel(FlowLevel);
1102  IsSimpleKeyAllowed = false;
1103  Token T;
1104  T.Kind = IsSequence ? Token::TK_FlowSequenceEnd
1105                      : Token::TK_FlowMappingEnd;
1106  T.Range = StringRef(Current, 1);
1107  skip(1);
1108  TokenQueue.push_back(T);
1109  if (FlowLevel)
1110    --FlowLevel;
1111  return true;
1112}
1113
1114bool Scanner::scanFlowEntry() {
1115  removeSimpleKeyCandidatesOnFlowLevel(FlowLevel);
1116  IsSimpleKeyAllowed = true;
1117  Token T;
1118  T.Kind = Token::TK_FlowEntry;
1119  T.Range = StringRef(Current, 1);
1120  skip(1);
1121  TokenQueue.push_back(T);
1122  return true;
1123}
1124
1125bool Scanner::scanBlockEntry() {
1126  rollIndent(Column, Token::TK_BlockSequenceStart, TokenQueue.end());
1127  removeSimpleKeyCandidatesOnFlowLevel(FlowLevel);
1128  IsSimpleKeyAllowed = true;
1129  Token T;
1130  T.Kind = Token::TK_BlockEntry;
1131  T.Range = StringRef(Current, 1);
1132  skip(1);
1133  TokenQueue.push_back(T);
1134  return true;
1135}
1136
1137bool Scanner::scanKey() {
1138  if (!FlowLevel)
1139    rollIndent(Column, Token::TK_BlockMappingStart, TokenQueue.end());
1140
1141  removeSimpleKeyCandidatesOnFlowLevel(FlowLevel);
1142  IsSimpleKeyAllowed = !FlowLevel;
1143
1144  Token T;
1145  T.Kind = Token::TK_Key;
1146  T.Range = StringRef(Current, 1);
1147  skip(1);
1148  TokenQueue.push_back(T);
1149  return true;
1150}
1151
1152bool Scanner::scanValue() {
1153  // If the previous token could have been a simple key, insert the key token
1154  // into the token queue.
1155  if (!SimpleKeys.empty()) {
1156    SimpleKey SK = SimpleKeys.pop_back_val();
1157    Token T;
1158    T.Kind = Token::TK_Key;
1159    T.Range = SK.Tok->Range;
1160    TokenQueueT::iterator i, e;
1161    for (i = TokenQueue.begin(), e = TokenQueue.end(); i != e; ++i) {
1162      if (i == SK.Tok)
1163        break;
1164    }
1165    assert(i != e && "SimpleKey not in token queue!");
1166    i = TokenQueue.insert(i, T);
1167
1168    // We may also need to add a Block-Mapping-Start token.
1169    rollIndent(SK.Column, Token::TK_BlockMappingStart, i);
1170
1171    IsSimpleKeyAllowed = false;
1172  } else {
1173    if (!FlowLevel)
1174      rollIndent(Column, Token::TK_BlockMappingStart, TokenQueue.end());
1175    IsSimpleKeyAllowed = !FlowLevel;
1176  }
1177
1178  Token T;
1179  T.Kind = Token::TK_Value;
1180  T.Range = StringRef(Current, 1);
1181  skip(1);
1182  TokenQueue.push_back(T);
1183  return true;
1184}
1185
1186// Forbidding inlining improves performance by roughly 20%.
1187// FIXME: Remove once llvm optimizes this to the faster version without hints.
1188LLVM_ATTRIBUTE_NOINLINE static bool
1189wasEscaped(StringRef::iterator First, StringRef::iterator Position);
1190
1191// Returns whether a character at 'Position' was escaped with a leading '\'.
1192// 'First' specifies the position of the first character in the string.
1193static bool wasEscaped(StringRef::iterator First,
1194                       StringRef::iterator Position) {
1195  assert(Position - 1 >= First);
1196  StringRef::iterator I = Position - 1;
1197  // We calculate the number of consecutive '\'s before the current position
1198  // by iterating backwards through our string.
1199  while (I >= First && *I == '\\') --I;
1200  // (Position - 1 - I) now contains the number of '\'s before the current
1201  // position. If it is odd, the character at 'Position' was escaped.
1202  return (Position - 1 - I) % 2 == 1;
1203}
1204
1205bool Scanner::scanFlowScalar(bool IsDoubleQuoted) {
1206  StringRef::iterator Start = Current;
1207  unsigned ColStart = Column;
1208  if (IsDoubleQuoted) {
1209    do {
1210      ++Current;
1211      while (Current != End && *Current != '"')
1212        ++Current;
1213      // Repeat until the previous character was not a '\' or was an escaped
1214      // backslash.
