• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * Copyright 2014 Google Inc. All rights reserved.
3  *
4  * Licensed under the Apache License, Version 2.0 (the "License");
5  * you may not use this file except in compliance with the License.
6  * You may obtain a copy of the License at
7  *
8  *     http://www.apache.org/licenses/LICENSE-2.0
9  *
10  * Unless required by applicable law or agreed to in writing, software
11  * distributed under the License is distributed on an "AS IS" BASIS,
12  * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13  * See the License for the specific language governing permissions and
14  * limitations under the License.
15  */
16 
17 #include <algorithm>
18 #include <list>
19 #include <string>
20 
21 #include <math.h>
22 
23 #include "flatbuffers/idl.h"
24 #include "flatbuffers/util.h"
25 
26 namespace flatbuffers {
27 
28 const double kPi = 3.14159265358979323846;
29 
30 const char *const kTypeNames[] = {
31 // clang-format off
32   #define FLATBUFFERS_TD(ENUM, IDLTYPE, \
33     CTYPE, JTYPE, GTYPE, NTYPE, PTYPE, RTYPE) \
34     IDLTYPE,
35     FLATBUFFERS_GEN_TYPES(FLATBUFFERS_TD)
36   #undef FLATBUFFERS_TD
37   // clang-format on
38   nullptr
39 };
40 
41 const char kTypeSizes[] = {
42 // clang-format off
43   #define FLATBUFFERS_TD(ENUM, IDLTYPE, \
44       CTYPE, JTYPE, GTYPE, NTYPE, PTYPE, RTYPE) \
45       sizeof(CTYPE),
46     FLATBUFFERS_GEN_TYPES(FLATBUFFERS_TD)
47   #undef FLATBUFFERS_TD
48   // clang-format on
49 };
50 
51 // The enums in the reflection schema should match the ones we use internally.
52 // Compare the last element to check if these go out of sync.
53 static_assert(BASE_TYPE_UNION == static_cast<BaseType>(reflection::Union),
54               "enums don't match");
55 
56 // Any parsing calls have to be wrapped in this macro, which automates
57 // handling of recursive error checking a bit. It will check the received
58 // CheckedError object, and return straight away on error.
59 #define ECHECK(call)           \
60   {                            \
61     auto ce = (call);          \
62     if (ce.Check()) return ce; \
63   }
64 
65 // These two functions are called hundreds of times below, so define a short
66 // form:
67 #define NEXT() ECHECK(Next())
68 #define EXPECT(tok) ECHECK(Expect(tok))
69 
ValidateUTF8(const std::string & str)70 static bool ValidateUTF8(const std::string &str) {
71   const char *s = &str[0];
72   const char *const sEnd = s + str.length();
73   while (s < sEnd) {
74     if (FromUTF8(&s) < 0) { return false; }
75   }
76   return true;
77 }
78 
79 // Convert an underscore_based_indentifier in to camelCase.
80 // Also uppercases the first character if first is true.
MakeCamel(const std::string & in,bool first)81 std::string MakeCamel(const std::string &in, bool first) {
82   std::string s;
83   for (size_t i = 0; i < in.length(); i++) {
84     if (!i && first)
85       s += static_cast<char>(toupper(in[0]));
86     else if (in[i] == '_' && i + 1 < in.length())
87       s += static_cast<char>(toupper(in[++i]));
88     else
89       s += in[i];
90   }
91   return s;
92 }
93 
DeserializeDoc(std::vector<std::string> & doc,const Vector<Offset<String>> * documentation)94 void DeserializeDoc( std::vector<std::string> &doc,
95                      const Vector<Offset<String>> *documentation) {
96   if (documentation == nullptr) return;
97   for (uoffset_t index = 0; index < documentation->size(); index++)
98     doc.push_back(documentation->Get(index)->str());
99 }
100 
Message(const std::string & msg)101 void Parser::Message(const std::string &msg) {
102   error_ = file_being_parsed_.length() ? AbsolutePath(file_being_parsed_) : "";
103   // clang-format off
104   #ifdef _WIN32  // MSVC alike
105     error_ +=
106         "(" + NumToString(line_) + ", " + NumToString(CursorPosition()) + ")";
107   #else  // gcc alike
108     if (file_being_parsed_.length()) error_ += ":";
109     error_ += NumToString(line_) + ": " + NumToString(CursorPosition());
110   #endif
111   // clang-format on
112   error_ += ": " + msg;
113 }
114 
Warning(const std::string & msg)115 void Parser::Warning(const std::string &msg) { Message("warning: " + msg); }
116 
Error(const std::string & msg)117 CheckedError Parser::Error(const std::string &msg) {
118   Message("error: " + msg);
119   return CheckedError(true);
120 }
121 
NoError()122 inline CheckedError NoError() { return CheckedError(false); }
123 
RecurseError()124 CheckedError Parser::RecurseError() {
125   return Error("maximum parsing recursion of " +
126                NumToString(FLATBUFFERS_MAX_PARSING_DEPTH) + " reached");
127 }
128 
Recurse(F f)129 template<typename F> CheckedError Parser::Recurse(F f) {
130   if (recurse_protection_counter >= (FLATBUFFERS_MAX_PARSING_DEPTH))
131     return RecurseError();
132   recurse_protection_counter++;
133   auto ce = f();
134   recurse_protection_counter--;
135   return ce;
136 }
137 
InvalidNumber(const char * number,const std::string & msg)138 CheckedError Parser::InvalidNumber(const char *number, const std::string &msg) {
139   return Error("invalid number: \"" + std::string(number) + "\"" + msg);
140 }
141 // atot: templated version of atoi/atof: convert a string to an instance of T.
142 template<typename T>
atot(const char * s,Parser & parser,T * val)143 inline CheckedError atot(const char *s, Parser &parser, T *val) {
144   auto done = StringToNumber(s, val);
145   if (done) return NoError();
146 
147   return parser.InvalidNumber(
148       s, (0 == *val)
149              ? ""
150              : (", constant does not fit [" +
151                 NumToString(flatbuffers::numeric_limits<T>::lowest()) + "; " +
152                 NumToString(flatbuffers::numeric_limits<T>::max()) + "]"));
153 }
154 template<>
atot(const char * s,Parser & parser,Offset<void> * val)155 inline CheckedError atot<Offset<void>>(const char *s, Parser &parser,
156                                        Offset<void> *val) {
157   (void)parser;
158   *val = Offset<void>(atoi(s));
159   return NoError();
160 }
161 
GetFullyQualifiedName(const std::string & name,size_t max_components) const162 std::string Namespace::GetFullyQualifiedName(const std::string &name,
163                                              size_t max_components) const {
164   // Early exit if we don't have a defined namespace.
165   if (components.empty() || !max_components) { return name; }
166   std::string stream_str;
167   for (size_t i = 0; i < std::min(components.size(), max_components); i++) {
168     if (i) { stream_str += '.'; }
169     stream_str += std::string(components[i]);
170   }
171   if (name.length()) {
172     stream_str += '.';
173     stream_str += name;
174   }
175   return stream_str;
176 }
177 
178 // Declare tokens we'll use. Single character tokens are represented by their
179 // ascii character code (e.g. '{'), others above 256.
180 // clang-format off
181 #define FLATBUFFERS_GEN_TOKENS(TD) \
182   TD(Eof, 256, "end of file") \
183   TD(StringConstant, 257, "string constant") \
184   TD(IntegerConstant, 258, "integer constant") \
185   TD(FloatConstant, 259, "float constant") \
186   TD(Identifier, 260, "identifier")
187 #ifdef __GNUC__
188 __extension__  // Stop GCC complaining about trailing comma with -Wpendantic.
189 #endif
190 enum {
191   #define FLATBUFFERS_TOKEN(NAME, VALUE, STRING) kToken ## NAME = VALUE,
192     FLATBUFFERS_GEN_TOKENS(FLATBUFFERS_TOKEN)
193   #undef FLATBUFFERS_TOKEN
194 };
195 
TokenToString(int t)196 static std::string TokenToString(int t) {
197   static const char * const tokens[] = {
198     #define FLATBUFFERS_TOKEN(NAME, VALUE, STRING) STRING,
199       FLATBUFFERS_GEN_TOKENS(FLATBUFFERS_TOKEN)
200     #undef FLATBUFFERS_TOKEN
201     #define FLATBUFFERS_TD(ENUM, IDLTYPE, \
202       CTYPE, JTYPE, GTYPE, NTYPE, PTYPE, RTYPE) \
203       IDLTYPE,
204       FLATBUFFERS_GEN_TYPES(FLATBUFFERS_TD)
205     #undef FLATBUFFERS_TD
206   };
207   if (t < 256) {  // A single ascii char token.
208     std::string s;
209     s.append(1, static_cast<char>(t));
210     return s;
211   } else {       // Other tokens.
212     return tokens[t - 256];
213   }
214 }
215 // clang-format on
216 
TokenToStringId(int t) const217 std::string Parser::TokenToStringId(int t) const {
218   return t == kTokenIdentifier ? attribute_ : TokenToString(t);
219 }
220 
221 // Parses exactly nibbles worth of hex digits into a number, or error.
ParseHexNum(int nibbles,uint64_t * val)222 CheckedError Parser::ParseHexNum(int nibbles, uint64_t *val) {
223   FLATBUFFERS_ASSERT(nibbles > 0);
224   for (int i = 0; i < nibbles; i++)
225     if (!is_xdigit(cursor_[i]))
226       return Error("escape code must be followed by " + NumToString(nibbles) +
227                    " hex digits");
228   std::string target(cursor_, cursor_ + nibbles);
229   *val = StringToUInt(target.c_str(), 16);
230   cursor_ += nibbles;
231   return NoError();
232 }
233 
SkipByteOrderMark()234 CheckedError Parser::SkipByteOrderMark() {
235   if (static_cast<unsigned char>(*cursor_) != 0xef) return NoError();
236   cursor_++;
237   if (static_cast<unsigned char>(*cursor_) != 0xbb)
238     return Error("invalid utf-8 byte order mark");
239   cursor_++;
240   if (static_cast<unsigned char>(*cursor_) != 0xbf)
241     return Error("invalid utf-8 byte order mark");
242   cursor_++;
243   return NoError();
244 }
245 
IsIdentifierStart(char c)246 static inline bool IsIdentifierStart(char c) {
247   return is_alpha(c) || (c == '_');
248 }
249 
Next()250 CheckedError Parser::Next() {
251   doc_comment_.clear();
252   bool seen_newline = cursor_ == source_;
253   attribute_.clear();
254   attr_is_trivial_ascii_string_ = true;
255   for (;;) {
256     char c = *cursor_++;
257     token_ = c;
258     switch (c) {
259       case '\0':
260         cursor_--;
261         token_ = kTokenEof;
262         return NoError();
263       case ' ':
264       case '\r':
265       case '\t': break;
266       case '\n':
267         MarkNewLine();
268         seen_newline = true;
269         break;
270       case '{':
271       case '}':
272       case '(':
273       case ')':
274       case '[':
275       case ']':
276       case ',':
277       case ':':
278       case ';':
279       case '=': return NoError();
280       case '\"':
281       case '\'': {
282         int unicode_high_surrogate = -1;
283 
284         while (*cursor_ != c) {
285           if (*cursor_ < ' ' && static_cast<signed char>(*cursor_) >= 0)
286             return Error("illegal character in string constant");
287           if (*cursor_ == '\\') {
288             attr_is_trivial_ascii_string_ = false;  // has escape sequence
289             cursor_++;
290             if (unicode_high_surrogate != -1 && *cursor_ != 'u') {
291               return Error(
292                   "illegal Unicode sequence (unpaired high surrogate)");
293             }
294             switch (*cursor_) {
295               case 'n':
296                 attribute_ += '\n';
297                 cursor_++;
298                 break;
299               case 't':
300                 attribute_ += '\t';
301                 cursor_++;
302                 break;
303               case 'r':
304                 attribute_ += '\r';
305                 cursor_++;
306                 break;
307               case 'b':
308                 attribute_ += '\b';
309                 cursor_++;
310                 break;
311               case 'f':
312                 attribute_ += '\f';
313                 cursor_++;
314                 break;
315               case '\"':
316                 attribute_ += '\"';
317                 cursor_++;
318                 break;
319               case '\'':
320                 attribute_ += '\'';
321                 cursor_++;
322                 break;
323               case '\\':
324                 attribute_ += '\\';
325                 cursor_++;
326                 break;
327               case '/':
328                 attribute_ += '/';
329                 cursor_++;
330                 break;
331               case 'x': {  // Not in the JSON standard
332                 cursor_++;
333                 uint64_t val;
334                 ECHECK(ParseHexNum(2, &val));
335                 attribute_ += static_cast<char>(val);
336                 break;
337               }
338               case 'u': {
339                 cursor_++;
340                 uint64_t val;
341                 ECHECK(ParseHexNum(4, &val));
342                 if (val >= 0xD800 && val <= 0xDBFF) {
343                   if (unicode_high_surrogate != -1) {
344                     return Error(
345                         "illegal Unicode sequence (multiple high surrogates)");
346                   } else {
347                     unicode_high_surrogate = static_cast<int>(val);
348                   }
349                 } else if (val >= 0xDC00 && val <= 0xDFFF) {
350                   if (unicode_high_surrogate == -1) {
351                     return Error(
352                         "illegal Unicode sequence (unpaired low surrogate)");
353                   } else {
354                     int code_point = 0x10000 +
355                                      ((unicode_high_surrogate & 0x03FF) << 10) +
356                                      (val & 0x03FF);
357                     ToUTF8(code_point, &attribute_);
358                     unicode_high_surrogate = -1;
359                   }
360                 } else {
361                   if (unicode_high_surrogate != -1) {
362                     return Error(
363                         "illegal Unicode sequence (unpaired high surrogate)");
364                   }
365                   ToUTF8(static_cast<int>(val), &attribute_);
366                 }
367                 break;
368               }
369               default: return Error("unknown escape code in string constant");
370             }
371           } else {  // printable chars + UTF-8 bytes
372             if (unicode_high_surrogate != -1) {
373               return Error(
374                   "illegal Unicode sequence (unpaired high surrogate)");
375             }
376             // reset if non-printable
377             attr_is_trivial_ascii_string_ &= check_in_range(*cursor_, ' ', '~');
378 
379             attribute_ += *cursor_++;
380           }
381         }
382         if (unicode_high_surrogate != -1) {
383           return Error("illegal Unicode sequence (unpaired high surrogate)");
384         }
385         cursor_++;
386         if (!attr_is_trivial_ascii_string_ && !opts.allow_non_utf8 &&
387             !ValidateUTF8(attribute_)) {
388           return Error("illegal UTF-8 sequence");
389         }
390         token_ = kTokenStringConstant;
391         return NoError();
392       }
393       case '/':
394         if (*cursor_ == '/') {
395           const char *start = ++cursor_;
396           while (*cursor_ && *cursor_ != '\n' && *cursor_ != '\r') cursor_++;
397           if (*start == '/') {  // documentation comment
398             if (!seen_newline)
399               return Error(
400                   "a documentation comment should be on a line on its own");
401             doc_comment_.push_back(std::string(start + 1, cursor_));
402           }
403           break;
404         } else if (*cursor_ == '*') {
405           cursor_++;
406           // TODO: make nested.
407           while (*cursor_ != '*' || cursor_[1] != '/') {
408             if (*cursor_ == '\n') MarkNewLine();
409             if (!*cursor_) return Error("end of file in comment");
410             cursor_++;
411           }
412           cursor_ += 2;
413           break;
414         }
415         FLATBUFFERS_FALLTHROUGH(); // else fall thru
416       default:
417         const auto has_sign = (c == '+') || (c == '-');
418         // '-'/'+' and following identifier - can be a predefined constant like:
419         // NAN, INF, PI, etc.
420         if (IsIdentifierStart(c) || (has_sign && IsIdentifierStart(*cursor_))) {
421           // Collect all chars of an identifier:
422           const char *start = cursor_ - 1;
423           while (IsIdentifierStart(*cursor_) || is_digit(*cursor_)) cursor_++;
424           attribute_.append(start, cursor_);
425           token_ = has_sign ? kTokenStringConstant : kTokenIdentifier;
426           return NoError();
427         }
428 
429         auto dot_lvl = (c == '.') ? 0 : 1;  // dot_lvl==0 <=> exactly one '.' seen
430         if (!dot_lvl && !is_digit(*cursor_)) return NoError(); // enum?
431         // Parser accepts hexadecimal-floating-literal (see C++ 5.13.4).
432         if (is_digit(c) || has_sign || !dot_lvl) {
433           const auto start = cursor_ - 1;
434           auto start_digits = !is_digit(c) ? cursor_ : cursor_ - 1;
435           if (!is_digit(c) && is_digit(*cursor_)){
436             start_digits = cursor_; // see digit in cursor_ position
437             c = *cursor_++;
438           }
439           // hex-float can't begind with '.'
440           auto use_hex = dot_lvl && (c == '0') && is_alpha_char(*cursor_, 'X');
441           if (use_hex) start_digits = ++cursor_;  // '0x' is the prefix, skip it
442           // Read an integer number or mantisa of float-point number.
443           do {
444             if (use_hex) {
445               while (is_xdigit(*cursor_)) cursor_++;
446             } else {
447               while (is_digit(*cursor_)) cursor_++;
448             }
449           } while ((*cursor_ == '.') && (++cursor_) && (--dot_lvl >= 0));
450           // Exponent of float-point number.
