• 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 <cmath>
19 #include <list>
20 #include <string>
21 #include <utility>
22 
23 #include "flatbuffers/idl.h"
24 #include "flatbuffers/util.h"
25 
26 namespace flatbuffers {
27 
28 // Reflects the version at the compiling time of binary(lib/dll/so).
FLATBUFFERS_VERSION()29 const char *FLATBUFFERS_VERSION() {
30   // clang-format off
31   return
32       FLATBUFFERS_STRING(FLATBUFFERS_VERSION_MAJOR) "."
33       FLATBUFFERS_STRING(FLATBUFFERS_VERSION_MINOR) "."
34       FLATBUFFERS_STRING(FLATBUFFERS_VERSION_REVISION);
35   // clang-format on
36 }
37 
38 const double kPi = 3.14159265358979323846;
39 
40 // clang-format off
41 const char *const kTypeNames[] = {
42   #define FLATBUFFERS_TD(ENUM, IDLTYPE, ...) \
43     IDLTYPE,
44     FLATBUFFERS_GEN_TYPES(FLATBUFFERS_TD)
45   #undef FLATBUFFERS_TD
46   nullptr
47 };
48 
49 const char kTypeSizes[] = {
50   #define FLATBUFFERS_TD(ENUM, IDLTYPE, CTYPE, ...) \
51     sizeof(CTYPE),
52     FLATBUFFERS_GEN_TYPES(FLATBUFFERS_TD)
53   #undef FLATBUFFERS_TD
54 };
55 // clang-format on
56 
57 // The enums in the reflection schema should match the ones we use internally.
58 // Compare the last element to check if these go out of sync.
59 static_assert(BASE_TYPE_UNION == static_cast<BaseType>(reflection::Union),
60               "enums don't match");
61 
62 // Any parsing calls have to be wrapped in this macro, which automates
63 // handling of recursive error checking a bit. It will check the received
64 // CheckedError object, and return straight away on error.
65 #define ECHECK(call)           \
66   {                            \
67     auto ce = (call);          \
68     if (ce.Check()) return ce; \
69   }
70 
71 // These two functions are called hundreds of times below, so define a short
72 // form:
73 #define NEXT() ECHECK(Next())
74 #define EXPECT(tok) ECHECK(Expect(tok))
75 
ValidateUTF8(const std::string & str)76 static bool ValidateUTF8(const std::string &str) {
77   const char *s = &str[0];
78   const char *const sEnd = s + str.length();
79   while (s < sEnd) {
80     if (FromUTF8(&s) < 0) { return false; }
81   }
82   return true;
83 }
84 
IsLowerSnakeCase(const std::string & str)85 static bool IsLowerSnakeCase(const std::string &str) {
86   for (size_t i = 0; i < str.length(); i++) {
87     char c = str[i];
88     if (!check_ascii_range(c, 'a', 'z') && !is_digit(c) && c != '_') {
89       return false;
90     }
91   }
92   return true;
93 }
94 
95 // Convert an underscore_based_identifier in to camelCase.
96 // Also uppercases the first character if first is true.
MakeCamel(const std::string & in,bool first)97 std::string MakeCamel(const std::string &in, bool first) {
98   std::string s;
99   for (size_t i = 0; i < in.length(); i++) {
100     if (!i && first)
101       s += CharToUpper(in[0]);
102     else if (in[i] == '_' && i + 1 < in.length())
103       s += CharToUpper(in[++i]);
104     else
105       s += in[i];
106   }
107   return s;
108 }
109 
110 // Convert an underscore_based_identifier in to screaming snake case.
MakeScreamingCamel(const std::string & in)111 std::string MakeScreamingCamel(const std::string &in) {
112   std::string s;
113   for (size_t i = 0; i < in.length(); i++) {
114     if (in[i] != '_')
115       s += CharToUpper(in[i]);
116     else
117       s += in[i];
118   }
119   return s;
120 }
121 
DeserializeDoc(std::vector<std::string> & doc,const Vector<Offset<String>> * documentation)122 void DeserializeDoc(std::vector<std::string> &doc,
123                     const Vector<Offset<String>> *documentation) {
124   if (documentation == nullptr) return;
125   for (uoffset_t index = 0; index < documentation->size(); index++)
126     doc.push_back(documentation->Get(index)->str());
127 }
128 
Message(const std::string & msg)129 void Parser::Message(const std::string &msg) {
130   if (!error_.empty()) error_ += "\n";  // log all warnings and errors
131   error_ += file_being_parsed_.length() ? AbsolutePath(file_being_parsed_) : "";
132   // clang-format off
133 
134   #ifdef _WIN32  // MSVC alike
135     error_ +=
136         "(" + NumToString(line_) + ", " + NumToString(CursorPosition()) + ")";
137   #else  // gcc alike
138     if (file_being_parsed_.length()) error_ += ":";
139     error_ += NumToString(line_) + ": " + NumToString(CursorPosition());
140   #endif
141   // clang-format on
142   error_ += ": " + msg;
143 }
144 
Warning(const std::string & msg)145 void Parser::Warning(const std::string &msg) {
146   if (!opts.no_warnings) Message("warning: " + msg);
147 }
148 
Error(const std::string & msg)149 CheckedError Parser::Error(const std::string &msg) {
150   Message("error: " + msg);
151   return CheckedError(true);
152 }
153 
NoError()154 inline CheckedError NoError() { return CheckedError(false); }
155 
RecurseError()156 CheckedError Parser::RecurseError() {
157   return Error("maximum parsing depth " + NumToString(parse_depth_counter_) +
158                " reached");
159 }
160 
161 class Parser::ParseDepthGuard {
162  public:
ParseDepthGuard(Parser * parser_not_null)163   explicit ParseDepthGuard(Parser *parser_not_null)
164       : parser_(*parser_not_null), caller_depth_(parser_.parse_depth_counter_) {
165     FLATBUFFERS_ASSERT(caller_depth_ <= (FLATBUFFERS_MAX_PARSING_DEPTH) &&
166                        "Check() must be called to prevent stack overflow");
167     parser_.parse_depth_counter_ += 1;
168   }
169 
~ParseDepthGuard()170   ~ParseDepthGuard() { parser_.parse_depth_counter_ -= 1; }
171 
Check()172   CheckedError Check() {
173     return caller_depth_ >= (FLATBUFFERS_MAX_PARSING_DEPTH)
174                ? parser_.RecurseError()
175                : CheckedError(false);
176   }
177 
178   FLATBUFFERS_DELETE_FUNC(ParseDepthGuard(const ParseDepthGuard &));
179   FLATBUFFERS_DELETE_FUNC(ParseDepthGuard &operator=(const ParseDepthGuard &));
180 
181  private:
182   Parser &parser_;
183   const int caller_depth_;
184 };
185 
TypeToIntervalString()186 template<typename T> std::string TypeToIntervalString() {
187   return "[" + NumToString((flatbuffers::numeric_limits<T>::lowest)()) + "; " +
188          NumToString((flatbuffers::numeric_limits<T>::max)()) + "]";
189 }
190 
191 // atot: template version of atoi/atof: convert a string to an instance of T.
192 template<typename T>
atot_scalar(const char * s,T * val,bool_constant<false>)193 bool atot_scalar(const char *s, T *val, bool_constant<false>) {
194   return StringToNumber(s, val);
195 }
196 
197 template<typename T>
atot_scalar(const char * s,T * val,bool_constant<true>)198 bool atot_scalar(const char *s, T *val, bool_constant<true>) {
199   // Normalize NaN parsed from fbs or json to unsigned NaN.
200   if (false == StringToNumber(s, val)) return false;
201   *val = (*val != *val) ? std::fabs(*val) : *val;
202   return true;
203 }
204 
atot(const char * s,Parser & parser,T * val)205 template<typename T> CheckedError atot(const char *s, Parser &parser, T *val) {
206   auto done = atot_scalar(s, val, bool_constant<is_floating_point<T>::value>());
207   if (done) return NoError();
208   if (0 == *val)
209     return parser.Error("invalid number: \"" + std::string(s) + "\"");
210   else
211     return parser.Error("invalid number: \"" + std::string(s) + "\"" +
212                         ", constant does not fit " + TypeToIntervalString<T>());
213 }
214 template<>
atot(const char * s,Parser & parser,Offset<void> * val)215 inline CheckedError atot<Offset<void>>(const char *s, Parser &parser,
216                                        Offset<void> *val) {
217   (void)parser;
218   *val = Offset<void>(atoi(s));
219   return NoError();
220 }
221 
GetFullyQualifiedName(const std::string & name,size_t max_components) const222 std::string Namespace::GetFullyQualifiedName(const std::string &name,
223                                              size_t max_components) const {
224   // Early exit if we don't have a defined namespace.
225   if (components.empty() || !max_components) { return name; }
226   std::string stream_str;
227   for (size_t i = 0; i < std::min(components.size(), max_components); i++) {
228     stream_str += components[i];
229     stream_str += '.';
230   }
231   if (!stream_str.empty()) stream_str.pop_back();
232   if (name.length()) {
233     stream_str += '.';
234     stream_str += name;
235   }
236   return stream_str;
237 }
238 
239 template<typename T>
LookupTableByName(const SymbolTable<T> & table,const std::string & name,const Namespace & current_namespace,size_t skip_top)240 T *LookupTableByName(const SymbolTable<T> &table, const std::string &name,
241                      const Namespace &current_namespace, size_t skip_top) {
242   const auto &components = current_namespace.components;
243   if (table.dict.empty()) return nullptr;
244   if (components.size() < skip_top) return nullptr;
245   const auto N = components.size() - skip_top;
246   std::string full_name;
247   for (size_t i = 0; i < N; i++) {
248     full_name += components[i];
249     full_name += '.';
250   }
251   for (size_t i = N; i > 0; i--) {
252     full_name += name;
253     auto obj = table.Lookup(full_name);
254     if (obj) return obj;
255     auto len = full_name.size() - components[i - 1].size() - 1 - name.size();
256     full_name.resize(len);
257   }
258   FLATBUFFERS_ASSERT(full_name.empty());
259   return table.Lookup(name);  // lookup in global namespace
260 }
261 
262 // Declare tokens we'll use. Single character tokens are represented by their
263 // ascii character code (e.g. '{'), others above 256.
264 // clang-format off
265 #define FLATBUFFERS_GEN_TOKENS(TD) \
266   TD(Eof, 256, "end of file") \
267   TD(StringConstant, 257, "string constant") \
268   TD(IntegerConstant, 258, "integer constant") \
269   TD(FloatConstant, 259, "float constant") \
270   TD(Identifier, 260, "identifier")
271 #ifdef __GNUC__
272 __extension__  // Stop GCC complaining about trailing comma with -Wpendantic.
273 #endif
274 enum {
275   #define FLATBUFFERS_TOKEN(NAME, VALUE, STRING) kToken ## NAME = VALUE,
276     FLATBUFFERS_GEN_TOKENS(FLATBUFFERS_TOKEN)
277   #undef FLATBUFFERS_TOKEN
278 };
279 
TokenToString(int t)280 static std::string TokenToString(int t) {
281   static const char * const tokens[] = {
282     #define FLATBUFFERS_TOKEN(NAME, VALUE, STRING) STRING,
283       FLATBUFFERS_GEN_TOKENS(FLATBUFFERS_TOKEN)
284     #undef FLATBUFFERS_TOKEN
285     #define FLATBUFFERS_TD(ENUM, IDLTYPE, ...) \
286       IDLTYPE,
287       FLATBUFFERS_GEN_TYPES(FLATBUFFERS_TD)
288     #undef FLATBUFFERS_TD
289   };
290   if (t < 256) {  // A single ascii char token.
291     std::string s;
292     s.append(1, static_cast<char>(t));
293     return s;
294   } else {       // Other tokens.
295     return tokens[t - 256];
296   }
297 }
298 // clang-format on
299 
TokenToStringId(int t) const300 std::string Parser::TokenToStringId(int t) const {
301   return t == kTokenIdentifier ? attribute_ : TokenToString(t);
302 }
303 
304 // Parses exactly nibbles worth of hex digits into a number, or error.
ParseHexNum(int nibbles,uint64_t * val)305 CheckedError Parser::ParseHexNum(int nibbles, uint64_t *val) {
306   FLATBUFFERS_ASSERT(nibbles > 0);
307   for (int i = 0; i < nibbles; i++)
308     if (!is_xdigit(cursor_[i]))
309       return Error("escape code must be followed by " + NumToString(nibbles) +
310                    " hex digits");
311   std::string target(cursor_, cursor_ + nibbles);
312   *val = StringToUInt(target.c_str(), 16);
313   cursor_ += nibbles;
314   return NoError();
315 }
316 
SkipByteOrderMark()317 CheckedError Parser::SkipByteOrderMark() {
318   if (static_cast<unsigned char>(*cursor_) != 0xef) return NoError();
319   cursor_++;
320   if (static_cast<unsigned char>(*cursor_) != 0xbb)
321     return Error("invalid utf-8 byte order mark");
322   cursor_++;
323   if (static_cast<unsigned char>(*cursor_) != 0xbf)
324     return Error("invalid utf-8 byte order mark");
325   cursor_++;
326   return NoError();
327 }
328 
IsIdentifierStart(char c)329 static inline bool IsIdentifierStart(char c) {
330   return is_alpha(c) || (c == '_');
331 }
332 
Next()333 CheckedError Parser::Next() {
334   doc_comment_.clear();
335   bool seen_newline = cursor_ == source_;
336   attribute_.clear();
337   attr_is_trivial_ascii_string_ = true;
338   for (;;) {
339     char c = *cursor_++;
340     token_ = c;
341     switch (c) {
342       case '\0':
343         cursor_--;
344         token_ = kTokenEof;
345         return NoError();
346       case ' ':
347       case '\r':
348       case '\t': break;
349       case '\n':
350         MarkNewLine();
351         seen_newline = true;
352         break;
353       case '{':
354       case '}':
355       case '(':
356       case ')':
357       case '[':
358       case ']':
359       case ',':
360       case ':':
361       case ';':
362       case '=': return NoError();
363       case '\"':
364       case '\'': {
365         int unicode_high_surrogate = -1;
366 
367         while (*cursor_ != c) {
368           if (*cursor_ < ' ' && static_cast<signed char>(*cursor_) >= 0)
369             return Error("illegal character in string constant");
370           if (*cursor_ == '\\') {
371             attr_is_trivial_ascii_string_ = false;  // has escape sequence
372             cursor_++;
373             if (unicode_high_surrogate != -1 && *cursor_ != 'u') {
374               return Error(
375                   "illegal Unicode sequence (unpaired high surrogate)");
376             }
377             switch (*cursor_) {
378               case 'n':
379                 attribute_ += '\n';
380                 cursor_++;
381                 break;
382               case 't':
383                 attribute_ += '\t';
384                 cursor_++;
385                 break;
386               case 'r':
387                 attribute_ += '\r';
388                 cursor_++;
389                 break;
390               case 'b':
391                 attribute_ += '\b';
392                 cursor_++;
393                 break;
394               case 'f':
395                 attribute_ += '\f';
396                 cursor_++;
397                 break;
398               case '\"':
399                 attribute_ += '\"';
400                 cursor_++;
401                 break;
402               case '\'':
403                 attribute_ += '\'';
404                 cursor_++;
405                 break;
406               case '\\':
407                 attribute_ += '\\';
408                 cursor_++;
409                 break;
410               case '/':
411                 attribute_ += '/';
412                 cursor_++;
413                 break;
414               case 'x': {  // Not in the JSON standard
415                 cursor_++;
416                 uint64_t val;
417                 ECHECK(ParseHexNum(2, &val));
418                 attribute_ += static_cast<char>(val);
419                 break;
420               }
421               case 'u': {
422                 cursor_++;
423                 uint64_t val;
424                 ECHECK(ParseHexNum(4, &val));
425                 if (val >= 0xD800 && val <= 0xDBFF) {
426                   if (unicode_high_surrogate != -1) {
427                     return Error(
428                         "illegal Unicode sequence (multiple high surrogates)");
429                   } else {
430                     unicode_high_surrogate = static_cast<int>(val);
431                   }
432                 } else if (val >= 0xDC00 && val <= 0xDFFF) {
433                   if (unicode_high_surrogate == -1) {
434                     return Error(
435                         "illegal Unicode sequence (unpaired low surrogate)");
436                   } else {
437                     int code_point = 0x10000 +
438                                      ((unicode_high_surrogate & 0x03FF) << 10) +
439                                      (val & 0x03FF);
440                     ToUTF8(code_point, &attribute_);
441                     unicode_high_surrogate = -1;
442                   }
443                 } else {
444                   if (unicode_high_surrogate != -1) {
445                     return Error(
446                         "illegal Unicode sequence (unpaired high surrogate)");
447                   }
448                   ToUTF8(static_cast<int>(val), &attribute_);
449                 }
450                 break;
451               }
452               default: return Error("unknown escape code in string constant");
453             }
454           } else {  // printable chars + UTF-8 bytes
455             if (unicode_high_surrogate != -1) {
456               return Error(
457                   "illegal Unicode sequence (unpaired high surrogate)");
458             }
459             // reset if non-printable
460             attr_is_trivial_ascii_string_ &=
461                 check_ascii_range(*cursor_, ' ', '~');
462 
463             attribute_ += *cursor_++;
464           }
465         }
466         if (unicode_high_surrogate != -1) {
467           return Error("illegal Unicode sequence (unpaired high surrogate)");
468         }
469         cursor_++;
470         if (!attr_is_trivial_ascii_string_ && !opts.allow_non_utf8 &&
471             !ValidateUTF8(attribute_)) {
472           return Error("illegal UTF-8 sequence");
473         }
474         token_ = kTokenStringConstant;
475         return NoError();
476       }
477       case '/':
478         if (*cursor_ == '/') {
479           const char *start = ++cursor_;
480           while (*cursor_ && *cursor_ != '\n' && *cursor_ != '\r') cursor_++;
481           if (*start == '/') {  // documentation comment
482             if (!seen_newline)
483               return Error(
484                   "a documentation comment should be on a line on its own");
485             doc_comment_.push_back(std::string(start + 1, cursor_));
486           }
487           break;
488         } else if (*cursor_ == '*') {
489           cursor_++;
490           // TODO: make nested.
491           while (*cursor_ != '*' || cursor_[1] != '/') {
492             if (*cursor_ == '\n') MarkNewLine();
493             if (!*cursor_) return Error("end of file in comment");
494             cursor_++;
495           }
496           cursor_ += 2;
497           break;
498         }
499         FLATBUFFERS_FALLTHROUGH();  // else fall thru
500       default:
501         if (IsIdentifierStart(c)) {
502           // Collect all chars of an identifier:
503           const char *start = cursor_ - 1;
504           while (IsIdentifierStart(*cursor_) || is_digit(*cursor_)) cursor_++;
505           attribute_.append(start, cursor_);
506           token_ = kTokenIdentifier;
507           return NoError();
508         }
509 
510         const auto has_sign = (c == '+') || (c == '-');
511         if (has_sign && IsIdentifierStart(*cursor_)) {
512           // '-'/'+' and following identifier - it could be a predefined
513           // constant. Return the sign in token_, see ParseSingleValue.
514           return NoError();
515         }
516 
517         auto dot_lvl =
518             (c == '.') ? 0 : 1;  // dot_lvl==0 <=> exactly one '.' seen
519         if (!dot_lvl && !is_digit(*cursor_)) return NoError();  // enum?
520         // Parser accepts hexadecimal-floating-literal (see C++ 5.13.4).
521         if (is_digit(c) || has_sign || !dot_lvl) {
522           const auto start = cursor_ - 1;
523           auto start_digits = !is_digit(c) ? cursor_ : cursor_ - 1;
524           if (!is_digit(c) && is_digit(*cursor_)) {
525             start_digits = cursor_;  // see digit in cursor_ position
526             c = *cursor_++;
527           }
528           // hex-float can't begind with '.'
529           auto use_hex = dot_lvl && (c == '0') && is_alpha_char(*cursor_, 'X');
530           if (use_hex) start_digits = ++cursor_;  // '0x' is the prefix, skip it
531           // Read an integer number or mantisa of float-point number.
532           do {
533             if (use_hex) {
534               while (is_xdigit(*cursor_)) cursor_++;
535             } else {
536               while (is_digit(*cursor_)) cursor_++;
537             }
538           } while ((*cursor_ == '.') && (++cursor_) && (--dot_lvl >= 0));
539           // Exponent of float-point number.
540           if ((dot_lvl >= 0) && (cursor_ > start_digits)) {
541             // The exponent suffix of hexadecimal float number is mandatory.
542             if (use_hex && !dot_lvl) start_digits = cursor_;
543             if ((use_hex && is_alpha_char(*cursor_, 'P')) ||
544                 is_alpha_char(*cursor_, 'E')) {
545               dot_lvl = 0;  // Emulate dot to signal about float-point number.
546               cursor_++;
547               if (*cursor_ == '+' || *cursor_ == '-') cursor_++;
548               start_digits = cursor_;  // the exponent-part has to have digits
549               // Exponent is decimal integer number
550               while (is_digit(*cursor_)) cursor_++;
551               if (*cursor_ == '.') {
552                 cursor_++;  // If see a dot treat it as part of invalid number.
553                 dot_lvl = -1;  // Fall thru to Error().
554               }
555             }
556           }
557           // Finalize.
558           if ((dot_lvl >= 0) && (cursor_ > start_digits)) {
559             attribute_.append(start, cursor_);
560             token_ = dot_lvl ? kTokenIntegerConstant : kTokenFloatConstant;
561             return NoError();
562           } else {
563             return Error("invalid number: " + std::string(start, cursor_));
564           }
565         }
566         std::string ch;
567         ch = c;
568         if (false == check_ascii_range(c, ' ', '~'))
569           ch = "code: " + NumToString(c);
570         return Error("illegal character: " + ch);
571     }
572   }
573 }
574 
575 // Check if a given token is next.
Is(int t) const576 bool Parser::Is(int t) const { return t == token_; }
577 
IsIdent(const char * id) const578 bool Parser::IsIdent(const char *id) const {
579   return token_ == kTokenIdentifier && attribute_ == id;
580 }
581 
582 // Expect a given token to be next, consume it, or error if not present.