1215    } while (   Current != End
1216             && *(Current - 1) == '\\'
1217             && wasEscaped(Start + 1, Current));
1218  } else {
1219    skip(1);
1220    while (true) {
1221      // Skip a ' followed by another '.
1222      if (Current + 1 < End && *Current == '\'' && *(Current + 1) == '\'') {
1223        skip(2);
1224        continue;
1225      } else if (*Current == '\'')
1226        break;
1227      StringRef::iterator i = skip_nb_char(Current);
1228      if (i == Current) {
1229        i = skip_b_break(Current);
1230        if (i == Current)
1231          break;
1232        Current = i;
1233        Column = 0;
1234        ++Line;
1235      } else {
1236        if (i == End)
1237          break;
1238        Current = i;
1239        ++Column;
1240      }
1241    }
1242  }
1243
1244  if (Current == End) {
1245    setError("Expected quote at end of scalar", Current);
1246    return false;
1247  }
1248
1249  skip(1); // Skip ending quote.
1250  Token T;
1251  T.Kind = Token::TK_Scalar;
1252  T.Range = StringRef(Start, Current - Start);
1253  TokenQueue.push_back(T);
1254
1255  saveSimpleKeyCandidate(TokenQueue.back(), ColStart, false);
1256
1257  IsSimpleKeyAllowed = false;
1258
1259  return true;
1260}
1261
1262bool Scanner::scanPlainScalar() {
1263  StringRef::iterator Start = Current;
1264  unsigned ColStart = Column;
1265  unsigned LeadingBlanks = 0;
1266  assert(Indent >= -1 && "Indent must be >= -1 !");
1267  unsigned indent = static_cast<unsigned>(Indent + 1);
1268  while (true) {
1269    if (*Current == '#')
1270      break;
1271
1272    while (!isBlankOrBreak(Current)) {
1273      if (  FlowLevel && *Current == ':'
1274          && !(isBlankOrBreak(Current + 1) || *(Current + 1) == ',')) {
1275        setError("Found unexpected ':' while scanning a plain scalar", Current);
1276        return false;
1277      }
1278
1279      // Check for the end of the plain scalar.
1280      if (  (*Current == ':' && isBlankOrBreak(Current + 1))
1281          || (  FlowLevel
1282          && (StringRef(Current, 1).find_first_of(",:?[]{}")
1283              != StringRef::npos)))
1284        break;
1285
1286      StringRef::iterator i = skip_nb_char(Current);
1287      if (i == Current)
1288        break;
1289      Current = i;
1290      ++Column;
1291    }
1292
1293    // Are we at the end?
1294    if (!isBlankOrBreak(Current))
1295      break;
1296
1297    // Eat blanks.
1298    StringRef::iterator Tmp = Current;
1299    while (isBlankOrBreak(Tmp)) {
1300      StringRef::iterator i = skip_s_white(Tmp);
1301      if (i != Tmp) {
1302        if (LeadingBlanks && (Column < indent) && *Tmp == '\t') {
1303          setError("Found invalid tab character in indentation", Tmp);
1304          return false;
1305        }
1306        Tmp = i;
1307        ++Column;
1308      } else {
1309        i = skip_b_break(Tmp);
1310        if (!LeadingBlanks)
1311          LeadingBlanks = 1;
1312        Tmp = i;
1313        Column = 0;
1314        ++Line;
1315      }
1316    }
1317
1318    if (!FlowLevel && Column < indent)
1319      break;
1320
1321    Current = Tmp;
1322  }
1323  if (Start == Current) {
1324    setError("Got empty plain scalar", Start);
1325    return false;
1326  }
1327  Token T;
1328  T.Kind = Token::TK_Scalar;
1329  T.Range = StringRef(Start, Current - Start);
1330  TokenQueue.push_back(T);
1331
1332  // Plain scalars can be simple keys.
1333  saveSimpleKeyCandidate(TokenQueue.back(), ColStart, false);
1334
1335  IsSimpleKeyAllowed = false;
1336
1337  return true;
1338}
1339
1340bool Scanner::scanAliasOrAnchor(bool IsAlias) {
1341  StringRef::iterator Start = Current;
1342  unsigned ColStart = Column;
1343  skip(1);
1344  while(true) {
1345    if (   *Current == '[' || *Current == ']'
1346        || *Current == '{' || *Current == '}'
1347        || *Current == ','
1348        || *Current == ':')
1349      break;
1350    StringRef::iterator i = skip_ns_char(Current);
1351    if (i == Current)
1352      break;
1353    Current = i;
1354    ++Column;
1355  }
1356
1357  if (Start == Current) {
1358    setError("Got empty alias or anchor", Start);
1359    return false;
1360  }
1361
1362  Token T;
1363  T.Kind = IsAlias ? Token::TK_Alias : Token::TK_Anchor;
1364  T.Range = StringRef(Start, Current - Start);
1365  TokenQueue.push_back(T);
1366
1367  // Alias and anchors can be simple keys.