451           if ((dot_lvl >= 0) && (cursor_ > start_digits)) {
452             // The exponent suffix of hexadecimal float number is mandatory.
453             if (use_hex && !dot_lvl) start_digits = cursor_;
454             if ((use_hex && is_alpha_char(*cursor_, 'P')) ||
455                 is_alpha_char(*cursor_, 'E')) {
456               dot_lvl = 0;  // Emulate dot to signal about float-point number.
457               cursor_++;
458               if (*cursor_ == '+' || *cursor_ == '-') cursor_++;
459               start_digits = cursor_;  // the exponent-part has to have digits
460               // Exponent is decimal integer number
461               while (is_digit(*cursor_)) cursor_++;
462               if (*cursor_ == '.') {
463                 cursor_++;  // If see a dot treat it as part of invalid number.
464                 dot_lvl = -1;  // Fall thru to Error().
465               }
466             }
467           }
468           // Finalize.
469           if ((dot_lvl >= 0) && (cursor_ > start_digits)) {
470             attribute_.append(start, cursor_);
471             token_ = dot_lvl ? kTokenIntegerConstant : kTokenFloatConstant;
472             return NoError();
473           } else {
474             return Error("invalid number: " + std::string(start, cursor_));
475           }
476         }
477         std::string ch;
478         ch = c;
479         if (false == check_in_range(c, ' ', '~')) ch = "code: " + NumToString(c);
480         return Error("illegal character: " + ch);
481     }
482   }
483 }
484 
485 // Check if a given token is next.
Is(int t) const486 bool Parser::Is(int t) const { return t == token_; }
487 
IsIdent(const char * id) const488 bool Parser::IsIdent(const char *id) const {
489   return token_ == kTokenIdentifier && attribute_ == id;
490 }
491 
492 // Expect a given token to be next, consume it, or error if not present.
Expect(int t)493 CheckedError Parser::Expect(int t) {
494   if (t != token_) {
495     return Error("expecting: " + TokenToString(t) +
496                  " instead got: " + TokenToStringId(token_));
497   }
498   NEXT();
499   return NoError();
500 }
501 
ParseNamespacing(std::string * id,std::string * last)502 CheckedError Parser::ParseNamespacing(std::string *id, std::string *last) {
503   while (Is('.')) {
504     NEXT();
505     *id += ".";
506     *id += attribute_;
507     if (last) *last = attribute_;
508     EXPECT(kTokenIdentifier);
509   }
510   return NoError();
511 }
512 
LookupEnum(const std::string & id)513 EnumDef *Parser::LookupEnum(const std::string &id) {
514   // Search thru parent namespaces.
515   for (int components = static_cast<int>(current_namespace_->components.size());
516        components >= 0; components--) {
517     auto ed = enums_.Lookup(
518         current_namespace_->GetFullyQualifiedName(id, components));
519     if (ed) return ed;
520   }
521   return nullptr;
522 }
523 
LookupStruct(const std::string & id) const524 StructDef *Parser::LookupStruct(const std::string &id) const {
525   auto sd = structs_.Lookup(id);
526   if (sd) sd->refcount++;
527   return sd;
528 }
529 
ParseTypeIdent(Type & type)530 CheckedError Parser::ParseTypeIdent(Type &type) {
531   std::string id = attribute_;
532   EXPECT(kTokenIdentifier);
533   ECHECK(ParseNamespacing(&id, nullptr));
534   auto enum_def = LookupEnum(id);
535   if (enum_def) {
536     type = enum_def->underlying_type;
537     if (enum_def->is_union) type.base_type = BASE_TYPE_UNION;
538   } else {
539     type.base_type = BASE_TYPE_STRUCT;
540     type.struct_def = LookupCreateStruct(id);
541   }
542   return NoError();
543 }
544 
545 // Parse any IDL type.
ParseType(Type & type)546 CheckedError Parser::ParseType(Type &type) {
547   if (token_ == kTokenIdentifier) {
548     if (IsIdent("bool")) {
549       type.base_type = BASE_TYPE_BOOL;
550       NEXT();
551     } else if (IsIdent("byte") || IsIdent("int8")) {
552       type.base_type = BASE_TYPE_CHAR;
553       NEXT();
554     } else if (IsIdent("ubyte") || IsIdent("uint8")) {
555       type.base_type = BASE_TYPE_UCHAR;
556       NEXT();
557     } else if (IsIdent("short") || IsIdent("int16")) {
558       type.base_type = BASE_TYPE_SHORT;
559       NEXT();
560     } else if (IsIdent("ushort") || IsIdent("uint16")) {
561       type.base_type = BASE_TYPE_USHORT;
562       NEXT();
563     } else if (IsIdent("int") || IsIdent("int32")) {
564       type.base_type = BASE_TYPE_INT;
565       NEXT();
566     } else if (IsIdent("uint") || IsIdent("uint32")) {
567       type.base_type = BASE_TYPE_UINT;
568       NEXT();
569     } else if (IsIdent("long") || IsIdent("int64")) {
570       type.base_type = BASE_TYPE_LONG;
571       NEXT();
572     } else if (IsIdent("ulong") || IsIdent("uint64")) {
573       type.base_type = BASE_TYPE_ULONG;
574       NEXT();
575     } else if (IsIdent("float") || IsIdent("float32")) {
576       type.base_type = BASE_TYPE_FLOAT;
577       NEXT();
578     } else if (IsIdent("double") || IsIdent("float64")) {
579       type.base_type = BASE_TYPE_DOUBLE;
580       NEXT();
581     } else if (IsIdent("string")) {
582       type.base_type = BASE_TYPE_STRING;
583       NEXT();
584     } else {
585       ECHECK(ParseTypeIdent(type));
586     }
587   } else if (token_ == '[') {
588     NEXT();
589     Type subtype;
590     ECHECK(Recurse([&]() { return ParseType(subtype); }));
591     if (subtype.base_type == BASE_TYPE_VECTOR) {
592       // We could support this, but it will complicate things, and it's
593       // easier to work around with a struct around the inner vector.
594       return Error("nested vector types not supported (wrap in table first).");
595     }
596     type = Type(BASE_TYPE_VECTOR, subtype.struct_def, subtype.enum_def);
597     type.element = subtype.base_type;
598     EXPECT(']');
599   } else {
600     return Error("illegal type syntax");
601   }
602   return NoError();
603 }
604 
AddField(StructDef & struct_def,const std::string & name,const Type & type,FieldDef ** dest)605 CheckedError Parser::AddField(StructDef &struct_def, const std::string &name,
606                               const Type &type, FieldDef **dest) {
607   auto &field = *new FieldDef();
608   field.value.offset =
609       FieldIndexToOffset(static_cast<voffset_t>(struct_def.fields.vec.size()));
610   field.name = name;
611   field.file = struct_def.file;
612   field.value.type = type;
613   if (struct_def.fixed) {  // statically compute the field offset
614     auto size = InlineSize(type);
615     auto alignment = InlineAlignment(type);
616     // structs_ need to have a predictable format, so we need to align to
617     // the largest scalar
618     struct_def.minalign = std::max(struct_def.minalign, alignment);
619     struct_def.PadLastField(alignment);
620     field.value.offset = static_cast<voffset_t>(struct_def.bytesize);
621     struct_def.bytesize += size;
622   }
623   if (struct_def.fields.Add(name, &field))
624     return Error("field already exists: " + name);
625   *dest = &field;
626   return NoError();
627 }
628 
ParseField(StructDef & struct_def)629 CheckedError Parser::ParseField(StructDef &struct_def) {
630   std::string name = attribute_;
631 
632   if (LookupStruct(name))
633     return Error("field name can not be the same as table/struct name");
634 
635   std::vector<std::string> dc = doc_comment_;
636   EXPECT(kTokenIdentifier);
637   EXPECT(':');
638   Type type;
639   ECHECK(ParseType(type));
640 
641   if (struct_def.fixed && !IsScalar(type.base_type) && !IsStruct(type))
642     return Error("structs_ may contain only scalar or struct fields");
643 
644   FieldDef *typefield = nullptr;
645   if (type.base_type == BASE_TYPE_UNION) {
646     // For union fields, add a second auto-generated field to hold the type,
647     // with a special suffix.
648     ECHECK(AddField(struct_def, name + UnionTypeFieldSuffix(),
649                     type.enum_def->underlying_type, &typefield));
650   } else if (type.base_type == BASE_TYPE_VECTOR &&
651              type.element == BASE_TYPE_UNION) {
652     // Only cpp, js and ts supports the union vector feature so far.
653     if (!SupportsVectorOfUnions()) {
654       return Error(
655           "Vectors of unions are not yet supported in all "
656           "the specified programming languages.");
657     }
658     // For vector of union fields, add a second auto-generated vector field to
659     // hold the types, with a special suffix.
660     Type union_vector(BASE_TYPE_VECTOR, nullptr, type.enum_def);
661     union_vector.element = BASE_TYPE_UTYPE;
662     ECHECK(AddField(struct_def, name + UnionTypeFieldSuffix(), union_vector,
663                     &typefield));
664   }
665 
666   FieldDef *field;
667   ECHECK(AddField(struct_def, name, type, &field));
668 
669   if (token_ == '=') {
670     NEXT();
671     if (!IsScalar(type.base_type) ||
672         (struct_def.fixed && field->value.constant != "0"))
673       return Error(
674             "default values currently only supported for scalars in tables");
675     ECHECK(ParseSingleValue(&field->name, field->value, true));
676   }
677   if (type.enum_def &&
678       !type.enum_def->is_union &&
679       !type.enum_def->attributes.Lookup("bit_flags") &&
680       !type.enum_def->ReverseLookup(StringToInt(
681                                       field->value.constant.c_str()))) {
682     return Error("default value of " + field->value.constant + " for field " +
683                  name + " is not part of enum " + type.enum_def->name);
684   }
685   // Append .0 if the value has not it (skip hex and scientific floats).
686   // This suffix needed for generated C++ code.
687   if (IsFloat(type.base_type)) {
688     auto &text = field->value.constant;
689     FLATBUFFERS_ASSERT(false == text.empty());
690     auto s = text.c_str();
691     while(*s == ' ') s++;
692     if (*s == '-' || *s == '+') s++;
693     // 1) A float constants (nan, inf, pi, etc) is a kind of identifier.
694     // 2) A float number needn't ".0" at the end if it has exponent.
695     if ((false == IsIdentifierStart(*s)) &&
696         (std::string::npos == field->value.constant.find_first_of(".eEpP"))) {
697       field->value.constant += ".0";
698     }
699   }
700 
701   if (type.enum_def && IsScalar(type.base_type) && !struct_def.fixed &&
702       !type.enum_def->attributes.Lookup("bit_flags") &&
703       !type.enum_def->ReverseLookup(StringToInt(
704                                       field->value.constant.c_str())))
705     Warning("enum " + type.enum_def->name +
706             " does not have a declaration for this field\'s default of " +
707             field->value.constant);
708 
709   field->doc_comment = dc;
710   ECHECK(ParseMetaData(&field->attributes));
711   field->deprecated = field->attributes.Lookup("deprecated") != nullptr;
712   auto hash_name = field->attributes.Lookup("hash");
713   if (hash_name) {
714     switch ((type.base_type == BASE_TYPE_VECTOR) ? type.element : type.base_type) {
715       case BASE_TYPE_SHORT:
716       case BASE_TYPE_USHORT: {
717         if (FindHashFunction16(hash_name->constant.c_str()) == nullptr)
718           return Error("Unknown hashing algorithm for 16 bit types: " +
719                        hash_name->constant);
720         break;
721       }
722       case BASE_TYPE_INT:
723       case BASE_TYPE_UINT: {
724         if (FindHashFunction32(hash_name->constant.c_str()) == nullptr)
725           return Error("Unknown hashing algorithm for 32 bit types: " +
726                        hash_name->constant);
727         break;
728       }
729       case BASE_TYPE_LONG:
730       case BASE_TYPE_ULONG: {
731         if (FindHashFunction64(hash_name->constant.c_str()) == nullptr)
732           return Error("Unknown hashing algorithm for 64 bit types: " +
733                        hash_name->constant);
734         break;
735       }
736       default:
737         return Error(
738             "only short, ushort, int, uint, long and ulong data types support hashing.");
739     }
740   }
741   auto cpp_type = field->attributes.Lookup("cpp_type");
742   if (cpp_type) {
743     if (!hash_name)
744       return Error("cpp_type can only be used with a hashed field");
745     /// forcing cpp_ptr_type to 'naked' if unset
746     auto cpp_ptr_type = field->attributes.Lookup("cpp_ptr_type");
747     if (!cpp_ptr_type) {
748       auto val = new Value();
749       val->type = cpp_type->type;
750       val->constant = "naked";
751       field->attributes.Add("cpp_ptr_type", val);
752     }
753   }
754   if (field->deprecated && struct_def.fixed)
755     return Error("can't deprecate fields in a struct");
756   field->required = field->attributes.Lookup("required") != nullptr;
757   if (field->required &&
758       (struct_def.fixed || IsScalar(type.base_type)))
759     return Error("only non-scalar fields in tables may be 'required'");
760   field->key = field->attributes.Lookup("key") != nullptr;
761   if (field->key) {
762     if (struct_def.has_key) return Error("only one field may be set as 'key'");
763     struct_def.has_key = true;
764     if (!IsScalar(type.base_type)) {
765       field->required = true;
766       if (type.base_type != BASE_TYPE_STRING)
767         return Error("'key' field must be string or scalar type");
768     }
769   }
770   field->shared = field->attributes.Lookup("shared") != nullptr;
771   if (field->shared && field->value.type.base_type != BASE_TYPE_STRING)
772     return Error("shared can only be defined on strings");
773 
774   auto field_native_custom_alloc =
775       field->attributes.Lookup("native_custom_alloc");
776   if (field_native_custom_alloc)
777     return Error(
778         "native_custom_alloc can only be used with a table or struct "
779         "definition");
780 
781   field->native_inline = field->attributes.Lookup("native_inline") != nullptr;
782   if (field->native_inline && !IsStruct(field->value.type))
783     return Error("native_inline can only be defined on structs");
784 
785   auto nested = field->attributes.Lookup("nested_flatbuffer");
786   if (nested) {
787     if (nested->type.base_type != BASE_TYPE_STRING)
788       return Error(
789           "nested_flatbuffer attribute must be a string (the root type)");
790     if (type.base_type != BASE_TYPE_VECTOR || type.element != BASE_TYPE_UCHAR)
791       return Error(
792           "nested_flatbuffer attribute may only apply to a vector of ubyte");
793     // This will cause an error if the root type of the nested flatbuffer
794     // wasn't defined elsewhere.
795     LookupCreateStruct(nested->constant);
796 
797     // Keep a pointer to StructDef in FieldDef to simplify re-use later
798     auto nested_qualified_name =
799         current_namespace_->GetFullyQualifiedName(nested->constant);
800     field->nested_flatbuffer = LookupStruct(nested_qualified_name);
801   }
802 
803   if (field->attributes.Lookup("flexbuffer")) {
804     field->flexbuffer = true;
805     uses_flexbuffers_ = true;
806     if (type.base_type != BASE_TYPE_VECTOR ||
807         type.element != BASE_TYPE_UCHAR)
808       return Error("flexbuffer attribute may only apply to a vector of ubyte");
809   }
810 
811   if (typefield) {
812     if (!IsScalar(typefield->value.type.base_type)) {
813       // this is a union vector field
814       typefield->required = field->required;
815     }
816     // If this field is a union, and it has a manually assigned id,
817     // the automatically added type field should have an id as well (of N - 1).
818     auto attr = field->attributes.Lookup("id");
819     if (attr) {
820       auto id = atoi(attr->constant.c_str());
821       auto val = new Value();
822       val->type = attr->type;
823       val->constant = NumToString(id - 1);
824       typefield->attributes.Add("id", val);
825     }
826   }
827 
828   EXPECT(';');
829   return NoError();
830 }
831 
ParseString(Value & val)832 CheckedError Parser::ParseString(Value &val) {
833   auto s = attribute_;
834   EXPECT(kTokenStringConstant);
835   val.constant = NumToString(builder_.CreateString(s).o);
836   return NoError();
837 }
838 
ParseComma()839 CheckedError Parser::ParseComma() {
840   if (!opts.protobuf_ascii_alike) EXPECT(',');
841   return NoError();
842 }
843 
ParseAnyValue(Value & val,FieldDef * field,size_t parent_fieldn,const StructDef * parent_struct_def)844 CheckedError Parser::ParseAnyValue(Value &val, FieldDef *field,
845                                    size_t parent_fieldn,
846                                    const StructDef *parent_struct_def) {
847   switch (val.type.base_type) {
848     case BASE_TYPE_UNION: {
849       FLATBUFFERS_ASSERT(field);
850       std::string constant;
851       // Find corresponding type field we may have already parsed.
852       for (auto elem = field_stack_.rbegin();
853            elem != field_stack_.rbegin() + parent_fieldn; ++elem) {
854         auto &type = elem->second->value.type;
855         if (type.base_type == BASE_TYPE_UTYPE &&
856             type.enum_def == val.type.enum_def) {
857           constant = elem->first.constant;
858           break;
859         }
860       }
861       if (constant.empty()) {
862         // We haven't seen the type field yet. Sadly a lot of JSON writers
863         // output these in alphabetical order, meaning it comes after this
864         // value. So we scan past the value to find it, then come back here.