Expect(int t)583 CheckedError Parser::Expect(int t) {
584   if (t != token_) {
585     return Error("expecting: " + TokenToString(t) +
586                  " instead got: " + TokenToStringId(token_));
587   }
588   NEXT();
589   return NoError();
590 }
591 
ParseNamespacing(std::string * id,std::string * last)592 CheckedError Parser::ParseNamespacing(std::string *id, std::string *last) {
593   while (Is('.')) {
594     NEXT();
595     *id += ".";
596     *id += attribute_;
597     if (last) *last = attribute_;
598     EXPECT(kTokenIdentifier);
599   }
600   return NoError();
601 }
602 
LookupEnum(const std::string & id)603 EnumDef *Parser::LookupEnum(const std::string &id) {
604   // Search thru parent namespaces.
605   return LookupTableByName(enums_, id, *current_namespace_, 0);
606 }
607 
LookupStruct(const std::string & id) const608 StructDef *Parser::LookupStruct(const std::string &id) const {
609   auto sd = structs_.Lookup(id);
610   if (sd) sd->refcount++;
611   return sd;
612 }
613 
LookupStructThruParentNamespaces(const std::string & id) const614 StructDef *Parser::LookupStructThruParentNamespaces(
615     const std::string &id) const {
616   auto sd = LookupTableByName(structs_, id, *current_namespace_, 1);
617   if (sd) sd->refcount++;
618   return sd;
619 }
620 
ParseTypeIdent(Type & type)621 CheckedError Parser::ParseTypeIdent(Type &type) {
622   std::string id = attribute_;
623   EXPECT(kTokenIdentifier);
624   ECHECK(ParseNamespacing(&id, nullptr));
625   auto enum_def = LookupEnum(id);
626   if (enum_def) {
627     type = enum_def->underlying_type;
628     if (enum_def->is_union) type.base_type = BASE_TYPE_UNION;
629   } else {
630     type.base_type = BASE_TYPE_STRUCT;
631     type.struct_def = LookupCreateStruct(id);
632   }
633   return NoError();
634 }
635 
636 // Parse any IDL type.
ParseType(Type & type)637 CheckedError Parser::ParseType(Type &type) {
638   if (token_ == kTokenIdentifier) {
639     if (IsIdent("bool")) {
640       type.base_type = BASE_TYPE_BOOL;
641       NEXT();
642     } else if (IsIdent("byte") || IsIdent("int8")) {
643       type.base_type = BASE_TYPE_CHAR;
644       NEXT();
645     } else if (IsIdent("ubyte") || IsIdent("uint8")) {
646       type.base_type = BASE_TYPE_UCHAR;
647       NEXT();
648     } else if (IsIdent("short") || IsIdent("int16")) {
649       type.base_type = BASE_TYPE_SHORT;
650       NEXT();
651     } else if (IsIdent("ushort") || IsIdent("uint16")) {
652       type.base_type = BASE_TYPE_USHORT;
653       NEXT();
654     } else if (IsIdent("int") || IsIdent("int32")) {
655       type.base_type = BASE_TYPE_INT;
656       NEXT();
657     } else if (IsIdent("uint") || IsIdent("uint32")) {
658       type.base_type = BASE_TYPE_UINT;
659       NEXT();
660     } else if (IsIdent("long") || IsIdent("int64")) {
661       type.base_type = BASE_TYPE_LONG;
662       NEXT();
663     } else if (IsIdent("ulong") || IsIdent("uint64")) {
664       type.base_type = BASE_TYPE_ULONG;
665       NEXT();
666     } else if (IsIdent("float") || IsIdent("float32")) {
667       type.base_type = BASE_TYPE_FLOAT;
668       NEXT();
669     } else if (IsIdent("double") || IsIdent("float64")) {
670       type.base_type = BASE_TYPE_DOUBLE;
671       NEXT();
672     } else if (IsIdent("string")) {
673       type.base_type = BASE_TYPE_STRING;
674       NEXT();
675     } else {
676       ECHECK(ParseTypeIdent(type));
677     }
678   } else if (token_ == '[') {
679     ParseDepthGuard depth_guard(this);
680     ECHECK(depth_guard.Check());
681     NEXT();
682     Type subtype;
683     ECHECK(ParseType(subtype));
684     if (IsSeries(subtype)) {
685       // We could support this, but it will complicate things, and it's
686       // easier to work around with a struct around the inner vector.
687       return Error("nested vector types not supported (wrap in table first)");
688     }
689     if (token_ == ':') {
690       NEXT();
691       if (token_ != kTokenIntegerConstant) {
692         return Error("length of fixed-length array must be an integer value");
693       }
694       uint16_t fixed_length = 0;
695       bool check = StringToNumber(attribute_.c_str(), &fixed_length);
696       if (!check || fixed_length < 1) {
697         return Error(
698             "length of fixed-length array must be positive and fit to "
699             "uint16_t type");
700       }
701       type = Type(BASE_TYPE_ARRAY, subtype.struct_def, subtype.enum_def,
702                   fixed_length);
703       NEXT();
704     } else {
705       type = Type(BASE_TYPE_VECTOR, subtype.struct_def, subtype.enum_def);
706     }
707     type.element = subtype.base_type;
708     EXPECT(']');
709   } else {
710     return Error("illegal type syntax");
711   }
712   return NoError();
713 }
714 
AddField(StructDef & struct_def,const std::string & name,const Type & type,FieldDef ** dest)715 CheckedError Parser::AddField(StructDef &struct_def, const std::string &name,
716                               const Type &type, FieldDef **dest) {
717   auto &field = *new FieldDef();
718   field.value.offset =
719       FieldIndexToOffset(static_cast<voffset_t>(struct_def.fields.vec.size()));
720   field.name = name;
721   field.file = struct_def.file;
722   field.value.type = type;
723   if (struct_def.fixed) {  // statically compute the field offset
724     auto size = InlineSize(type);
725     auto alignment = InlineAlignment(type);
726     // structs_ need to have a predictable format, so we need to align to
727     // the largest scalar
728     struct_def.minalign = std::max(struct_def.minalign, alignment);
729     struct_def.PadLastField(alignment);
730     field.value.offset = static_cast<voffset_t>(struct_def.bytesize);
731     struct_def.bytesize += size;
732   }
733   if (struct_def.fields.Add(name, &field))
734     return Error("field already exists: " + name);
735   *dest = &field;
736   return NoError();
737 }
738 
ParseField(StructDef & struct_def)739 CheckedError Parser::ParseField(StructDef &struct_def) {
740   std::string name = attribute_;
741 
742   if (LookupCreateStruct(name, false, false))
743     return Error("field name can not be the same as table/struct name");
744 
745   if (!IsLowerSnakeCase(name)) {
746     Warning("field names should be lowercase snake_case, got: " + name);
747   }
748 
749   std::vector<std::string> dc = doc_comment_;
750   EXPECT(kTokenIdentifier);
751   EXPECT(':');
752   Type type;
753   ECHECK(ParseType(type));
754 
755   if (struct_def.fixed) {
756     auto valid = IsScalar(type.base_type) || IsStruct(type);
757     if (!valid && IsArray(type)) {
758       const auto &elem_type = type.VectorType();
759       valid |= IsScalar(elem_type.base_type) || IsStruct(elem_type);
760     }
761     if (!valid)
762       return Error("structs may contain only scalar or struct fields");
763   }
764 
765   if (!struct_def.fixed && IsArray(type))
766     return Error("fixed-length array in table must be wrapped in struct");
767 
768   if (IsArray(type)) {
769     advanced_features_ |= reflection::AdvancedArrayFeatures;
770     if (!SupportsAdvancedArrayFeatures()) {
771       return Error(
772           "Arrays are not yet supported in all "
773           "the specified programming languages.");
774     }
775   }
776 
777   FieldDef *typefield = nullptr;
778   if (type.base_type == BASE_TYPE_UNION) {
779     // For union fields, add a second auto-generated field to hold the type,
780     // with a special suffix.
781     ECHECK(AddField(struct_def, name + UnionTypeFieldSuffix(),
782                     type.enum_def->underlying_type, &typefield));
783   } else if (IsVector(type) && type.element == BASE_TYPE_UNION) {
784     advanced_features_ |= reflection::AdvancedUnionFeatures;
785     // Only cpp, js and ts supports the union vector feature so far.
786     if (!SupportsAdvancedUnionFeatures()) {
787       return Error(
788           "Vectors of unions are not yet supported in at least one of "
789           "the specified programming languages.");
790     }
791     // For vector of union fields, add a second auto-generated vector field to
792     // hold the types, with a special suffix.
793     Type union_vector(BASE_TYPE_VECTOR, nullptr, type.enum_def);
794     union_vector.element = BASE_TYPE_UTYPE;
795     ECHECK(AddField(struct_def, name + UnionTypeFieldSuffix(), union_vector,
796                     &typefield));
797   }
798 
799   FieldDef *field;
800   ECHECK(AddField(struct_def, name, type, &field));
801 
802   if (token_ == '=') {
803     NEXT();
804     ECHECK(ParseSingleValue(&field->name, field->value, true));
805     if (IsStruct(type) || (struct_def.fixed && field->value.constant != "0"))
806       return Error(
807           "default values are not supported for struct fields, table fields, "
808           "or in structs.");
809     if (IsString(type) || IsVector(type)) {
810       advanced_features_ |= reflection::DefaultVectorsAndStrings;
811       if (field->value.constant != "0" && field->value.constant != "null" &&
812           !SupportsDefaultVectorsAndStrings()) {
813         return Error(
814             "Default values for strings and vectors are not supported in one "
815             "of the specified programming languages");
816       }
817     }
818 
819     if (IsVector(type) && field->value.constant != "0" &&
820         field->value.constant != "[]") {
821       return Error("The only supported default for vectors is `[]`.");
822     }
823   }
824 
825   // Append .0 if the value has not it (skip hex and scientific floats).
826   // This suffix needed for generated C++ code.
827   if (IsFloat(type.base_type)) {
828     auto &text = field->value.constant;
829     FLATBUFFERS_ASSERT(false == text.empty());
830     auto s = text.c_str();
831     while (*s == ' ') s++;
832     if (*s == '-' || *s == '+') s++;
833     // 1) A float constants (nan, inf, pi, etc) is a kind of identifier.
834     // 2) A float number needn't ".0" at the end if it has exponent.
835     if ((false == IsIdentifierStart(*s)) &&
836         (std::string::npos == field->value.constant.find_first_of(".eEpP"))) {
837       field->value.constant += ".0";
838     }
839   }
840 
841   field->doc_comment = dc;
842   ECHECK(ParseMetaData(&field->attributes));
843   field->deprecated = field->attributes.Lookup("deprecated") != nullptr;
844   auto hash_name = field->attributes.Lookup("hash");
845   if (hash_name) {
846     switch ((IsVector(type)) ? type.element : type.base_type) {
847       case BASE_TYPE_SHORT:
848       case BASE_TYPE_USHORT: {
849         if (FindHashFunction16(hash_name->constant.c_str()) == nullptr)
850           return Error("Unknown hashing algorithm for 16 bit types: " +
851                        hash_name->constant);
852         break;
853       }
854       case BASE_TYPE_INT:
855       case BASE_TYPE_UINT: {
856         if (FindHashFunction32(hash_name->constant.c_str()) == nullptr)
857           return Error("Unknown hashing algorithm for 32 bit types: " +
858                        hash_name->constant);
859         break;
860       }
861       case BASE_TYPE_LONG:
862       case BASE_TYPE_ULONG: {
863         if (FindHashFunction64(hash_name->constant.c_str()) == nullptr)
864           return Error("Unknown hashing algorithm for 64 bit types: " +
865                        hash_name->constant);
866         break;
867       }
868       default:
869         return Error(
870             "only short, ushort, int, uint, long and ulong data types support "
871             "hashing.");
872     }
873   }
874 
875   // For historical convenience reasons, string keys are assumed required.
876   // Scalars are kDefault unless otherwise specified.
877   // Nonscalars are kOptional unless required;
878   field->key = field->attributes.Lookup("key") != nullptr;
879   const bool required = field->attributes.Lookup("required") != nullptr ||
880                         (IsString(type) && field->key);
881   const bool default_str_or_vec =
882       ((IsString(type) || IsVector(type)) && field->value.constant != "0");
883   const bool optional = IsScalar(type.base_type)
884                             ? (field->value.constant == "null")
885                             : !(required || default_str_or_vec);
886   if (required && optional) {
887     return Error("Fields cannot be both optional and required.");
888   }
889   field->presence = FieldDef::MakeFieldPresence(optional, required);
890 
891   if (required && (struct_def.fixed || IsScalar(type.base_type))) {
892     return Error("only non-scalar fields in tables may be 'required'");
893   }
894   if (field->key) {
895     if (struct_def.has_key) return Error("only one field may be set as 'key'");
896     struct_def.has_key = true;
897     if (!IsScalar(type.base_type) && !IsString(type)) {
898       return Error("'key' field must be string or scalar type");
899     }
900   }
901 
902   if (field->IsScalarOptional()) {
903     advanced_features_ |= reflection::OptionalScalars;
904     if (type.enum_def && type.enum_def->Lookup("null")) {
905       FLATBUFFERS_ASSERT(IsInteger(type.base_type));
906       return Error(
907           "the default 'null' is reserved for declaring optional scalar "
908           "fields, it conflicts with declaration of enum '" +
909           type.enum_def->name + "'.");
910     }
911     if (field->attributes.Lookup("key")) {
912       return Error(
913           "only a non-optional scalar field can be used as a 'key' field");
914     }
915     if (!SupportsOptionalScalars()) {
916       return Error(
917           "Optional scalars are not yet supported in at least one the of "
918           "the specified programming languages.");
919     }
920   }
921 
922   if (type.enum_def) {
923     // Verify the enum's type and default value.
924     const std::string &constant = field->value.constant;
925     if (type.base_type == BASE_TYPE_UNION) {
926       if (constant != "0") { return Error("Union defaults must be NONE"); }
927     } else if (IsVector(type)) {
928       if (constant != "0" && constant != "[]") {
929         return Error("Vector defaults may only be `[]`.");
930       }
931     } else if (IsArray(type)) {
932       if (constant != "0") {
933         return Error("Array defaults are not supported yet.");
934       }
935     } else {
936       if (!IsInteger(type.base_type)) {
937         return Error("Enums must have integer base types");
938       }
939       // Optional and bitflags enums may have default constants that are not
940       // their specified variants.
941       if (!field->IsOptional() &&
942           type.enum_def->attributes.Lookup("bit_flags") == nullptr) {
943         if (type.enum_def->FindByValue(constant) == nullptr) {
944           return Error("default value of `" + constant + "` for " + "field `" +
945                        name + "` is not part of enum `" + type.enum_def->name +
946                        "`.");
947         }
948       }
949     }
950   }
951 
952   if (field->deprecated && struct_def.fixed)
953     return Error("can't deprecate fields in a struct");
954 
955   auto cpp_type = field->attributes.Lookup("cpp_type");
956   if (cpp_type) {
957     if (!hash_name)
958       return Error("cpp_type can only be used with a hashed field");
959     /// forcing cpp_ptr_type to 'naked' if unset
960     auto cpp_ptr_type = field->attributes.Lookup("cpp_ptr_type");
961     if (!cpp_ptr_type) {
962       auto val = new Value();
963       val->type = cpp_type->type;
964       val->constant = "naked";
965       field->attributes.Add("cpp_ptr_type", val);
966     }
967   }
968 
969   field->shared = field->attributes.Lookup("shared") != nullptr;
970   if (field->shared && field->value.type.base_type != BASE_TYPE_STRING)
971     return Error("shared can only be defined on strings");
972 
973   auto field_native_custom_alloc =
974       field->attributes.Lookup("native_custom_alloc");
975   if (field_native_custom_alloc)
976     return Error(
977         "native_custom_alloc can only be used with a table or struct "
978         "definition");
979 
980   field->native_inline = field->attributes.Lookup("native_inline") != nullptr;
981   if (field->native_inline && !IsStruct(field->value.type))
982     return Error("native_inline can only be defined on structs");
983 
984   auto nested = field->attributes.Lookup("nested_flatbuffer");
985   if (nested) {
986     if (nested->type.base_type != BASE_TYPE_STRING)
987       return Error(
988           "nested_flatbuffer attribute must be a string (the root type)");
989     if (type.base_type != BASE_TYPE_VECTOR || type.element != BASE_TYPE_UCHAR)
990       return Error(
991           "nested_flatbuffer attribute may only apply to a vector of ubyte");
992     // This will cause an error if the root type of the nested flatbuffer
993     // wasn't defined elsewhere.
994     field->nested_flatbuffer = LookupCreateStruct(nested->constant);
995   }
996 
997   if (field->attributes.Lookup("flexbuffer")) {
998     field->flexbuffer = true;
999     uses_flexbuffers_ = true;
1000     if (type.base_type != BASE_TYPE_VECTOR || type.element != BASE_TYPE_UCHAR)
1001       return Error("flexbuffer attribute may only apply to a vector of ubyte");
1002   }
1003 
1004   if (typefield) {
1005     if (!IsScalar(typefield->value.type.base_type)) {
1006       // this is a union vector field
1007       typefield->presence = field->presence;
1008     }
1009     // If this field is a union, and it has a manually assigned id,
1010     // the automatically added type field should have an id as well (of N - 1).
1011     auto attr = field->attributes.Lookup("id");
1012     if (attr) {
1013       const auto &id_str = attr->constant;
1014       voffset_t id = 0;
1015       const auto done = !atot(id_str.c_str(), *this, &id).Check();
1016       if (done && id > 0) {
1017         auto val = new Value();
1018         val->type = attr->type;
1019         val->constant = NumToString(id - 1);
1020         typefield->attributes.Add("id", val);
1021       } else {
1022         return Error(
1023             "a union type effectively adds two fields with non-negative ids, "
1024             "its id must be that of the second field (the first field is "
1025             "the type field and not explicitly declared in the schema);\n"
1026             "field: " +
1027             field->name + ", id: " + id_str);
1028       }
1029     }
1030     // if this field is a union that is deprecated,
1031     // the automatically added type field should be deprecated as well
1032     if (field->deprecated) { typefield->deprecated = true; }
1033   }
1034 
1035   EXPECT(';');
1036   return NoError();
1037 }
1038 
ParseString(Value & val,bool use_string_pooling)1039 CheckedError Parser::ParseString(Value &val, bool use_string_pooling) {
1040   auto s = attribute_;
1041   EXPECT(kTokenStringConstant);
1042   if (use_string_pooling) {
1043     val.constant = NumToString(builder_.CreateSharedString(s).o);
1044   } else {
1045     val.constant = NumToString(builder_.CreateString(s).o);
1046   }
1047   return NoError();
1048 }
1049 
ParseComma()1050 CheckedError Parser::ParseComma() {
1051   if (!opts.protobuf_ascii_alike) EXPECT(',');
1052   return NoError();
1053 }
1054 
ParseAnyValue(Value & val,FieldDef * field,size_t parent_fieldn,const StructDef * parent_struct_def,uoffset_t count,bool inside_vector)1055 CheckedError Parser::ParseAnyValue(Value &val, FieldDef *field,
1056                                    size_t parent_fieldn,
1057                                    const StructDef *parent_struct_def,
1058                                    uoffset_t count, bool inside_vector) {
1059   switch (val.type.base_type) {
1060     case BASE_TYPE_UNION: {
1061       FLATBUFFERS_ASSERT(field);
1062       std::string constant;
1063       Vector<uint8_t> *vector_of_union_types = nullptr;
1064       // Find corresponding type field we may have already parsed.
1065       for (auto elem = field_stack_.rbegin() + count;
1066            elem != field_stack_.rbegin() + parent_fieldn + count; ++elem) {
1067         auto &type = elem->second->value.type;
1068         if (type.enum_def == val.type.enum_def) {
1069           if (inside_vector) {
1070             if (IsVector(type) && type.element == BASE_TYPE_UTYPE) {
1071               // Vector of union type field.
1072               uoffset_t offset;
1073               ECHECK(atot(elem->first.constant.c_str(), *this, &offset));
1074               vector_of_union_types = reinterpret_cast<Vector<uint8_t> *>(
1075                   builder_.GetCurrentBufferPointer() + builder_.GetSize() -
1076                   offset);
1077               break;
1078             }
1079           } else {
1080             if (type.base_type == BASE_TYPE_UTYPE) {
1081               // Union type field.
1082               constant = elem->first.constant;
1083               break;
1084             }
1085           }
1086         }
1087       }
1088       if (constant.empty() && !inside_vector) {
1089         // We haven't seen the type field yet. Sadly a lot of JSON writers
1090         // output these in alphabetical order, meaning it comes after this
1091         // value. So we scan past the value to find it, then come back here.
1092         // We currently don't do this for vectors of unions because the
1093         // scanning/serialization logic would get very complicated.
1094         auto type_name = field->name + UnionTypeFieldSuffix();
1095         FLATBUFFERS_ASSERT(parent_struct_def);
1096         auto type_field = parent_struct_def->fields.Lookup(type_name);
1097         FLATBUFFERS_ASSERT(type_field);  // Guaranteed by ParseField().
1098         // Remember where we are in the source file, so we can come back here.