1368  saveSimpleKeyCandidate(TokenQueue.back(), ColStart, false);
1369
1370  IsSimpleKeyAllowed = false;
1371
1372  return true;
1373}
1374
1375bool Scanner::scanBlockScalar(bool IsLiteral) {
1376  StringRef::iterator Start = Current;
1377  skip(1); // Eat | or >
1378  while(true) {
1379    StringRef::iterator i = skip_nb_char(Current);
1380    if (i == Current) {
1381      if (Column == 0)
1382        break;
1383      i = skip_b_break(Current);
1384      if (i != Current) {
1385        // We got a line break.
1386        Column = 0;
1387        ++Line;
1388        Current = i;
1389        continue;
1390      } else {
1391        // There was an error, which should already have been printed out.
1392        return false;
1393      }
1394    }
1395    Current = i;
1396    ++Column;
1397  }
1398
1399  if (Start == Current) {
1400    setError("Got empty block scalar", Start);
1401    return false;
1402  }
1403
1404  Token T;
1405  T.Kind = Token::TK_Scalar;
1406  T.Range = StringRef(Start, Current - Start);
1407  TokenQueue.push_back(T);
1408  return true;
1409}
1410
1411bool Scanner::scanTag() {
1412  StringRef::iterator Start = Current;
1413  unsigned ColStart = Column;
1414  skip(1); // Eat !.
1415  if (Current == End || isBlankOrBreak(Current)); // An empty tag.
1416  else if (*Current == '<') {
1417    skip(1);
1418    scan_ns_uri_char();
1419    if (!consume('>'))
1420      return false;
1421  } else {
1422    // FIXME: Actually parse the c-ns-shorthand-tag rule.
1423    Current = skip_while(&Scanner::skip_ns_char, Current);
1424  }
1425
1426  Token T;
1427  T.Kind = Token::TK_Tag;
1428  T.Range = StringRef(Start, Current - Start);
1429  TokenQueue.push_back(T);
1430
1431  // Tags can be simple keys.
1432  saveSimpleKeyCandidate(TokenQueue.back(), ColStart, false);
1433
1434  IsSimpleKeyAllowed = false;
1435
1436  return true;
1437}
1438
1439bool Scanner::fetchMoreTokens() {
1440  if (IsStartOfStream)
1441    return scanStreamStart();
1442
1443  scanToNextToken();
1444
1445  if (Current == End)
1446    return scanStreamEnd();
1447
1448  removeStaleSimpleKeyCandidates();
1449
1450  unrollIndent(Column);
1451
1452  if (Column == 0 && *Current == '%')
1453    return scanDirective();
1454
1455  if (Column == 0 && Current + 4 <= End
1456      && *Current == '-'
1457      && *(Current + 1) == '-'
1458      && *(Current + 2) == '-'
1459      && (Current + 3 == End || isBlankOrBreak(Current + 3)))
1460    return scanDocumentIndicator(true);
1461
1462  if (Column == 0 && Current + 4 <= End
1463      && *Current == '.'
1464      && *(Current + 1) == '.'
1465      && *(Current + 2) == '.'
1466      && (Current + 3 == End || isBlankOrBreak(Current + 3)))
1467    return scanDocumentIndicator(false);
1468
1469  if (*Current == '[')
1470    return scanFlowCollectionStart(true);
1471
1472  if (*Current == '{')
1473    return scanFlowCollectionStart(false);
1474
1475  if (*Current == ']')
1476    return scanFlowCollectionEnd(true);
1477
1478  if (*Current == '}')
1479    return scanFlowCollectionEnd(false);
1480
1481  if (*Current == ',')
1482    return scanFlowEntry();
1483
1484  if (*Current == '-' && isBlankOrBreak(Current + 1))
1485    return scanBlockEntry();
1486
1487  if (*Current == '?' && (FlowLevel || isBlankOrBreak(Current + 1)))
1488    return scanKey();
1489
1490  if (*Current == ':' && (FlowLevel || isBlankOrBreak(Current + 1)))
1491    return scanValue();
1492
1493  if (*Current == '*')
1494    return scanAliasOrAnchor(true);
1495
1496  if (*Current == '&')
1497    return scanAliasOrAnchor(false);
1498
1499  if (*Current == '!')