865         auto type_name = field->name + UnionTypeFieldSuffix();
866         FLATBUFFERS_ASSERT(parent_struct_def);
867         auto type_field = parent_struct_def->fields.Lookup(type_name);
868         FLATBUFFERS_ASSERT(type_field);  // Guaranteed by ParseField().
869         // Remember where we are in the source file, so we can come back here.
870         auto backup = *static_cast<ParserState *>(this);
871         ECHECK(SkipAnyJsonValue());  // The table.
872         ECHECK(ParseComma());
873         auto next_name = attribute_;
874         if (Is(kTokenStringConstant)) {
875           NEXT();
876         } else {
877           EXPECT(kTokenIdentifier);
878         }
879         if (next_name != type_name)
880           return Error("missing type field after this union value: " +
881                        type_name);
882         EXPECT(':');
883         Value type_val = type_field->value;
884         ECHECK(ParseAnyValue(type_val, type_field, 0, nullptr));
885         constant = type_val.constant;
886         // Got the information we needed, now rewind:
887         *static_cast<ParserState *>(this) = backup;
888       }
889       uint8_t enum_idx;
890       ECHECK(atot(constant.c_str(), *this, &enum_idx));
891       auto enum_val = val.type.enum_def->ReverseLookup(enum_idx);
892       if (!enum_val) return Error("illegal type id for: " + field->name);
893       if (enum_val->union_type.base_type == BASE_TYPE_STRUCT) {
894         ECHECK(ParseTable(*enum_val->union_type.struct_def, &val.constant,
895                           nullptr));
896         if (enum_val->union_type.struct_def->fixed) {
897           // All BASE_TYPE_UNION values are offsets, so turn this into one.
898           SerializeStruct(*enum_val->union_type.struct_def, val);
899           builder_.ClearOffsets();
900           val.constant = NumToString(builder_.GetSize());
901         }
902       } else if (enum_val->union_type.base_type == BASE_TYPE_STRING) {
903         ECHECK(ParseString(val));
904       } else {
905         FLATBUFFERS_ASSERT(false);
906       }
907       break;
908     }
909     case BASE_TYPE_STRUCT:
910       ECHECK(ParseTable(*val.type.struct_def, &val.constant, nullptr));
911       break;
912     case BASE_TYPE_STRING: {
913       ECHECK(ParseString(val));
914       break;
915     }
916     case BASE_TYPE_VECTOR: {
917       uoffset_t off;
918       ECHECK(ParseVector(val.type.VectorType(), &off));
919       val.constant = NumToString(off);
920       break;
921     }
922     case BASE_TYPE_INT:
923     case BASE_TYPE_UINT:
924     case BASE_TYPE_LONG:
925     case BASE_TYPE_ULONG: {
926       if (field && field->attributes.Lookup("hash") &&
927           (token_ == kTokenIdentifier || token_ == kTokenStringConstant)) {
928         ECHECK(ParseHash(val, field));
929       } else {
930         ECHECK(ParseSingleValue(field ? &field->name : nullptr, val, false));
931       }
932       break;
933     }
934     default:
935       ECHECK(ParseSingleValue(field ? &field->name : nullptr, val, false));
936       break;
937   }
938   return NoError();
939 }
940 
SerializeStruct(const StructDef & struct_def,const Value & val)941 void Parser::SerializeStruct(const StructDef &struct_def, const Value &val) {
942   FLATBUFFERS_ASSERT(val.constant.length() == struct_def.bytesize);
943   builder_.Align(struct_def.minalign);
944   builder_.PushBytes(reinterpret_cast<const uint8_t *>(val.constant.c_str()),
945                      struct_def.bytesize);
946   builder_.AddStructOffset(val.offset, builder_.GetSize());
947 }
948 
949 template <typename F>
ParseTableDelimiters(size_t & fieldn,const StructDef * struct_def,F body)950 CheckedError Parser::ParseTableDelimiters(size_t &fieldn,
951                                           const StructDef *struct_def,
952                                           F body) {
953   // We allow tables both as JSON object{ .. } with field names
954   // or vector[..] with all fields in order
955   char terminator = '}';
956   bool is_nested_vector = struct_def && Is('[');
957   if (is_nested_vector) {
958     NEXT();
959     terminator = ']';
960   } else {
961     EXPECT('{');
962   }
963   for (;;) {
964     if ((!opts.strict_json || !fieldn) && Is(terminator)) break;
965     std::string name;
966     if (is_nested_vector) {
967       if (fieldn >= struct_def->fields.vec.size()) {
968         return Error("too many unnamed fields in nested array");
969       }
970       name = struct_def->fields.vec[fieldn]->name;
971     } else {
972       name = attribute_;
973       if (Is(kTokenStringConstant)) {
974         NEXT();
975       } else {
976         EXPECT(opts.strict_json ? kTokenStringConstant : kTokenIdentifier);
977       }
978       if (!opts.protobuf_ascii_alike || !(Is('{') || Is('['))) EXPECT(':');
979     }
980     ECHECK(body(name, fieldn, struct_def));
981     if (Is(terminator)) break;
982     ECHECK(ParseComma());
983   }
984   NEXT();
985   if (is_nested_vector && fieldn != struct_def->fields.vec.size()) {
986     return Error("wrong number of unnamed fields in table vector");
987   }
988   return NoError();
989 }
990 
ParseTable(const StructDef & struct_def,std::string * value,uoffset_t * ovalue)991 CheckedError Parser::ParseTable(const StructDef &struct_def, std::string *value,
992                                 uoffset_t *ovalue) {
993   size_t fieldn_outer = 0;
994   auto err = ParseTableDelimiters(
995       fieldn_outer, &struct_def,
996       [&](const std::string &name, size_t &fieldn,
997           const StructDef *struct_def_inner) -> CheckedError {
998         if (name == "$schema") {
999           ECHECK(Expect(kTokenStringConstant));
1000           return NoError();
1001         }
1002         auto field = struct_def_inner->fields.Lookup(name);
1003         if (!field) {
1004           if (!opts.skip_unexpected_fields_in_json) {
1005             return Error("unknown field: " + name);
1006           } else {
1007             ECHECK(SkipAnyJsonValue());
1008           }
1009         } else {
1010           if (IsIdent("null") && !IsScalar(field->value.type.base_type)) {
1011             ECHECK(Next());  // Ignore this field.
1012           } else {
1013             Value val = field->value;
1014             if (field->flexbuffer) {
1015               flexbuffers::Builder builder(1024,
1016                                            flexbuffers::BUILDER_FLAG_SHARE_ALL);
1017               ECHECK(ParseFlexBufferValue(&builder));
1018               builder.Finish();
1019               // Force alignment for nested flexbuffer
1020               builder_.ForceVectorAlignment(builder.GetSize(), sizeof(uint8_t),
1021                                             sizeof(largest_scalar_t));
1022               auto off = builder_.CreateVector(builder.GetBuffer());
1023               val.constant = NumToString(off.o);
1024             } else if (field->nested_flatbuffer) {
1025               ECHECK(
1026                   ParseNestedFlatbuffer(val, field, fieldn, struct_def_inner));
1027             } else {
1028               ECHECK(Recurse([&]() {
1029                 return ParseAnyValue(val, field, fieldn, struct_def_inner);
1030               }));
1031             }
1032             // Hardcoded insertion-sort with error-check.
1033             // If fields are specified in order, then this loop exits
1034             // immediately.
1035             auto elem = field_stack_.rbegin();
1036             for (; elem != field_stack_.rbegin() + fieldn; ++elem) {
1037               auto existing_field = elem->second;
1038               if (existing_field == field)
1039                 return Error("field set more than once: " + field->name);
1040               if (existing_field->value.offset < field->value.offset) break;
1041             }
1042             // Note: elem points to before the insertion point, thus .base()
1043             // points to the correct spot.
1044             field_stack_.insert(elem.base(), std::make_pair(val, field));
1045             fieldn++;
1046           }
1047         }
1048         return NoError();
1049       });
1050   ECHECK(err);
1051 
1052   // Check if all required fields are parsed.
1053   for (auto field_it = struct_def.fields.vec.begin();
1054        field_it != struct_def.fields.vec.end(); ++field_it) {
1055     auto required_field = *field_it;
1056     if (!required_field->required) { continue; }
1057     bool found = false;
1058     for (auto pf_it = field_stack_.end() - fieldn_outer;
1059          pf_it != field_stack_.end(); ++pf_it) {
1060       auto parsed_field = pf_it->second;
1061       if (parsed_field == required_field) {
1062         found = true;
1063         break;
1064       }
1065     }
1066     if (!found) {
1067       return Error("required field is missing: " + required_field->name +
1068                    " in " + struct_def.name);
1069     }
1070   }
1071 
1072   if (struct_def.fixed && fieldn_outer != struct_def.fields.vec.size())
1073     return Error("struct: wrong number of initializers: " + struct_def.name);
1074 
1075   auto start = struct_def.fixed ? builder_.StartStruct(struct_def.minalign)
1076                                 : builder_.StartTable();
1077 
1078   for (size_t size = struct_def.sortbysize ? sizeof(largest_scalar_t) : 1; size;
1079        size /= 2) {
1080     // Go through elements in reverse, since we're building the data backwards.
1081     for (auto it = field_stack_.rbegin();
1082          it != field_stack_.rbegin() + fieldn_outer; ++it) {
1083       auto &field_value = it->first;
1084       auto field = it->second;
1085       if (!struct_def.sortbysize ||
1086           size == SizeOf(field_value.type.base_type)) {
1087         switch (field_value.type.base_type) {
1088           // clang-format off
1089           #define FLATBUFFERS_TD(ENUM, IDLTYPE, \
1090             CTYPE, JTYPE, GTYPE, NTYPE, PTYPE, RTYPE) \
1091             case BASE_TYPE_ ## ENUM: \
1092               builder_.Pad(field->padding); \
1093               if (struct_def.fixed) { \
1094                 CTYPE val; \
1095                 ECHECK(atot(field_value.constant.c_str(), *this, &val)); \
1096                 builder_.PushElement(val); \
1097               } else { \
1098                 CTYPE val, valdef; \
1099                 ECHECK(atot(field_value.constant.c_str(), *this, &val)); \
1100                 ECHECK(atot(field->value.constant.c_str(), *this, &valdef)); \
1101                 builder_.AddElement(field_value.offset, val, valdef); \
1102               } \
1103               break;
1104             FLATBUFFERS_GEN_TYPES_SCALAR(FLATBUFFERS_TD);
1105           #undef FLATBUFFERS_TD
1106           #define FLATBUFFERS_TD(ENUM, IDLTYPE, \
1107             CTYPE, JTYPE, GTYPE, NTYPE, PTYPE, RTYPE) \
1108             case BASE_TYPE_ ## ENUM: \
1109               builder_.Pad(field->padding); \
1110               if (IsStruct(field->value.type)) { \
1111                 SerializeStruct(*field->value.type.struct_def, field_value); \
1112               } else { \
1113                 CTYPE val; \
1114                 ECHECK(atot(field_value.constant.c_str(), *this, &val)); \
1115                 builder_.AddOffset(field_value.offset, val); \
1116               } \
1117               break;
1118             FLATBUFFERS_GEN_TYPES_POINTER(FLATBUFFERS_TD);
1119           #undef FLATBUFFERS_TD
1120           // clang-format on
1121         }
1122       }
1123     }
1124   }
1125   for (size_t i = 0; i < fieldn_outer; i++) field_stack_.pop_back();
1126 
1127   if (struct_def.fixed) {
1128     builder_.ClearOffsets();
1129     builder_.EndStruct();
1130     FLATBUFFERS_ASSERT(value);
1131     // Temporarily store this struct in the value string, since it is to
1132     // be serialized in-place elsewhere.
1133     value->assign(
1134         reinterpret_cast<const char *>(builder_.GetCurrentBufferPointer()),
1135         struct_def.bytesize);
1136     builder_.PopBytes(struct_def.bytesize);
1137     FLATBUFFERS_ASSERT(!ovalue);
1138   } else {
1139     auto val = builder_.EndTable(start);
1140     if (ovalue) *ovalue = val;
1141     if (value) *value = NumToString(val);
1142   }
1143   return NoError();
1144 }
1145 
1146 template <typename F>
ParseVectorDelimiters(size_t & count,F body)1147 CheckedError Parser::ParseVectorDelimiters(size_t &count, F body) {
1148   EXPECT('[');
1149   for (;;) {
1150     if ((!opts.strict_json || !count) && Is(']')) break;
1151     ECHECK(body(count));
1152     count++;
1153     if (Is(']')) break;
1154     ECHECK(ParseComma());
1155   }
1156   NEXT();
1157   return NoError();
1158 }
1159 
ParseVector(const Type & type,uoffset_t * ovalue)1160 CheckedError Parser::ParseVector(const Type &type, uoffset_t *ovalue) {
1161   size_t count = 0;
1162   auto err = ParseVectorDelimiters(count, [&](size_t &) -> CheckedError {
1163     Value val;
1164     val.type = type;
1165     ECHECK(Recurse([&]() { return ParseAnyValue(val, nullptr, 0, nullptr); }));
1166     field_stack_.push_back(std::make_pair(val, nullptr));
1167     return NoError();
1168   });
1169   ECHECK(err);
1170 
1171   builder_.StartVector(count * InlineSize(type) / InlineAlignment(type),
1172                        InlineAlignment(type));
1173   for (size_t i = 0; i < count; i++) {
1174     // start at the back, since we're building the data backwards.
1175     auto &val = field_stack_.back().first;
1176     switch (val.type.base_type) {
1177       // clang-format off
1178       #define FLATBUFFERS_TD(ENUM, IDLTYPE, \
1179         CTYPE, JTYPE, GTYPE, NTYPE, PTYPE, RTYPE) \
1180         case BASE_TYPE_ ## ENUM: \
1181           if (IsStruct(val.type)) SerializeStruct(*val.type.struct_def, val); \
1182           else { \
1183              CTYPE elem; \
1184              ECHECK(atot(val.constant.c_str(), *this, &elem)); \
1185              builder_.PushElement(elem); \
1186           } \
1187           break;
1188         FLATBUFFERS_GEN_TYPES(FLATBUFFERS_TD)
1189       #undef FLATBUFFERS_TD
1190       // clang-format on
1191     }
1192     field_stack_.pop_back();
1193   }
1194 
1195   builder_.ClearOffsets();
1196   *ovalue = builder_.EndVector(count);
1197   return NoError();
1198 }
1199 
ParseNestedFlatbuffer(Value & val,FieldDef * field,size_t fieldn,const StructDef * parent_struct_def)1200 CheckedError Parser::ParseNestedFlatbuffer(Value &val, FieldDef *field,
1201                                            size_t fieldn,
1202                                            const StructDef *parent_struct_def) {
1203   if (token_ == '[') {  // backwards compat for 'legacy' ubyte buffers
1204     ECHECK(ParseAnyValue(val, field, fieldn, parent_struct_def));
1205   } else {
1206     auto cursor_at_value_begin = cursor_;
1207     ECHECK(SkipAnyJsonValue());
1208     std::string substring(cursor_at_value_begin - 1, cursor_ - 1);
1209 
1210     // Create and initialize new parser
1211     Parser nested_parser;
1212     FLATBUFFERS_ASSERT(field->nested_flatbuffer);
1213     nested_parser.root_struct_def_ = field->nested_flatbuffer;
1214     nested_parser.enums_ = enums_;
1215     nested_parser.opts = opts;
1216     nested_parser.uses_flexbuffers_ = uses_flexbuffers_;
1217 
1218     // Parse JSON substring into new flatbuffer builder using nested_parser
1219     bool ok = nested_parser.Parse(substring.c_str(), nullptr, nullptr);
1220 
1221     // Clean nested_parser to avoid deleting the elements in
1222     // the SymbolTables on destruction
1223     nested_parser.enums_.dict.clear();
1224     nested_parser.enums_.vec.clear();
1225 
1226     if (!ok) {
1227       ECHECK(Error(nested_parser.error_));
1228     }
1229     // Force alignment for nested flatbuffer
1230     builder_.ForceVectorAlignment(nested_parser.builder_.GetSize(), sizeof(uint8_t),
1231                                   nested_parser.builder_.GetBufferMinAlignment());
1232 
1233     auto off = builder_.CreateVector(nested_parser.builder_.GetBufferPointer(),
1234                                      nested_parser.builder_.GetSize());
1235     val.constant = NumToString(off.o);
1236   }
1237   return NoError();
1238 }
1239 
ParseMetaData(SymbolTable<Value> * attributes)1240 CheckedError Parser::ParseMetaData(SymbolTable<Value> *attributes) {
1241   if (Is('(')) {
1242     NEXT();
1243     for (;;) {
1244       auto name = attribute_;
1245       if (false == (Is(kTokenIdentifier) || Is(kTokenStringConstant)))
1246         return Error("attribute name must be either identifier or string: " +
1247           name);
1248       if (known_attributes_.find(name) == known_attributes_.end())
1249         return Error("user define attributes must be declared before use: " +
1250                      name);
1251       NEXT();
1252       auto e = new Value();
1253       attributes->Add(name, e);
1254       if (Is(':')) {
1255         NEXT();
1256         ECHECK(ParseSingleValue(&name, *e, true));
1257       }
1258       if (Is(')')) {
1259         NEXT();
1260         break;
1261       }
1262       EXPECT(',');
1263     }
1264   }
1265   return NoError();
1266 }
1267 
TryTypedValue(const std::string * name,int dtoken,bool check,Value & e,BaseType req,bool * destmatch)1268 CheckedError Parser::TryTypedValue(const std::string *name, int dtoken,
1269                                    bool check, Value &e, BaseType req,
1270                                    bool *destmatch) {
1271   bool match = dtoken == token_;
1272   if (match) {
1273     FLATBUFFERS_ASSERT(*destmatch == false);
1274     *destmatch = true;
1275     e.constant = attribute_;
1276     // Check token match
1277     if (!check) {
1278       if (e.type.base_type == BASE_TYPE_NONE) {
1279         e.type.base_type = req;
1280       } else {
1281         return Error(
1282             std::string("type mismatch: expecting: ") +
1283             kTypeNames[e.type.base_type] + ", found: " + kTypeNames[req] +
1284             ", name: " + (name ? *name : "") + ", value: " + e.constant);
1285       }
1286     }
1287     // The exponent suffix of hexadecimal float-point number is mandatory.