1099         auto backup = *static_cast<ParserState *>(this);
1100         ECHECK(SkipAnyJsonValue());  // The table.
1101         ECHECK(ParseComma());
1102         auto next_name = attribute_;
1103         if (Is(kTokenStringConstant)) {
1104           NEXT();
1105         } else {
1106           EXPECT(kTokenIdentifier);
1107         }
1108         if (next_name == type_name) {
1109           EXPECT(':');
1110           ParseDepthGuard depth_guard(this);
1111           ECHECK(depth_guard.Check());
1112           Value type_val = type_field->value;
1113           ECHECK(ParseAnyValue(type_val, type_field, 0, nullptr, 0));
1114           constant = type_val.constant;
1115           // Got the information we needed, now rewind:
1116           *static_cast<ParserState *>(this) = backup;
1117         }
1118       }
1119       if (constant.empty() && !vector_of_union_types) {
1120         return Error("missing type field for this union value: " + field->name);
1121       }
1122       uint8_t enum_idx;
1123       if (vector_of_union_types) {
1124         enum_idx = vector_of_union_types->Get(count);
1125       } else {
1126         ECHECK(atot(constant.c_str(), *this, &enum_idx));
1127       }
1128       auto enum_val = val.type.enum_def->ReverseLookup(enum_idx, true);
1129       if (!enum_val) return Error("illegal type id for: " + field->name);
1130       if (enum_val->union_type.base_type == BASE_TYPE_STRUCT) {
1131         ECHECK(ParseTable(*enum_val->union_type.struct_def, &val.constant,
1132                           nullptr));
1133         if (enum_val->union_type.struct_def->fixed) {
1134           // All BASE_TYPE_UNION values are offsets, so turn this into one.
1135           SerializeStruct(*enum_val->union_type.struct_def, val);
1136           builder_.ClearOffsets();
1137           val.constant = NumToString(builder_.GetSize());
1138         }
1139       } else if (IsString(enum_val->union_type)) {
1140         ECHECK(ParseString(val, field->shared));
1141       } else {
1142         FLATBUFFERS_ASSERT(false);
1143       }
1144       break;
1145     }
1146     case BASE_TYPE_STRUCT:
1147       ECHECK(ParseTable(*val.type.struct_def, &val.constant, nullptr));
1148       break;
1149     case BASE_TYPE_STRING: {
1150       ECHECK(ParseString(val, field->shared));
1151       break;
1152     }
1153     case BASE_TYPE_VECTOR: {
1154       uoffset_t off;
1155       ECHECK(ParseVector(val.type.VectorType(), &off, field, parent_fieldn));
1156       val.constant = NumToString(off);
1157       break;
1158     }
1159     case BASE_TYPE_ARRAY: {
1160       ECHECK(ParseArray(val));
1161       break;
1162     }
1163     case BASE_TYPE_INT:
1164     case BASE_TYPE_UINT:
1165     case BASE_TYPE_LONG:
1166     case BASE_TYPE_ULONG: {
1167       if (field && field->attributes.Lookup("hash") &&
1168           (token_ == kTokenIdentifier || token_ == kTokenStringConstant)) {
1169         ECHECK(ParseHash(val, field));
1170       } else {
1171         ECHECK(ParseSingleValue(field ? &field->name : nullptr, val, false));
1172       }
1173       break;
1174     }
1175     default:
1176       ECHECK(ParseSingleValue(field ? &field->name : nullptr, val, false));
1177       break;
1178   }
1179   return NoError();
1180 }
1181 
SerializeStruct(const StructDef & struct_def,const Value & val)1182 void Parser::SerializeStruct(const StructDef &struct_def, const Value &val) {
1183   SerializeStruct(builder_, struct_def, val);
1184 }
1185 
SerializeStruct(FlatBufferBuilder & builder,const StructDef & struct_def,const Value & val)1186 void Parser::SerializeStruct(FlatBufferBuilder &builder,
1187                              const StructDef &struct_def, const Value &val) {
1188   FLATBUFFERS_ASSERT(val.constant.length() == struct_def.bytesize);
1189   builder.Align(struct_def.minalign);
1190   builder.PushBytes(reinterpret_cast<const uint8_t *>(val.constant.c_str()),
1191                     struct_def.bytesize);
1192   builder.AddStructOffset(val.offset, builder.GetSize());
1193 }
1194 
1195 template<typename F>
ParseTableDelimiters(size_t & fieldn,const StructDef * struct_def,F body)1196 CheckedError Parser::ParseTableDelimiters(size_t &fieldn,
1197                                           const StructDef *struct_def, F body) {
1198   // We allow tables both as JSON object{ .. } with field names
1199   // or vector[..] with all fields in order
1200   char terminator = '}';
1201   bool is_nested_vector = struct_def && Is('[');
1202   if (is_nested_vector) {
1203     NEXT();
1204     terminator = ']';
1205   } else {
1206     EXPECT('{');
1207   }
1208   for (;;) {
1209     if ((!opts.strict_json || !fieldn) && Is(terminator)) break;
1210     std::string name;
1211     if (is_nested_vector) {
1212       if (fieldn >= struct_def->fields.vec.size()) {
1213         return Error("too many unnamed fields in nested array");
1214       }
1215       name = struct_def->fields.vec[fieldn]->name;
1216     } else {
1217       name = attribute_;
1218       if (Is(kTokenStringConstant)) {
1219         NEXT();
1220       } else {
1221         EXPECT(opts.strict_json ? kTokenStringConstant : kTokenIdentifier);
1222       }
1223       if (!opts.protobuf_ascii_alike || !(Is('{') || Is('['))) EXPECT(':');
1224     }
1225     ECHECK(body(name, fieldn, struct_def));
1226     if (Is(terminator)) break;
1227     ECHECK(ParseComma());
1228   }
1229   NEXT();
1230   if (is_nested_vector && fieldn != struct_def->fields.vec.size()) {
1231     return Error("wrong number of unnamed fields in table vector");
1232   }
1233   return NoError();
1234 }
1235 
ParseTable(const StructDef & struct_def,std::string * value,uoffset_t * ovalue)1236 CheckedError Parser::ParseTable(const StructDef &struct_def, std::string *value,
1237                                 uoffset_t *ovalue) {
1238   ParseDepthGuard depth_guard(this);
1239   ECHECK(depth_guard.Check());
1240 
1241   size_t fieldn_outer = 0;
1242   auto err = ParseTableDelimiters(
1243       fieldn_outer, &struct_def,
1244       [&](const std::string &name, size_t &fieldn,
1245           const StructDef *struct_def_inner) -> CheckedError {
1246         if (name == "$schema") {
1247           ECHECK(Expect(kTokenStringConstant));
1248           return NoError();
1249         }
1250         auto field = struct_def_inner->fields.Lookup(name);
1251         if (!field) {
1252           if (!opts.skip_unexpected_fields_in_json) {
1253             return Error("unknown field: " + name);
1254           } else {
1255             ECHECK(SkipAnyJsonValue());
1256           }
1257         } else {
1258           if (IsIdent("null") && !IsScalar(field->value.type.base_type)) {
1259             ECHECK(Next());  // Ignore this field.
1260           } else {
1261             Value val = field->value;
1262             if (field->flexbuffer) {
1263               flexbuffers::Builder builder(1024,
1264                                            flexbuffers::BUILDER_FLAG_SHARE_ALL);
1265               ECHECK(ParseFlexBufferValue(&builder));
1266               builder.Finish();
1267               // Force alignment for nested flexbuffer
1268               builder_.ForceVectorAlignment(builder.GetSize(), sizeof(uint8_t),
1269                                             sizeof(largest_scalar_t));
1270               auto off = builder_.CreateVector(builder.GetBuffer());
1271               val.constant = NumToString(off.o);
1272             } else if (field->nested_flatbuffer) {
1273               ECHECK(
1274                   ParseNestedFlatbuffer(val, field, fieldn, struct_def_inner));
1275             } else {
1276               ECHECK(ParseAnyValue(val, field, fieldn, struct_def_inner, 0));
1277             }
1278             // Hardcoded insertion-sort with error-check.
1279             // If fields are specified in order, then this loop exits
1280             // immediately.
1281             auto elem = field_stack_.rbegin();
1282             for (; elem != field_stack_.rbegin() + fieldn; ++elem) {
1283               auto existing_field = elem->second;
1284               if (existing_field == field)
1285                 return Error("field set more than once: " + field->name);
1286               if (existing_field->value.offset < field->value.offset) break;
1287             }
1288             // Note: elem points to before the insertion point, thus .base()
1289             // points to the correct spot.
1290             field_stack_.insert(elem.base(), std::make_pair(val, field));
1291             fieldn++;
1292           }
1293         }
1294         return NoError();
1295       });
1296   ECHECK(err);
1297 
1298   // Check if all required fields are parsed.
1299   for (auto field_it = struct_def.fields.vec.begin();
1300        field_it != struct_def.fields.vec.end(); ++field_it) {
1301     auto required_field = *field_it;
1302     if (!required_field->IsRequired()) { continue; }
1303     bool found = false;
1304     for (auto pf_it = field_stack_.end() - fieldn_outer;
1305          pf_it != field_stack_.end(); ++pf_it) {
1306       auto parsed_field = pf_it->second;
1307       if (parsed_field == required_field) {
1308         found = true;
1309         break;
1310       }
1311     }
1312     if (!found) {
1313       return Error("required field is missing: " + required_field->name +
1314                    " in " + struct_def.name);
1315     }
1316   }
1317 
1318   if (struct_def.fixed && fieldn_outer != struct_def.fields.vec.size())
1319     return Error("struct: wrong number of initializers: " + struct_def.name);
1320 
1321   auto start = struct_def.fixed ? builder_.StartStruct(struct_def.minalign)
1322                                 : builder_.StartTable();
1323 
1324   for (size_t size = struct_def.sortbysize ? sizeof(largest_scalar_t) : 1; size;
1325        size /= 2) {
1326     // Go through elements in reverse, since we're building the data backwards.
1327     for (auto it = field_stack_.rbegin();
1328          it != field_stack_.rbegin() + fieldn_outer; ++it) {
1329       auto &field_value = it->first;
1330       auto field = it->second;
1331       if (!struct_def.sortbysize ||
1332           size == SizeOf(field_value.type.base_type)) {
1333         switch (field_value.type.base_type) {
1334           // clang-format off
1335           #define FLATBUFFERS_TD(ENUM, IDLTYPE, CTYPE, ...) \
1336             case BASE_TYPE_ ## ENUM: \
1337               builder_.Pad(field->padding); \
1338               if (struct_def.fixed) { \
1339                 CTYPE val; \
1340                 ECHECK(atot(field_value.constant.c_str(), *this, &val)); \
1341                 builder_.PushElement(val); \
1342               } else { \
1343                 CTYPE val, valdef; \
1344                 ECHECK(atot(field_value.constant.c_str(), *this, &val)); \
1345                 ECHECK(atot(field->value.constant.c_str(), *this, &valdef)); \
1346                 builder_.AddElement(field_value.offset, val, valdef); \
1347               } \
1348               break;
1349             FLATBUFFERS_GEN_TYPES_SCALAR(FLATBUFFERS_TD)
1350           #undef FLATBUFFERS_TD
1351           #define FLATBUFFERS_TD(ENUM, IDLTYPE, CTYPE, ...) \
1352             case BASE_TYPE_ ## ENUM: \
1353               builder_.Pad(field->padding); \
1354               if (IsStruct(field->value.type)) { \
1355                 SerializeStruct(*field->value.type.struct_def, field_value); \
1356               } else { \
1357                 CTYPE val; \
1358                 ECHECK(atot(field_value.constant.c_str(), *this, &val)); \
1359                 builder_.AddOffset(field_value.offset, val); \
1360               } \
1361               break;
1362             FLATBUFFERS_GEN_TYPES_POINTER(FLATBUFFERS_TD)
1363           #undef FLATBUFFERS_TD
1364             case BASE_TYPE_ARRAY:
1365               builder_.Pad(field->padding);
1366               builder_.PushBytes(
1367                 reinterpret_cast<const uint8_t*>(field_value.constant.c_str()),
1368                 InlineSize(field_value.type));
1369               break;
1370             // clang-format on
1371         }
1372       }
1373     }
1374   }
1375   for (size_t i = 0; i < fieldn_outer; i++) field_stack_.pop_back();
1376 
1377   if (struct_def.fixed) {
1378     builder_.ClearOffsets();
1379     builder_.EndStruct();
1380     FLATBUFFERS_ASSERT(value);
1381     // Temporarily store this struct in the value string, since it is to
1382     // be serialized in-place elsewhere.
1383     value->assign(
1384         reinterpret_cast<const char *>(builder_.GetCurrentBufferPointer()),
1385         struct_def.bytesize);
1386     builder_.PopBytes(struct_def.bytesize);
1387     FLATBUFFERS_ASSERT(!ovalue);
1388   } else {
1389     auto val = builder_.EndTable(start);
1390     if (ovalue) *ovalue = val;
1391     if (value) *value = NumToString(val);
1392   }
1393   return NoError();
1394 }
1395 
1396 template<typename F>
ParseVectorDelimiters(uoffset_t & count,F body)1397 CheckedError Parser::ParseVectorDelimiters(uoffset_t &count, F body) {
1398   EXPECT('[');
1399   for (;;) {
1400     if ((!opts.strict_json || !count) && Is(']')) break;
1401     ECHECK(body(count));
1402     count++;
1403     if (Is(']')) break;
1404     ECHECK(ParseComma());
1405   }
1406   NEXT();
1407   return NoError();
1408 }
1409 
CompareSerializedScalars(const uint8_t * a,const uint8_t * b,const FieldDef & key)1410 static bool CompareSerializedScalars(const uint8_t *a, const uint8_t *b,
1411                                      const FieldDef &key) {
1412   switch (key.value.type.base_type) {
1413 #define FLATBUFFERS_TD(ENUM, IDLTYPE, CTYPE, ...)                       \
1414   case BASE_TYPE_##ENUM: {                                              \
1415     CTYPE def = static_cast<CTYPE>(0);                                  \
1416     if (!a || !b) { StringToNumber(key.value.constant.c_str(), &def); } \
1417     const auto av = a ? ReadScalar<CTYPE>(a) : def;                     \
1418     const auto bv = b ? ReadScalar<CTYPE>(b) : def;                     \
1419     return av < bv;                                                     \
1420   }
1421     FLATBUFFERS_GEN_TYPES_SCALAR(FLATBUFFERS_TD)
1422 #undef FLATBUFFERS_TD
1423     default: {
1424       FLATBUFFERS_ASSERT(false && "scalar type expected");
1425       return false;
1426     }
1427   }
1428 }
1429 
CompareTablesByScalarKey(const Offset<Table> * _a,const Offset<Table> * _b,const FieldDef & key)1430 static bool CompareTablesByScalarKey(const Offset<Table> *_a,
1431                                      const Offset<Table> *_b,
1432                                      const FieldDef &key) {
1433   const voffset_t offset = key.value.offset;
1434   // Indirect offset pointer to table pointer.
1435   auto a = reinterpret_cast<const uint8_t *>(_a) + ReadScalar<uoffset_t>(_a);
1436   auto b = reinterpret_cast<const uint8_t *>(_b) + ReadScalar<uoffset_t>(_b);
1437   // Fetch field address from table.
1438   a = reinterpret_cast<const Table *>(a)->GetAddressOf(offset);
1439   b = reinterpret_cast<const Table *>(b)->GetAddressOf(offset);
1440   return CompareSerializedScalars(a, b, key);
1441 }
1442 
CompareTablesByStringKey(const Offset<Table> * _a,const Offset<Table> * _b,const FieldDef & key)1443 static bool CompareTablesByStringKey(const Offset<Table> *_a,
1444                                      const Offset<Table> *_b,
1445                                      const FieldDef &key) {
1446   const voffset_t offset = key.value.offset;
1447   // Indirect offset pointer to table pointer.
1448   auto a = reinterpret_cast<const uint8_t *>(_a) + ReadScalar<uoffset_t>(_a);
1449   auto b = reinterpret_cast<const uint8_t *>(_b) + ReadScalar<uoffset_t>(_b);
1450   // Fetch field address from table.
1451   a = reinterpret_cast<const Table *>(a)->GetAddressOf(offset);
1452   b = reinterpret_cast<const Table *>(b)->GetAddressOf(offset);
1453   if (a && b) {
1454     // Indirect offset pointer to string pointer.
1455     a += ReadScalar<uoffset_t>(a);
1456     b += ReadScalar<uoffset_t>(b);
1457     return *reinterpret_cast<const String *>(a) <
1458            *reinterpret_cast<const String *>(b);
1459   } else {
1460     return a ? true : false;
1461   }
1462 }
1463 
SwapSerializedTables(Offset<Table> * a,Offset<Table> * b)1464 static void SwapSerializedTables(Offset<Table> *a, Offset<Table> *b) {
1465   // These are serialized offsets, so are relative where they are
1466   // stored in memory, so compute the distance between these pointers:
1467   ptrdiff_t diff = (b - a) * sizeof(Offset<Table>);
1468   FLATBUFFERS_ASSERT(diff >= 0);  // Guaranteed by SimpleQsort.
1469   auto udiff = static_cast<uoffset_t>(diff);
1470   a->o = EndianScalar(ReadScalar<uoffset_t>(a) - udiff);
1471   b->o = EndianScalar(ReadScalar<uoffset_t>(b) + udiff);
1472   std::swap(*a, *b);
1473 }
1474 
1475 // See below for why we need our own sort :(
1476 template<typename T, typename F, typename S>
SimpleQsort(T * begin,T * end,size_t width,F comparator,S swapper)1477 void SimpleQsort(T *begin, T *end, size_t width, F comparator, S swapper) {
1478   if (end - begin <= static_cast<ptrdiff_t>(width)) return;
1479   auto l = begin + width;
1480   auto r = end;
1481   while (l < r) {
1482     if (comparator(begin, l)) {
1483       r -= width;
1484       swapper(l, r);
1485     } else {
1486       l += width;
1487     }
1488   }
1489   l -= width;
1490   swapper(begin, l);
1491   SimpleQsort(begin, l, width, comparator, swapper);
1492   SimpleQsort(r, end, width, comparator, swapper);
1493 }
1494 
ParseAlignAttribute(const std::string & align_constant,size_t min_align,size_t * align)1495 CheckedError Parser::ParseAlignAttribute(const std::string &align_constant,
1496                                          size_t min_align, size_t *align) {
1497   // Use uint8_t to avoid problems with size_t==`unsigned long` on LP64.
1498   uint8_t align_value;
1499   if (StringToNumber(align_constant.c_str(), &align_value) &&
1500       VerifyAlignmentRequirements(static_cast<size_t>(align_value),
1501                                   min_align)) {
1502     *align = align_value;
1503     return NoError();
1504   }
1505   return Error("unexpected force_align value '" + align_constant +
1506                "', alignment must be a power of two integer ranging from the "
1507                "type\'s natural alignment " +
1508                NumToString(min_align) + " to " +
1509                NumToString(FLATBUFFERS_MAX_ALIGNMENT));
1510 }
1511 
ParseVector(const Type & type,uoffset_t * ovalue,FieldDef * field,size_t fieldn)1512 CheckedError Parser::ParseVector(const Type &type, uoffset_t *ovalue,
1513                                  FieldDef *field, size_t fieldn) {
1514   uoffset_t count = 0;
1515   auto err = ParseVectorDelimiters(count, [&](uoffset_t &) -> CheckedError {
1516     Value val;
1517     val.type = type;
1518     ECHECK(ParseAnyValue(val, field, fieldn, nullptr, count, true));
1519     field_stack_.push_back(std::make_pair(val, nullptr));
1520     return NoError();
1521   });
1522   ECHECK(err);
1523 
1524   const size_t len = count * InlineSize(type) / InlineAlignment(type);
1525   const size_t elemsize = InlineAlignment(type);
1526   const auto force_align = field->attributes.Lookup("force_align");
1527   if (force_align) {
1528     size_t align;
1529     ECHECK(ParseAlignAttribute(force_align->constant, 1, &align));
1530     if (align > 1) { builder_.ForceVectorAlignment(len, elemsize, align); }
1531   }
1532 
1533   builder_.StartVector(len, elemsize);
1534   for (uoffset_t i = 0; i < count; i++) {
1535     // start at the back, since we're building the data backwards.
1536     auto &val = field_stack_.back().first;
1537     switch (val.type.base_type) {
1538       // clang-format off
1539       #define FLATBUFFERS_TD(ENUM, IDLTYPE, CTYPE,...) \
1540         case BASE_TYPE_ ## ENUM: \
1541           if (IsStruct(val.type)) SerializeStruct(*val.type.struct_def, val); \
1542           else { \
1543              CTYPE elem; \
1544              ECHECK(atot(val.constant.c_str(), *this, &elem)); \
1545              builder_.PushElement(elem); \
1546           } \
1547           break;
1548         FLATBUFFERS_GEN_TYPES(FLATBUFFERS_TD)
1549       #undef FLATBUFFERS_TD
1550       // clang-format on
1551     }
1552     field_stack_.pop_back();
1553   }
1554 
1555   builder_.ClearOffsets();
1556   *ovalue = builder_.EndVector(count);
1557 
1558   if (type.base_type == BASE_TYPE_STRUCT && type.struct_def->has_key) {
1559     // We should sort this vector. Find the key first.
1560     const FieldDef *key = nullptr;
1561     for (auto it = type.struct_def->fields.vec.begin();
1562          it != type.struct_def->fields.vec.end(); ++it) {
1563       if ((*it)->key) {
1564         key = (*it);
1565         break;
1566       }
1567     }
1568     FLATBUFFERS_ASSERT(key);
1569     // Now sort it.
1570     // We can't use std::sort because for structs the size is not known at
1571     // compile time, and for tables our iterators dereference offsets, so can't
1572     // be used to swap elements.
1573     // And we can't use C qsort either, since that would force use to use
1574     // globals, making parsing thread-unsafe.
1575     // So for now, we use SimpleQsort above.
1576     // TODO: replace with something better, preferably not recursive.
1577 
1578     if (type.struct_def->fixed) {
1579       const voffset_t offset = key->value.offset;
1580       const size_t struct_size = type.struct_def->bytesize;
1581       auto v =
1582           reinterpret_cast<VectorOfAny *>(builder_.GetCurrentBufferPointer());
1583       SimpleQsort<uint8_t>(
1584           v->Data(), v->Data() + v->size() * type.struct_def->bytesize,
1585           type.struct_def->bytesize,
1586           [offset, key](const uint8_t *a, const uint8_t *b) -> bool {
1587             return CompareSerializedScalars(a + offset, b + offset, *key);
1588           },
1589           [struct_size](uint8_t *a, uint8_t *b) {
1590             // FIXME: faster?