1500    return scanTag();
1501
1502  if (*Current == '|' && !FlowLevel)
1503    return scanBlockScalar(true);
1504
1505  if (*Current == '>' && !FlowLevel)
1506    return scanBlockScalar(false);
1507
1508  if (*Current == '\'')
1509    return scanFlowScalar(false);
1510
1511  if (*Current == '"')
1512    return scanFlowScalar(true);
1513
1514  // Get a plain scalar.
1515  StringRef FirstChar(Current, 1);
1516  if (!(isBlankOrBreak(Current)
1517        || FirstChar.find_first_of("-?:,[]{}#&*!|>'\"%@`") != StringRef::npos)
1518      || (*Current == '-' && !isBlankOrBreak(Current + 1))
1519      || (!FlowLevel && (*Current == '?' || *Current == ':')
1520          && isBlankOrBreak(Current + 1))
1521      || (!FlowLevel && *Current == ':'
1522                      && Current + 2 < End
1523                      && *(Current + 1) == ':'
1524                      && !isBlankOrBreak(Current + 2)))
1525    return scanPlainScalar();
1526
1527  setError("Unrecognized character while tokenizing.");
1528  return false;
1529}
1530
1531Stream::Stream(StringRef Input, SourceMgr &SM)
1532  : scanner(new Scanner(Input, SM))
1533  , CurrentDoc(0) {}
1534
1535Stream::~Stream() {}
1536
1537bool Stream::failed() { return scanner->failed(); }
1538
1539void Stream::printError(Node *N, const Twine &Msg) {
1540  SmallVector<SMRange, 1> Ranges;
1541  Ranges.push_back(N->getSourceRange());
1542  scanner->printError( N->getSourceRange().Start
1543                     , SourceMgr::DK_Error
1544                     , Msg
1545                     , Ranges);
1546}
1547
1548void Stream::handleYAMLDirective(const Token &t) {
1549  // TODO: Ensure version is 1.x.
1550}
1551
1552document_iterator Stream::begin() {
1553  if (CurrentDoc)
1554    report_fatal_error("Can only iterate over the stream once");
1555
1556  // Skip Stream-Start.
1557  scanner->getNext();
1558
1559  CurrentDoc.reset(new Document(*this));
1560  return document_iterator(CurrentDoc);
1561}
1562
1563document_iterator Stream::end() {
1564  return document_iterator();
1565}
1566
1567void Stream::skip() {
1568  for (document_iterator i = begin(), e = end(); i != e; ++i)
1569    i->skip();
1570}
1571
1572Node::Node(unsigned int Type, OwningPtr<Document> &D, StringRef A)
1573  : Doc(D)
1574  , TypeID(Type)
1575  , Anchor(A) {
1576  SMLoc Start = SMLoc::getFromPointer(peekNext().Range.begin());
1577  SourceRange = SMRange(Start, Start);
1578}
1579
1580Token &Node::peekNext() {
1581  return Doc->peekNext();
1582}
1583
1584Token Node::getNext() {
1585  return Doc->getNext();
1586}
1587
1588Node *Node::parseBlockNode() {
1589  return Doc->parseBlockNode();
1590}
1591
1592BumpPtrAllocator &Node::getAllocator() {
1593  return Doc->NodeAllocator;
1594}
1595
1596void Node::setError(const Twine &Msg, Token &Tok) const {
1597  Doc->setError(Msg, Tok);
1598}
1599
1600bool Node::failed() const {
1601  return Doc->failed();
1602}
1603
1604
1605
1606StringRef ScalarNode::getValue(SmallVectorImpl<char> &Storage) const {
1607  // TODO: Handle newlines properly. We need to remove leading whitespace.
1608  if (Value[0] == '"') { // Double quoted.
1609    // Pull off the leading and trailing "s.
1610    StringRef UnquotedValue = Value.substr(1, Value.size() - 2);
1611    // Search for characters that would require unescaping the value.
1612    StringRef::size_type i = UnquotedValue.find_first_of("\\\r\n");
1613    if (i != StringRef::npos)
1614      return unescapeDoubleQuoted(UnquotedValue, i, Storage);
1615    return UnquotedValue;
1616  } else if (Value[0] == '\'') { // Single quoted.
1617    // Pull off the leading and trailing 's.
1618    StringRef UnquotedValue = Value.substr(1, Value.size() - 2);
1619    StringRef::size_type i = UnquotedValue.find('\'');
1620    if (i != StringRef::npos) {
1621      // We're going to need Storage.