1288     // A hex-integer constant is forbidden as an initializer of float number.
1289     if ((kTokenFloatConstant != dtoken) && IsFloat(e.type.base_type)) {
1290       const auto &s = e.constant;
1291       const auto k = s.find_first_of("0123456789.");
1292       if ((std::string::npos != k) && (s.length() > (k + 1)) &&
1293           (s.at(k) == '0' && is_alpha_char(s.at(k + 1), 'X')) &&
1294           (std::string::npos == s.find_first_of("pP", k + 2))) {
1295         return Error(
1296             "invalid number, the exponent suffix of hexadecimal "
1297             "floating-point literals is mandatory: \"" +
1298             s + "\"");
1299       }
1300     }
1301 
1302     NEXT();
1303   }
1304   return NoError();
1305 }
1306 
ParseEnumFromString(Type & type,int64_t * result)1307 CheckedError Parser::ParseEnumFromString(Type &type, int64_t *result) {
1308   *result = 0;
1309   // Parse one or more enum identifiers, separated by spaces.
1310   const char *next = attribute_.c_str();
1311   do {
1312     const char *divider = strchr(next, ' ');
1313     std::string word;
1314     if (divider) {
1315       word = std::string(next, divider);
1316       next = divider + strspn(divider, " ");
1317     } else {
1318       word = next;
1319       next += word.length();
1320     }
1321     if (type.enum_def) {  // The field has an enum type
1322       auto enum_val = type.enum_def->vals.Lookup(word);
1323       if (!enum_val)
1324         return Error("unknown enum value: " + word +
1325                      ", for enum: " + type.enum_def->name);
1326       *result |= enum_val->value;
1327     } else {  // No enum type, probably integral field.
1328       if (!IsInteger(type.base_type))
1329         return Error("not a valid value for this field: " + word);
1330       // TODO: could check if its a valid number constant here.
1331       const char *dot = strrchr(word.c_str(), '.');
1332       if (!dot)
1333         return Error("enum values need to be qualified by an enum type");
1334       std::string enum_def_str(word.c_str(), dot);
1335       std::string enum_val_str(dot + 1, word.c_str() + word.length());
1336       auto enum_def = LookupEnum(enum_def_str);
1337       if (!enum_def) return Error("unknown enum: " + enum_def_str);
1338       auto enum_val = enum_def->vals.Lookup(enum_val_str);
1339       if (!enum_val) return Error("unknown enum value: " + enum_val_str);
1340       *result |= enum_val->value;
1341     }
1342   } while (*next);
1343   return NoError();
1344 }
1345 
ParseHash(Value & e,FieldDef * field)1346 CheckedError Parser::ParseHash(Value &e, FieldDef *field) {
1347   FLATBUFFERS_ASSERT(field);
1348   Value *hash_name = field->attributes.Lookup("hash");
1349   switch (e.type.base_type) {
1350     case BASE_TYPE_SHORT: {
1351       auto hash = FindHashFunction16(hash_name->constant.c_str());
1352       int16_t hashed_value = static_cast<int16_t>(hash(attribute_.c_str()));
1353       e.constant = NumToString(hashed_value);
1354       break;
1355     }
1356     case BASE_TYPE_USHORT: {
1357       auto hash = FindHashFunction16(hash_name->constant.c_str());
1358       uint16_t hashed_value = hash(attribute_.c_str());
1359       e.constant = NumToString(hashed_value);
1360       break;
1361     }
1362     case BASE_TYPE_INT: {
1363       auto hash = FindHashFunction32(hash_name->constant.c_str());
1364       int32_t hashed_value = static_cast<int32_t>(hash(attribute_.c_str()));
1365       e.constant = NumToString(hashed_value);
1366       break;
1367     }
1368     case BASE_TYPE_UINT: {
1369       auto hash = FindHashFunction32(hash_name->constant.c_str());
1370       uint32_t hashed_value = hash(attribute_.c_str());
1371       e.constant = NumToString(hashed_value);
1372       break;
1373     }
1374     case BASE_TYPE_LONG: {
1375       auto hash = FindHashFunction64(hash_name->constant.c_str());
1376       int64_t hashed_value = static_cast<int64_t>(hash(attribute_.c_str()));
1377       e.constant = NumToString(hashed_value);
1378       break;
1379     }
1380     case BASE_TYPE_ULONG: {
1381       auto hash = FindHashFunction64(hash_name->constant.c_str());
1382       uint64_t hashed_value = hash(attribute_.c_str());
1383       e.constant = NumToString(hashed_value);
1384       break;
1385     }
1386     default: FLATBUFFERS_ASSERT(0);
1387   }
1388   NEXT();
1389   return NoError();
1390 }
1391 
TokenError()1392 CheckedError Parser::TokenError() {
1393   return Error("cannot parse value starting with: " + TokenToStringId(token_));
1394 }
1395 
ParseSingleValue(const std::string * name,Value & e,bool check_now)1396 CheckedError Parser::ParseSingleValue(const std::string *name, Value &e,
1397                                       bool check_now) {
1398   // First see if this could be a conversion function:
1399   if (token_ == kTokenIdentifier && *cursor_ == '(') {
1400     // todo: Extract processing of conversion functions to ParseFunction.
1401     const auto functionname = attribute_;
1402     if (!IsFloat(e.type.base_type)) {
1403       return Error(functionname + ": type of argument mismatch, expecting: " +
1404                    kTypeNames[BASE_TYPE_DOUBLE] +
1405                    ", found: " + kTypeNames[e.type.base_type] +
1406                    ", name: " + (name ? *name : "") + ", value: " + e.constant);
1407     }
1408     NEXT();
1409     EXPECT('(');
1410     ECHECK(Recurse([&]() { return ParseSingleValue(name, e, false); }));
1411     EXPECT(')');
1412     // calculate with double precision
1413     double x, y = 0.0;
1414     ECHECK(atot(e.constant.c_str(), *this, &x));
1415     auto func_match = false;
1416     // clang-format off
1417     #define FLATBUFFERS_FN_DOUBLE(name, op) \
1418       if (!func_match && functionname == name) { y = op; func_match = true; }
1419     FLATBUFFERS_FN_DOUBLE("deg", x / kPi * 180);
1420     FLATBUFFERS_FN_DOUBLE("rad", x * kPi / 180);
1421     FLATBUFFERS_FN_DOUBLE("sin", sin(x));
1422     FLATBUFFERS_FN_DOUBLE("cos", cos(x));
1423     FLATBUFFERS_FN_DOUBLE("tan", tan(x));
1424     FLATBUFFERS_FN_DOUBLE("asin", asin(x));
1425     FLATBUFFERS_FN_DOUBLE("acos", acos(x));
1426     FLATBUFFERS_FN_DOUBLE("atan", atan(x));
1427     // TODO(wvo): add more useful conversion functions here.
1428     #undef FLATBUFFERS_FN_DOUBLE
1429     // clang-format on
1430     if (true != func_match) {
1431       return Error(std::string("Unknown conversion function: ") + functionname +
1432                    ", field name: " + (name ? *name : "") +
1433                    ", value: " + e.constant);
1434     }
1435     e.constant = NumToString(y);
1436     return NoError();
1437   }
1438 
1439   auto match = false;
1440   // clang-format off
1441   #define TRY_ECHECK(force, dtoken, check, req)    \
1442     if (!match && ((check) || IsConstTrue(force))) \
1443     ECHECK(TryTypedValue(name, dtoken, check, e, req, &match))
1444   // clang-format on
1445 
1446   if (token_ == kTokenStringConstant || token_ == kTokenIdentifier) {
1447     const auto kTokenStringOrIdent = token_;
1448     // The string type is a most probable type, check it first.
1449     TRY_ECHECK(false, kTokenStringConstant,
1450                e.type.base_type == BASE_TYPE_STRING, BASE_TYPE_STRING);
1451 
1452     // avoid escaped and non-ascii in the string
1453     if ((token_ == kTokenStringConstant) && IsScalar(e.type.base_type) &&
1454         !attr_is_trivial_ascii_string_) {
1455       return Error(
1456           std::string("type mismatch or invalid value, an initializer of "
1457                       "non-string field must be trivial ASCII string: type: ") +
1458           kTypeNames[e.type.base_type] + ", name: " + (name ? *name : "") +
1459           ", value: " + attribute_);
1460     }
1461 
1462     // A boolean as true/false. Boolean as Integer check below.
1463     if (!match && IsBool(e.type.base_type)) {
1464       auto is_true = attribute_ == "true";
1465       if (is_true || attribute_ == "false") {
1466         attribute_ = is_true ? "1" : "0";
1467         // accepts both kTokenStringConstant and kTokenIdentifier
1468         TRY_ECHECK(false, kTokenStringOrIdent, IsBool(e.type.base_type),
1469                    BASE_TYPE_BOOL);
1470       }
1471     }
1472     // Check if this could be a string/identifier enum value.
1473     // Enum can have only true integer base type.
1474     if (!match && IsInteger(e.type.base_type) && !IsBool(e.type.base_type) &&
1475         IsIdentifierStart(*attribute_.c_str())) {
1476       int64_t val;
1477       ECHECK(ParseEnumFromString(e.type, &val));
1478       e.constant = NumToString(val);
1479       NEXT();
1480       match = true;
1481     }
1482     // float/integer number in string
1483     if ((token_ == kTokenStringConstant) && IsScalar(e.type.base_type)) {
1484       // remove trailing whitespaces from attribute_
1485       auto last = attribute_.find_last_not_of(' ');
1486       if (std::string::npos != last)  // has non-whitespace
1487         attribute_.resize(last + 1);
1488     }
1489     // Float numbers or nan, inf, pi, etc.
1490     TRY_ECHECK(false, kTokenStringOrIdent, IsFloat(e.type.base_type),
1491                BASE_TYPE_FLOAT);
1492     // An integer constant in string.
1493     TRY_ECHECK(false, kTokenStringOrIdent, IsInteger(e.type.base_type),
1494                BASE_TYPE_INT);
1495     // Unknown tokens will be interpreted as string type.
1496     TRY_ECHECK(true, kTokenStringConstant, e.type.base_type == BASE_TYPE_STRING,
1497                BASE_TYPE_STRING);
1498   } else {
1499     // Try a float number.
1500     TRY_ECHECK(false, kTokenFloatConstant, IsFloat(e.type.base_type),
1501                BASE_TYPE_FLOAT);
1502     // Integer token can init any scalar (integer of float).
1503     TRY_ECHECK(true, kTokenIntegerConstant, IsScalar(e.type.base_type),
1504                BASE_TYPE_INT);
1505   }
1506   #undef TRY_ECHECK
1507 
1508   if (!match) return TokenError();
1509 
1510   // The check_now flag must be true when parse a fbs-schema.
1511   // This flag forces to check default scalar values or metadata of field.
1512   // For JSON parser the flag should be false.
1513   // If it is set for JSON each value will be checked twice (see ParseTable).
1514   if (check_now && IsScalar(e.type.base_type)) {
1515     // "re-pack" an integer scalar to remove any ambiguities like leading zeros
1516     // which can be treated as octal-literal (idl_gen_cpp/GenDefaultConstant).
1517     const auto repack = IsInteger(e.type.base_type);
1518     switch (e.type.base_type) {
1519     // clang-format off
1520     #define FLATBUFFERS_TD(ENUM, IDLTYPE, \
1521             CTYPE, JTYPE, GTYPE, NTYPE, PTYPE, RTYPE) \
1522             case BASE_TYPE_ ## ENUM: {\
1523                 CTYPE val; \
1524                 ECHECK(atot(e.constant.c_str(), *this, &val)); \
1525                 if(repack) e.constant = NumToString(val); \
1526               break; }
1527     FLATBUFFERS_GEN_TYPES_SCALAR(FLATBUFFERS_TD);
1528     #undef FLATBUFFERS_TD
1529     default: break;
1530     // clang-format on
1531     }
1532   }
1533   return NoError();
1534 }
1535 
LookupCreateStruct(const std::string & name,bool create_if_new,bool definition)1536 StructDef *Parser::LookupCreateStruct(const std::string &name,
1537                                       bool create_if_new, bool definition) {
1538   std::string qualified_name = current_namespace_->GetFullyQualifiedName(name);
1539   // See if it exists pre-declared by an unqualified use.
1540   auto struct_def = LookupStruct(name);
1541   if (struct_def && struct_def->predecl) {
1542     if (definition) {
1543       // Make sure it has the current namespace, and is registered under its
1544       // qualified name.
1545       struct_def->defined_namespace = current_namespace_;
1546       structs_.Move(name, qualified_name);
1547     }
1548     return struct_def;
1549   }
1550   // See if it exists pre-declared by an qualified use.
1551   struct_def = LookupStruct(qualified_name);
1552   if (struct_def && struct_def->predecl) {
1553     if (definition) {
1554       // Make sure it has the current namespace.
1555       struct_def->defined_namespace = current_namespace_;
1556     }
1557     return struct_def;
1558   }
1559   if (!definition) {
1560     // Search thru parent namespaces.
1561     for (size_t components = current_namespace_->components.size();
1562          components && !struct_def; components--) {
1563       struct_def = LookupStruct(
1564           current_namespace_->GetFullyQualifiedName(name, components - 1));
1565     }
1566   }
1567   if (!struct_def && create_if_new) {
1568     struct_def = new StructDef();
1569     if (definition) {
1570       structs_.Add(qualified_name, struct_def);
1571       struct_def->name = name;
1572       struct_def->defined_namespace = current_namespace_;
1573     } else {
1574       // Not a definition.
1575       // Rather than failing, we create a "pre declared" StructDef, due to
1576       // circular references, and check for errors at the end of parsing.
1577       // It is defined in the current namespace, as the best guess what the
1578       // final namespace will be.
1579       structs_.Add(name, struct_def);
1580       struct_def->name = name;
1581       struct_def->defined_namespace = current_namespace_;
1582       struct_def->original_location.reset(
1583           new std::string(file_being_parsed_ + ":" + NumToString(line_)));
1584     }
1585   }
1586   return struct_def;
1587 }
1588 
ParseEnum(bool is_union,EnumDef ** dest)1589 CheckedError Parser::ParseEnum(bool is_union, EnumDef **dest) {
1590   std::vector<std::string> enum_comment = doc_comment_;
1591   NEXT();
1592   std::string enum_name = attribute_;
1593   EXPECT(kTokenIdentifier);
1594   EnumDef *enum_def;
1595   ECHECK(StartEnum(enum_name, is_union, &enum_def));
1596   enum_def->doc_comment = enum_comment;
1597   if (!is_union && !opts.proto_mode) {
1598     // Give specialized error message, since this type spec used to
1599     // be optional in the first FlatBuffers release.
1600     if (!Is(':')) {
1601       return Error(
1602           "must specify the underlying integer type for this"
1603           " enum (e.g. \': short\', which was the default).");
1604     } else {
1605       NEXT();
1606     }
1607     // Specify the integer type underlying this enum.
1608     ECHECK(ParseType(enum_def->underlying_type));
1609     if (!IsInteger(enum_def->underlying_type.base_type) ||
1610         IsBool(enum_def->underlying_type.base_type))
1611       return Error("underlying enum type must be integral");
1612     // Make this type refer back to the enum it was derived from.
1613     enum_def->underlying_type.enum_def = enum_def;
1614   }
1615   ECHECK(ParseMetaData(&enum_def->attributes));
1616   EXPECT('{');
1617   if (is_union) enum_def->vals.Add("NONE", new EnumVal("NONE", 0));
1618   std::set<std::pair<BaseType, StructDef*>> union_types;
1619   for (;;) {
1620     if (opts.proto_mode && attribute_ == "option") {
1621       ECHECK(ParseProtoOption());
1622     } else {
1623       auto value_name = attribute_;
1624       auto full_name = value_name;
1625       std::vector<std::string> value_comment = doc_comment_;
1626       EXPECT(kTokenIdentifier);
1627       if (is_union) {
1628         ECHECK(ParseNamespacing(&full_name, &value_name));
1629         if (opts.union_value_namespacing) {
1630           // Since we can't namespace the actual enum identifiers, turn
1631           // namespace parts into part of the identifier.