1591             for (size_t i = 0; i < struct_size; i++) { std::swap(a[i], b[i]); }
1592           });
1593     } else {
1594       auto v = reinterpret_cast<Vector<Offset<Table>> *>(
1595           builder_.GetCurrentBufferPointer());
1596       // Here also can't use std::sort. We do have an iterator type for it,
1597       // but it is non-standard as it will dereference the offsets, and thus
1598       // can't be used to swap elements.
1599       if (key->value.type.base_type == BASE_TYPE_STRING) {
1600         SimpleQsort<Offset<Table>>(
1601             v->data(), v->data() + v->size(), 1,
1602             [key](const Offset<Table> *_a, const Offset<Table> *_b) -> bool {
1603               return CompareTablesByStringKey(_a, _b, *key);
1604             },
1605             SwapSerializedTables);
1606       } else {
1607         SimpleQsort<Offset<Table>>(
1608             v->data(), v->data() + v->size(), 1,
1609             [key](const Offset<Table> *_a, const Offset<Table> *_b) -> bool {
1610               return CompareTablesByScalarKey(_a, _b, *key);
1611             },
1612             SwapSerializedTables);
1613       }
1614     }
1615   }
1616   return NoError();
1617 }
1618 
ParseArray(Value & array)1619 CheckedError Parser::ParseArray(Value &array) {
1620   std::vector<Value> stack;
1621   FlatBufferBuilder builder;
1622   const auto &type = array.type.VectorType();
1623   auto length = array.type.fixed_length;
1624   uoffset_t count = 0;
1625   auto err = ParseVectorDelimiters(count, [&](uoffset_t &) -> CheckedError {
1626     vector_emplace_back(&stack, Value());
1627     auto &val = stack.back();
1628     val.type = type;
1629     if (IsStruct(type)) {
1630       ECHECK(ParseTable(*val.type.struct_def, &val.constant, nullptr));
1631     } else {
1632       ECHECK(ParseSingleValue(nullptr, val, false));
1633     }
1634     return NoError();
1635   });
1636   ECHECK(err);
1637   if (length != count) return Error("Fixed-length array size is incorrect.");
1638 
1639   for (auto it = stack.rbegin(); it != stack.rend(); ++it) {
1640     auto &val = *it;
1641     // clang-format off
1642     switch (val.type.base_type) {
1643       #define FLATBUFFERS_TD(ENUM, IDLTYPE, CTYPE, ...) \
1644         case BASE_TYPE_ ## ENUM: \
1645           if (IsStruct(val.type)) { \
1646             SerializeStruct(builder, *val.type.struct_def, val); \
1647           } else { \
1648             CTYPE elem; \
1649             ECHECK(atot(val.constant.c_str(), *this, &elem)); \
1650             builder.PushElement(elem); \
1651           } \
1652         break;
1653         FLATBUFFERS_GEN_TYPES(FLATBUFFERS_TD)
1654       #undef FLATBUFFERS_TD
1655       default: FLATBUFFERS_ASSERT(0);
1656     }
1657     // clang-format on
1658   }
1659 
1660   array.constant.assign(
1661       reinterpret_cast<const char *>(builder.GetCurrentBufferPointer()),
1662       InlineSize(array.type));
1663   return NoError();
1664 }
1665 
ParseNestedFlatbuffer(Value & val,FieldDef * field,size_t fieldn,const StructDef * parent_struct_def)1666 CheckedError Parser::ParseNestedFlatbuffer(Value &val, FieldDef *field,
1667                                            size_t fieldn,
1668                                            const StructDef *parent_struct_def) {
1669   if (token_ == '[') {  // backwards compat for 'legacy' ubyte buffers
1670     ECHECK(ParseAnyValue(val, field, fieldn, parent_struct_def, 0));
1671   } else {
1672     auto cursor_at_value_begin = cursor_;
1673     ECHECK(SkipAnyJsonValue());
1674     std::string substring(cursor_at_value_begin - 1, cursor_ - 1);
1675 
1676     // Create and initialize new parser
1677     Parser nested_parser;
1678     FLATBUFFERS_ASSERT(field->nested_flatbuffer);
1679     nested_parser.root_struct_def_ = field->nested_flatbuffer;
1680     nested_parser.enums_ = enums_;
1681     nested_parser.opts = opts;
1682     nested_parser.uses_flexbuffers_ = uses_flexbuffers_;
1683     nested_parser.parse_depth_counter_ = parse_depth_counter_;
1684     // Parse JSON substring into new flatbuffer builder using nested_parser
1685     bool ok = nested_parser.Parse(substring.c_str(), nullptr, nullptr);
1686 
1687     // Clean nested_parser to avoid deleting the elements in
1688     // the SymbolTables on destruction
1689     nested_parser.enums_.dict.clear();
1690     nested_parser.enums_.vec.clear();
1691 
1692     if (!ok) { ECHECK(Error(nested_parser.error_)); }
1693     // Force alignment for nested flatbuffer
1694     builder_.ForceVectorAlignment(
1695         nested_parser.builder_.GetSize(), sizeof(uint8_t),
1696         nested_parser.builder_.GetBufferMinAlignment());
1697 
1698     auto off = builder_.CreateVector(nested_parser.builder_.GetBufferPointer(),
1699                                      nested_parser.builder_.GetSize());
1700     val.constant = NumToString(off.o);
1701   }
1702   return NoError();
1703 }
1704 
ParseMetaData(SymbolTable<Value> * attributes)1705 CheckedError Parser::ParseMetaData(SymbolTable<Value> *attributes) {
1706   if (Is('(')) {
1707     NEXT();
1708     for (;;) {
1709       auto name = attribute_;
1710       if (false == (Is(kTokenIdentifier) || Is(kTokenStringConstant)))
1711         return Error("attribute name must be either identifier or string: " +
1712                      name);
1713       if (known_attributes_.find(name) == known_attributes_.end())
1714         return Error("user define attributes must be declared before use: " +
1715                      name);
1716       NEXT();
1717       auto e = new Value();
1718       if (attributes->Add(name, e)) Warning("attribute already found: " + name);
1719       if (Is(':')) {
1720         NEXT();
1721         ECHECK(ParseSingleValue(&name, *e, true));
1722       }
1723       if (Is(')')) {
1724         NEXT();
1725         break;
1726       }
1727       EXPECT(',');
1728     }
1729   }
1730   return NoError();
1731 }
1732 
ParseEnumFromString(const Type & type,std::string * result)1733 CheckedError Parser::ParseEnumFromString(const Type &type,
1734                                          std::string *result) {
1735   const auto base_type =
1736       type.enum_def ? type.enum_def->underlying_type.base_type : type.base_type;
1737   if (!IsInteger(base_type)) return Error("not a valid value for this field");
1738   uint64_t u64 = 0;
1739   for (size_t pos = 0; pos != std::string::npos;) {
1740     const auto delim = attribute_.find_first_of(' ', pos);
1741     const auto last = (std::string::npos == delim);
1742     auto word = attribute_.substr(pos, !last ? delim - pos : std::string::npos);
1743     pos = !last ? delim + 1 : std::string::npos;
1744     const EnumVal *ev = nullptr;
1745     if (type.enum_def) {
1746       ev = type.enum_def->Lookup(word);
1747     } else {
1748       auto dot = word.find_first_of('.');
1749       if (std::string::npos == dot)
1750         return Error("enum values need to be qualified by an enum type");
1751       auto enum_def_str = word.substr(0, dot);
1752       const auto enum_def = LookupEnum(enum_def_str);
1753       if (!enum_def) return Error("unknown enum: " + enum_def_str);
1754       auto enum_val_str = word.substr(dot + 1);
1755       ev = enum_def->Lookup(enum_val_str);
1756     }
1757     if (!ev) return Error("unknown enum value: " + word);
1758     u64 |= ev->GetAsUInt64();
1759   }
1760   *result = IsUnsigned(base_type) ? NumToString(u64)
1761                                   : NumToString(static_cast<int64_t>(u64));
1762   return NoError();
1763 }
1764 
ParseHash(Value & e,FieldDef * field)1765 CheckedError Parser::ParseHash(Value &e, FieldDef *field) {
1766   FLATBUFFERS_ASSERT(field);
1767   Value *hash_name = field->attributes.Lookup("hash");
1768   switch (e.type.base_type) {
1769     case BASE_TYPE_SHORT: {
1770       auto hash = FindHashFunction16(hash_name->constant.c_str());
1771       int16_t hashed_value = static_cast<int16_t>(hash(attribute_.c_str()));
1772       e.constant = NumToString(hashed_value);
1773       break;
1774     }
1775     case BASE_TYPE_USHORT: {
1776       auto hash = FindHashFunction16(hash_name->constant.c_str());
1777       uint16_t hashed_value = hash(attribute_.c_str());
1778       e.constant = NumToString(hashed_value);
1779       break;
1780     }
1781     case BASE_TYPE_INT: {
1782       auto hash = FindHashFunction32(hash_name->constant.c_str());
1783       int32_t hashed_value = static_cast<int32_t>(hash(attribute_.c_str()));
1784       e.constant = NumToString(hashed_value);
1785       break;
1786     }
1787     case BASE_TYPE_UINT: {
1788       auto hash = FindHashFunction32(hash_name->constant.c_str());
1789       uint32_t hashed_value = hash(attribute_.c_str());
1790       e.constant = NumToString(hashed_value);
1791       break;
1792     }
1793     case BASE_TYPE_LONG: {
1794       auto hash = FindHashFunction64(hash_name->constant.c_str());
1795       int64_t hashed_value = static_cast<int64_t>(hash(attribute_.c_str()));
1796       e.constant = NumToString(hashed_value);
1797       break;
1798     }
1799     case BASE_TYPE_ULONG: {
1800       auto hash = FindHashFunction64(hash_name->constant.c_str());
1801       uint64_t hashed_value = hash(attribute_.c_str());
1802       e.constant = NumToString(hashed_value);
1803       break;
1804     }
1805     default: FLATBUFFERS_ASSERT(0);
1806   }
1807   NEXT();
1808   return NoError();
1809 }
1810 
TokenError()1811 CheckedError Parser::TokenError() {
1812   return Error("cannot parse value starting with: " + TokenToStringId(token_));
1813 }
1814 
1815 // Re-pack helper (ParseSingleValue) to normalize defaults of scalars.
SingleValueRepack(Value & e,T val)1816 template<typename T> inline void SingleValueRepack(Value &e, T val) {
1817   // Remove leading zeros.
1818   if (IsInteger(e.type.base_type)) { e.constant = NumToString(val); }
1819 }
1820 #if defined(FLATBUFFERS_HAS_NEW_STRTOD) && (FLATBUFFERS_HAS_NEW_STRTOD > 0)
1821 // Normalize defaults NaN to unsigned quiet-NaN(0) if value was parsed from
1822 // hex-float literal.
SingleValueRepack(Value & e,float val)1823 static inline void SingleValueRepack(Value &e, float val) {
1824   if (val != val) e.constant = "nan";
1825 }
SingleValueRepack(Value & e,double val)1826 static inline void SingleValueRepack(Value &e, double val) {
1827   if (val != val) e.constant = "nan";
1828 }
1829 #endif
1830 
ParseFunction(const std::string * name,Value & e)1831 CheckedError Parser::ParseFunction(const std::string *name, Value &e) {
1832   ParseDepthGuard depth_guard(this);
1833   ECHECK(depth_guard.Check());
1834 
1835   // Copy name, attribute will be changed on NEXT().
1836   const auto functionname = attribute_;
1837   if (!IsFloat(e.type.base_type)) {
1838     return Error(functionname + ": type of argument mismatch, expecting: " +
1839                  kTypeNames[BASE_TYPE_DOUBLE] +
1840                  ", found: " + kTypeNames[e.type.base_type] +
1841                  ", name: " + (name ? *name : "") + ", value: " + e.constant);
1842   }
1843   NEXT();
1844   EXPECT('(');
1845   ECHECK(ParseSingleValue(name, e, false));
1846   EXPECT(')');
1847   // calculate with double precision
1848   double x, y = 0.0;
1849   ECHECK(atot(e.constant.c_str(), *this, &x));
1850   // clang-format off
1851   auto func_match = false;
1852   #define FLATBUFFERS_FN_DOUBLE(name, op) \
1853     if (!func_match && functionname == name) { y = op; func_match = true; }
1854   FLATBUFFERS_FN_DOUBLE("deg", x / kPi * 180);
1855   FLATBUFFERS_FN_DOUBLE("rad", x * kPi / 180);
1856   FLATBUFFERS_FN_DOUBLE("sin", sin(x));
1857   FLATBUFFERS_FN_DOUBLE("cos", cos(x));
1858   FLATBUFFERS_FN_DOUBLE("tan", tan(x));
1859   FLATBUFFERS_FN_DOUBLE("asin", asin(x));
1860   FLATBUFFERS_FN_DOUBLE("acos", acos(x));
1861   FLATBUFFERS_FN_DOUBLE("atan", atan(x));
1862   // TODO(wvo): add more useful conversion functions here.
1863   #undef FLATBUFFERS_FN_DOUBLE
1864   // clang-format on
1865   if (true != func_match) {
1866     return Error(std::string("Unknown conversion function: ") + functionname +
1867                  ", field name: " + (name ? *name : "") +
1868                  ", value: " + e.constant);
1869   }
1870   e.constant = NumToString(y);
1871   return NoError();
1872 }
1873 
TryTypedValue(const std::string * name,int dtoken,bool check,Value & e,BaseType req,bool * destmatch)1874 CheckedError Parser::TryTypedValue(const std::string *name, int dtoken,
1875                                    bool check, Value &e, BaseType req,
1876                                    bool *destmatch) {
1877   FLATBUFFERS_ASSERT(*destmatch == false && dtoken == token_);
1878   *destmatch = true;
1879   e.constant = attribute_;
1880   // Check token match
1881   if (!check) {
1882     if (e.type.base_type == BASE_TYPE_NONE) {
1883       e.type.base_type = req;
1884     } else {
1885       return Error(std::string("type mismatch: expecting: ") +
1886                    kTypeNames[e.type.base_type] +
1887                    ", found: " + kTypeNames[req] +
1888                    ", name: " + (name ? *name : "") + ", value: " + e.constant);
1889     }
1890   }
1891   // The exponent suffix of hexadecimal float-point number is mandatory.
1892   // A hex-integer constant is forbidden as an initializer of float number.
1893   if ((kTokenFloatConstant != dtoken) && IsFloat(e.type.base_type)) {
1894     const auto &s = e.constant;
1895     const auto k = s.find_first_of("0123456789.");
1896     if ((std::string::npos != k) && (s.length() > (k + 1)) &&
1897         (s[k] == '0' && is_alpha_char(s[k + 1], 'X')) &&
1898         (std::string::npos == s.find_first_of("pP", k + 2))) {
1899       return Error(
1900           "invalid number, the exponent suffix of hexadecimal "
1901           "floating-point literals is mandatory: \"" +
1902           s + "\"");
1903     }
1904   }
1905   NEXT();
1906   return NoError();
1907 }
1908 
ParseSingleValue(const std::string * name,Value & e,bool check_now)1909 CheckedError Parser::ParseSingleValue(const std::string *name, Value &e,
1910                                       bool check_now) {
1911   if (token_ == '+' || token_ == '-') {
1912     const char sign = static_cast<char>(token_);
1913     // Get an indentifier: NAN, INF, or function name like cos/sin/deg.
1914     NEXT();
1915     if (token_ != kTokenIdentifier) return Error("constant name expected");
1916     attribute_.insert(0, 1, sign);
1917   }
1918 
1919   const auto in_type = e.type.base_type;
1920   const auto is_tok_ident = (token_ == kTokenIdentifier);
1921   const auto is_tok_string = (token_ == kTokenStringConstant);
1922 
1923   // First see if this could be a conversion function.
1924   if (is_tok_ident && *cursor_ == '(') { return ParseFunction(name, e); }
1925 
1926   // clang-format off
1927   auto match = false;
1928 
1929   #define IF_ECHECK_(force, dtoken, check, req)    \
1930     if (!match && ((dtoken) == token_) && ((check) || IsConstTrue(force))) \
1931       ECHECK(TryTypedValue(name, dtoken, check, e, req, &match))
1932   #define TRY_ECHECK(dtoken, check, req) IF_ECHECK_(false, dtoken, check, req)
1933   #define FORCE_ECHECK(dtoken, check, req) IF_ECHECK_(true, dtoken, check, req)
1934   // clang-format on
1935 
1936   if (is_tok_ident || is_tok_string) {
1937     const auto kTokenStringOrIdent = token_;
1938     // The string type is a most probable type, check it first.
1939     TRY_ECHECK(kTokenStringConstant, in_type == BASE_TYPE_STRING,
1940                BASE_TYPE_STRING);
1941 
1942     // avoid escaped and non-ascii in the string
1943     if (!match && is_tok_string && IsScalar(in_type) &&
1944         !attr_is_trivial_ascii_string_) {
1945       return Error(
1946           std::string("type mismatch or invalid value, an initializer of "
1947                       "non-string field must be trivial ASCII string: type: ") +
1948           kTypeNames[in_type] + ", name: " + (name ? *name : "") +
1949           ", value: " + attribute_);
1950     }
1951 
1952     // A boolean as true/false. Boolean as Integer check below.
1953     if (!match && IsBool(in_type)) {
1954       auto is_true = attribute_ == "true";
1955       if (is_true || attribute_ == "false") {
1956         attribute_ = is_true ? "1" : "0";
1957         // accepts both kTokenStringConstant and kTokenIdentifier
1958         TRY_ECHECK(kTokenStringOrIdent, IsBool(in_type), BASE_TYPE_BOOL);
1959       }
1960     }
1961     // Check for optional scalars.
1962     if (!match && IsScalar(in_type) && attribute_ == "null") {
1963       e.constant = "null";
1964       NEXT();
1965       match = true;
1966     }
1967     // Check if this could be a string/identifier enum value.
1968     // Enum can have only true integer base type.
1969     if (!match && IsInteger(in_type) && !IsBool(in_type) &&
1970         IsIdentifierStart(*attribute_.c_str())) {
1971       ECHECK(ParseEnumFromString(e.type, &e.constant));
1972       NEXT();
1973       match = true;
1974     }
1975     // Parse a float/integer number from the string.
1976     // A "scalar-in-string" value needs extra checks.
1977     if (!match && is_tok_string && IsScalar(in_type)) {
1978       // Strip trailing whitespaces from attribute_.
1979       auto last_non_ws = attribute_.find_last_not_of(' ');
1980       if (std::string::npos != last_non_ws) attribute_.resize(last_non_ws + 1);
1981       if (IsFloat(e.type.base_type)) {
1982         // The functions strtod() and strtof() accept both 'nan' and
1983         // 'nan(number)' literals. While 'nan(number)' is rejected by the parser
1984         // as an unsupported function if is_tok_ident is true.
1985         if (attribute_.find_last_of(')') != std::string::npos) {
1986           return Error("invalid number: " + attribute_);
1987         }
1988       }
1989     }
1990     // Float numbers or nan, inf, pi, etc.
1991     TRY_ECHECK(kTokenStringOrIdent, IsFloat(in_type), BASE_TYPE_FLOAT);
1992     // An integer constant in string.
1993     TRY_ECHECK(kTokenStringOrIdent, IsInteger(in_type), BASE_TYPE_INT);
1994     // Unknown tokens will be interpreted as string type.
1995     // An attribute value may be a scalar or string constant.
1996     FORCE_ECHECK(kTokenStringConstant, in_type == BASE_TYPE_STRING,
1997                  BASE_TYPE_STRING);
1998   } else {
1999     // Try a float number.
2000     TRY_ECHECK(kTokenFloatConstant, IsFloat(in_type), BASE_TYPE_FLOAT);
2001     // Integer token can init any scalar (integer of float).
2002     FORCE_ECHECK(kTokenIntegerConstant, IsScalar(in_type), BASE_TYPE_INT);
2003   }
2004   // Match empty vectors for default-empty-vectors.
2005   if (!match && IsVector(e.type) && token_ == '[') {
2006     NEXT();
2007     if (token_ != ']') { return Error("Expected `]` in vector default"); }
2008     NEXT();
2009     match = true;
2010     e.constant = "[]";
2011   }
2012 
2013 #undef FORCE_ECHECK
2014 #undef TRY_ECHECK
2015 #undef IF_ECHECK_
2016 
2017   if (!match) {
2018     std::string msg;
2019     msg += "Cannot assign token starting with '" + TokenToStringId(token_) +
2020            "' to value of <" + std::string(kTypeNames[in_type]) + "> type.";
2021     return Error(msg);
2022   }
2023   const auto match_type = e.type.base_type;  // may differ from in_type
2024   // The check_now flag must be true when parse a fbs-schema.
2025   // This flag forces to check default scalar values or metadata of field.
2026   // For JSON parser the flag should be false.
2027   // If it is set for JSON each value will be checked twice (see ParseTable).
2028   // Special case 'null' since atot can't handle that.
2029   if (check_now && IsScalar(match_type) && e.constant != "null") {
2030     // clang-format off
2031     switch (match_type) {
2032     #define FLATBUFFERS_TD(ENUM, IDLTYPE, CTYPE, ...) \
2033       case BASE_TYPE_ ## ENUM: {\
2034           CTYPE val; \
2035           ECHECK(atot(e.constant.c_str(), *this, &val)); \
2036           SingleValueRepack(e, val); \
2037         break; }
2038     FLATBUFFERS_GEN_TYPES_SCALAR(FLATBUFFERS_TD)
2039     #undef FLATBUFFERS_TD
2040     default: break;
2041     }
2042     // clang-format on
2043   }
2044   return NoError();
2045 }
2046 
LookupCreateStruct(const std::string & name,bool create_if_new,bool definition)2047 StructDef *Parser::LookupCreateStruct(const std::string &name,
2048                                       bool create_if_new, bool definition) {
2049   std::string qualified_name = current_namespace_->GetFullyQualifiedName(name);
2050   // See if it exists pre-declared by an unqualified use.
2051   auto struct_def = LookupStruct(name);
2052   if (struct_def && struct_def->predecl) {
2053     if (definition) {
2054       // Make sure it has the current namespace, and is registered under its
2055       // qualified name.