1622      Storage.clear();
1623      Storage.reserve(UnquotedValue.size());
1624      for (; i != StringRef::npos; i = UnquotedValue.find('\'')) {
1625        StringRef Valid(UnquotedValue.begin(), i);
1626        Storage.insert(Storage.end(), Valid.begin(), Valid.end());
1627        Storage.push_back('\'');
1628        UnquotedValue = UnquotedValue.substr(i + 2);
1629      }
1630      Storage.insert(Storage.end(), UnquotedValue.begin(), UnquotedValue.end());
1631      return StringRef(Storage.begin(), Storage.size());
1632    }
1633    return UnquotedValue;
1634  }
1635  // Plain or block.
1636  return Value.rtrim(" ");
1637}
1638
1639StringRef ScalarNode::unescapeDoubleQuoted( StringRef UnquotedValue
1640                                          , StringRef::size_type i
1641                                          , SmallVectorImpl<char> &Storage)
1642                                          const {
1643  // Use Storage to build proper value.
1644  Storage.clear();
1645  Storage.reserve(UnquotedValue.size());
1646  for (; i != StringRef::npos; i = UnquotedValue.find_first_of("\\\r\n")) {
1647    // Insert all previous chars into Storage.
1648    StringRef Valid(UnquotedValue.begin(), i);
1649    Storage.insert(Storage.end(), Valid.begin(), Valid.end());
1650    // Chop off inserted chars.
1651    UnquotedValue = UnquotedValue.substr(i);
1652
1653    assert(!UnquotedValue.empty() && "Can't be empty!");
1654
1655    // Parse escape or line break.
1656    switch (UnquotedValue[0]) {
1657    case '\r':
1658    case '\n':
1659      Storage.push_back('\n');
1660      if (   UnquotedValue.size() > 1
1661          && (UnquotedValue[1] == '\r' || UnquotedValue[1] == '\n'))
1662        UnquotedValue = UnquotedValue.substr(1);
1663      UnquotedValue = UnquotedValue.substr(1);
1664      break;
1665    default:
1666      if (UnquotedValue.size() == 1)
1667        // TODO: Report error.
1668        break;
1669      UnquotedValue = UnquotedValue.substr(1);
1670      switch (UnquotedValue[0]) {
1671      default: {
1672          Token T;
1673          T.Range = StringRef(UnquotedValue.begin(), 1);
1674          setError("Unrecognized escape code!", T);
1675          return "";
1676        }
1677      case '\r':
1678      case '\n':
1679        // Remove the new line.
1680        if (   UnquotedValue.size() > 1
1681            && (UnquotedValue[1] == '\r' || UnquotedValue[1] == '\n'))
1682          UnquotedValue = UnquotedValue.substr(1);
1683        // If this was just a single byte newline, it will get skipped
1684        // below.
1685        break;
1686      case '0':
1687        Storage.push_back(0x00);
1688        break;
1689      case 'a':
1690        Storage.push_back(0x07);
1691        break;
1692      case 'b':
1693        Storage.push_back(0x08);
1694        break;
1695      case 't':
1696      case 0x09:
1697        Storage.push_back(0x09);
1698        break;
1699      case 'n':
1700        Storage.push_back(0x0A);
1701        break;
1702      case 'v':
1703        Storage.push_back(0x0B);
1704        break;
1705      case 'f':
1706        Storage.push_back(0x0C);
1707        break;
1708      case 'r':
1709        Storage.push_back(0x0D);
1710        break;
1711      case 'e':
1712        Storage.push_back(0x1B);
1713        break;
1714      case ' ':
1715        Storage.push_back(0x20);
1716        break;
1717      case '"':
1718        Storage.push_back(0x22);
1719        break;
1720      case '/':
1721        Storage.push_back(0x2F);
1722        break;
1723      case '\\':
1724        Storage.push_back(0x5C);
1725        break;
1726      case 'N':
1727        encodeUTF8(0x85, Storage);
1728        break;
1729      case '_':
1730        encodeUTF8(0xA0, Storage);
1731        break;
1732      case 'L':
1733        encodeUTF8(0x2028, Storage);
1734        break;
1735      case 'P':
1736        encodeUTF8(0x2029, Storage);
1737        break;
1738      case 'x': {
1739          if (UnquotedValue.size() < 3)
1740            // TODO: Report error.
1741            break;
1742          unsigned int UnicodeScalarValue;
1743          if (UnquotedValue.substr(1, 2).getAsInteger(16, UnicodeScalarValue))
1744            // TODO: Report error.