1632           value_name = full_name;
1633           std::replace(value_name.begin(), value_name.end(), '.', '_');
1634         }
1635       }
1636       auto prevsize = enum_def->vals.vec.size();
1637       auto prevvalue = prevsize > 0 ? enum_def->vals.vec.back()->value : 0;
1638       auto &ev = *new EnumVal(value_name, 0);
1639       if (enum_def->vals.Add(value_name, &ev))
1640         return Error("enum value already exists: " + value_name);
1641       ev.doc_comment = value_comment;
1642       if (is_union) {
1643         if (Is(':')) {
1644           NEXT();
1645           ECHECK(ParseType(ev.union_type));
1646           if (ev.union_type.base_type != BASE_TYPE_STRUCT &&
1647               ev.union_type.base_type != BASE_TYPE_STRING)
1648             return Error("union value type may only be table/struct/string");
1649         } else {
1650           ev.union_type = Type(BASE_TYPE_STRUCT, LookupCreateStruct(full_name));
1651         }
1652         if (!enum_def->uses_multiple_type_instances) {
1653           auto union_type_key = std::make_pair(ev.union_type.base_type, ev.union_type.struct_def);
1654           if (union_types.count(union_type_key) > 0) {
1655             enum_def->uses_multiple_type_instances = true;
1656           } else {
1657             union_types.insert(union_type_key);
1658           }
1659         }
1660       }
1661       if (Is('=')) {
1662         NEXT();
1663         ECHECK(atot(attribute_.c_str(), *this, &ev.value));
1664         EXPECT(kTokenIntegerConstant);
1665         if (!opts.proto_mode && prevsize &&
1666             enum_def->vals.vec[prevsize - 1]->value >= ev.value)
1667           return Error("enum values must be specified in ascending order");
1668       } else if (prevsize == 0) {
1669         // already set to zero
1670       } else if (prevvalue != flatbuffers::numeric_limits<int64_t>::max()) {
1671         ev.value = prevvalue + 1;
1672       } else {
1673         return Error("enum value overflows");
1674       }
1675 
1676       // Check that value fits into the underlying type.
1677       switch (enum_def->underlying_type.base_type) {
1678         // clang-format off
1679         #define FLATBUFFERS_TD(ENUM, IDLTYPE, CTYPE, JTYPE, GTYPE, NTYPE, \
1680                                PTYPE, RTYPE)                              \
1681           case BASE_TYPE_##ENUM: {                                        \
1682             int64_t min_value = static_cast<int64_t>(                     \
1683               flatbuffers::numeric_limits<CTYPE>::lowest());              \
1684             int64_t max_value = static_cast<int64_t>(                     \
1685               flatbuffers::numeric_limits<CTYPE>::max());                 \
1686             if (ev.value < min_value || ev.value > max_value) {           \
1687               return Error(                                               \
1688                 "enum value does not fit [" +  NumToString(min_value) +   \
1689                 "; " + NumToString(max_value) + "]");                     \
1690             }                                                             \
1691             break;                                                        \
1692           }
1693         FLATBUFFERS_GEN_TYPES_SCALAR(FLATBUFFERS_TD);
1694         #undef FLATBUFFERS_TD
1695         default: break;
1696         // clang-format on
1697       }
1698 
1699       if (opts.proto_mode && Is('[')) {
1700         NEXT();
1701         // ignore attributes on enums.
1702         while (token_ != ']') NEXT();
1703         NEXT();
1704       }
1705     }
1706     if (!Is(opts.proto_mode ? ';' : ',')) break;
1707     NEXT();
1708     if (Is('}')) break;
1709   }
1710   EXPECT('}');
1711   if (enum_def->attributes.Lookup("bit_flags")) {
1712     for (auto it = enum_def->vals.vec.begin(); it != enum_def->vals.vec.end();
1713          ++it) {
1714       if (static_cast<size_t>((*it)->value) >=
1715           SizeOf(enum_def->underlying_type.base_type) * 8)
1716         return Error("bit flag out of range of underlying integral type");
1717       (*it)->value = 1LL << (*it)->value;
1718     }
1719   }
1720   if (dest) *dest = enum_def;
1721   types_.Add(current_namespace_->GetFullyQualifiedName(enum_def->name),
1722              new Type(BASE_TYPE_UNION, nullptr, enum_def));
1723   return NoError();
1724 }
1725 
StartStruct(const std::string & name,StructDef ** dest)1726 CheckedError Parser::StartStruct(const std::string &name, StructDef **dest) {
1727   auto &struct_def = *LookupCreateStruct(name, true, true);
1728   if (!struct_def.predecl) return Error("datatype already exists: " + name);
1729   struct_def.predecl = false;
1730   struct_def.name = name;
1731   struct_def.file = file_being_parsed_;
1732   // Move this struct to the back of the vector just in case it was predeclared,
1733   // to preserve declaration order.
1734   *std::remove(structs_.vec.begin(), structs_.vec.end(), &struct_def) =
1735       &struct_def;
1736   *dest = &struct_def;
1737   return NoError();
1738 }
1739 
CheckClash(std::vector<FieldDef * > & fields,StructDef * struct_def,const char * suffix,BaseType basetype)1740 CheckedError Parser::CheckClash(std::vector<FieldDef *> &fields,
1741                                 StructDef *struct_def, const char *suffix,
1742                                 BaseType basetype) {
1743   auto len = strlen(suffix);
1744   for (auto it = fields.begin(); it != fields.end(); ++it) {
1745     auto &fname = (*it)->name;
1746     if (fname.length() > len &&
1747         fname.compare(fname.length() - len, len, suffix) == 0 &&
1748         (*it)->value.type.base_type != BASE_TYPE_UTYPE) {
1749       auto field =
1750           struct_def->fields.Lookup(fname.substr(0, fname.length() - len));
1751       if (field && field->value.type.base_type == basetype)
1752         return Error("Field " + fname +
1753                      " would clash with generated functions for field " +
1754                      field->name);
1755     }
1756   }
1757   return NoError();
1758 }
1759 
SupportsVectorOfUnions() const1760 bool Parser::SupportsVectorOfUnions() const {
1761   return opts.lang_to_generate != 0 &&
1762          (opts.lang_to_generate & ~(IDLOptions::kCpp | IDLOptions::kJs |
1763                                     IDLOptions::kTs | IDLOptions::kPhp |
1764                                     IDLOptions::kJava | IDLOptions::kCSharp)) == 0;
1765 }
1766 
UniqueNamespace(Namespace * ns)1767 Namespace *Parser::UniqueNamespace(Namespace *ns) {
1768   for (auto it = namespaces_.begin(); it != namespaces_.end(); ++it) {
1769     if (ns->components == (*it)->components) {
1770       delete ns;
1771       return *it;
1772     }
1773   }
1774   namespaces_.push_back(ns);
1775   return ns;
1776 }
1777 
UnqualifiedName(std::string full_qualified_name)1778 std::string Parser::UnqualifiedName(std::string full_qualified_name) {
1779   Namespace *ns = new Namespace();
1780 
1781   std::size_t current, previous = 0;
1782   current = full_qualified_name.find('.');
1783   while (current != std::string::npos) {
1784     ns->components.push_back(
1785         full_qualified_name.substr(previous, current - previous));
1786     previous = current + 1;
1787     current = full_qualified_name.find('.', previous);
1788   }
1789   current_namespace_ = UniqueNamespace(ns);
1790   return full_qualified_name.substr(previous, current - previous);
1791 }
1792 
compareFieldDefs(const FieldDef * a,const FieldDef * b)1793 static bool compareFieldDefs(const FieldDef *a, const FieldDef *b) {
1794   auto a_id = atoi(a->attributes.Lookup("id")->constant.c_str());
1795   auto b_id = atoi(b->attributes.Lookup("id")->constant.c_str());
1796   return a_id < b_id;
1797 }
1798 
ParseDecl()1799 CheckedError Parser::ParseDecl() {
1800   std::vector<std::string> dc = doc_comment_;
1801   bool fixed = IsIdent("struct");
1802   if (!fixed && !IsIdent("table")) return Error("declaration expected");
1803   NEXT();
1804   std::string name = attribute_;
1805   EXPECT(kTokenIdentifier);
1806   StructDef *struct_def;
1807   ECHECK(StartStruct(name, &struct_def));
1808   struct_def->doc_comment = dc;
1809   struct_def->fixed = fixed;
1810   ECHECK(ParseMetaData(&struct_def->attributes));
1811   struct_def->sortbysize =
1812       struct_def->attributes.Lookup("original_order") == nullptr && !fixed;
1813   EXPECT('{');
1814   while (token_ != '}') ECHECK(ParseField(*struct_def));
1815   auto force_align = struct_def->attributes.Lookup("force_align");
1816   if (fixed) {
1817     if (force_align) {
1818       auto align = static_cast<size_t>(atoi(force_align->constant.c_str()));
1819       if (force_align->type.base_type != BASE_TYPE_INT ||
1820           align < struct_def->minalign || align > FLATBUFFERS_MAX_ALIGNMENT ||
1821           align & (align - 1))
1822         return Error(
1823             "force_align must be a power of two integer ranging from the"
1824             "struct\'s natural alignment to " +
1825             NumToString(FLATBUFFERS_MAX_ALIGNMENT));
1826       struct_def->minalign = align;
1827     }
1828     if (!struct_def->bytesize) return Error("size 0 structs not allowed");
1829   }
1830   struct_def->PadLastField(struct_def->minalign);
1831   // Check if this is a table that has manual id assignments
1832   auto &fields = struct_def->fields.vec;
1833   if (!fixed && fields.size()) {
1834     size_t num_id_fields = 0;
1835     for (auto it = fields.begin(); it != fields.end(); ++it) {
1836       if ((*it)->attributes.Lookup("id")) num_id_fields++;
1837     }
1838     // If any fields have ids..
1839     if (num_id_fields) {
1840       // Then all fields must have them.
1841       if (num_id_fields != fields.size())
1842         return Error(
1843             "either all fields or no fields must have an 'id' attribute");
1844       // Simply sort by id, then the fields are the same as if no ids had
1845       // been specified.
1846       std::sort(fields.begin(), fields.end(), compareFieldDefs);
1847       // Verify we have a contiguous set, and reassign vtable offsets.
1848       for (int i = 0; i < static_cast<int>(fields.size()); i++) {
1849         if (i != atoi(fields[i]->attributes.Lookup("id")->constant.c_str()))
1850           return Error("field id\'s must be consecutive from 0, id " +
1851                        NumToString(i) + " missing or set twice");
1852         fields[i]->value.offset = FieldIndexToOffset(static_cast<voffset_t>(i));
1853       }
1854     }
1855   }
1856 
1857   ECHECK(
1858       CheckClash(fields, struct_def, UnionTypeFieldSuffix(), BASE_TYPE_UNION));
1859   ECHECK(CheckClash(fields, struct_def, "Type", BASE_TYPE_UNION));
1860   ECHECK(CheckClash(fields, struct_def, "_length", BASE_TYPE_VECTOR));
1861   ECHECK(CheckClash(fields, struct_def, "Length", BASE_TYPE_VECTOR));
1862   ECHECK(CheckClash(fields, struct_def, "_byte_vector", BASE_TYPE_STRING));
1863   ECHECK(CheckClash(fields, struct_def, "ByteVector", BASE_TYPE_STRING));
1864   EXPECT('}');
1865   types_.Add(current_namespace_->GetFullyQualifiedName(struct_def->name),
1866              new Type(BASE_TYPE_STRUCT, struct_def, nullptr));
1867   return NoError();
1868 }
1869 
ParseService()1870 CheckedError Parser::ParseService() {
1871   std::vector<std::string> service_comment = doc_comment_;
1872   NEXT();
1873   auto service_name = attribute_;
1874   EXPECT(kTokenIdentifier);
1875   auto &service_def = *new ServiceDef();
1876   service_def.name = service_name;
1877   service_def.file = file_being_parsed_;
1878   service_def.doc_comment = service_comment;
1879   service_def.defined_namespace = current_namespace_;
1880   if (services_.Add(current_namespace_->GetFullyQualifiedName(service_name),
1881                     &service_def))
1882     return Error("service already exists: " + service_name);
1883   ECHECK(ParseMetaData(&service_def.attributes));
1884   EXPECT('{');
1885   do {
1886     std::vector<std::string> doc_comment = doc_comment_;
1887     auto rpc_name = attribute_;
1888     EXPECT(kTokenIdentifier);
1889     EXPECT('(');
1890     Type reqtype, resptype;
1891     ECHECK(ParseTypeIdent(reqtype));
1892     EXPECT(')');
1893     EXPECT(':');
1894     ECHECK(ParseTypeIdent(resptype));
1895     if (reqtype.base_type != BASE_TYPE_STRUCT || reqtype.struct_def->fixed ||
1896         resptype.base_type != BASE_TYPE_STRUCT || resptype.struct_def->fixed)
1897       return Error("rpc request and response types must be tables");
1898     auto &rpc = *new RPCCall();
1899     rpc.name = rpc_name;
1900     rpc.request = reqtype.struct_def;
1901     rpc.response = resptype.struct_def;
1902     rpc.doc_comment = doc_comment;
1903     if (service_def.calls.Add(rpc_name, &rpc))
1904       return Error("rpc already exists: " + rpc_name);
1905     ECHECK(ParseMetaData(&rpc.attributes));
1906     EXPECT(';');
1907   } while (token_ != '}');
1908   NEXT();
1909   return NoError();
1910 }
1911 
SetRootType(const char * name)1912 bool Parser::SetRootType(const char *name) {
1913   root_struct_def_ = LookupStruct(name);
1914   if (!root_struct_def_)
1915     root_struct_def_ =
1916         LookupStruct(current_namespace_->GetFullyQualifiedName(name));
1917   return root_struct_def_ != nullptr;
1918 }
1919 
MarkGenerated()1920 void Parser::MarkGenerated() {
1921   // This function marks all existing definitions as having already
1922   // been generated, which signals no code for included files should be
1923   // generated.
1924   for (auto it = enums_.vec.begin(); it != enums_.vec.end(); ++it) {
1925     (*it)->generated = true;
1926   }
1927   for (auto it = structs_.vec.begin(); it != structs_.vec.end(); ++it) {
1928     if (!(*it)->predecl) { (*it)->generated = true; }
1929   }
1930   for (auto it = services_.vec.begin(); it != services_.vec.end(); ++it) {
1931     (*it)->generated = true;
1932   }
1933 }
1934 
ParseNamespace()1935 CheckedError Parser::ParseNamespace() {
1936   NEXT();
1937   auto ns = new Namespace();
1938   namespaces_.push_back(ns);  // Store it here to not leak upon error.
1939   if (token_ != ';') {
1940     for (;;) {
1941       ns->components.push_back(attribute_);
1942       EXPECT(kTokenIdentifier);
1943       if (Is('.')) NEXT() else break;
1944     }
1945   }
1946   namespaces_.pop_back();
1947   current_namespace_ = UniqueNamespace(ns);
1948   EXPECT(';');
1949   return NoError();
1950 }
1951 
compareEnumVals(const EnumVal * a,const EnumVal * b)1952 static bool compareEnumVals(const EnumVal *a, const EnumVal *b) {
1953   return a->value < b->value;
1954 }
1955 
1956 // Best effort parsing of .proto declarations, with the aim to turn them
1957 // in the closest corresponding FlatBuffer equivalent.
1958 // We parse everything as identifiers instead of keywords, since we don't
1959 // want protobuf keywords to become invalid identifiers in FlatBuffers.
ParseProtoDecl()1960 CheckedError Parser::ParseProtoDecl() {
1961   bool isextend = IsIdent("extend");
1962   if (IsIdent("package")) {
1963     // These are identical in syntax to FlatBuffer's namespace decl.
1964     ECHECK(ParseNamespace());
1965   } else if (IsIdent("message") || isextend) {
1966     std::vector<std::string> struct_comment = doc_comment_;
1967     NEXT();
1968     StructDef *struct_def = nullptr;
1969     Namespace *parent_namespace = nullptr;
1970     if (isextend) {
1971       if (Is('.')) NEXT();  // qualified names may start with a . ?
1972       auto id = attribute_;
1973       EXPECT(kTokenIdentifier);
1974       ECHECK(ParseNamespacing(&id, nullptr));
1975       struct_def = LookupCreateStruct(id, false);
1976       if (!struct_def)
1977         return Error("cannot extend unknown message type: " + id);
1978     } else {
1979       std::string name = attribute_;
1980       EXPECT(kTokenIdentifier);
1981       ECHECK(StartStruct(name, &struct_def));
1982       // Since message definitions can be nested, we create a new namespace.
1983       auto ns = new Namespace();
1984       // Copy of current namespace.
1985       *ns = *current_namespace_;
1986       // But with current message name.
1987       ns->components.push_back(name);
1988       ns->from_table++;
1989       parent_namespace = current_namespace_;
1990       current_namespace_ = UniqueNamespace(ns);
1991     }
1992     struct_def->doc_comment = struct_comment;
1993     ECHECK(ParseProtoFields(struct_def, isextend, false));
1994     if (!isextend) { current_namespace_ = parent_namespace; }
1995     if (Is(';')) NEXT();
1996   } else if (IsIdent("enum")) {
1997     // These are almost the same, just with different terminator:
1998     EnumDef *enum_def;
1999     ECHECK(ParseEnum(false, &enum_def));
2000     if (Is(';')) NEXT();
2001     // Protobuf allows them to be specified in any order, so sort afterwards.
2002     auto &v = enum_def->vals.vec;
2003     std::sort(v.begin(), v.end(), compareEnumVals);
2004 
2005     // Temp: remove any duplicates, as .fbs files can't handle them.