2056       struct_def->defined_namespace = current_namespace_;
2057       structs_.Move(name, qualified_name);
2058     }
2059     return struct_def;
2060   }
2061   // See if it exists pre-declared by an qualified use.
2062   struct_def = LookupStruct(qualified_name);
2063   if (struct_def && struct_def->predecl) {
2064     if (definition) {
2065       // Make sure it has the current namespace.
2066       struct_def->defined_namespace = current_namespace_;
2067     }
2068     return struct_def;
2069   }
2070   if (!definition && !struct_def) {
2071     struct_def = LookupStructThruParentNamespaces(name);
2072   }
2073   if (!struct_def && create_if_new) {
2074     struct_def = new StructDef();
2075     if (definition) {
2076       structs_.Add(qualified_name, struct_def);
2077       struct_def->name = name;
2078       struct_def->defined_namespace = current_namespace_;
2079     } else {
2080       // Not a definition.
2081       // Rather than failing, we create a "pre declared" StructDef, due to
2082       // circular references, and check for errors at the end of parsing.
2083       // It is defined in the current namespace, as the best guess what the
2084       // final namespace will be.
2085       structs_.Add(name, struct_def);
2086       struct_def->name = name;
2087       struct_def->defined_namespace = current_namespace_;
2088       struct_def->original_location.reset(
2089           new std::string(file_being_parsed_ + ":" + NumToString(line_)));
2090     }
2091   }
2092   return struct_def;
2093 }
2094 
MinValue() const2095 const EnumVal *EnumDef::MinValue() const {
2096   return vals.vec.empty() ? nullptr : vals.vec.front();
2097 }
MaxValue() const2098 const EnumVal *EnumDef::MaxValue() const {
2099   return vals.vec.empty() ? nullptr : vals.vec.back();
2100 }
2101 
EnumDistanceImpl(T e1,T e2)2102 template<typename T> static uint64_t EnumDistanceImpl(T e1, T e2) {
2103   if (e1 < e2) { std::swap(e1, e2); }  // use std for scalars
2104   // Signed overflow may occur, use unsigned calculation.
2105   // The unsigned overflow is well-defined by C++ standard (modulo 2^n).
2106   return static_cast<uint64_t>(e1) - static_cast<uint64_t>(e2);
2107 }
2108 
Distance(const EnumVal * v1,const EnumVal * v2) const2109 uint64_t EnumDef::Distance(const EnumVal *v1, const EnumVal *v2) const {
2110   return IsUInt64() ? EnumDistanceImpl(v1->GetAsUInt64(), v2->GetAsUInt64())
2111                     : EnumDistanceImpl(v1->GetAsInt64(), v2->GetAsInt64());
2112 }
2113 
AllFlags() const2114 std::string EnumDef::AllFlags() const {
2115   FLATBUFFERS_ASSERT(attributes.Lookup("bit_flags"));
2116   uint64_t u64 = 0;
2117   for (auto it = Vals().begin(); it != Vals().end(); ++it) {
2118     u64 |= (*it)->GetAsUInt64();
2119   }
2120   return IsUInt64() ? NumToString(u64) : NumToString(static_cast<int64_t>(u64));
2121 }
2122 
ReverseLookup(int64_t enum_idx,bool skip_union_default) const2123 EnumVal *EnumDef::ReverseLookup(int64_t enum_idx,
2124                                 bool skip_union_default) const {
2125   auto skip_first = static_cast<int>(is_union && skip_union_default);
2126   for (auto it = Vals().begin() + skip_first; it != Vals().end(); ++it) {
2127     if ((*it)->GetAsInt64() == enum_idx) { return *it; }
2128   }
2129   return nullptr;
2130 }
2131 
FindByValue(const std::string & constant) const2132 EnumVal *EnumDef::FindByValue(const std::string &constant) const {
2133   int64_t i64;
2134   auto done = false;
2135   if (IsUInt64()) {
2136     uint64_t u64;  // avoid reinterpret_cast of pointers
2137     done = StringToNumber(constant.c_str(), &u64);
2138     i64 = static_cast<int64_t>(u64);
2139   } else {
2140     done = StringToNumber(constant.c_str(), &i64);
2141   }
2142   FLATBUFFERS_ASSERT(done);
2143   if (!done) return nullptr;
2144   return ReverseLookup(i64, false);
2145 }
2146 
SortByValue()2147 void EnumDef::SortByValue() {
2148   auto &v = vals.vec;
2149   if (IsUInt64())
2150     std::sort(v.begin(), v.end(), [](const EnumVal *e1, const EnumVal *e2) {
2151       return e1->GetAsUInt64() < e2->GetAsUInt64();
2152     });
2153   else
2154     std::sort(v.begin(), v.end(), [](const EnumVal *e1, const EnumVal *e2) {
2155       return e1->GetAsInt64() < e2->GetAsInt64();
2156     });
2157 }
2158 
RemoveDuplicates()2159 void EnumDef::RemoveDuplicates() {
2160   // This method depends form SymbolTable implementation!
2161   // 1) vals.vec - owner (raw pointer)
2162   // 2) vals.dict - access map
2163   auto first = vals.vec.begin();
2164   auto last = vals.vec.end();
2165   if (first == last) return;
2166   auto result = first;
2167   while (++first != last) {
2168     if ((*result)->value != (*first)->value) {
2169       *(++result) = *first;
2170     } else {
2171       auto ev = *first;
2172       for (auto it = vals.dict.begin(); it != vals.dict.end(); ++it) {
2173         if (it->second == ev) it->second = *result;  // reassign
2174       }
2175       delete ev;  // delete enum value
2176       *first = nullptr;
2177     }
2178   }
2179   vals.vec.erase(++result, last);
2180 }
2181 
ChangeEnumValue(EnumVal * ev,T new_value)2182 template<typename T> void EnumDef::ChangeEnumValue(EnumVal *ev, T new_value) {
2183   ev->value = static_cast<int64_t>(new_value);
2184 }
2185 
2186 namespace EnumHelper {
2187 template<BaseType E> struct EnumValType { typedef int64_t type; };
2188 template<> struct EnumValType<BASE_TYPE_ULONG> { typedef uint64_t type; };
2189 }  // namespace EnumHelper
2190 
2191 struct EnumValBuilder {
CreateEnumeratorflatbuffers::EnumValBuilder2192   EnumVal *CreateEnumerator(const std::string &ev_name) {
2193     FLATBUFFERS_ASSERT(!temp);
2194     auto first = enum_def.vals.vec.empty();
2195     user_value = first;
2196     temp = new EnumVal(ev_name, first ? 0 : enum_def.vals.vec.back()->value);
2197     return temp;
2198   }
2199 
CreateEnumeratorflatbuffers::EnumValBuilder2200   EnumVal *CreateEnumerator(const std::string &ev_name, int64_t val) {
2201     FLATBUFFERS_ASSERT(!temp);
2202     user_value = true;
2203     temp = new EnumVal(ev_name, val);
2204     return temp;
2205   }
2206 
AcceptEnumeratorflatbuffers::EnumValBuilder2207   FLATBUFFERS_CHECKED_ERROR AcceptEnumerator(const std::string &name) {
2208     FLATBUFFERS_ASSERT(temp);
2209     ECHECK(ValidateValue(&temp->value, false == user_value));
2210     FLATBUFFERS_ASSERT((temp->union_type.enum_def == nullptr) ||
2211                        (temp->union_type.enum_def == &enum_def));
2212     auto not_unique = enum_def.vals.Add(name, temp);
2213     temp = nullptr;
2214     if (not_unique) return parser.Error("enum value already exists: " + name);
2215     return NoError();
2216   }
2217 
AcceptEnumeratorflatbuffers::EnumValBuilder2218   FLATBUFFERS_CHECKED_ERROR AcceptEnumerator() {
2219     return AcceptEnumerator(temp->name);
2220   }
2221 
AssignEnumeratorValueflatbuffers::EnumValBuilder2222   FLATBUFFERS_CHECKED_ERROR AssignEnumeratorValue(const std::string &value) {
2223     user_value = true;
2224     auto fit = false;
2225     if (enum_def.IsUInt64()) {
2226       uint64_t u64;
2227       fit = StringToNumber(value.c_str(), &u64);
2228       temp->value = static_cast<int64_t>(u64);  // well-defined since C++20.
2229     } else {
2230       int64_t i64;
2231       fit = StringToNumber(value.c_str(), &i64);
2232       temp->value = i64;
2233     }
2234     if (!fit) return parser.Error("enum value does not fit, \"" + value + "\"");
2235     return NoError();
2236   }
2237 
2238   template<BaseType E, typename CTYPE>
ValidateImplflatbuffers::EnumValBuilder2239   inline FLATBUFFERS_CHECKED_ERROR ValidateImpl(int64_t *ev, int m) {
2240     typedef typename EnumHelper::EnumValType<E>::type T;  // int64_t or uint64_t
2241     static_assert(sizeof(T) == sizeof(int64_t), "invalid EnumValType");
2242     const auto v = static_cast<T>(*ev);
2243     auto up = static_cast<T>((flatbuffers::numeric_limits<CTYPE>::max)());
2244     auto dn = static_cast<T>((flatbuffers::numeric_limits<CTYPE>::lowest)());
2245     if (v < dn || v > (up - m)) {
2246       return parser.Error("enum value does not fit, \"" + NumToString(v) +
2247                           (m ? " + 1\"" : "\"") + " out of " +
2248                           TypeToIntervalString<CTYPE>());
2249     }
2250     *ev = static_cast<int64_t>(v + m);  // well-defined since C++20.
2251     return NoError();
2252   }
2253 
ValidateValueflatbuffers::EnumValBuilder2254   FLATBUFFERS_CHECKED_ERROR ValidateValue(int64_t *ev, bool next) {
2255     // clang-format off
2256     switch (enum_def.underlying_type.base_type) {
2257     #define FLATBUFFERS_TD(ENUM, IDLTYPE, CTYPE, ...)                   \
2258       case BASE_TYPE_##ENUM: {                                          \
2259         if (!IsInteger(BASE_TYPE_##ENUM)) break;                        \
2260         return ValidateImpl<BASE_TYPE_##ENUM, CTYPE>(ev, next ? 1 : 0); \
2261       }
2262       FLATBUFFERS_GEN_TYPES_SCALAR(FLATBUFFERS_TD)
2263     #undef FLATBUFFERS_TD
2264     default: break;
2265     }
2266     // clang-format on
2267     return parser.Error("fatal: invalid enum underlying type");
2268   }
2269 
EnumValBuilderflatbuffers::EnumValBuilder2270   EnumValBuilder(Parser &_parser, EnumDef &_enum_def)
2271       : parser(_parser),
2272         enum_def(_enum_def),
2273         temp(nullptr),
2274         user_value(false) {}
2275 
~EnumValBuilderflatbuffers::EnumValBuilder2276   ~EnumValBuilder() { delete temp; }
2277 
2278   Parser &parser;
2279   EnumDef &enum_def;
2280   EnumVal *temp;
2281   bool user_value;
2282 };
2283 
ParseEnum(const bool is_union,EnumDef ** dest)2284 CheckedError Parser::ParseEnum(const bool is_union, EnumDef **dest) {
2285   std::vector<std::string> enum_comment = doc_comment_;
2286   NEXT();
2287   std::string enum_name = attribute_;
2288   EXPECT(kTokenIdentifier);
2289   EnumDef *enum_def;
2290   ECHECK(StartEnum(enum_name, is_union, &enum_def));
2291   enum_def->doc_comment = enum_comment;
2292   if (!is_union && !opts.proto_mode) {
2293     // Give specialized error message, since this type spec used to
2294     // be optional in the first FlatBuffers release.
2295     if (!Is(':')) {
2296       return Error(
2297           "must specify the underlying integer type for this"
2298           " enum (e.g. \': short\', which was the default).");
2299     } else {
2300       NEXT();
2301     }
2302     // Specify the integer type underlying this enum.
2303     ECHECK(ParseType(enum_def->underlying_type));
2304     if (!IsInteger(enum_def->underlying_type.base_type) ||
2305         IsBool(enum_def->underlying_type.base_type))
2306       return Error("underlying enum type must be integral");
2307     // Make this type refer back to the enum it was derived from.
2308     enum_def->underlying_type.enum_def = enum_def;
2309   }
2310   ECHECK(ParseMetaData(&enum_def->attributes));
2311   const auto underlying_type = enum_def->underlying_type.base_type;
2312   if (enum_def->attributes.Lookup("bit_flags") &&
2313       !IsUnsigned(underlying_type)) {
2314     // todo: Convert to the Error in the future?
2315     Warning("underlying type of bit_flags enum must be unsigned");
2316   }
2317   EnumValBuilder evb(*this, *enum_def);
2318   EXPECT('{');
2319   // A lot of code generatos expect that an enum is not-empty.
2320   if ((is_union || Is('}')) && !opts.proto_mode) {
2321     evb.CreateEnumerator("NONE");
2322     ECHECK(evb.AcceptEnumerator());
2323   }
2324   std::set<std::pair<BaseType, StructDef *>> union_types;
2325   while (!Is('}')) {
2326     if (opts.proto_mode && attribute_ == "option") {
2327       ECHECK(ParseProtoOption());
2328     } else {
2329       auto &ev = *evb.CreateEnumerator(attribute_);
2330       auto full_name = ev.name;
2331       ev.doc_comment = doc_comment_;
2332       EXPECT(kTokenIdentifier);
2333       if (is_union) {
2334         ECHECK(ParseNamespacing(&full_name, &ev.name));
2335         if (opts.union_value_namespacing) {
2336           // Since we can't namespace the actual enum identifiers, turn
2337           // namespace parts into part of the identifier.
2338           ev.name = full_name;
2339           std::replace(ev.name.begin(), ev.name.end(), '.', '_');
2340         }
2341         if (Is(':')) {
2342           NEXT();
2343           ECHECK(ParseType(ev.union_type));
2344           if (ev.union_type.base_type != BASE_TYPE_STRUCT &&
2345               ev.union_type.base_type != BASE_TYPE_STRING)
2346             return Error("union value type may only be table/struct/string");
2347         } else {
2348           ev.union_type = Type(BASE_TYPE_STRUCT, LookupCreateStruct(full_name));
2349         }
2350         if (!enum_def->uses_multiple_type_instances) {
2351           auto ins = union_types.insert(std::make_pair(
2352               ev.union_type.base_type, ev.union_type.struct_def));
2353           enum_def->uses_multiple_type_instances = (false == ins.second);
2354         }
2355       }
2356 
2357       if (Is('=')) {
2358         NEXT();
2359         ECHECK(evb.AssignEnumeratorValue(attribute_));
2360         EXPECT(kTokenIntegerConstant);
2361       }
2362 
2363       ECHECK(evb.AcceptEnumerator());
2364 
2365       if (opts.proto_mode && Is('[')) {
2366         NEXT();
2367         // ignore attributes on enums.
2368         while (token_ != ']') NEXT();
2369         NEXT();
2370       }
2371     }
2372     if (!Is(opts.proto_mode ? ';' : ',')) break;
2373     NEXT();
2374   }
2375   EXPECT('}');
2376 
2377   // At this point, the enum can be empty if input is invalid proto-file.
2378   if (!enum_def->size())
2379     return Error("incomplete enum declaration, values not found");
2380 
2381   if (enum_def->attributes.Lookup("bit_flags")) {
2382     const auto base_width = static_cast<uint64_t>(8 * SizeOf(underlying_type));
2383     for (auto it = enum_def->Vals().begin(); it != enum_def->Vals().end();
2384          ++it) {
2385       auto ev = *it;
2386       const auto u = ev->GetAsUInt64();
2387       // Stop manipulations with the sign.
2388       if (!IsUnsigned(underlying_type) && u == (base_width - 1))
2389         return Error("underlying type of bit_flags enum must be unsigned");
2390       if (u >= base_width)
2391         return Error("bit flag out of range of underlying integral type");
2392       enum_def->ChangeEnumValue(ev, 1ULL << u);
2393     }
2394   }
2395 
2396   enum_def->SortByValue();  // Must be sorted to use MinValue/MaxValue.
2397 
2398   // Ensure enum value uniqueness.
2399   auto prev_it = enum_def->Vals().begin();
2400   for (auto it = prev_it + 1; it != enum_def->Vals().end(); ++it) {
2401     auto prev_ev = *prev_it;
2402     auto ev = *it;
2403     if (prev_ev->GetAsUInt64() == ev->GetAsUInt64())
2404       return Error("all enum values must be unique: " + prev_ev->name +
2405                    " and " + ev->name + " are both " +
2406                    NumToString(ev->GetAsInt64()));
2407   }
2408 
2409   if (dest) *dest = enum_def;
2410   types_.Add(current_namespace_->GetFullyQualifiedName(enum_def->name),
2411              new Type(BASE_TYPE_UNION, nullptr, enum_def));
2412   return NoError();
2413 }
2414 
StartStruct(const std::string & name,StructDef ** dest)2415 CheckedError Parser::StartStruct(const std::string &name, StructDef **dest) {
2416   auto &struct_def = *LookupCreateStruct(name, true, true);
2417   if (!struct_def.predecl) return Error("datatype already exists: " + name);
2418   struct_def.predecl = false;
2419   struct_def.name = name;
2420   struct_def.file = file_being_parsed_;
2421   // Move this struct to the back of the vector just in case it was predeclared,
2422   // to preserve declaration order.
2423   *std::remove(structs_.vec.begin(), structs_.vec.end(), &struct_def) =
2424       &struct_def;
2425   *dest = &struct_def;
2426   return NoError();
2427 }
2428 
CheckClash(std::vector<FieldDef * > & fields,StructDef * struct_def,const char * suffix,BaseType basetype)2429 CheckedError Parser::CheckClash(std::vector<FieldDef *> &fields,
2430                                 StructDef *struct_def, const char *suffix,
2431                                 BaseType basetype) {
2432   auto len = strlen(suffix);
2433   for (auto it = fields.begin(); it != fields.end(); ++it) {
2434     auto &fname = (*it)->name;
2435     if (fname.length() > len &&
2436         fname.compare(fname.length() - len, len, suffix) == 0 &&
2437         (*it)->value.type.base_type != BASE_TYPE_UTYPE) {
2438       auto field =
2439           struct_def->fields.Lookup(fname.substr(0, fname.length() - len));
2440       if (field && field->value.type.base_type == basetype)
2441         return Error("Field " + fname +
2442                      " would clash with generated functions for field " +
2443                      field->name);
2444     }
2445   }
2446   return NoError();
2447 }
2448 
SupportsOptionalScalars(const flatbuffers::IDLOptions & opts)2449 bool Parser::SupportsOptionalScalars(const flatbuffers::IDLOptions &opts) {
2450   static FLATBUFFERS_CONSTEXPR unsigned long supported_langs =
2451       IDLOptions::kRust | IDLOptions::kSwift | IDLOptions::kLobster |
2452       IDLOptions::kKotlin | IDLOptions::kCpp | IDLOptions::kJava |
2453       IDLOptions::kCSharp | IDLOptions::kTs | IDLOptions::kBinary;
2454   unsigned long langs = opts.lang_to_generate;
2455   return (langs > 0 && langs < IDLOptions::kMAX) && !(langs & ~supported_langs);
2456 }
SupportsOptionalScalars() const2457 bool Parser::SupportsOptionalScalars() const {
2458   // Check in general if a language isn't specified.
2459   return opts.lang_to_generate == 0 || SupportsOptionalScalars(opts);
2460 }
2461 
SupportsDefaultVectorsAndStrings() const2462 bool Parser::SupportsDefaultVectorsAndStrings() const {
2463   static FLATBUFFERS_CONSTEXPR unsigned long supported_langs =
2464       IDLOptions::kRust | IDLOptions::kSwift;
2465   return !(opts.lang_to_generate & ~supported_langs);
2466 }
2467 
SupportsAdvancedUnionFeatures() const2468 bool Parser::SupportsAdvancedUnionFeatures() const {
2469   return opts.lang_to_generate != 0 &&
2470          (opts.lang_to_generate &
2471           ~(IDLOptions::kCpp | IDLOptions::kTs | IDLOptions::kPhp |
2472             IDLOptions::kJava | IDLOptions::kCSharp | IDLOptions::kKotlin |
2473             IDLOptions::kBinary | IDLOptions::kSwift)) == 0;
2474 }
2475 
SupportsAdvancedArrayFeatures() const2476 bool Parser::SupportsAdvancedArrayFeatures() const {
2477   return (opts.lang_to_generate &
2478           ~(IDLOptions::kCpp | IDLOptions::kPython | IDLOptions::kJava |
2479             IDLOptions::kCSharp | IDLOptions::kJsonSchema | IDLOptions::kJson |
2480             IDLOptions::kBinary | IDLOptions::kRust)) == 0;
2481 }
2482 
UniqueNamespace(Namespace * ns)2483 Namespace *Parser::UniqueNamespace(Namespace *ns) {
2484   for (auto it = namespaces_.begin(); it != namespaces_.end(); ++it) {
2485     if (ns->components == (*it)->components) {
2486       delete ns;
2487       return *it;
2488     }
2489   }
2490   namespaces_.push_back(ns);
2491   return ns;
2492 }
2493 
UnqualifiedName(const std::string & full_qualified_name)2494 std::string Parser::UnqualifiedName(const std::string &full_qualified_name) {
2495   Namespace *ns = new Namespace();
2496 
2497   std::size_t current, previous = 0;
2498   current = full_qualified_name.find('.');
2499   while (current != std::string::npos) {
2500     ns->components.push_back(
2501         full_qualified_name.substr(previous, current - previous));
2502     previous = current + 1;
2503     current = full_qualified_name.find('.', previous);
2504   }
2505   current_namespace_ = UniqueNamespace(ns);
2506   return full_qualified_name.substr(previous, current - previous);
2507 }
2508 
compareFieldDefs(const FieldDef * a,const FieldDef * b)2509 static bool compareFieldDefs(const FieldDef *a, const FieldDef *b) {
2510   auto a_id = atoi(a->attributes.Lookup("id")->constant.c_str());
2511   auto b_id = atoi(b->attributes.Lookup("id")->constant.c_str());
2512   return a_id < b_id;
2513 }
2514 
ParseDecl()2515 CheckedError Parser::ParseDecl() {
2516   std::vector<std::string> dc = doc_comment_;
2517   bool fixed = IsIdent("struct");
2518   if (!fixed && !IsIdent("table")) return Error("declaration expected");
2519   NEXT();
2520   std::string name = attribute_;
2521   EXPECT(kTokenIdentifier);
2522   StructDef *struct_def;
2523   ECHECK(StartStruct(name, &struct_def));
2524   struct_def->doc_comment = dc;
2525   struct_def->fixed = fixed;
2526   ECHECK(ParseMetaData(&struct_def->attributes));
2527   struct_def->sortbysize =
2528       struct_def->attributes.Lookup("original_order") == nullptr && !fixed;
2529   EXPECT('{');
2530   while (token_ != '}') ECHECK(ParseField(*struct_def));
2531   if (fixed) {
2532     const auto force_align = struct_def->attributes.Lookup("force_align");
2533     if (force_align) {
2534       size_t align;
2535       ECHECK(ParseAlignAttribute(force_align->constant, struct_def->minalign,
2536                                  &align));
2537       struct_def->minalign = align;
2538     }
2539     if (!struct_def->bytesize) return Error("size 0 structs not allowed");
2540   }
2541   struct_def->PadLastField(struct_def->minalign);
2542   // Check if this is a table that has manual id assignments
2543   auto &fields = struct_def->fields.vec;
2544   if (!fixed && fields.size()) {
2545     size_t num_id_fields = 0;
2546     for (auto it = fields.begin(); it != fields.end(); ++it) {
2547       if ((*it)->attributes.Lookup("id")) num_id_fields++;
2548     }
2549     // If any fields have ids..