1745            UnicodeScalarValue = 0xFFFD;
1746          encodeUTF8(UnicodeScalarValue, Storage);
1747          UnquotedValue = UnquotedValue.substr(2);
1748          break;
1749        }
1750      case 'u': {
1751          if (UnquotedValue.size() < 5)
1752            // TODO: Report error.
1753            break;
1754          unsigned int UnicodeScalarValue;
1755          if (UnquotedValue.substr(1, 4).getAsInteger(16, UnicodeScalarValue))
1756            // TODO: Report error.
1757            UnicodeScalarValue = 0xFFFD;
1758          encodeUTF8(UnicodeScalarValue, Storage);
1759          UnquotedValue = UnquotedValue.substr(4);
1760          break;
1761        }
1762      case 'U': {
1763          if (UnquotedValue.size() < 9)
1764            // TODO: Report error.
1765            break;
1766          unsigned int UnicodeScalarValue;
1767          if (UnquotedValue.substr(1, 8).getAsInteger(16, UnicodeScalarValue))
1768            // TODO: Report error.
1769            UnicodeScalarValue = 0xFFFD;
1770          encodeUTF8(UnicodeScalarValue, Storage);
1771          UnquotedValue = UnquotedValue.substr(8);
1772          break;
1773        }
1774      }
1775      UnquotedValue = UnquotedValue.substr(1);
1776    }
1777  }
1778  Storage.insert(Storage.end(), UnquotedValue.begin(), UnquotedValue.end());
1779  return StringRef(Storage.begin(), Storage.size());
1780}
1781
1782Node *KeyValueNode::getKey() {
1783  if (Key)
1784    return Key;
1785  // Handle implicit null keys.
1786  {
1787    Token &t = peekNext();
1788    if (   t.Kind == Token::TK_BlockEnd
1789        || t.Kind == Token::TK_Value
1790        || t.Kind == Token::TK_Error) {
1791      return Key = new (getAllocator()) NullNode(Doc);
1792    }
1793    if (t.Kind == Token::TK_Key)
1794      getNext(); // skip TK_Key.
1795  }
1796
1797  // Handle explicit null keys.
1798  Token &t = peekNext();
1799  if (t.Kind == Token::TK_BlockEnd || t.Kind == Token::TK_Value) {
1800    return Key = new (getAllocator()) NullNode(Doc);
1801  }
1802
1803  // We've got a normal key.
1804  return Key = parseBlockNode();
1805}
1806
1807Node *KeyValueNode::getValue() {
1808  if (Value)
1809    return Value;
1810  getKey()->skip();
1811  if (failed())
1812    return Value = new (getAllocator()) NullNode(Doc);
1813
1814  // Handle implicit null values.
1815  {
1816    Token &t = peekNext();
1817    if (   t.Kind == Token::TK_BlockEnd
1818        || t.Kind == Token::TK_FlowMappingEnd
1819        || t.Kind == Token::TK_Key
1820        || t.Kind == Token::TK_FlowEntry
1821        || t.Kind == Token::TK_Error) {
1822      return Value = new (getAllocator()) NullNode(Doc);
1823    }
1824
1825    if (t.Kind != Token::TK_Value) {
1826      setError("Unexpected token in Key Value.", t);
1827      return Value = new (getAllocator()) NullNode(Doc);
1828    }
1829    getNext(); // skip TK_Value.
1830  }
1831
1832  // Handle explicit null values.
1833  Token &t = peekNext();
1834  if (t.Kind == Token::TK_BlockEnd || t.Kind == Token::TK_Key) {
1835    return Value = new (getAllocator()) NullNode(Doc);
1836  }
1837
1838  // We got a normal value.
1839  return Value = parseBlockNode();
1840}
1841
1842void MappingNode::increment() {
1843  if (failed()) {
1844    IsAtEnd = true;
1845    CurrentEntry = 0;
1846    return;
1847  }
1848  if (CurrentEntry) {
1849    CurrentEntry->skip();
1850    if (Type == MT_Inline) {
1851      IsAtEnd = true;
1852      CurrentEntry = 0;
1853      return;
1854    }
1855  }
1856  Token T = peekNext();
1857  if (T.Kind == Token::TK_Key || T.Kind == Token::TK_Scalar) {
1858    // KeyValueNode eats the TK_Key. That way it can detect null keys.