2006     for (auto it = v.begin(); it != v.end();) {
2007       if (it != v.begin() && it[0]->value == it[-1]->value)
2008         it = v.erase(it);
2009       else
2010         ++it;
2011     }
2012   } else if (IsIdent("syntax")) {  // Skip these.
2013     NEXT();
2014     EXPECT('=');
2015     EXPECT(kTokenStringConstant);
2016     EXPECT(';');
2017   } else if (IsIdent("option")) {  // Skip these.
2018     ECHECK(ParseProtoOption());
2019     EXPECT(';');
2020   } else if (IsIdent("service")) {  // Skip these.
2021     NEXT();
2022     EXPECT(kTokenIdentifier);
2023     ECHECK(ParseProtoCurliesOrIdent());
2024   } else {
2025     return Error("don\'t know how to parse .proto declaration starting with " +
2026                  TokenToStringId(token_));
2027   }
2028   return NoError();
2029 }
2030 
StartEnum(const std::string & enum_name,bool is_union,EnumDef ** dest)2031 CheckedError Parser::StartEnum(const std::string &enum_name, bool is_union,
2032                                EnumDef **dest) {
2033   auto &enum_def = *new EnumDef();
2034   enum_def.name = enum_name;
2035   enum_def.file = file_being_parsed_;
2036   enum_def.doc_comment = doc_comment_;
2037   enum_def.is_union = is_union;
2038   enum_def.defined_namespace = current_namespace_;
2039   if (enums_.Add(current_namespace_->GetFullyQualifiedName(enum_name),
2040                  &enum_def))
2041     return Error("enum already exists: " + enum_name);
2042   enum_def.underlying_type.base_type = is_union ? BASE_TYPE_UTYPE
2043                                                 : BASE_TYPE_INT;
2044   enum_def.underlying_type.enum_def = &enum_def;
2045   if (dest) *dest = &enum_def;
2046   return NoError();
2047 }
2048 
ParseProtoFields(StructDef * struct_def,bool isextend,bool inside_oneof)2049 CheckedError Parser::ParseProtoFields(StructDef *struct_def, bool isextend,
2050                                       bool inside_oneof) {
2051   EXPECT('{');
2052   while (token_ != '}') {
2053     if (IsIdent("message") || IsIdent("extend") || IsIdent("enum")) {
2054       // Nested declarations.
2055       ECHECK(ParseProtoDecl());
2056     } else if (IsIdent("extensions")) {  // Skip these.
2057       NEXT();
2058       EXPECT(kTokenIntegerConstant);
2059       if (Is(kTokenIdentifier)) {
2060         NEXT();  // to
2061         NEXT();  // num
2062       }
2063       EXPECT(';');
2064     } else if (IsIdent("option")) {  // Skip these.
2065       ECHECK(ParseProtoOption());
2066       EXPECT(';');
2067     } else if (IsIdent("reserved")) {  // Skip these.
2068       NEXT();
2069       while (!Is(';')) { NEXT(); }  // A variety of formats, just skip.
2070       NEXT();
2071     } else {
2072       std::vector<std::string> field_comment = doc_comment_;
2073       // Parse the qualifier.
2074       bool required = false;
2075       bool repeated = false;
2076       bool oneof = false;
2077       if (!inside_oneof) {
2078         if (IsIdent("optional")) {
2079           // This is the default.
2080           NEXT();
2081         } else if (IsIdent("required")) {
2082           required = true;
2083           NEXT();
2084         } else if (IsIdent("repeated")) {
2085           repeated = true;
2086           NEXT();
2087         } else if (IsIdent("oneof")) {
2088           oneof = true;
2089           NEXT();
2090         } else {
2091           // can't error, proto3 allows decls without any of the above.
2092         }
2093       }
2094       StructDef *anonymous_struct = nullptr;
2095       EnumDef *oneof_union = nullptr;
2096       Type type;
2097       if (IsIdent("group") || oneof) {
2098         if (!oneof) NEXT();
2099         if (oneof && opts.proto_oneof_union) {
2100           auto name = MakeCamel(attribute_, true) + "Union";
2101           ECHECK(StartEnum(name, true, &oneof_union));
2102           type = Type(BASE_TYPE_UNION, nullptr, oneof_union);
2103         } else {
2104           auto name = "Anonymous" + NumToString(anonymous_counter++);
2105           ECHECK(StartStruct(name, &anonymous_struct));
2106           type = Type(BASE_TYPE_STRUCT, anonymous_struct);
2107         }
2108       } else {
2109         ECHECK(ParseTypeFromProtoType(&type));
2110       }
2111       // Repeated elements get mapped to a vector.
2112       if (repeated) {
2113         type.element = type.base_type;
2114         type.base_type = BASE_TYPE_VECTOR;
2115         if (type.element == BASE_TYPE_VECTOR) {
2116           // We have a vector or vectors, which FlatBuffers doesn't support.
2117           // For now make it a vector of string (since the source is likely
2118           // "repeated bytes").
2119           // TODO(wvo): A better solution would be to wrap this in a table.
2120           type.element = BASE_TYPE_STRING;
2121         }
2122       }
2123       std::string name = attribute_;
2124       EXPECT(kTokenIdentifier);
2125       if (!oneof) {
2126         // Parse the field id. Since we're just translating schemas, not
2127         // any kind of binary compatibility, we can safely ignore these, and
2128         // assign our own.
2129         EXPECT('=');
2130         EXPECT(kTokenIntegerConstant);
2131       }
2132       FieldDef *field = nullptr;
2133       if (isextend) {
2134         // We allow a field to be re-defined when extending.
2135         // TODO: are there situations where that is problematic?
2136         field = struct_def->fields.Lookup(name);
2137       }
2138       if (!field) ECHECK(AddField(*struct_def, name, type, &field));
2139       field->doc_comment = field_comment;
2140       if (!IsScalar(type.base_type)) field->required = required;
2141       // See if there's a default specified.
2142       if (Is('[')) {
2143         NEXT();
2144         for (;;) {
2145           auto key = attribute_;
2146           ECHECK(ParseProtoKey());
2147           EXPECT('=');
2148           auto val = attribute_;
2149           ECHECK(ParseProtoCurliesOrIdent());
2150           if (key == "default") {
2151             // Temp: skip non-numeric defaults (enums).
2152             auto numeric = strpbrk(val.c_str(), "0123456789-+.");
2153             if (IsScalar(type.base_type) && numeric == val.c_str())
2154               field->value.constant = val;
2155           } else if (key == "deprecated") {
2156             field->deprecated = val == "true";
2157           }
2158           if (!Is(',')) break;
2159           NEXT();
2160         }
2161         EXPECT(']');
2162       }
2163       if (anonymous_struct) {
2164         ECHECK(ParseProtoFields(anonymous_struct, false, oneof));
2165         if (Is(';')) NEXT();
2166       } else if (oneof_union) {
2167         // Parse into a temporary StructDef, then transfer fields into an
2168         // EnumDef describing the oneof as a union.
2169         StructDef oneof_struct;
2170         ECHECK(ParseProtoFields(&oneof_struct, false, oneof));
2171         if (Is(';')) NEXT();
2172         for (auto field_it = oneof_struct.fields.vec.begin();
2173              field_it != oneof_struct.fields.vec.end(); ++field_it) {
2174           const auto &oneof_field = **field_it;
2175           const auto &oneof_type = oneof_field.value.type;
2176           if (oneof_type.base_type != BASE_TYPE_STRUCT ||
2177               !oneof_type.struct_def || oneof_type.struct_def->fixed)
2178             return Error("oneof '" + name +
2179                 "' cannot be mapped to a union because member '" +
2180                 oneof_field.name + "' is not a table type.");
2181           auto enum_val = new EnumVal(oneof_type.struct_def->name,
2182                                       oneof_union->vals.vec.size());
2183           enum_val->union_type = oneof_type;
2184           enum_val->doc_comment = oneof_field.doc_comment;
2185           oneof_union->vals.Add(oneof_field.name, enum_val);
2186         }
2187       } else {
2188         EXPECT(';');
2189       }
2190     }
2191   }
2192   NEXT();
2193   return NoError();
2194 }
2195 
ParseProtoKey()2196 CheckedError Parser::ParseProtoKey() {
2197   if (token_ == '(') {
2198     NEXT();
2199     // Skip "(a.b)" style custom attributes.
2200     while (token_ == '.' || token_ == kTokenIdentifier) NEXT();
2201     EXPECT(')');
2202     while (Is('.')) {
2203       NEXT();
2204       EXPECT(kTokenIdentifier);
2205     }
2206   } else {
2207     EXPECT(kTokenIdentifier);
2208   }
2209   return NoError();
2210 }
2211 
ParseProtoCurliesOrIdent()2212 CheckedError Parser::ParseProtoCurliesOrIdent() {
2213   if (Is('{')) {
2214     NEXT();
2215     for (int nesting = 1; nesting;) {
2216       if (token_ == '{')
2217         nesting++;
2218       else if (token_ == '}')
2219         nesting--;
2220       NEXT();
2221     }
2222   } else {
2223     NEXT();  // Any single token.
2224   }
2225   return NoError();
2226 }
2227 
ParseProtoOption()2228 CheckedError Parser::ParseProtoOption() {
2229   NEXT();
2230   ECHECK(ParseProtoKey());
2231   EXPECT('=');
2232   ECHECK(ParseProtoCurliesOrIdent());
2233   return NoError();
2234 }
2235 
2236 // Parse a protobuf type, and map it to the corresponding FlatBuffer one.
ParseTypeFromProtoType(Type * type)2237 CheckedError Parser::ParseTypeFromProtoType(Type *type) {
2238   struct type_lookup {
2239     const char *proto_type;
2240     BaseType fb_type, element;
2241   };
2242   static type_lookup lookup[] = {
2243     { "float", BASE_TYPE_FLOAT, BASE_TYPE_NONE },
2244     { "double", BASE_TYPE_DOUBLE, BASE_TYPE_NONE },
2245     { "int32", BASE_TYPE_INT, BASE_TYPE_NONE },
2246     { "int64", BASE_TYPE_LONG, BASE_TYPE_NONE },
2247     { "uint32", BASE_TYPE_UINT, BASE_TYPE_NONE },
2248     { "uint64", BASE_TYPE_ULONG, BASE_TYPE_NONE },
2249     { "sint32", BASE_TYPE_INT, BASE_TYPE_NONE },
2250     { "sint64", BASE_TYPE_LONG, BASE_TYPE_NONE },
2251     { "fixed32", BASE_TYPE_UINT, BASE_TYPE_NONE },
2252     { "fixed64", BASE_TYPE_ULONG, BASE_TYPE_NONE },
2253     { "sfixed32", BASE_TYPE_INT, BASE_TYPE_NONE },
2254     { "sfixed64", BASE_TYPE_LONG, BASE_TYPE_NONE },
2255     { "bool", BASE_TYPE_BOOL, BASE_TYPE_NONE },
2256     { "string", BASE_TYPE_STRING, BASE_TYPE_NONE },
2257     { "bytes", BASE_TYPE_VECTOR, BASE_TYPE_UCHAR },
2258     { nullptr, BASE_TYPE_NONE, BASE_TYPE_NONE }
2259   };
2260   for (auto tl = lookup; tl->proto_type; tl++) {
2261     if (attribute_ == tl->proto_type) {
2262       type->base_type = tl->fb_type;
2263       type->element = tl->element;
2264       NEXT();
2265       return NoError();
2266     }
2267   }
2268   if (Is('.')) NEXT();  // qualified names may start with a . ?
2269   ECHECK(ParseTypeIdent(*type));
2270   return NoError();
2271 }
2272 
SkipAnyJsonValue()2273 CheckedError Parser::SkipAnyJsonValue() {
2274   switch (token_) {
2275     case '{': {
2276       size_t fieldn_outer = 0;
2277       return ParseTableDelimiters(
2278           fieldn_outer, nullptr,
2279           [&](const std::string &, size_t &fieldn,
2280               const StructDef *) -> CheckedError {
2281             ECHECK(Recurse([&]() { return SkipAnyJsonValue(); }));
2282             fieldn++;
2283             return NoError();
2284           });
2285     }
2286     case '[': {
2287       size_t count = 0;
2288       return ParseVectorDelimiters(count, [&](size_t &) -> CheckedError {
2289         return Recurse([&]() { return SkipAnyJsonValue(); });
2290       });
2291     }
2292     case kTokenStringConstant:
2293     case kTokenIntegerConstant:
2294     case kTokenFloatConstant: NEXT(); break;
2295     default:
2296       if (IsIdent("true") || IsIdent("false") || IsIdent("null")) {
2297         NEXT();
2298       } else
2299         return TokenError();
2300   }
2301   return NoError();
2302 }
2303 
ParseFlexBufferValue(flexbuffers::Builder * builder)2304 CheckedError Parser::ParseFlexBufferValue(flexbuffers::Builder *builder) {
2305   switch (token_) {
2306     case '{': {
2307       auto start = builder->StartMap();
2308       size_t fieldn_outer = 0;
2309       auto err =
2310           ParseTableDelimiters(fieldn_outer, nullptr,
2311                                [&](const std::string &name, size_t &fieldn,
2312                                    const StructDef *) -> CheckedError {
2313                                  builder->Key(name);
2314                                  ECHECK(ParseFlexBufferValue(builder));
2315                                  fieldn++;
2316                                  return NoError();
2317                                });
2318       ECHECK(err);
2319       builder->EndMap(start);
2320       break;
2321     }
2322     case '[': {
2323       auto start = builder->StartVector();
2324       size_t count = 0;
2325       ECHECK(ParseVectorDelimiters(count, [&](size_t &) -> CheckedError {
2326         return ParseFlexBufferValue(builder);
2327       }));
2328       builder->EndVector(start, false, false);
2329       break;
2330     }
2331     case kTokenStringConstant:
2332       builder->String(attribute_);
2333       EXPECT(kTokenStringConstant);
2334       break;
2335     case kTokenIntegerConstant:
2336       builder->Int(StringToInt(attribute_.c_str()));
2337       EXPECT(kTokenIntegerConstant);
2338       break;
2339     case kTokenFloatConstant:
2340       builder->Double(strtod(attribute_.c_str(), nullptr));
2341       EXPECT(kTokenFloatConstant);
2342       break;
2343     default:
2344       if (IsIdent("true")) {
2345         builder->Bool(true);
2346         NEXT();
2347       } else if (IsIdent("false")) {
2348         builder->Bool(false);
2349         NEXT();
2350       } else if (IsIdent("null")) {
2351         builder->Null();
2352         NEXT();
2353       } else
2354         return TokenError();
2355   }
2356   return NoError();
2357 }
2358 
ParseFlexBuffer(const char * source,const char * source_filename,flexbuffers::Builder * builder)2359 bool Parser::ParseFlexBuffer(const char *source, const char *source_filename,
2360                              flexbuffers::Builder *builder) {
2361   auto ok = !StartParseFile(source, source_filename).Check() &&
2362             !ParseFlexBufferValue(builder).Check();
2363   if (ok) builder->Finish();
2364   return ok;
2365 }
2366 
Parse(const char * source,const char ** include_paths,const char * source_filename)2367 bool Parser::Parse(const char *source, const char **include_paths,
2368                    const char *source_filename) {
2369   FLATBUFFERS_ASSERT(0 == recurse_protection_counter);
2370   auto r = !ParseRoot(source, include_paths, source_filename).Check();
2371   FLATBUFFERS_ASSERT(0 == recurse_protection_counter);
2372   return r;
2373 }
2374 
StartParseFile(const char * source,const char * source_filename)2375 CheckedError Parser::StartParseFile(const char *source,
2376                                     const char *source_filename) {
2377   file_being_parsed_ = source_filename ? source_filename : "";
2378   source_ = source;
2379   ResetState(source_);
2380   error_.clear();
2381   ECHECK(SkipByteOrderMark());
2382   NEXT();
2383   if (Is(kTokenEof)) return Error("input file is empty");
2384   return NoError();
2385 }
2386 
ParseRoot(const char * source,const char ** include_paths,const char * source_filename)2387 CheckedError Parser::ParseRoot(const char *source, const char **include_paths,
2388                                const char *source_filename) {
2389   ECHECK(DoParse(source, include_paths, source_filename, nullptr));
2390 
2391   // Check that all types were defined.
2392   for (auto it = structs_.vec.begin(); it != structs_.vec.end();) {
2393     auto &struct_def = **it;
2394     if (struct_def.predecl) {
2395       if (opts.proto_mode) {
2396         // Protos allow enums to be used before declaration, so check if that
2397         // is the case here.
2398         EnumDef *enum_def = nullptr;
2399         for (size_t components =
2400                  struct_def.defined_namespace->components.size() + 1;
2401              components && !enum_def; components--) {
2402           auto qualified_name =
2403               struct_def.defined_namespace->GetFullyQualifiedName(
2404                   struct_def.name, components - 1);
2405           enum_def = LookupEnum(qualified_name);
2406         }
2407         if (enum_def) {
2408           // This is pretty slow, but a simple solution for now.