2550     if (num_id_fields || opts.require_explicit_ids) {
2551       // Then all fields must have them.
2552       if (num_id_fields != fields.size()) {
2553         if (opts.require_explicit_ids) {
2554           return Error(
2555               "all fields must have an 'id' attribute when "
2556               "--require-explicit-ids is used");
2557         } else {
2558           return Error(
2559               "either all fields or no fields must have an 'id' attribute");
2560         }
2561       }
2562       // Simply sort by id, then the fields are the same as if no ids had
2563       // been specified.
2564       std::sort(fields.begin(), fields.end(), compareFieldDefs);
2565       // Verify we have a contiguous set, and reassign vtable offsets.
2566       FLATBUFFERS_ASSERT(fields.size() <=
2567                          flatbuffers::numeric_limits<voffset_t>::max());
2568       for (voffset_t i = 0; i < static_cast<voffset_t>(fields.size()); i++) {
2569         auto &field = *fields[i];
2570         const auto &id_str = field.attributes.Lookup("id")->constant;
2571         // Metadata values have a dynamic type, they can be `float`, 'int', or
2572         // 'string`.
2573         // The FieldIndexToOffset(i) expects the voffset_t so `id` is limited by
2574         // this type.
2575         voffset_t id = 0;
2576         const auto done = !atot(id_str.c_str(), *this, &id).Check();
2577         if (!done)
2578           return Error("field id\'s must be non-negative number, field: " +
2579                        field.name + ", id: " + id_str);
2580         if (i != id)
2581           return Error("field id\'s must be consecutive from 0, id " +
2582                        NumToString(i) + " missing or set twice, field: " +
2583                        field.name + ", id: " + id_str);
2584         field.value.offset = FieldIndexToOffset(i);
2585       }
2586     }
2587   }
2588 
2589   ECHECK(
2590       CheckClash(fields, struct_def, UnionTypeFieldSuffix(), BASE_TYPE_UNION));
2591   ECHECK(CheckClash(fields, struct_def, "Type", BASE_TYPE_UNION));
2592   ECHECK(CheckClash(fields, struct_def, "_length", BASE_TYPE_VECTOR));
2593   ECHECK(CheckClash(fields, struct_def, "Length", BASE_TYPE_VECTOR));
2594   ECHECK(CheckClash(fields, struct_def, "_byte_vector", BASE_TYPE_STRING));
2595   ECHECK(CheckClash(fields, struct_def, "ByteVector", BASE_TYPE_STRING));
2596   EXPECT('}');
2597   types_.Add(current_namespace_->GetFullyQualifiedName(struct_def->name),
2598              new Type(BASE_TYPE_STRUCT, struct_def, nullptr));
2599   return NoError();
2600 }
2601 
ParseService()2602 CheckedError Parser::ParseService() {
2603   std::vector<std::string> service_comment = doc_comment_;
2604   NEXT();
2605   auto service_name = attribute_;
2606   EXPECT(kTokenIdentifier);
2607   auto &service_def = *new ServiceDef();
2608   service_def.name = service_name;
2609   service_def.file = file_being_parsed_;
2610   service_def.doc_comment = service_comment;
2611   service_def.defined_namespace = current_namespace_;
2612   if (services_.Add(current_namespace_->GetFullyQualifiedName(service_name),
2613                     &service_def))
2614     return Error("service already exists: " + service_name);
2615   ECHECK(ParseMetaData(&service_def.attributes));
2616   EXPECT('{');
2617   do {
2618     std::vector<std::string> doc_comment = doc_comment_;
2619     auto rpc_name = attribute_;
2620     EXPECT(kTokenIdentifier);
2621     EXPECT('(');
2622     Type reqtype, resptype;
2623     ECHECK(ParseTypeIdent(reqtype));
2624     EXPECT(')');
2625     EXPECT(':');
2626     ECHECK(ParseTypeIdent(resptype));
2627     if (reqtype.base_type != BASE_TYPE_STRUCT || reqtype.struct_def->fixed ||
2628         resptype.base_type != BASE_TYPE_STRUCT || resptype.struct_def->fixed)
2629       return Error("rpc request and response types must be tables");
2630     auto &rpc = *new RPCCall();
2631     rpc.name = rpc_name;
2632     rpc.request = reqtype.struct_def;
2633     rpc.response = resptype.struct_def;
2634     rpc.doc_comment = doc_comment;
2635     if (service_def.calls.Add(rpc_name, &rpc))
2636       return Error("rpc already exists: " + rpc_name);
2637     ECHECK(ParseMetaData(&rpc.attributes));
2638     EXPECT(';');
2639   } while (token_ != '}');
2640   NEXT();
2641   return NoError();
2642 }
2643 
SetRootType(const char * name)2644 bool Parser::SetRootType(const char *name) {
2645   root_struct_def_ = LookupStruct(name);
2646   if (!root_struct_def_)
2647     root_struct_def_ =
2648         LookupStruct(current_namespace_->GetFullyQualifiedName(name));
2649   return root_struct_def_ != nullptr;
2650 }
2651 
MarkGenerated()2652 void Parser::MarkGenerated() {
2653   // This function marks all existing definitions as having already
2654   // been generated, which signals no code for included files should be
2655   // generated.
2656   for (auto it = enums_.vec.begin(); it != enums_.vec.end(); ++it) {
2657     (*it)->generated = true;
2658   }
2659   for (auto it = structs_.vec.begin(); it != structs_.vec.end(); ++it) {
2660     if (!(*it)->predecl) { (*it)->generated = true; }
2661   }
2662   for (auto it = services_.vec.begin(); it != services_.vec.end(); ++it) {
2663     (*it)->generated = true;
2664   }
2665 }
2666 
ParseNamespace()2667 CheckedError Parser::ParseNamespace() {
2668   NEXT();
2669   auto ns = new Namespace();
2670   namespaces_.push_back(ns);  // Store it here to not leak upon error.
2671   if (token_ != ';') {
2672     for (;;) {
2673       ns->components.push_back(attribute_);
2674       EXPECT(kTokenIdentifier);
2675       if (Is('.')) NEXT() else break;
2676     }
2677   }
2678   namespaces_.pop_back();
2679   current_namespace_ = UniqueNamespace(ns);
2680   EXPECT(';');
2681   return NoError();
2682 }
2683 
2684 // Best effort parsing of .proto declarations, with the aim to turn them
2685 // in the closest corresponding FlatBuffer equivalent.
2686 // We parse everything as identifiers instead of keywords, since we don't
2687 // want protobuf keywords to become invalid identifiers in FlatBuffers.
ParseProtoDecl()2688 CheckedError Parser::ParseProtoDecl() {
2689   bool isextend = IsIdent("extend");
2690   if (IsIdent("package")) {
2691     // These are identical in syntax to FlatBuffer's namespace decl.
2692     ECHECK(ParseNamespace());
2693   } else if (IsIdent("message") || isextend) {
2694     std::vector<std::string> struct_comment = doc_comment_;
2695     NEXT();
2696     StructDef *struct_def = nullptr;
2697     Namespace *parent_namespace = nullptr;
2698     if (isextend) {
2699       if (Is('.')) NEXT();  // qualified names may start with a . ?
2700       auto id = attribute_;
2701       EXPECT(kTokenIdentifier);
2702       ECHECK(ParseNamespacing(&id, nullptr));
2703       struct_def = LookupCreateStruct(id, false);
2704       if (!struct_def)
2705         return Error("cannot extend unknown message type: " + id);
2706     } else {
2707       std::string name = attribute_;
2708       EXPECT(kTokenIdentifier);
2709       ECHECK(StartStruct(name, &struct_def));
2710       // Since message definitions can be nested, we create a new namespace.
2711       auto ns = new Namespace();
2712       // Copy of current namespace.
2713       *ns = *current_namespace_;
2714       // But with current message name.
2715       ns->components.push_back(name);
2716       ns->from_table++;
2717       parent_namespace = current_namespace_;
2718       current_namespace_ = UniqueNamespace(ns);
2719     }
2720     struct_def->doc_comment = struct_comment;
2721     ECHECK(ParseProtoFields(struct_def, isextend, false));
2722     if (!isextend) { current_namespace_ = parent_namespace; }
2723     if (Is(';')) NEXT();
2724   } else if (IsIdent("enum")) {
2725     // These are almost the same, just with different terminator:
2726     EnumDef *enum_def;
2727     ECHECK(ParseEnum(false, &enum_def));
2728     if (Is(';')) NEXT();
2729     // Temp: remove any duplicates, as .fbs files can't handle them.
2730     enum_def->RemoveDuplicates();
2731   } else if (IsIdent("syntax")) {  // Skip these.
2732     NEXT();
2733     EXPECT('=');
2734     EXPECT(kTokenStringConstant);
2735     EXPECT(';');
2736   } else if (IsIdent("option")) {  // Skip these.
2737     ECHECK(ParseProtoOption());
2738     EXPECT(';');
2739   } else if (IsIdent("service")) {  // Skip these.
2740     NEXT();
2741     EXPECT(kTokenIdentifier);
2742     ECHECK(ParseProtoCurliesOrIdent());
2743   } else {
2744     return Error("don\'t know how to parse .proto declaration starting with " +
2745                  TokenToStringId(token_));
2746   }
2747   return NoError();
2748 }
2749 
StartEnum(const std::string & enum_name,bool is_union,EnumDef ** dest)2750 CheckedError Parser::StartEnum(const std::string &enum_name, bool is_union,
2751                                EnumDef **dest) {
2752   auto &enum_def = *new EnumDef();
2753   enum_def.name = enum_name;
2754   enum_def.file = file_being_parsed_;
2755   enum_def.doc_comment = doc_comment_;
2756   enum_def.is_union = is_union;
2757   enum_def.defined_namespace = current_namespace_;
2758   if (enums_.Add(current_namespace_->GetFullyQualifiedName(enum_name),
2759                  &enum_def))
2760     return Error("enum already exists: " + enum_name);
2761   enum_def.underlying_type.base_type =
2762       is_union ? BASE_TYPE_UTYPE : BASE_TYPE_INT;
2763   enum_def.underlying_type.enum_def = &enum_def;
2764   if (dest) *dest = &enum_def;
2765   return NoError();
2766 }
2767 
ParseProtoFields(StructDef * struct_def,bool isextend,bool inside_oneof)2768 CheckedError Parser::ParseProtoFields(StructDef *struct_def, bool isextend,
2769                                       bool inside_oneof) {
2770   EXPECT('{');
2771   while (token_ != '}') {
2772     if (IsIdent("message") || IsIdent("extend") || IsIdent("enum")) {
2773       // Nested declarations.
2774       ECHECK(ParseProtoDecl());
2775     } else if (IsIdent("extensions")) {  // Skip these.
2776       NEXT();
2777       EXPECT(kTokenIntegerConstant);
2778       if (Is(kTokenIdentifier)) {
2779         NEXT();  // to
2780         NEXT();  // num
2781       }
2782       EXPECT(';');
2783     } else if (IsIdent("option")) {  // Skip these.
2784       ECHECK(ParseProtoOption());
2785       EXPECT(';');
2786     } else if (IsIdent("reserved")) {  // Skip these.
2787       NEXT();
2788       while (!Is(';')) { NEXT(); }  // A variety of formats, just skip.
2789       NEXT();
2790     } else {
2791       std::vector<std::string> field_comment = doc_comment_;
2792       // Parse the qualifier.
2793       bool required = false;
2794       bool repeated = false;
2795       bool oneof = false;
2796       if (!inside_oneof) {
2797         if (IsIdent("optional")) {
2798           // This is the default.
2799           NEXT();
2800         } else if (IsIdent("required")) {
2801           required = true;
2802           NEXT();
2803         } else if (IsIdent("repeated")) {
2804           repeated = true;
2805           NEXT();
2806         } else if (IsIdent("oneof")) {
2807           oneof = true;
2808           NEXT();
2809         } else {
2810           // can't error, proto3 allows decls without any of the above.
2811         }
2812       }
2813       StructDef *anonymous_struct = nullptr;
2814       EnumDef *oneof_union = nullptr;
2815       Type type;
2816       if (IsIdent("group") || oneof) {
2817         if (!oneof) NEXT();
2818         if (oneof && opts.proto_oneof_union) {
2819           auto name = MakeCamel(attribute_, true) + "Union";
2820           ECHECK(StartEnum(name, true, &oneof_union));
2821           type = Type(BASE_TYPE_UNION, nullptr, oneof_union);
2822         } else {
2823           auto name = "Anonymous" + NumToString(anonymous_counter_++);
2824           ECHECK(StartStruct(name, &anonymous_struct));
2825           type = Type(BASE_TYPE_STRUCT, anonymous_struct);
2826         }
2827       } else {
2828         ECHECK(ParseTypeFromProtoType(&type));
2829       }
2830       // Repeated elements get mapped to a vector.
2831       if (repeated) {
2832         type.element = type.base_type;
2833         type.base_type = BASE_TYPE_VECTOR;
2834         if (type.element == BASE_TYPE_VECTOR) {
2835           // We have a vector or vectors, which FlatBuffers doesn't support.
2836           // For now make it a vector of string (since the source is likely
2837           // "repeated bytes").
2838           // TODO(wvo): A better solution would be to wrap this in a table.
2839           type.element = BASE_TYPE_STRING;
2840         }
2841       }
2842       std::string name = attribute_;
2843       EXPECT(kTokenIdentifier);
2844       if (!oneof) {
2845         // Parse the field id. Since we're just translating schemas, not
2846         // any kind of binary compatibility, we can safely ignore these, and
2847         // assign our own.
2848         EXPECT('=');
2849         EXPECT(kTokenIntegerConstant);
2850       }
2851       FieldDef *field = nullptr;
2852       if (isextend) {
2853         // We allow a field to be re-defined when extending.
2854         // TODO: are there situations where that is problematic?
2855         field = struct_def->fields.Lookup(name);
2856       }
2857       if (!field) ECHECK(AddField(*struct_def, name, type, &field));
2858       field->doc_comment = field_comment;
2859       if (!IsScalar(type.base_type) && required) {
2860         field->presence = FieldDef::kRequired;
2861       }
2862       // See if there's a default specified.
2863       if (Is('[')) {
2864         NEXT();
2865         for (;;) {
2866           auto key = attribute_;
2867           ECHECK(ParseProtoKey());
2868           EXPECT('=');
2869           auto val = attribute_;
2870           ECHECK(ParseProtoCurliesOrIdent());
2871           if (key == "default") {
2872             // Temp: skip non-numeric and non-boolean defaults (enums).
2873             auto numeric = strpbrk(val.c_str(), "0123456789-+.");
2874             if (IsScalar(type.base_type) && numeric == val.c_str()) {
2875               field->value.constant = val;
2876             } else if (val == "true") {
2877               field->value.constant = val;
2878             }  // "false" is default, no need to handle explicitly.
2879           } else if (key == "deprecated") {
2880             field->deprecated = val == "true";
2881           }
2882           if (!Is(',')) break;
2883           NEXT();
2884         }
2885         EXPECT(']');
2886       }
2887       if (anonymous_struct) {
2888         ECHECK(ParseProtoFields(anonymous_struct, false, oneof));
2889         if (Is(';')) NEXT();
2890       } else if (oneof_union) {
2891         // Parse into a temporary StructDef, then transfer fields into an
2892         // EnumDef describing the oneof as a union.
2893         StructDef oneof_struct;
2894         ECHECK(ParseProtoFields(&oneof_struct, false, oneof));
2895         if (Is(';')) NEXT();
2896         for (auto field_it = oneof_struct.fields.vec.begin();
2897              field_it != oneof_struct.fields.vec.end(); ++field_it) {
2898           const auto &oneof_field = **field_it;
2899           const auto &oneof_type = oneof_field.value.type;
2900           if (oneof_type.base_type != BASE_TYPE_STRUCT ||
2901               !oneof_type.struct_def || oneof_type.struct_def->fixed)
2902             return Error("oneof '" + name +
2903                          "' cannot be mapped to a union because member '" +
2904                          oneof_field.name + "' is not a table type.");
2905           EnumValBuilder evb(*this, *oneof_union);
2906           auto ev = evb.CreateEnumerator(oneof_type.struct_def->name);
2907           ev->union_type = oneof_type;
2908           ev->doc_comment = oneof_field.doc_comment;
2909           ECHECK(evb.AcceptEnumerator(oneof_field.name));
2910         }
2911       } else {
2912         EXPECT(';');
2913       }
2914     }
2915   }
2916   NEXT();
2917   return NoError();
2918 }
2919 
ParseProtoKey()2920 CheckedError Parser::ParseProtoKey() {
2921   if (token_ == '(') {
2922     NEXT();
2923     // Skip "(a.b)" style custom attributes.
2924     while (token_ == '.' || token_ == kTokenIdentifier) NEXT();
2925     EXPECT(')');
2926     while (Is('.')) {
2927       NEXT();
2928       EXPECT(kTokenIdentifier);
2929     }
2930   } else {
2931     EXPECT(kTokenIdentifier);
2932   }
2933   return NoError();
2934 }
2935 
ParseProtoCurliesOrIdent()2936 CheckedError Parser::ParseProtoCurliesOrIdent() {
2937   if (Is('{')) {
2938     NEXT();
2939     for (int nesting = 1; nesting;) {
2940       if (token_ == '{')
2941         nesting++;
2942       else if (token_ == '}')
2943         nesting--;
2944       NEXT();
2945     }
2946   } else {
2947     NEXT();  // Any single token.
2948   }
2949   return NoError();
2950 }
2951 
ParseProtoOption()2952 CheckedError Parser::ParseProtoOption() {
2953   NEXT();
2954   ECHECK(ParseProtoKey());
2955   EXPECT('=');
2956   ECHECK(ParseProtoCurliesOrIdent());
2957   return NoError();
2958 }
2959 
2960 // Parse a protobuf type, and map it to the corresponding FlatBuffer one.
ParseTypeFromProtoType(Type * type)2961 CheckedError Parser::ParseTypeFromProtoType(Type *type) {
2962   struct type_lookup {
2963     const char *proto_type;
2964     BaseType fb_type, element;
2965   };
2966   static type_lookup lookup[] = {
2967     { "float", BASE_TYPE_FLOAT, BASE_TYPE_NONE },
2968     { "double", BASE_TYPE_DOUBLE, BASE_TYPE_NONE },
2969     { "int32", BASE_TYPE_INT, BASE_TYPE_NONE },
2970     { "int64", BASE_TYPE_LONG, BASE_TYPE_NONE },
2971     { "uint32", BASE_TYPE_UINT, BASE_TYPE_NONE },
2972     { "uint64", BASE_TYPE_ULONG, BASE_TYPE_NONE },
2973     { "sint32", BASE_TYPE_INT, BASE_TYPE_NONE },
2974     { "sint64", BASE_TYPE_LONG, BASE_TYPE_NONE },
2975     { "fixed32", BASE_TYPE_UINT, BASE_TYPE_NONE },
2976     { "fixed64", BASE_TYPE_ULONG, BASE_TYPE_NONE },
2977     { "sfixed32", BASE_TYPE_INT, BASE_TYPE_NONE },
2978     { "sfixed64", BASE_TYPE_LONG, BASE_TYPE_NONE },
2979     { "bool", BASE_TYPE_BOOL, BASE_TYPE_NONE },
2980     { "string", BASE_TYPE_STRING, BASE_TYPE_NONE },
2981     { "bytes", BASE_TYPE_VECTOR, BASE_TYPE_UCHAR },
2982     { nullptr, BASE_TYPE_NONE, BASE_TYPE_NONE }
2983   };
2984   for (auto tl = lookup; tl->proto_type; tl++) {
2985     if (attribute_ == tl->proto_type) {
2986       type->base_type = tl->fb_type;
2987       type->element = tl->element;
2988       NEXT();
2989       return NoError();
2990     }
2991   }
2992   if (Is('.')) NEXT();  // qualified names may start with a . ?