1859    CurrentEntry = new (getAllocator()) KeyValueNode(Doc);
1860  } else if (Type == MT_Block) {
1861    switch (T.Kind) {
1862    case Token::TK_BlockEnd:
1863      getNext();
1864      IsAtEnd = true;
1865      CurrentEntry = 0;
1866      break;
1867    default:
1868      setError("Unexpected token. Expected Key or Block End", T);
1869    case Token::TK_Error:
1870      IsAtEnd = true;
1871      CurrentEntry = 0;
1872    }
1873  } else {
1874    switch (T.Kind) {
1875    case Token::TK_FlowEntry:
1876      // Eat the flow entry and recurse.
1877      getNext();
1878      return increment();
1879    case Token::TK_FlowMappingEnd:
1880      getNext();
1881    case Token::TK_Error:
1882      // Set this to end iterator.
1883      IsAtEnd = true;
1884      CurrentEntry = 0;
1885      break;
1886    default:
1887      setError( "Unexpected token. Expected Key, Flow Entry, or Flow "
1888                "Mapping End."
1889              , T);
1890      IsAtEnd = true;
1891      CurrentEntry = 0;
1892    }
1893  }
1894}
1895
1896void SequenceNode::increment() {
1897  if (failed()) {
1898    IsAtEnd = true;
1899    CurrentEntry = 0;
1900    return;
1901  }
1902  if (CurrentEntry)
1903    CurrentEntry->skip();
1904  Token T = peekNext();
1905  if (SeqType == ST_Block) {
1906    switch (T.Kind) {
1907    case Token::TK_BlockEntry:
1908      getNext();
1909      CurrentEntry = parseBlockNode();
1910      if (CurrentEntry == 0) { // An error occurred.
1911        IsAtEnd = true;
1912        CurrentEntry = 0;
1913      }
1914      break;
1915    case Token::TK_BlockEnd:
1916      getNext();
1917      IsAtEnd = true;
1918      CurrentEntry = 0;
1919      break;
1920    default:
1921      setError( "Unexpected token. Expected Block Entry or Block End."
1922              , T);
1923    case Token::TK_Error:
1924      IsAtEnd = true;
1925      CurrentEntry = 0;
1926    }
1927  } else if (SeqType == ST_Indentless) {
1928    switch (T.Kind) {
1929    case Token::TK_BlockEntry:
1930      getNext();
1931      CurrentEntry = parseBlockNode();
1932      if (CurrentEntry == 0) { // An error occurred.
1933        IsAtEnd = true;
1934        CurrentEntry = 0;
1935      }
1936      break;
1937    default:
1938    case Token::TK_Error:
1939      IsAtEnd = true;
1940      CurrentEntry = 0;
1941    }
1942  } else if (SeqType == ST_Flow) {
1943    switch (T.Kind) {
1944    case Token::TK_FlowEntry:
1945      // Eat the flow entry and recurse.
1946      getNext();
1947      WasPreviousTokenFlowEntry = true;
1948      return increment();
1949    case Token::TK_FlowSequenceEnd:
1950      getNext();
1951    case Token::TK_Error:
1952      // Set this to end iterator.
1953      IsAtEnd = true;
1954      CurrentEntry = 0;
1955      break;
1956    case Token::TK_StreamEnd:
1957    case Token::TK_DocumentEnd:
1958    case Token::TK_DocumentStart:
1959      setError("Could not find closing ]!", T);
1960      // Set this to end iterator.
1961      IsAtEnd = true;
1962      CurrentEntry = 0;
1963      break;
1964    default:
1965      if (!WasPreviousTokenFlowEntry) {
1966        setError("Expected , between entries!", T);
1967        IsAtEnd = true;
1968        CurrentEntry = 0;
1969        break;
1970      }
1971      // Otherwise it must be a flow entry.
1972      CurrentEntry = parseBlockNode();
1973      if (!CurrentEntry) {
1974        IsAtEnd = true;
1975      }
1976      WasPreviousTokenFlowEntry = false;
1977      break;
1978    }
1979  }
1980}
1981
1982Document::Document(Stream &S) : stream(S), Root(0) {
1983  if (parseDirectives())
1984    expectToken(Token::TK_DocumentStart);
1985  Token &T = peekNext();
1986  if (T.Kind == Token::TK_DocumentStart)
1987    getNext();
1988}
1989
1990bool Document::skip()  {
1991  if (stream.scanner->failed())
1992    return false;
1993  if (!Root)
1994    getRoot();
1995  Root->skip();
1996  Token &T = peekNext();
1997  if (T.Kind == Token::TK_StreamEnd)
1998    return false;
1999  if (T.Kind == Token::TK_DocumentEnd) {
2000    getNext();
2001    return skip();
2002  }
2003  return true;
2004}
2005
2006Token &Document::peekNext() {
2007  return stream.scanner->peekNext();
2008}
2009
2010Token Document::getNext() {
2011  return stream.scanner->getNext();
2012}
2013
2014void Document::setError(const Twine &Message, Token &Location) const {
2015  stream.scanner->setError(Message, Location.Range.begin());
2016}
2017
2018bool Document::failed() const {
2019  return stream.scanner->failed();
2020}
2021
2022Node *Document::parseBlockNode() {
2023  Token T = peekNext();
2024  // Handle properties.