2409           auto initial_count = struct_def.refcount;
2410           for (auto struct_it = structs_.vec.begin();
2411                struct_it != structs_.vec.end(); ++struct_it) {
2412             auto &sd = **struct_it;
2413             for (auto field_it = sd.fields.vec.begin();
2414                  field_it != sd.fields.vec.end(); ++field_it) {
2415               auto &field = **field_it;
2416               if (field.value.type.struct_def == &struct_def) {
2417                 field.value.type.struct_def = nullptr;
2418                 field.value.type.enum_def = enum_def;
2419                 auto &bt = field.value.type.base_type == BASE_TYPE_VECTOR
2420                                ? field.value.type.element
2421                                : field.value.type.base_type;
2422                 FLATBUFFERS_ASSERT(bt == BASE_TYPE_STRUCT);
2423                 bt = enum_def->underlying_type.base_type;
2424                 struct_def.refcount--;
2425                 enum_def->refcount++;
2426               }
2427             }
2428           }
2429           if (struct_def.refcount)
2430             return Error("internal: " + NumToString(struct_def.refcount) + "/" +
2431                          NumToString(initial_count) +
2432                          " use(s) of pre-declaration enum not accounted for: " +
2433                          enum_def->name);
2434           structs_.dict.erase(structs_.dict.find(struct_def.name));
2435           it = structs_.vec.erase(it);
2436           delete &struct_def;
2437           continue;  // Skip error.
2438         }
2439       }
2440       auto err = "type referenced but not defined (check namespace): " +
2441                  struct_def.name;
2442       if (struct_def.original_location)
2443         err += ", originally at: " + *struct_def.original_location;
2444       return Error(err);
2445     }
2446     ++it;
2447   }
2448 
2449   // This check has to happen here and not earlier, because only now do we
2450   // know for sure what the type of these are.
2451   for (auto it = enums_.vec.begin(); it != enums_.vec.end(); ++it) {
2452     auto &enum_def = **it;
2453     if (enum_def.is_union) {
2454       for (auto val_it = enum_def.vals.vec.begin();
2455            val_it != enum_def.vals.vec.end(); ++val_it) {
2456         auto &val = **val_it;
2457         if (!SupportsVectorOfUnions() && val.union_type.struct_def &&
2458             val.union_type.struct_def->fixed)
2459           return Error(
2460               "only tables can be union elements in the generated language: " +
2461               val.name);
2462       }
2463     }
2464   }
2465   return NoError();
2466 }
2467 
DoParse(const char * source,const char ** include_paths,const char * source_filename,const char * include_filename)2468 CheckedError Parser::DoParse(const char *source, const char **include_paths,
2469                              const char *source_filename,
2470                              const char *include_filename) {
2471   if (source_filename) {
2472     if (included_files_.find(source_filename) == included_files_.end()) {
2473       included_files_[source_filename] =
2474           include_filename ? include_filename : "";
2475       files_included_per_file_[source_filename] = std::set<std::string>();
2476     } else {
2477       return NoError();
2478     }
2479   }
2480   if (!include_paths) {
2481     static const char *current_directory[] = { "", nullptr };
2482     include_paths = current_directory;
2483   }
2484   field_stack_.clear();
2485   builder_.Clear();
2486   // Start with a blank namespace just in case this file doesn't have one.
2487   current_namespace_ = empty_namespace_;
2488 
2489   ECHECK(StartParseFile(source, source_filename));
2490 
2491   // Includes must come before type declarations:
2492   for (;;) {
2493     // Parse pre-include proto statements if any:
2494     if (opts.proto_mode && (attribute_ == "option" || attribute_ == "syntax" ||
2495                             attribute_ == "package")) {
2496       ECHECK(ParseProtoDecl());
2497     } else if (IsIdent("native_include")) {
2498       NEXT();
2499       vector_emplace_back(&native_included_files_, attribute_);
2500       EXPECT(kTokenStringConstant);
2501       EXPECT(';');
2502     } else if (IsIdent("include") || (opts.proto_mode && IsIdent("import"))) {
2503       NEXT();
2504       if (opts.proto_mode && attribute_ == "public") NEXT();
2505       auto name = flatbuffers::PosixPath(attribute_.c_str());
2506       EXPECT(kTokenStringConstant);
2507       // Look for the file in include_paths.
2508       std::string filepath;
2509       for (auto paths = include_paths; paths && *paths; paths++) {
2510         filepath = flatbuffers::ConCatPathFileName(*paths, name);
2511         if (FileExists(filepath.c_str())) break;
2512       }
2513       if (filepath.empty())
2514         return Error("unable to locate include file: " + name);
2515       if (source_filename)
2516         files_included_per_file_[source_filename].insert(filepath);
2517       if (included_files_.find(filepath) == included_files_.end()) {
2518         // We found an include file that we have not parsed yet.
2519         // Load it and parse it.
2520         std::string contents;
2521         if (!LoadFile(filepath.c_str(), true, &contents))
2522           return Error("unable to load include file: " + name);
2523         ECHECK(DoParse(contents.c_str(), include_paths, filepath.c_str(),
2524                        name.c_str()));
2525         // We generally do not want to output code for any included files:
2526         if (!opts.generate_all) MarkGenerated();
2527         // Reset these just in case the included file had them, and the
2528         // parent doesn't.
2529         root_struct_def_ = nullptr;
2530         file_identifier_.clear();
2531         file_extension_.clear();
2532         // This is the easiest way to continue this file after an include:
2533         // instead of saving and restoring all the state, we simply start the
2534         // file anew. This will cause it to encounter the same include
2535         // statement again, but this time it will skip it, because it was
2536         // entered into included_files_.
2537         // This is recursive, but only go as deep as the number of include
2538         // statements.
2539         if (source_filename) {
2540           included_files_.erase(source_filename);
2541         }
2542         return DoParse(source, include_paths, source_filename,
2543                        include_filename);
2544       }
2545       EXPECT(';');
2546     } else {
2547       break;
2548     }
2549   }
2550   // Now parse all other kinds of declarations:
2551   while (token_ != kTokenEof) {
2552     if (opts.proto_mode) {
2553       ECHECK(ParseProtoDecl());
2554     } else if (IsIdent("namespace")) {
2555       ECHECK(ParseNamespace());
2556     } else if (token_ == '{') {
2557       if (!root_struct_def_)
2558         return Error("no root type set to parse json with");
2559       if (builder_.GetSize()) {
2560         return Error("cannot have more than one json object in a file");
2561       }
2562       uoffset_t toff;
2563       ECHECK(ParseTable(*root_struct_def_, nullptr, &toff));
2564       if (opts.size_prefixed) {
2565         builder_.FinishSizePrefixed(Offset<Table>(toff), file_identifier_.length()
2566                                                              ? file_identifier_.c_str()
2567                                                              : nullptr);
2568       } else {
2569         builder_.Finish(Offset<Table>(toff), file_identifier_.length()
2570                                                  ? file_identifier_.c_str()
2571                                                  : nullptr);
2572       }
2573       // Check that JSON file doesn't contain more objects or IDL directives.
2574       // Comments after JSON are allowed.
2575       EXPECT(kTokenEof);
2576     } else if (IsIdent("enum")) {
2577       ECHECK(ParseEnum(false, nullptr));
2578     } else if (IsIdent("union")) {
2579       ECHECK(ParseEnum(true, nullptr));
2580     } else if (IsIdent("root_type")) {
2581       NEXT();
2582       auto root_type = attribute_;
2583       EXPECT(kTokenIdentifier);
2584       ECHECK(ParseNamespacing(&root_type, nullptr));
2585       if (opts.root_type.empty()) {
2586         if (!SetRootType(root_type.c_str()))
2587           return Error("unknown root type: " + root_type);
2588         if (root_struct_def_->fixed)
2589           return Error("root type must be a table");
2590       }
2591       EXPECT(';');
2592     } else if (IsIdent("file_identifier")) {
2593       NEXT();
2594       file_identifier_ = attribute_;
2595       EXPECT(kTokenStringConstant);
2596       if (file_identifier_.length() != FlatBufferBuilder::kFileIdentifierLength)
2597         return Error("file_identifier must be exactly " +
2598                      NumToString(FlatBufferBuilder::kFileIdentifierLength) +
2599                      " characters");
2600       EXPECT(';');
2601     } else if (IsIdent("file_extension")) {
2602       NEXT();
2603       file_extension_ = attribute_;
2604       EXPECT(kTokenStringConstant);
2605       EXPECT(';');
2606     } else if (IsIdent("include")) {
2607       return Error("includes must come before declarations");
2608     } else if (IsIdent("attribute")) {
2609       NEXT();
2610       auto name = attribute_;
2611       if (Is(kTokenIdentifier)) {
2612         NEXT();
2613       } else {
2614         EXPECT(kTokenStringConstant);
2615       }
2616       EXPECT(';');
2617       known_attributes_[name] = false;
2618     } else if (IsIdent("rpc_service")) {
2619       ECHECK(ParseService());
2620     } else {
2621       ECHECK(ParseDecl());
2622     }
2623   }
2624   return NoError();
2625 }
2626 
GetIncludedFilesRecursive(const std::string & file_name) const2627 std::set<std::string> Parser::GetIncludedFilesRecursive(
2628     const std::string &file_name) const {
2629   std::set<std::string> included_files;
2630   std::list<std::string> to_process;
2631 
2632   if (file_name.empty()) return included_files;
2633   to_process.push_back(file_name);
2634 
2635   while (!to_process.empty()) {
2636     std::string current = to_process.front();
2637     to_process.pop_front();
2638     included_files.insert(current);
2639 
2640     // Workaround the lack of const accessor in C++98 maps.
2641     auto &new_files =
2642         (*const_cast<std::map<std::string, std::set<std::string>> *>(
2643             &files_included_per_file_))[current];
2644     for (auto it = new_files.begin(); it != new_files.end(); ++it) {
2645       if (included_files.find(*it) == included_files.end())
2646         to_process.push_back(*it);
2647     }
2648   }
2649 
2650   return included_files;
2651 }
2652 
2653 // Schema serialization functionality:
2654 
compareName(const T * a,const T * b)2655 template<typename T> bool compareName(const T *a, const T *b) {
2656   return a->defined_namespace->GetFullyQualifiedName(a->name) <
2657          b->defined_namespace->GetFullyQualifiedName(b->name);
2658 }
2659 
AssignIndices(const std::vector<T * > & defvec)2660 template<typename T> void AssignIndices(const std::vector<T *> &defvec) {
2661   // Pre-sort these vectors, such that we can set the correct indices for them.
2662   auto vec = defvec;
2663   std::sort(vec.begin(), vec.end(), compareName<T>);
2664   for (int i = 0; i < static_cast<int>(vec.size()); i++) vec[i]->index = i;
2665 }
2666 
Serialize()2667 void Parser::Serialize() {
2668   builder_.Clear();
2669   AssignIndices(structs_.vec);
2670   AssignIndices(enums_.vec);
2671   std::vector<Offset<reflection::Object>> object_offsets;
2672   for (auto it = structs_.vec.begin(); it != structs_.vec.end(); ++it) {
2673     auto offset = (*it)->Serialize(&builder_, *this);
2674     object_offsets.push_back(offset);
2675     (*it)->serialized_location = offset.o;
2676   }
2677   std::vector<Offset<reflection::Enum>> enum_offsets;
2678   for (auto it = enums_.vec.begin(); it != enums_.vec.end(); ++it) {
2679     auto offset = (*it)->Serialize(&builder_, *this);
2680     enum_offsets.push_back(offset);
2681     (*it)->serialized_location = offset.o;
2682   }
2683   std::vector<Offset<reflection::Service>> service_offsets;
2684   for (auto it = services_.vec.begin(); it != services_.vec.end(); ++it) {
2685     auto offset = (*it)->Serialize(&builder_, *this);
2686     service_offsets.push_back(offset);
2687     (*it)->serialized_location = offset.o;
2688   }
2689   auto objs__ = builder_.CreateVectorOfSortedTables(&object_offsets);
2690   auto enum__ = builder_.CreateVectorOfSortedTables(&enum_offsets);
2691   auto fiid__ = builder_.CreateString(file_identifier_);
2692   auto fext__ = builder_.CreateString(file_extension_);
2693   auto serv__ = builder_.CreateVectorOfSortedTables(&service_offsets);
2694   auto schema_offset =
2695       reflection::CreateSchema(builder_, objs__, enum__, fiid__, fext__,
2696         (root_struct_def_ ? root_struct_def_->serialized_location : 0),
2697         serv__);
2698   if (opts.size_prefixed) {
2699     builder_.FinishSizePrefixed(schema_offset, reflection::SchemaIdentifier());
2700   } else {
2701     builder_.Finish(schema_offset, reflection::SchemaIdentifier());
2702   }
2703 }
2704 
GetNamespace(const std::string & qualified_name,std::vector<Namespace * > & namespaces,std::map<std::string,Namespace * > & namespaces_index)2705 static Namespace *GetNamespace(
2706     const std::string &qualified_name, std::vector<Namespace *> &namespaces,
2707     std::map<std::string, Namespace *> &namespaces_index) {
2708   size_t dot = qualified_name.find_last_of('.');
2709   std::string namespace_name = (dot != std::string::npos)
2710                                    ? std::string(qualified_name.c_str(), dot)
2711                                    : "";
2712   Namespace *&ns = namespaces_index[namespace_name];
2713 
2714   if (!ns) {
2715     ns = new Namespace();
2716     namespaces.push_back(ns);
2717 
2718     size_t pos = 0;
2719 
2720     for (;;) {
2721       dot = qualified_name.find('.', pos);
2722       if (dot == std::string::npos) { break; }
2723       ns->components.push_back(qualified_name.substr(pos, dot-pos));
2724       pos = dot + 1;
2725     }
2726   }
2727 
2728   return ns;
2729 }
2730 
Serialize(FlatBufferBuilder * builder,const Parser & parser) const2731 Offset<reflection::Object> StructDef::Serialize(FlatBufferBuilder *builder,
2732                                                 const Parser &parser) const {
2733   std::vector<Offset<reflection::Field>> field_offsets;
2734   for (auto it = fields.vec.begin(); it != fields.vec.end(); ++it) {
2735     field_offsets.push_back((*it)->Serialize(
2736         builder, static_cast<uint16_t>(it - fields.vec.begin()), parser));
2737   }
2738   auto qualified_name = defined_namespace->GetFullyQualifiedName(name);
2739   auto name__ = builder->CreateString(qualified_name);
2740   auto flds__ = builder->CreateVectorOfSortedTables(&field_offsets);
2741   auto attr__ = SerializeAttributes(builder, parser);
2742   auto docs__ = parser.opts.binary_schema_comments
2743                 ? builder->CreateVectorOfStrings(doc_comment)
2744                 : 0;
2745   return reflection::CreateObject(*builder, name__, flds__, fixed,
2746                                   static_cast<int>(minalign),
2747                                   static_cast<int>(bytesize),
2748                                   attr__, docs__);
2749 }
2750 
Deserialize(Parser & parser,const reflection::Object * object)2751 bool StructDef::Deserialize(Parser &parser, const reflection::Object *object) {
2752   if (!DeserializeAttributes(parser, object->attributes()))
2753     return false;
2754   DeserializeDoc(doc_comment, object->documentation());
2755   name = parser.UnqualifiedName(object->name()->str());
2756   fixed = object->is_struct();
2757   minalign = object->minalign();
2758   predecl = false;
2759   sortbysize = attributes.Lookup("original_order") == nullptr && !fixed;
2760   std::vector<uoffset_t> indexes =
2761     std::vector<uoffset_t>(object->fields()->size());
2762   for (uoffset_t i = 0; i < object->fields()->size(); i++)
2763     indexes[object->fields()->Get(i)->id()] = i;
2764   for (size_t i = 0; i < indexes.size(); i++) {
2765     auto field = object->fields()->Get(indexes[i]);
2766     auto field_def = new FieldDef();
2767     if (!field_def->Deserialize(parser, field) ||
2768         fields.Add(field_def->name, field_def)) {
2769       delete field_def;
2770       return false;
2771     }
2772     if (fixed) {
2773       // Recompute padding since that's currently not serialized.
2774       auto size = InlineSize(field_def->value.type);
2775       auto next_field =
2776           i + 1 < indexes.size()
2777           ? object->fields()->Get(indexes[i+1])
2778           : nullptr;
2779       bytesize += size;
2780       field_def->padding =
2781           next_field ? (next_field->offset() - field_def->value.offset) - size
2782                      : PaddingBytes(bytesize, minalign);
2783       bytesize += field_def->padding;
2784     }
2785   }
2786   FLATBUFFERS_ASSERT(static_cast<int>(bytesize) == object->bytesize());
2787   return true;
2788 }
2789 
Serialize(FlatBufferBuilder * builder,uint16_t id,const Parser & parser) const2790 Offset<reflection::Field> FieldDef::Serialize(FlatBufferBuilder *builder,
2791                                               uint16_t id,
2792                                               const Parser &parser) const {
2793   auto name__ = builder->CreateString(name);
2794   auto type__ = value.type.Serialize(builder);
2795   auto attr__ = SerializeAttributes(builder, parser);
2796   auto docs__ = parser.opts.binary_schema_comments
2797                 ? builder->CreateVectorOfStrings(doc_comment)
2798                 : 0;
2799   return reflection::CreateField(*builder, name__, type__, id, value.offset,
2800       // Is uint64>max(int64) tested?
2801       IsInteger(value.type.base_type) ? StringToInt(value.constant.c_str()) : 0,
2802       // result may be platform-dependent if underlying is float (not double)
2803       IsFloat(value.type.base_type) ? strtod(value.constant.c_str(), nullptr)
2804                                     : 0.0,
2805       deprecated, required, key, attr__, docs__);
2806   // TODO: value.constant is almost always "0", we could save quite a bit of
2807   // space by sharing it. Same for common values of value.type.