2993   ECHECK(ParseTypeIdent(*type));
2994   return NoError();
2995 }
2996 
SkipAnyJsonValue()2997 CheckedError Parser::SkipAnyJsonValue() {
2998   ParseDepthGuard depth_guard(this);
2999   ECHECK(depth_guard.Check());
3000 
3001   switch (token_) {
3002     case '{': {
3003       size_t fieldn_outer = 0;
3004       return ParseTableDelimiters(fieldn_outer, nullptr,
3005                                   [&](const std::string &, size_t &fieldn,
3006                                       const StructDef *) -> CheckedError {
3007                                     ECHECK(SkipAnyJsonValue());
3008                                     fieldn++;
3009                                     return NoError();
3010                                   });
3011     }
3012     case '[': {
3013       uoffset_t count = 0;
3014       return ParseVectorDelimiters(count, [&](uoffset_t &) -> CheckedError {
3015         return SkipAnyJsonValue();
3016       });
3017     }
3018     case kTokenStringConstant:
3019     case kTokenIntegerConstant:
3020     case kTokenFloatConstant: NEXT(); break;
3021     default:
3022       if (IsIdent("true") || IsIdent("false") || IsIdent("null")) {
3023         NEXT();
3024       } else
3025         return TokenError();
3026   }
3027   return NoError();
3028 }
3029 
ParseFlexBufferNumericConstant(flexbuffers::Builder * builder)3030 CheckedError Parser::ParseFlexBufferNumericConstant(
3031     flexbuffers::Builder *builder) {
3032   double d;
3033   if (!StringToNumber(attribute_.c_str(), &d))
3034     return Error("unexpected floating-point constant: " + attribute_);
3035   builder->Double(d);
3036   return NoError();
3037 }
3038 
ParseFlexBufferValue(flexbuffers::Builder * builder)3039 CheckedError Parser::ParseFlexBufferValue(flexbuffers::Builder *builder) {
3040   ParseDepthGuard depth_guard(this);
3041   ECHECK(depth_guard.Check());
3042 
3043   switch (token_) {
3044     case '{': {
3045       auto start = builder->StartMap();
3046       size_t fieldn_outer = 0;
3047       auto err =
3048           ParseTableDelimiters(fieldn_outer, nullptr,
3049                                [&](const std::string &name, size_t &fieldn,
3050                                    const StructDef *) -> CheckedError {
3051                                  builder->Key(name);
3052                                  ECHECK(ParseFlexBufferValue(builder));
3053                                  fieldn++;
3054                                  return NoError();
3055                                });
3056       ECHECK(err);
3057       builder->EndMap(start);
3058       if (builder->HasDuplicateKeys())
3059         return Error("FlexBuffers map has duplicate keys");
3060       break;
3061     }
3062     case '[': {
3063       auto start = builder->StartVector();
3064       uoffset_t count = 0;
3065       ECHECK(ParseVectorDelimiters(count, [&](uoffset_t &) -> CheckedError {
3066         return ParseFlexBufferValue(builder);
3067       }));
3068       builder->EndVector(start, false, false);
3069       break;
3070     }
3071     case kTokenStringConstant:
3072       builder->String(attribute_);
3073       EXPECT(kTokenStringConstant);
3074       break;
3075     case kTokenIntegerConstant:
3076       builder->Int(StringToInt(attribute_.c_str()));
3077       EXPECT(kTokenIntegerConstant);
3078       break;
3079     case kTokenFloatConstant: {
3080       double d;
3081       StringToNumber(attribute_.c_str(), &d);
3082       builder->Double(d);
3083       EXPECT(kTokenFloatConstant);
3084       break;
3085     }
3086     case '-':
3087     case '+': {
3088       // `[-+]?(nan|inf|infinity)`, see ParseSingleValue().
3089       const auto sign = static_cast<char>(token_);
3090       NEXT();
3091       if (token_ != kTokenIdentifier)
3092         return Error("floating-point constant expected");
3093       attribute_.insert(0, 1, sign);
3094       ECHECK(ParseFlexBufferNumericConstant(builder));
3095       NEXT();
3096       break;
3097     }
3098     default:
3099       if (IsIdent("true")) {
3100         builder->Bool(true);
3101         NEXT();
3102       } else if (IsIdent("false")) {
3103         builder->Bool(false);
3104         NEXT();
3105       } else if (IsIdent("null")) {
3106         builder->Null();
3107         NEXT();
3108       } else if (IsIdent("inf") || IsIdent("infinity") || IsIdent("nan")) {
3109         ECHECK(ParseFlexBufferNumericConstant(builder));
3110         NEXT();
3111       } else
3112         return TokenError();
3113   }
3114   return NoError();
3115 }
3116 
ParseFlexBuffer(const char * source,const char * source_filename,flexbuffers::Builder * builder)3117 bool Parser::ParseFlexBuffer(const char *source, const char *source_filename,
3118                              flexbuffers::Builder *builder) {
3119   const auto initial_depth = parse_depth_counter_;
3120   (void)initial_depth;
3121   auto ok = !StartParseFile(source, source_filename).Check() &&
3122             !ParseFlexBufferValue(builder).Check();
3123   if (ok) builder->Finish();
3124   FLATBUFFERS_ASSERT(initial_depth == parse_depth_counter_);
3125   return ok;
3126 }
3127 
Parse(const char * source,const char ** include_paths,const char * source_filename)3128 bool Parser::Parse(const char *source, const char **include_paths,
3129                    const char *source_filename) {
3130   const auto initial_depth = parse_depth_counter_;
3131   (void)initial_depth;
3132   bool r;
3133 
3134   if (opts.use_flexbuffers) {
3135     r = ParseFlexBuffer(source, source_filename, &flex_builder_);
3136   } else {
3137     r = !ParseRoot(source, include_paths, source_filename).Check();
3138   }
3139   FLATBUFFERS_ASSERT(initial_depth == parse_depth_counter_);
3140   return r;
3141 }
3142 
ParseJson(const char * json,const char * json_filename)3143 bool Parser::ParseJson(const char *json, const char *json_filename) {
3144   const auto initial_depth = parse_depth_counter_;
3145   (void)initial_depth;
3146   builder_.Clear();
3147   const auto done =
3148       !StartParseFile(json, json_filename).Check() && !DoParseJson().Check();
3149   FLATBUFFERS_ASSERT(initial_depth == parse_depth_counter_);
3150   return done;
3151 }
3152 
StartParseFile(const char * source,const char * source_filename)3153 CheckedError Parser::StartParseFile(const char *source,
3154                                     const char *source_filename) {
3155   file_being_parsed_ = source_filename ? source_filename : "";
3156   source_ = source;
3157   ResetState(source_);
3158   error_.clear();
3159   ECHECK(SkipByteOrderMark());
3160   NEXT();
3161   if (Is(kTokenEof)) return Error("input file is empty");
3162   return NoError();
3163 }
3164 
ParseRoot(const char * source,const char ** include_paths,const char * source_filename)3165 CheckedError Parser::ParseRoot(const char *source, const char **include_paths,
3166                                const char *source_filename) {
3167   ECHECK(DoParse(source, include_paths, source_filename, nullptr));
3168 
3169   // Check that all types were defined.
3170   for (auto it = structs_.vec.begin(); it != structs_.vec.end();) {
3171     auto &struct_def = **it;
3172     if (struct_def.predecl) {
3173       if (opts.proto_mode) {
3174         // Protos allow enums to be used before declaration, so check if that
3175         // is the case here.
3176         EnumDef *enum_def = nullptr;
3177         for (size_t components =
3178                  struct_def.defined_namespace->components.size() + 1;
3179              components && !enum_def; components--) {
3180           auto qualified_name =
3181               struct_def.defined_namespace->GetFullyQualifiedName(
3182                   struct_def.name, components - 1);
3183           enum_def = LookupEnum(qualified_name);
3184         }
3185         if (enum_def) {
3186           // This is pretty slow, but a simple solution for now.
3187           auto initial_count = struct_def.refcount;
3188           for (auto struct_it = structs_.vec.begin();
3189                struct_it != structs_.vec.end(); ++struct_it) {
3190             auto &sd = **struct_it;
3191             for (auto field_it = sd.fields.vec.begin();
3192                  field_it != sd.fields.vec.end(); ++field_it) {
3193               auto &field = **field_it;
3194               if (field.value.type.struct_def == &struct_def) {
3195                 field.value.type.struct_def = nullptr;
3196                 field.value.type.enum_def = enum_def;
3197                 auto &bt = IsVector(field.value.type)
3198                                ? field.value.type.element
3199                                : field.value.type.base_type;
3200                 FLATBUFFERS_ASSERT(bt == BASE_TYPE_STRUCT);
3201                 bt = enum_def->underlying_type.base_type;
3202                 struct_def.refcount--;
3203                 enum_def->refcount++;
3204               }
3205             }
3206           }
3207           if (struct_def.refcount)
3208             return Error("internal: " + NumToString(struct_def.refcount) + "/" +
3209                          NumToString(initial_count) +
3210                          " use(s) of pre-declaration enum not accounted for: " +
3211                          enum_def->name);
3212           structs_.dict.erase(structs_.dict.find(struct_def.name));
3213           it = structs_.vec.erase(it);
3214           delete &struct_def;
3215           continue;  // Skip error.
3216         }
3217       }
3218       auto err = "type referenced but not defined (check namespace): " +
3219                  struct_def.name;
3220       if (struct_def.original_location)
3221         err += ", originally at: " + *struct_def.original_location;
3222       return Error(err);
3223     }
3224     ++it;
3225   }
3226 
3227   // This check has to happen here and not earlier, because only now do we
3228   // know for sure what the type of these are.
3229   for (auto it = enums_.vec.begin(); it != enums_.vec.end(); ++it) {
3230     auto &enum_def = **it;
3231     if (enum_def.is_union) {
3232       for (auto val_it = enum_def.Vals().begin();
3233            val_it != enum_def.Vals().end(); ++val_it) {
3234         auto &val = **val_it;
3235         if (!SupportsAdvancedUnionFeatures() &&
3236             (IsStruct(val.union_type) || IsString(val.union_type)))
3237           return Error(
3238               "only tables can be union elements in the generated language: " +
3239               val.name);
3240       }
3241     }
3242   }
3243   // Parse JSON object only if the scheme has been parsed.
3244   if (token_ == '{') { ECHECK(DoParseJson()); }
3245   EXPECT(kTokenEof);
3246   return NoError();
3247 }
3248 
3249 // Generate a unique hash for a file based on its name and contents (if any).
HashFile(const char * source_filename,const char * source)3250 static uint64_t HashFile(const char *source_filename, const char *source) {
3251   uint64_t hash = 0;
3252 
3253   if (source_filename)
3254     hash = HashFnv1a<uint64_t>(StripPath(source_filename).c_str());
3255 
3256   if (source && *source) hash ^= HashFnv1a<uint64_t>(source);
3257 
3258   return hash;
3259 }
3260 
DoParse(const char * source,const char ** include_paths,const char * source_filename,const char * include_filename)3261 CheckedError Parser::DoParse(const char *source, const char **include_paths,
3262                              const char *source_filename,
3263                              const char *include_filename) {
3264   uint64_t source_hash = 0;
3265   if (source_filename) {
3266     // If the file is in-memory, don't include its contents in the hash as we
3267     // won't be able to load them later.
3268     if (FileExists(source_filename))
3269       source_hash = HashFile(source_filename, source);
3270     else
3271       source_hash = HashFile(source_filename, nullptr);
3272 
3273     if (included_files_.find(source_hash) == included_files_.end()) {
3274       included_files_[source_hash] = include_filename ? include_filename : "";
3275       files_included_per_file_[source_filename] = std::set<std::string>();
3276     } else {
3277       return NoError();
3278     }
3279   }
3280   if (!include_paths) {
3281     static const char *current_directory[] = { "", nullptr };
3282     include_paths = current_directory;
3283   }
3284   field_stack_.clear();
3285   builder_.Clear();
3286   // Start with a blank namespace just in case this file doesn't have one.
3287   current_namespace_ = empty_namespace_;
3288 
3289   ECHECK(StartParseFile(source, source_filename));
3290 
3291   // Includes must come before type declarations:
3292   for (;;) {
3293     // Parse pre-include proto statements if any:
3294     if (opts.proto_mode && (attribute_ == "option" || attribute_ == "syntax" ||
3295                             attribute_ == "package")) {
3296       ECHECK(ParseProtoDecl());
3297     } else if (IsIdent("native_include")) {
3298       NEXT();
3299       vector_emplace_back(&native_included_files_, attribute_);
3300       EXPECT(kTokenStringConstant);
3301       EXPECT(';');
3302     } else if (IsIdent("include") || (opts.proto_mode && IsIdent("import"))) {
3303       NEXT();
3304       if (opts.proto_mode && attribute_ == "public") NEXT();
3305       auto name = flatbuffers::PosixPath(attribute_.c_str());
3306       EXPECT(kTokenStringConstant);
3307       // Look for the file relative to the directory of the current file.
3308       std::string filepath;
3309       if (source_filename) {
3310         auto source_file_directory =
3311             flatbuffers::StripFileName(source_filename);
3312         filepath = flatbuffers::ConCatPathFileName(source_file_directory, name);
3313       }
3314       if (filepath.empty() || !FileExists(filepath.c_str())) {
3315         // Look for the file in include_paths.
3316         for (auto paths = include_paths; paths && *paths; paths++) {
3317           filepath = flatbuffers::ConCatPathFileName(*paths, name);
3318           if (FileExists(filepath.c_str())) break;
3319         }
3320       }
3321       if (filepath.empty())
3322         return Error("unable to locate include file: " + name);
3323       if (source_filename)
3324         files_included_per_file_[source_filename].insert(filepath);
3325 
3326       std::string contents;
3327       bool file_loaded = LoadFile(filepath.c_str(), true, &contents);
3328       if (included_files_.find(HashFile(filepath.c_str(), contents.c_str())) ==
3329           included_files_.end()) {
3330         // We found an include file that we have not parsed yet.
3331         // Parse it.
3332         if (!file_loaded) return Error("unable to load include file: " + name);
3333         ECHECK(DoParse(contents.c_str(), include_paths, filepath.c_str(),
3334                        name.c_str()));
3335         // We generally do not want to output code for any included files:
3336         if (!opts.generate_all) MarkGenerated();
3337         // Reset these just in case the included file had them, and the
3338         // parent doesn't.
3339         root_struct_def_ = nullptr;
3340         file_identifier_.clear();
3341         file_extension_.clear();
3342         // This is the easiest way to continue this file after an include:
3343         // instead of saving and restoring all the state, we simply start the
3344         // file anew. This will cause it to encounter the same include
3345         // statement again, but this time it will skip it, because it was
3346         // entered into included_files_.
3347         // This is recursive, but only go as deep as the number of include
3348         // statements.
3349         included_files_.erase(source_hash);
3350         return DoParse(source, include_paths, source_filename,
3351                        include_filename);
3352       }
3353       EXPECT(';');
3354     } else {
3355       break;
3356     }
3357   }
3358   // Now parse all other kinds of declarations:
3359   while (token_ != kTokenEof) {
3360     if (opts.proto_mode) {
3361       ECHECK(ParseProtoDecl());
3362     } else if (IsIdent("namespace")) {
3363       ECHECK(ParseNamespace());
3364     } else if (token_ == '{') {
3365       return NoError();
3366     } else if (IsIdent("enum")) {
3367       ECHECK(ParseEnum(false, nullptr));
3368     } else if (IsIdent("union")) {
3369       ECHECK(ParseEnum(true, nullptr));
3370     } else if (IsIdent("root_type")) {
3371       NEXT();
3372       auto root_type = attribute_;
3373       EXPECT(kTokenIdentifier);
3374       ECHECK(ParseNamespacing(&root_type, nullptr));
3375       if (opts.root_type.empty()) {
3376         if (!SetRootType(root_type.c_str()))
3377           return Error("unknown root type: " + root_type);
3378         if (root_struct_def_->fixed) return Error("root type must be a table");
3379       }
3380       EXPECT(';');
3381     } else if (IsIdent("file_identifier")) {
3382       NEXT();
3383       file_identifier_ = attribute_;
3384       EXPECT(kTokenStringConstant);
3385       if (file_identifier_.length() != FlatBufferBuilder::kFileIdentifierLength)
3386         return Error("file_identifier must be exactly " +
3387                      NumToString(FlatBufferBuilder::kFileIdentifierLength) +
3388                      " characters");
3389       EXPECT(';');
3390     } else if (IsIdent("file_extension")) {
3391       NEXT();
3392       file_extension_ = attribute_;
3393       EXPECT(kTokenStringConstant);
3394       EXPECT(';');
3395     } else if (IsIdent("include")) {
3396       return Error("includes must come before declarations");
3397     } else if (IsIdent("attribute")) {
3398       NEXT();
3399       auto name = attribute_;
3400       if (Is(kTokenIdentifier)) {
3401         NEXT();
3402       } else {
3403         EXPECT(kTokenStringConstant);
3404       }
3405       EXPECT(';');
3406       known_attributes_[name] = false;
3407     } else if (IsIdent("rpc_service")) {
3408       ECHECK(ParseService());
3409     } else {
3410       ECHECK(ParseDecl());
3411     }
3412   }
3413   return NoError();
3414 }
3415 
DoParseJson()3416 CheckedError Parser::DoParseJson() {
3417   if (token_ != '{') {
3418     EXPECT('{');
3419   } else {
3420     if (!root_struct_def_) return Error("no root type set to parse json with");
3421     if (builder_.GetSize()) {
3422       return Error("cannot have more than one json object in a file");
3423     }
3424     uoffset_t toff;
3425     ECHECK(ParseTable(*root_struct_def_, nullptr, &toff));
3426     if (opts.size_prefixed) {
3427       builder_.FinishSizePrefixed(
3428           Offset<Table>(toff),
3429           file_identifier_.length() ? file_identifier_.c_str() : nullptr);
3430     } else {
3431       builder_.Finish(Offset<Table>(toff), file_identifier_.length()
3432                                                ? file_identifier_.c_str()
3433                                                : nullptr);
3434     }
3435   }
3436   // Check that JSON file doesn't contain more objects or IDL directives.
3437   // Comments after JSON are allowed.
3438   EXPECT(kTokenEof);
3439   return NoError();
3440 }
3441 
GetIncludedFilesRecursive(const std::string & file_name) const3442 std::set<std::string> Parser::GetIncludedFilesRecursive(
3443     const std::string &file_name) const {
3444   std::set<std::string> included_files;
3445   std::list<std::string> to_process;
3446 
3447   if (file_name.empty()) return included_files;
3448   to_process.push_back(file_name);
3449 
3450   while (!to_process.empty()) {
3451     std::string current = to_process.front();
3452     to_process.pop_front();
3453     included_files.insert(current);
3454 
3455     // Workaround the lack of const accessor in C++98 maps.
3456     auto &new_files =
3457         (*const_cast<std::map<std::string, std::set<std::string>> *>(
3458             &files_included_per_file_))[current];
3459     for (auto it = new_files.begin(); it != new_files.end(); ++it) {
3460       if (included_files.find(*it) == included_files.end())
3461         to_process.push_back(*it);
3462     }
3463   }
3464 
3465   return included_files;
3466 }
3467 
3468 // Schema serialization functionality:
3469 
compareName(const T * a,const T * b)3470 template<typename T> bool compareName(const T *a, const T *b) {
3471   return a->defined_namespace->GetFullyQualifiedName(a->name) <
3472          b->defined_namespace->GetFullyQualifiedName(b->name);
3473 }
3474 
AssignIndices(const std::vector<T * > & defvec)3475 template<typename T> void AssignIndices(const std::vector<T *> &defvec) {
3476   // Pre-sort these vectors, such that we can set the correct indices for them.
3477   auto vec = defvec;
3478   std::sort(vec.begin(), vec.end(), compareName<T>);
3479   for (int i = 0; i < static_cast<int>(vec.size()); i++) vec[i]->index = i;
3480 }
3481 
Serialize()3482 void Parser::Serialize() {
3483   builder_.Clear();
3484   AssignIndices(structs_.vec);
3485   AssignIndices(enums_.vec);
3486   std::vector<Offset<reflection::Object>> object_offsets;
3487   for (auto it = structs_.vec.begin(); it != structs_.vec.end(); ++it) {
3488     auto offset = (*it)->Serialize(&builder_, *this);
3489     object_offsets.push_back(offset);
3490     (*it)->serialized_location = offset.o;
3491   }
3492   std::vector<Offset<reflection::Enum>> enum_offsets;
3493   for (auto it = enums_.vec.begin(); it != enums_.vec.end(); ++it) {
3494     auto offset = (*it)->Serialize(&builder_, *this);
3495     enum_offsets.push_back(offset);
3496     (*it)->serialized_location = offset.o;
3497   }
3498   std::vector<Offset<reflection::Service>> service_offsets;
3499   for (auto it = services_.vec.begin(); it != services_.vec.end(); ++it) {
3500     auto offset = (*it)->Serialize(&builder_, *this);
3501     service_offsets.push_back(offset);
3502     (*it)->serialized_location = offset.o;
3503   }
3504   auto objs__ = builder_.CreateVectorOfSortedTables(&object_offsets);
3505   auto enum__ = builder_.CreateVectorOfSortedTables(&enum_offsets);
3506   auto fiid__ = builder_.CreateString(file_identifier_);
3507   auto fext__ = builder_.CreateString(file_extension_);
3508   auto serv__ = builder_.CreateVectorOfSortedTables(&service_offsets);
3509   auto schema_offset = reflection::CreateSchema(
3510       builder_, objs__, enum__, fiid__, fext__,
3511       (root_struct_def_ ? root_struct_def_->serialized_location : 0), serv__,
3512       static_cast<reflection::AdvancedFeatures>(advanced_features_));
3513   if (opts.size_prefixed) {
3514     builder_.FinishSizePrefixed(schema_offset, reflection::SchemaIdentifier());
3515   } else {
3516     builder_.Finish(schema_offset, reflection::SchemaIdentifier());
3517   }
3518 }
3519 
GetNamespace(const std::string & qualified_name,std::vector<Namespace * > & namespaces,std::map<std::string,Namespace * > & namespaces_index)3520 static Namespace *GetNamespace(
3521     const std::string &qualified_name, std::vector<Namespace *> &namespaces,
3522     std::map<std::string, Namespace *> &namespaces_index) {
3523   size_t dot = qualified_name.find_last_of('.');
3524   std::string namespace_name = (dot != std::string::npos)
3525                                    ? std::string(qualified_name.c_str(), dot)
3526                                    : "";
3527   Namespace *&ns = namespaces_index[namespace_name];
3528 
3529   if (!ns) {
3530     ns = new Namespace();
3531     namespaces.push_back(ns);
3532 
3533     size_t pos = 0;
3534 
3535     for (;;) {
3536       dot = qualified_name.find('.', pos);
3537       if (dot == std::string::npos) { break; }
3538       ns->components.push_back(qualified_name.substr(pos, dot - pos));
3539       pos = dot + 1;
3540     }
3541   }
3542 
3543   return ns;
3544 }
3545 
Serialize(FlatBufferBuilder * builder,const Parser & parser) const3546 Offset<reflection::Object> StructDef::Serialize(FlatBufferBuilder *builder,
3547                                                 const Parser &parser) const {
3548   std::vector<Offset<reflection::Field>> field_offsets;
3549   for (auto it = fields.vec.begin(); it != fields.vec.end(); ++it) {
3550     field_offsets.push_back((*it)->Serialize(
3551         builder, static_cast<uint16_t>(it - fields.vec.begin()), parser));
3552   }
3553   auto qualified_name = defined_namespace->GetFullyQualifiedName(name);
3554   auto name__ = builder->CreateString(qualified_name);
3555   auto flds__ = builder->CreateVectorOfSortedTables(&field_offsets);
3556   auto attr__ = SerializeAttributes(builder, parser);
3557   auto docs__ = parser.opts.binary_schema_comments
3558                     ? builder->CreateVectorOfStrings(doc_comment)
3559                     : 0;
3560   return reflection::CreateObject(*builder, name__, flds__, fixed,
3561                                   static_cast<int>(minalign),
3562                                   static_cast<int>(bytesize), attr__, docs__);
3563 }
3564 
Deserialize(Parser & parser,const reflection::Object * object)3565 bool StructDef::Deserialize(Parser &parser, const reflection::Object *object) {
3566   if (!DeserializeAttributes(parser, object->attributes())) return false;
3567   DeserializeDoc(doc_comment, object->documentation());
3568   name = parser.UnqualifiedName(object->name()->str());
3569   predecl = false;
3570   sortbysize = attributes.Lookup("original_order") == nullptr && !fixed;
3571   const auto &of = *(object->fields());
3572   auto indexes = std::vector<uoffset_t>(of.size());
3573   for (uoffset_t i = 0; i < of.size(); i++) indexes[of.Get(i)->id()] = i;
3574   size_t tmp_struct_size = 0;
3575   for (size_t i = 0; i < indexes.size(); i++) {
3576     auto field = of.Get(indexes[i]);
3577     auto field_def = new FieldDef();
3578     if (!field_def->Deserialize(parser, field) ||
3579         fields.Add(field_def->name, field_def)) {
3580       delete field_def;
3581       return false;
3582     }
3583     if (fixed) {
3584       // Recompute padding since that's currently not serialized.