2025  Token AnchorInfo;
2026parse_property:
2027  switch (T.Kind) {
2028  case Token::TK_Alias:
2029    getNext();
2030    return new (NodeAllocator) AliasNode(stream.CurrentDoc, T.Range.substr(1));
2031  case Token::TK_Anchor:
2032    if (AnchorInfo.Kind == Token::TK_Anchor) {
2033      setError("Already encountered an anchor for this node!", T);
2034      return 0;
2035    }
2036    AnchorInfo = getNext(); // Consume TK_Anchor.
2037    T = peekNext();
2038    goto parse_property;
2039  case Token::TK_Tag:
2040    getNext(); // Skip TK_Tag.
2041    T = peekNext();
2042    goto parse_property;
2043  default:
2044    break;
2045  }
2046
2047  switch (T.Kind) {
2048  case Token::TK_BlockEntry:
2049    // We got an unindented BlockEntry sequence. This is not terminated with
2050    // a BlockEnd.
2051    // Don't eat the TK_BlockEntry, SequenceNode needs it.
2052    return new (NodeAllocator) SequenceNode( stream.CurrentDoc
2053                                           , AnchorInfo.Range.substr(1)
2054                                           , SequenceNode::ST_Indentless);
2055  case Token::TK_BlockSequenceStart:
2056    getNext();
2057    return new (NodeAllocator)
2058      SequenceNode( stream.CurrentDoc
2059                  , AnchorInfo.Range.substr(1)
2060                  , SequenceNode::ST_Block);
2061  case Token::TK_BlockMappingStart:
2062    getNext();
2063    return new (NodeAllocator)
2064      MappingNode( stream.CurrentDoc
2065                 , AnchorInfo.Range.substr(1)
2066                 , MappingNode::MT_Block);
2067  case Token::TK_FlowSequenceStart:
2068    getNext();
2069    return new (NodeAllocator)
2070      SequenceNode( stream.CurrentDoc
2071                  , AnchorInfo.Range.substr(1)
2072                  , SequenceNode::ST_Flow);
2073  case Token::TK_FlowMappingStart:
2074    getNext();
2075    return new (NodeAllocator)
2076      MappingNode( stream.CurrentDoc
2077                 , AnchorInfo.Range.substr(1)
2078                 , MappingNode::MT_Flow);
2079  case Token::TK_Scalar:
2080    getNext();
2081    return new (NodeAllocator)
2082      ScalarNode( stream.CurrentDoc
2083                , AnchorInfo.Range.substr(1)
2084                , T.Range);
2085  case Token::TK_Key:
2086    // Don't eat the TK_Key, KeyValueNode expects it.
2087    return new (NodeAllocator)
2088      MappingNode( stream.CurrentDoc
2089                 , AnchorInfo.Range.substr(1)
2090                 , MappingNode::MT_Inline);
2091  case Token::TK_DocumentStart:
2092  case Token::TK_DocumentEnd:
2093  case Token::TK_StreamEnd:
2094  default:
2095    // TODO: Properly handle tags. "[!!str ]" should resolve to !!str "", not
2096    //       !!null null.
2097    return new (NodeAllocator) NullNode(stream.CurrentDoc);
2098  case Token::TK_Error:
2099    return 0;
2100  }
2101  llvm_unreachable("Control flow shouldn't reach here.");
2102  return 0;
2103}
2104
2105bool Document::parseDirectives() {
2106  bool isDirective = false;
2107  while (true) {
2108    Token T = peekNext();
2109    if (T.Kind == Token::TK_TagDirective) {
2110      handleTagDirective(getNext());
2111      isDirective = true;
2112    } else if (T.Kind == Token::TK_VersionDirective) {
2113      stream.handleYAMLDirective(getNext());
2114      isDirective = true;
2115    } else
2116      break;
2117  }
2118  return isDirective;
2119}
2120
2121bool Document::expectToken(int TK) {
2122  Token T = getNext();
2123  if (T.Kind != TK) {
2124    setError("Unexpected token", T);
2125    return false;
2126  }
2127  return true;
2128}
2129