2808 }
2809 
Deserialize(Parser & parser,const reflection::Field * field)2810 bool FieldDef::Deserialize(Parser &parser, const reflection::Field *field) {
2811   name = parser.UnqualifiedName(field->name()->str());
2812   defined_namespace = parser.current_namespace_;
2813   if (!value.type.Deserialize(parser, field->type()))
2814     return false;
2815   value.offset = field->offset();
2816   if (IsInteger(value.type.base_type)) {
2817     value.constant = NumToString(field->default_integer());
2818   } else if (IsFloat(value.type.base_type)) {
2819     value.constant = FloatToString(field->default_real(), 16);
2820     size_t last_zero = value.constant.find_last_not_of('0');
2821     if (last_zero != std::string::npos && last_zero != 0) {
2822       value.constant.erase(last_zero, std::string::npos);
2823     }
2824   }
2825   deprecated = field->deprecated();
2826   required = field->required();
2827   key = field->key();
2828   if (!DeserializeAttributes(parser, field->attributes()))
2829     return false;
2830   // TODO: this should probably be handled by a separate attribute
2831   if (attributes.Lookup("flexbuffer")) {
2832     flexbuffer = true;
2833     parser.uses_flexbuffers_ = true;
2834     if (value.type.base_type != BASE_TYPE_VECTOR ||
2835         value.type.element != BASE_TYPE_UCHAR)
2836       return false;
2837   }
2838   DeserializeDoc(doc_comment, field->documentation());
2839   return true;
2840 }
2841 
Serialize(FlatBufferBuilder * builder,const Parser & parser) const2842 Offset<reflection::RPCCall> RPCCall::Serialize(FlatBufferBuilder *builder,
2843                                                const Parser &parser) const {
2844   auto name__ = builder->CreateString(name);
2845   auto attr__ = SerializeAttributes(builder, parser);
2846   auto docs__ = parser.opts.binary_schema_comments
2847                 ? builder->CreateVectorOfStrings(doc_comment)
2848                 : 0;
2849   return reflection::CreateRPCCall(*builder, name__,
2850                                    request->serialized_location,
2851                                    response->serialized_location,
2852                                    attr__, docs__);
2853 }
2854 
Deserialize(Parser & parser,const reflection::RPCCall * call)2855 bool RPCCall::Deserialize(Parser &parser, const reflection::RPCCall *call) {
2856   name = call->name()->str();
2857   if (!DeserializeAttributes(parser, call->attributes()))
2858     return false;
2859   DeserializeDoc(doc_comment, call->documentation());
2860   request = parser.structs_.Lookup(call->request()->name()->str());
2861   response = parser.structs_.Lookup(call->response()->name()->str());
2862   if (!request || !response) { return false; }
2863   return true;
2864 }
2865 
Serialize(FlatBufferBuilder * builder,const Parser & parser) const2866 Offset<reflection::Service> ServiceDef::Serialize(FlatBufferBuilder *builder,
2867                                                   const Parser &parser) const {
2868   std::vector<Offset<reflection::RPCCall>> servicecall_offsets;
2869   for (auto it = calls.vec.begin(); it != calls.vec.end(); ++it) {
2870     servicecall_offsets.push_back((*it)->Serialize(builder, parser));
2871   }
2872   auto qualified_name = defined_namespace->GetFullyQualifiedName(name);
2873   auto name__ = builder->CreateString(qualified_name);
2874   auto call__ = builder->CreateVector(servicecall_offsets);
2875   auto attr__ = SerializeAttributes(builder, parser);
2876   auto docs__ = parser.opts.binary_schema_comments
2877                 ? builder->CreateVectorOfStrings(doc_comment)
2878                 : 0;
2879   return reflection::CreateService(*builder, name__, call__, attr__, docs__);
2880 }
2881 
Deserialize(Parser & parser,const reflection::Service * service)2882 bool ServiceDef::Deserialize(Parser &parser,
2883                              const reflection::Service *service) {
2884   name = parser.UnqualifiedName(service->name()->str());
2885   if (service->calls()) {
2886     for (uoffset_t i = 0; i < service->calls()->size(); ++i) {
2887       auto call = new RPCCall();
2888       if (!call->Deserialize(parser, service->calls()->Get(i)) ||
2889           calls.Add(call->name, call)) {
2890         delete call;
2891         return false;
2892       }
2893     }
2894   }
2895   if (!DeserializeAttributes(parser, service->attributes()))
2896     return false;
2897   DeserializeDoc(doc_comment, service->documentation());
2898   return true;
2899 }
2900 
Serialize(FlatBufferBuilder * builder,const Parser & parser) const2901 Offset<reflection::Enum> EnumDef::Serialize(FlatBufferBuilder *builder,
2902                                             const Parser &parser) const {
2903   std::vector<Offset<reflection::EnumVal>> enumval_offsets;
2904   for (auto it = vals.vec.begin(); it != vals.vec.end(); ++it) {
2905     enumval_offsets.push_back((*it)->Serialize(builder, parser));
2906   }
2907   auto qualified_name = defined_namespace->GetFullyQualifiedName(name);
2908   auto name__ = builder->CreateString(qualified_name);
2909   auto vals__ = builder->CreateVector(enumval_offsets);
2910   auto type__ = underlying_type.Serialize(builder);
2911   auto attr__ = SerializeAttributes(builder, parser);
2912   auto docs__ = parser.opts.binary_schema_comments
2913                 ? builder->CreateVectorOfStrings(doc_comment)
2914                 : 0;
2915   return reflection::CreateEnum(*builder, name__, vals__, is_union, type__,
2916                                 attr__, docs__);
2917 }
2918 
Deserialize(Parser & parser,const reflection::Enum * _enum)2919 bool EnumDef::Deserialize(Parser &parser, const reflection::Enum *_enum) {
2920   name = parser.UnqualifiedName(_enum->name()->str());
2921   for (uoffset_t i = 0; i < _enum->values()->size(); ++i) {
2922     auto val = new EnumVal();
2923     if (!val->Deserialize(parser, _enum->values()->Get(i)) ||
2924         vals.Add(val->name, val)) {
2925       delete val;
2926       return false;
2927     }
2928   }
2929   is_union = _enum->is_union();
2930   if (!underlying_type.Deserialize(parser, _enum->underlying_type())) {
2931     return false;
2932   }
2933   if (!DeserializeAttributes(parser, _enum->attributes()))
2934     return false;
2935   DeserializeDoc(doc_comment, _enum->documentation());
2936   return true;
2937 }
2938 
Serialize(FlatBufferBuilder * builder,const Parser & parser) const2939 Offset<reflection::EnumVal> EnumVal::Serialize(FlatBufferBuilder *builder,
2940                                                const Parser &parser) const {
2941   auto name__ = builder->CreateString(name);
2942   auto type__ = union_type.Serialize(builder);
2943   auto docs__ = parser.opts.binary_schema_comments
2944                 ? builder->CreateVectorOfStrings(doc_comment)
2945                 : 0;
2946   return reflection::CreateEnumVal(*builder, name__, value,
2947       union_type.struct_def ? union_type.struct_def->serialized_location : 0,
2948       type__, docs__);
2949 }
2950 
Deserialize(const Parser & parser,const reflection::EnumVal * val)2951 bool EnumVal::Deserialize(const Parser &parser,
2952                           const reflection::EnumVal *val) {
2953   name = val->name()->str();
2954   value = val->value();
2955   if (!union_type.Deserialize(parser, val->union_type()))
2956     return false;
2957   DeserializeDoc(doc_comment, val->documentation());
2958   return true;
2959 }
2960 
Serialize(FlatBufferBuilder * builder) const2961 Offset<reflection::Type> Type::Serialize(FlatBufferBuilder *builder) const {
2962   return reflection::CreateType(
2963       *builder,
2964       static_cast<reflection::BaseType>(base_type),
2965       static_cast<reflection::BaseType>(element),
2966       struct_def ? struct_def->index : (enum_def ? enum_def->index : -1));
2967 }
2968 
Deserialize(const Parser & parser,const reflection::Type * type)2969 bool Type::Deserialize(const Parser &parser, const reflection::Type *type) {
2970   if (type == nullptr) return true;
2971   base_type = static_cast<BaseType>(type->base_type());
2972   element = static_cast<BaseType>(type->element());
2973   if (type->index() >= 0) {
2974     if (type->base_type() == reflection::Obj ||
2975         (type->base_type() == reflection::Vector &&
2976          type->element() == reflection::Obj)) {
2977       if (static_cast<size_t>(type->index()) < parser.structs_.vec.size()) {
2978         struct_def = parser.structs_.vec[type->index()];
2979         struct_def->refcount++;
2980       } else {
2981         return false;
2982       }
2983     } else {
2984       if (static_cast<size_t>(type->index()) < parser.enums_.vec.size()) {
2985         enum_def = parser.enums_.vec[type->index()];
2986       } else {
2987         return false;
2988       }
2989     }
2990   }
2991   return true;
2992 }
2993 
2994 flatbuffers::Offset<
2995     flatbuffers::Vector<flatbuffers::Offset<reflection::KeyValue>>>
SerializeAttributes(FlatBufferBuilder * builder,const Parser & parser) const2996 Definition::SerializeAttributes(FlatBufferBuilder *builder,
2997                                 const Parser &parser) const {
2998   std::vector<flatbuffers::Offset<reflection::KeyValue>> attrs;
2999   for (auto kv = attributes.dict.begin(); kv != attributes.dict.end(); ++kv) {
3000     auto it = parser.known_attributes_.find(kv->first);
3001     FLATBUFFERS_ASSERT(it != parser.known_attributes_.end());
3002     if (parser.opts.binary_schema_builtins || !it->second) {
3003       auto key = builder->CreateString(kv->first);
3004       auto val = builder->CreateString(kv->second->constant);
3005       attrs.push_back(reflection::CreateKeyValue(*builder, key, val));
3006     }
3007   }
3008   if (attrs.size()) {
3009     return builder->CreateVectorOfSortedTables(&attrs);
3010   } else {
3011     return 0;
3012   }
3013 }
3014 
DeserializeAttributes(Parser & parser,const Vector<Offset<reflection::KeyValue>> * attrs)3015 bool Definition::DeserializeAttributes(
3016     Parser &parser, const Vector<Offset<reflection::KeyValue>> *attrs) {
3017   if (attrs == nullptr)
3018     return true;
3019   for (uoffset_t i = 0; i < attrs->size(); ++i) {
3020     auto kv = attrs->Get(i);
3021     auto value = new Value();
3022     if (kv->value()) { value->constant = kv->value()->str(); }
3023     if (attributes.Add(kv->key()->str(), value)) {
3024       delete value;
3025       return false;
3026     }
3027     parser.known_attributes_[kv->key()->str()];
3028   }
3029   return true;
3030 }
3031 
3032 /************************************************************************/
3033 /* DESERIALIZATION                                                      */
3034 /************************************************************************/
Deserialize(const uint8_t * buf,const size_t size)3035 bool Parser::Deserialize(const uint8_t *buf, const size_t size) {
3036   flatbuffers::Verifier verifier(reinterpret_cast<const uint8_t *>(buf), size);
3037   bool size_prefixed = false;
3038   if(!reflection::SchemaBufferHasIdentifier(buf)) {
3039     if (!flatbuffers::BufferHasIdentifier(buf, reflection::SchemaIdentifier(),
3040                                           true))
3041       return false;
3042     else
3043       size_prefixed = true;
3044   }
3045   auto verify_fn = size_prefixed ? &reflection::VerifySizePrefixedSchemaBuffer
3046                                  : &reflection::VerifySchemaBuffer;
3047   if (!verify_fn(verifier)) {
3048     return false;
3049   }
3050   auto schema = size_prefixed ? reflection::GetSizePrefixedSchema(buf)
3051                               : reflection::GetSchema(buf);
3052   return Deserialize(schema);
3053 }
3054 
Deserialize(const reflection::Schema * schema)3055 bool Parser::Deserialize(const reflection::Schema *schema) {
3056   file_identifier_ = schema->file_ident() ? schema->file_ident()->str() : "";
3057   file_extension_ = schema->file_ext() ? schema->file_ext()->str() : "";
3058   std::map<std::string, Namespace *> namespaces_index;
3059 
3060   // Create defs without deserializing so references from fields to structs and
3061   // enums can be resolved.
3062   for (auto it = schema->objects()->begin(); it != schema->objects()->end();
3063        ++it) {
3064     auto struct_def = new StructDef();
3065     if (structs_.Add(it->name()->str(), struct_def)) {
3066       delete struct_def;
3067       return false;
3068     }
3069     auto type = new Type(BASE_TYPE_STRUCT, struct_def, nullptr);
3070     if (types_.Add(it->name()->str(), type)) {
3071       delete type;
3072       return false;
3073     }
3074   }
3075   for (auto it = schema->enums()->begin(); it != schema->enums()->end(); ++it) {
3076     auto enum_def = new EnumDef();
3077     if (enums_.Add(it->name()->str(), enum_def)) {
3078       delete enum_def;
3079       return false;
3080     }
3081     auto type = new Type(BASE_TYPE_UNION, nullptr, enum_def);
3082     if (types_.Add(it->name()->str(), type)) {
3083       delete type;
3084       return false;
3085     }
3086   }
3087 
3088   // Now fields can refer to structs and enums by index.
3089   for (auto it = schema->objects()->begin(); it != schema->objects()->end();
3090        ++it) {
3091     std::string qualified_name = it->name()->str();
3092     auto struct_def = structs_.Lookup(qualified_name);
3093     struct_def->defined_namespace =
3094         GetNamespace(qualified_name, namespaces_, namespaces_index);
3095     if (!struct_def->Deserialize(*this, * it)) { return false; }
3096     if (schema->root_table() == *it) { root_struct_def_ = struct_def; }
3097   }
3098   for (auto it = schema->enums()->begin(); it != schema->enums()->end(); ++it) {
3099     std::string qualified_name = it->name()->str();
3100     auto enum_def = enums_.Lookup(qualified_name);
3101     enum_def->defined_namespace =
3102         GetNamespace(qualified_name, namespaces_, namespaces_index);
3103     if (!enum_def->Deserialize(*this, *it)) { return false; }
3104   }
3105 
3106   if (schema->services()) {
3107     for (auto it = schema->services()->begin(); it != schema->services()->end();
3108          ++it) {
3109       std::string qualified_name = it->name()->str();
3110       auto service_def = new ServiceDef();
3111       service_def->defined_namespace =
3112           GetNamespace(qualified_name, namespaces_, namespaces_index);
3113       if (!service_def->Deserialize(*this, *it) ||
3114           services_.Add(qualified_name, service_def)) {
3115         delete service_def;
3116         return false;
3117       }
3118     }
3119   }
3120 
3121   return true;
3122 }
3123 
ConformTo(const Parser & base)3124 std::string Parser::ConformTo(const Parser &base) {
3125   for (auto sit = structs_.vec.begin(); sit != structs_.vec.end(); ++sit) {
3126     auto &struct_def = **sit;
3127     auto qualified_name =
3128         struct_def.defined_namespace->GetFullyQualifiedName(struct_def.name);
3129     auto struct_def_base = base.LookupStruct(qualified_name);
3130     if (!struct_def_base) continue;
3131     for (auto fit = struct_def.fields.vec.begin();
3132          fit != struct_def.fields.vec.end(); ++fit) {
3133       auto &field = **fit;
3134       auto field_base = struct_def_base->fields.Lookup(field.name);
3135       if (field_base) {
3136         if (field.value.offset != field_base->value.offset)
3137           return "offsets differ for field: " + field.name;
3138         if (field.value.constant != field_base->value.constant)
3139           return "defaults differ for field: " + field.name;
3140         if (!EqualByName(field.value.type, field_base->value.type))
3141           return "types differ for field: " + field.name;
3142       } else {
3143         // Doesn't have to exist, deleting fields is fine.
3144         // But we should check if there is a field that has the same offset
3145         // but is incompatible (in the case of field renaming).
3146         for (auto fbit = struct_def_base->fields.vec.begin();
3147              fbit != struct_def_base->fields.vec.end(); ++fbit) {
3148           field_base = *fbit;
3149           if (field.value.offset == field_base->value.offset) {
3150             if (!EqualByName(field.value.type, field_base->value.type))
3151               return "field renamed to different type: " + field.name;
3152             break;
3153           }
3154         }
3155       }
3156     }
3157   }
3158   for (auto eit = enums_.vec.begin(); eit != enums_.vec.end(); ++eit) {
3159     auto &enum_def = **eit;
3160     auto qualified_name =
3161         enum_def.defined_namespace->GetFullyQualifiedName(enum_def.name);
3162     auto enum_def_base = base.enums_.Lookup(qualified_name);
3163     if (!enum_def_base) continue;
3164     for (auto evit = enum_def.vals.vec.begin(); evit != enum_def.vals.vec.end();
3165          ++evit) {
3166       auto &enum_val = **evit;
3167       auto enum_val_base = enum_def_base->vals.Lookup(enum_val.name);
3168       if (enum_val_base) {
3169         if (enum_val.value != enum_val_base->value)
3170           return "values differ for enum: " + enum_val.name;
3171       }
3172     }
3173   }
3174   return "";
3175 }
3176 
3177 }  // namespace flatbuffers
3178