3585       auto size = InlineSize(field_def->value.type);
3586       auto next_field =
3587           i + 1 < indexes.size() ? of.Get(indexes[i + 1]) : nullptr;
3588       tmp_struct_size += size;
3589       field_def->padding =
3590           next_field ? (next_field->offset() - field_def->value.offset) - size
3591                      : PaddingBytes(tmp_struct_size, minalign);
3592       tmp_struct_size += field_def->padding;
3593     }
3594   }
3595   FLATBUFFERS_ASSERT(static_cast<int>(tmp_struct_size) == object->bytesize());
3596   return true;
3597 }
3598 
Serialize(FlatBufferBuilder * builder,uint16_t id,const Parser & parser) const3599 Offset<reflection::Field> FieldDef::Serialize(FlatBufferBuilder *builder,
3600                                               uint16_t id,
3601                                               const Parser &parser) const {
3602   auto name__ = builder->CreateString(name);
3603   auto type__ = value.type.Serialize(builder);
3604   auto attr__ = SerializeAttributes(builder, parser);
3605   auto docs__ = parser.opts.binary_schema_comments
3606                     ? builder->CreateVectorOfStrings(doc_comment)
3607                     : 0;
3608   double d;
3609   StringToNumber(value.constant.c_str(), &d);
3610   return reflection::CreateField(
3611       *builder, name__, type__, id, value.offset,
3612       // Is uint64>max(int64) tested?
3613       IsInteger(value.type.base_type) ? StringToInt(value.constant.c_str()) : 0,
3614       // result may be platform-dependent if underlying is float (not double)
3615       IsFloat(value.type.base_type) ? d : 0.0, deprecated, IsRequired(), key,
3616       attr__, docs__, IsOptional());
3617   // TODO: value.constant is almost always "0", we could save quite a bit of
3618   // space by sharing it. Same for common values of value.type.
3619 }
3620 
Deserialize(Parser & parser,const reflection::Field * field)3621 bool FieldDef::Deserialize(Parser &parser, const reflection::Field *field) {
3622   name = field->name()->str();
3623   defined_namespace = parser.current_namespace_;
3624   if (!value.type.Deserialize(parser, field->type())) return false;
3625   value.offset = field->offset();
3626   if (IsInteger(value.type.base_type)) {
3627     value.constant = NumToString(field->default_integer());
3628   } else if (IsFloat(value.type.base_type)) {
3629     value.constant = FloatToString(field->default_real(), 16);
3630   }
3631   presence = FieldDef::MakeFieldPresence(field->optional(), field->required());
3632   key = field->key();
3633   if (!DeserializeAttributes(parser, field->attributes())) return false;
3634   // TODO: this should probably be handled by a separate attribute
3635   if (attributes.Lookup("flexbuffer")) {
3636     flexbuffer = true;
3637     parser.uses_flexbuffers_ = true;
3638     if (value.type.base_type != BASE_TYPE_VECTOR ||
3639         value.type.element != BASE_TYPE_UCHAR)
3640       return false;
3641   }
3642   if (auto nested = attributes.Lookup("nested_flatbuffer")) {
3643     auto nested_qualified_name =
3644         parser.current_namespace_->GetFullyQualifiedName(nested->constant);
3645     nested_flatbuffer = parser.LookupStruct(nested_qualified_name);
3646     if (!nested_flatbuffer) return false;
3647   }
3648   shared = attributes.Lookup("shared") != nullptr;
3649   DeserializeDoc(doc_comment, field->documentation());
3650   return true;
3651 }
3652 
Serialize(FlatBufferBuilder * builder,const Parser & parser) const3653 Offset<reflection::RPCCall> RPCCall::Serialize(FlatBufferBuilder *builder,
3654                                                const Parser &parser) const {
3655   auto name__ = builder->CreateString(name);
3656   auto attr__ = SerializeAttributes(builder, parser);
3657   auto docs__ = parser.opts.binary_schema_comments
3658                     ? builder->CreateVectorOfStrings(doc_comment)
3659                     : 0;
3660   return reflection::CreateRPCCall(
3661       *builder, name__, request->serialized_location,
3662       response->serialized_location, attr__, docs__);
3663 }
3664 
Deserialize(Parser & parser,const reflection::RPCCall * call)3665 bool RPCCall::Deserialize(Parser &parser, const reflection::RPCCall *call) {
3666   name = call->name()->str();
3667   if (!DeserializeAttributes(parser, call->attributes())) return false;
3668   DeserializeDoc(doc_comment, call->documentation());
3669   request = parser.structs_.Lookup(call->request()->name()->str());
3670   response = parser.structs_.Lookup(call->response()->name()->str());
3671   if (!request || !response) { return false; }
3672   return true;
3673 }
3674 
Serialize(FlatBufferBuilder * builder,const Parser & parser) const3675 Offset<reflection::Service> ServiceDef::Serialize(FlatBufferBuilder *builder,
3676                                                   const Parser &parser) const {
3677   std::vector<Offset<reflection::RPCCall>> servicecall_offsets;
3678   for (auto it = calls.vec.begin(); it != calls.vec.end(); ++it) {
3679     servicecall_offsets.push_back((*it)->Serialize(builder, parser));
3680   }
3681   auto qualified_name = defined_namespace->GetFullyQualifiedName(name);
3682   auto name__ = builder->CreateString(qualified_name);
3683   auto call__ = builder->CreateVector(servicecall_offsets);
3684   auto attr__ = SerializeAttributes(builder, parser);
3685   auto docs__ = parser.opts.binary_schema_comments
3686                     ? builder->CreateVectorOfStrings(doc_comment)
3687                     : 0;
3688   return reflection::CreateService(*builder, name__, call__, attr__, docs__);
3689 }
3690 
Deserialize(Parser & parser,const reflection::Service * service)3691 bool ServiceDef::Deserialize(Parser &parser,
3692                              const reflection::Service *service) {
3693   name = parser.UnqualifiedName(service->name()->str());
3694   if (service->calls()) {
3695     for (uoffset_t i = 0; i < service->calls()->size(); ++i) {
3696       auto call = new RPCCall();
3697       if (!call->Deserialize(parser, service->calls()->Get(i)) ||
3698           calls.Add(call->name, call)) {
3699         delete call;
3700         return false;
3701       }
3702     }
3703   }
3704   if (!DeserializeAttributes(parser, service->attributes())) return false;
3705   DeserializeDoc(doc_comment, service->documentation());
3706   return true;
3707 }
3708 
Serialize(FlatBufferBuilder * builder,const Parser & parser) const3709 Offset<reflection::Enum> EnumDef::Serialize(FlatBufferBuilder *builder,
3710                                             const Parser &parser) const {
3711   std::vector<Offset<reflection::EnumVal>> enumval_offsets;
3712   for (auto it = vals.vec.begin(); it != vals.vec.end(); ++it) {
3713     enumval_offsets.push_back((*it)->Serialize(builder, parser));
3714   }
3715   auto qualified_name = defined_namespace->GetFullyQualifiedName(name);
3716   auto name__ = builder->CreateString(qualified_name);
3717   auto vals__ = builder->CreateVector(enumval_offsets);
3718   auto type__ = underlying_type.Serialize(builder);
3719   auto attr__ = SerializeAttributes(builder, parser);
3720   auto docs__ = parser.opts.binary_schema_comments
3721                     ? builder->CreateVectorOfStrings(doc_comment)
3722                     : 0;
3723   return reflection::CreateEnum(*builder, name__, vals__, is_union, type__,
3724                                 attr__, docs__);
3725 }
3726 
Deserialize(Parser & parser,const reflection::Enum * _enum)3727 bool EnumDef::Deserialize(Parser &parser, const reflection::Enum *_enum) {
3728   name = parser.UnqualifiedName(_enum->name()->str());
3729   for (uoffset_t i = 0; i < _enum->values()->size(); ++i) {
3730     auto val = new EnumVal();
3731     if (!val->Deserialize(parser, _enum->values()->Get(i)) ||
3732         vals.Add(val->name, val)) {
3733       delete val;
3734       return false;
3735     }
3736   }
3737   is_union = _enum->is_union();
3738   if (!underlying_type.Deserialize(parser, _enum->underlying_type())) {
3739     return false;
3740   }
3741   if (!DeserializeAttributes(parser, _enum->attributes())) return false;
3742   DeserializeDoc(doc_comment, _enum->documentation());
3743   return true;
3744 }
3745 
Serialize(FlatBufferBuilder * builder,const Parser & parser) const3746 Offset<reflection::EnumVal> EnumVal::Serialize(FlatBufferBuilder *builder,
3747                                                const Parser &parser) const {
3748   auto name__ = builder->CreateString(name);
3749   auto type__ = union_type.Serialize(builder);
3750   auto docs__ = parser.opts.binary_schema_comments
3751                     ? builder->CreateVectorOfStrings(doc_comment)
3752                     : 0;
3753   return reflection::CreateEnumVal(
3754       *builder, name__, value,
3755       union_type.struct_def ? union_type.struct_def->serialized_location : 0,
3756       type__, docs__);
3757 }
3758 
Deserialize(const Parser & parser,const reflection::EnumVal * val)3759 bool EnumVal::Deserialize(const Parser &parser,
3760                           const reflection::EnumVal *val) {
3761   name = val->name()->str();
3762   value = val->value();
3763   if (!union_type.Deserialize(parser, val->union_type())) return false;
3764   DeserializeDoc(doc_comment, val->documentation());
3765   return true;
3766 }
3767 
Serialize(FlatBufferBuilder * builder) const3768 Offset<reflection::Type> Type::Serialize(FlatBufferBuilder *builder) const {
3769   return reflection::CreateType(
3770       *builder, static_cast<reflection::BaseType>(base_type),
3771       static_cast<reflection::BaseType>(element),
3772       struct_def ? struct_def->index : (enum_def ? enum_def->index : -1),
3773       fixed_length);
3774 }
3775 
Deserialize(const Parser & parser,const reflection::Type * type)3776 bool Type::Deserialize(const Parser &parser, const reflection::Type *type) {
3777   if (type == nullptr) return true;
3778   base_type = static_cast<BaseType>(type->base_type());
3779   element = static_cast<BaseType>(type->element());
3780   fixed_length = type->fixed_length();
3781   if (type->index() >= 0) {
3782     bool is_series = type->base_type() == reflection::Vector ||
3783                      type->base_type() == reflection::Array;
3784     if (type->base_type() == reflection::Obj ||
3785         (is_series && type->element() == reflection::Obj)) {
3786       if (static_cast<size_t>(type->index()) < parser.structs_.vec.size()) {
3787         struct_def = parser.structs_.vec[type->index()];
3788         struct_def->refcount++;
3789       } else {
3790         return false;
3791       }
3792     } else {
3793       if (static_cast<size_t>(type->index()) < parser.enums_.vec.size()) {
3794         enum_def = parser.enums_.vec[type->index()];
3795       } else {
3796         return false;
3797       }
3798     }
3799   }
3800   return true;
3801 }
3802 
3803 flatbuffers::Offset<
3804     flatbuffers::Vector<flatbuffers::Offset<reflection::KeyValue>>>
SerializeAttributes(FlatBufferBuilder * builder,const Parser & parser) const3805 Definition::SerializeAttributes(FlatBufferBuilder *builder,
3806                                 const Parser &parser) const {
3807   std::vector<flatbuffers::Offset<reflection::KeyValue>> attrs;
3808   for (auto kv = attributes.dict.begin(); kv != attributes.dict.end(); ++kv) {
3809     auto it = parser.known_attributes_.find(kv->first);
3810     FLATBUFFERS_ASSERT(it != parser.known_attributes_.end());
3811     if (parser.opts.binary_schema_builtins || !it->second) {
3812       auto key = builder->CreateString(kv->first);
3813       auto val = builder->CreateString(kv->second->constant);
3814       attrs.push_back(reflection::CreateKeyValue(*builder, key, val));
3815     }
3816   }
3817   if (attrs.size()) {
3818     return builder->CreateVectorOfSortedTables(&attrs);
3819   } else {
3820     return 0;
3821   }
3822 }
3823 
DeserializeAttributes(Parser & parser,const Vector<Offset<reflection::KeyValue>> * attrs)3824 bool Definition::DeserializeAttributes(
3825     Parser &parser, const Vector<Offset<reflection::KeyValue>> *attrs) {
3826   if (attrs == nullptr) return true;
3827   for (uoffset_t i = 0; i < attrs->size(); ++i) {
3828     auto kv = attrs->Get(i);
3829     auto value = new Value();
3830     if (kv->value()) { value->constant = kv->value()->str(); }
3831     if (attributes.Add(kv->key()->str(), value)) {
3832       delete value;
3833       return false;
3834     }
3835     parser.known_attributes_[kv->key()->str()];
3836   }
3837   return true;
3838 }
3839 
3840 /************************************************************************/
3841 /* DESERIALIZATION                                                      */
3842 /************************************************************************/
Deserialize(const uint8_t * buf,const size_t size)3843 bool Parser::Deserialize(const uint8_t *buf, const size_t size) {
3844   flatbuffers::Verifier verifier(reinterpret_cast<const uint8_t *>(buf), size);
3845   bool size_prefixed = false;
3846   if (!reflection::SchemaBufferHasIdentifier(buf)) {
3847     if (!flatbuffers::BufferHasIdentifier(buf, reflection::SchemaIdentifier(),
3848                                           true))
3849       return false;
3850     else
3851       size_prefixed = true;
3852   }
3853   auto verify_fn = size_prefixed ? &reflection::VerifySizePrefixedSchemaBuffer
3854                                  : &reflection::VerifySchemaBuffer;
3855   if (!verify_fn(verifier)) { return false; }
3856   auto schema = size_prefixed ? reflection::GetSizePrefixedSchema(buf)
3857                               : reflection::GetSchema(buf);
3858   return Deserialize(schema);
3859 }
3860 
Deserialize(const reflection::Schema * schema)3861 bool Parser::Deserialize(const reflection::Schema *schema) {
3862   file_identifier_ = schema->file_ident() ? schema->file_ident()->str() : "";
3863   file_extension_ = schema->file_ext() ? schema->file_ext()->str() : "";
3864   std::map<std::string, Namespace *> namespaces_index;
3865 
3866   // Create defs without deserializing so references from fields to structs and
3867   // enums can be resolved.
3868   for (auto it = schema->objects()->begin(); it != schema->objects()->end();
3869        ++it) {
3870     auto struct_def = new StructDef();
3871     struct_def->bytesize = it->bytesize();
3872     struct_def->fixed = it->is_struct();
3873     struct_def->minalign = it->minalign();
3874     if (structs_.Add(it->name()->str(), struct_def)) {
3875       delete struct_def;
3876       return false;
3877     }
3878     auto type = new Type(BASE_TYPE_STRUCT, struct_def, nullptr);
3879     if (types_.Add(it->name()->str(), type)) {
3880       delete type;
3881       return false;
3882     }
3883   }
3884   for (auto it = schema->enums()->begin(); it != schema->enums()->end(); ++it) {
3885     auto enum_def = new EnumDef();
3886     if (enums_.Add(it->name()->str(), enum_def)) {
3887       delete enum_def;
3888       return false;
3889     }
3890     auto type = new Type(BASE_TYPE_UNION, nullptr, enum_def);
3891     if (types_.Add(it->name()->str(), type)) {
3892       delete type;
3893       return false;
3894     }
3895   }
3896 
3897   // Now fields can refer to structs and enums by index.
3898   for (auto it = schema->objects()->begin(); it != schema->objects()->end();
3899        ++it) {
3900     std::string qualified_name = it->name()->str();
3901     auto struct_def = structs_.Lookup(qualified_name);
3902     struct_def->defined_namespace =
3903         GetNamespace(qualified_name, namespaces_, namespaces_index);
3904     if (!struct_def->Deserialize(*this, *it)) { return false; }
3905     if (schema->root_table() == *it) { root_struct_def_ = struct_def; }
3906   }
3907   for (auto it = schema->enums()->begin(); it != schema->enums()->end(); ++it) {
3908     std::string qualified_name = it->name()->str();
3909     auto enum_def = enums_.Lookup(qualified_name);
3910     enum_def->defined_namespace =
3911         GetNamespace(qualified_name, namespaces_, namespaces_index);
3912     if (!enum_def->Deserialize(*this, *it)) { return false; }
3913   }
3914 
3915   if (schema->services()) {
3916     for (auto it = schema->services()->begin(); it != schema->services()->end();
3917          ++it) {
3918       std::string qualified_name = it->name()->str();
3919       auto service_def = new ServiceDef();
3920       service_def->defined_namespace =
3921           GetNamespace(qualified_name, namespaces_, namespaces_index);
3922       if (!service_def->Deserialize(*this, *it) ||
3923           services_.Add(qualified_name, service_def)) {
3924         delete service_def;
3925         return false;
3926       }
3927     }
3928   }
3929   advanced_features_ = schema->advanced_features();
3930   return true;
3931 }
3932 
ConformTo(const Parser & base)3933 std::string Parser::ConformTo(const Parser &base) {
3934   for (auto sit = structs_.vec.begin(); sit != structs_.vec.end(); ++sit) {
3935     auto &struct_def = **sit;
3936     auto qualified_name =
3937         struct_def.defined_namespace->GetFullyQualifiedName(struct_def.name);
3938     auto struct_def_base = base.LookupStruct(qualified_name);
3939     if (!struct_def_base) continue;
3940     for (auto fit = struct_def.fields.vec.begin();
3941          fit != struct_def.fields.vec.end(); ++fit) {
3942       auto &field = **fit;
3943       auto field_base = struct_def_base->fields.Lookup(field.name);
3944       if (field_base) {
3945         if (field.value.offset != field_base->value.offset)
3946           return "offsets differ for field: " + field.name;
3947         if (field.value.constant != field_base->value.constant)
3948           return "defaults differ for field: " + field.name;
3949         if (!EqualByName(field.value.type, field_base->value.type))
3950           return "types differ for field: " + field.name;
3951       } else {
3952         // Doesn't have to exist, deleting fields is fine.
3953         // But we should check if there is a field that has the same offset
3954         // but is incompatible (in the case of field renaming).
3955         for (auto fbit = struct_def_base->fields.vec.begin();
3956              fbit != struct_def_base->fields.vec.end(); ++fbit) {
3957           field_base = *fbit;
3958           if (field.value.offset == field_base->value.offset) {
3959             if (!EqualByName(field.value.type, field_base->value.type))
3960               return "field renamed to different type: " + field.name;
3961             break;
3962           }
3963         }
3964       }
3965     }
3966   }
3967   for (auto eit = enums_.vec.begin(); eit != enums_.vec.end(); ++eit) {
3968     auto &enum_def = **eit;
3969     auto qualified_name =
3970         enum_def.defined_namespace->GetFullyQualifiedName(enum_def.name);
3971     auto enum_def_base = base.enums_.Lookup(qualified_name);
3972     if (!enum_def_base) continue;
3973     for (auto evit = enum_def.Vals().begin(); evit != enum_def.Vals().end();
3974          ++evit) {
3975       auto &enum_val = **evit;
3976       auto enum_val_base = enum_def_base->Lookup(enum_val.name);
3977       if (enum_val_base) {
3978         if (enum_val != *enum_val_base)
3979           return "values differ for enum: " + enum_val.name;
3980       }
3981     }
3982   }
3983   return "";
3984 }
3985 
3986 }  // namespace flatbuffers
3987