blob: 76b7219f5413ce79f0f7eb375d0ad9719bdd02a3 [file] [log] [blame]
temporal40ee5512008-07-10 02:12:20 +00001// Protocol Buffers - Google's data interchange format
kenton@google.com24bf56f2008-09-24 20:31:01 +00002// Copyright 2008 Google Inc. All rights reserved.
temporal40ee5512008-07-10 02:12:20 +00003// http://code.google.com/p/protobuf/
4//
kenton@google.com24bf56f2008-09-24 20:31:01 +00005// Redistribution and use in source and binary forms, with or without
6// modification, are permitted provided that the following conditions are
7// met:
temporal40ee5512008-07-10 02:12:20 +00008//
kenton@google.com24bf56f2008-09-24 20:31:01 +00009// * Redistributions of source code must retain the above copyright
10// notice, this list of conditions and the following disclaimer.
11// * Redistributions in binary form must reproduce the above
12// copyright notice, this list of conditions and the following disclaimer
13// in the documentation and/or other materials provided with the
14// distribution.
15// * Neither the name of Google Inc. nor the names of its
16// contributors may be used to endorse or promote products derived from
17// this software without specific prior written permission.
temporal40ee5512008-07-10 02:12:20 +000018//
kenton@google.com24bf56f2008-09-24 20:31:01 +000019// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
20// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
21// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
22// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
23// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
24// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
25// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
26// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
27// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
28// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
29// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
temporal40ee5512008-07-10 02:12:20 +000030
31// Author: kenton@google.com (Kenton Varda)
32// Based on original Protocol Buffers design by
33// Sanjay Ghemawat, Jeff Dean, and others.
34
35#include <google/protobuf/stubs/hash.h>
36#include <map>
37#include <set>
kenton@google.com24bf56f2008-09-24 20:31:01 +000038#include <vector>
temporal40ee5512008-07-10 02:12:20 +000039#include <algorithm>
kenton@google.com684d45b2009-12-19 04:50:00 +000040#include <limits>
temporal40ee5512008-07-10 02:12:20 +000041
42#include <google/protobuf/descriptor.h>
43#include <google/protobuf/descriptor_database.h>
44#include <google/protobuf/descriptor.pb.h>
liujisi@google.com33165fe2010-11-02 13:14:58 +000045#include <google/protobuf/dynamic_message.h>
temporal40ee5512008-07-10 02:12:20 +000046#include <google/protobuf/text_format.h>
kenton@google.com24bf56f2008-09-24 20:31:01 +000047#include <google/protobuf/unknown_field_set.h>
48#include <google/protobuf/wire_format.h>
49#include <google/protobuf/io/coded_stream.h>
liujisi@google.com33165fe2010-11-02 13:14:58 +000050#include <google/protobuf/io/tokenizer.h>
kenton@google.com24bf56f2008-09-24 20:31:01 +000051#include <google/protobuf/io/zero_copy_stream_impl.h>
temporal40ee5512008-07-10 02:12:20 +000052#include <google/protobuf/stubs/common.h>
kenton@google.com63e646b2009-05-06 19:27:03 +000053#include <google/protobuf/stubs/once.h>
temporal40ee5512008-07-10 02:12:20 +000054#include <google/protobuf/stubs/strutil.h>
55#include <google/protobuf/stubs/substitute.h>
56#include <google/protobuf/stubs/map-util.h>
57#include <google/protobuf/stubs/stl_util-inl.h>
58
59#undef PACKAGE // autoheader #defines this. :(
60
61namespace google {
62namespace protobuf {
63
64const FieldDescriptor::CppType
65FieldDescriptor::kTypeToCppTypeMap[MAX_TYPE + 1] = {
66 static_cast<CppType>(0), // 0 is reserved for errors
67
68 CPPTYPE_DOUBLE, // TYPE_DOUBLE
69 CPPTYPE_FLOAT, // TYPE_FLOAT
70 CPPTYPE_INT64, // TYPE_INT64
71 CPPTYPE_UINT64, // TYPE_UINT64
72 CPPTYPE_INT32, // TYPE_INT32
73 CPPTYPE_UINT64, // TYPE_FIXED64
74 CPPTYPE_UINT32, // TYPE_FIXED32
75 CPPTYPE_BOOL, // TYPE_BOOL
76 CPPTYPE_STRING, // TYPE_STRING
77 CPPTYPE_MESSAGE, // TYPE_GROUP
78 CPPTYPE_MESSAGE, // TYPE_MESSAGE
79 CPPTYPE_STRING, // TYPE_BYTES
80 CPPTYPE_UINT32, // TYPE_UINT32
81 CPPTYPE_ENUM, // TYPE_ENUM
82 CPPTYPE_INT32, // TYPE_SFIXED32
83 CPPTYPE_INT64, // TYPE_SFIXED64
84 CPPTYPE_INT32, // TYPE_SINT32
85 CPPTYPE_INT64, // TYPE_SINT64
86};
87
88const char * const FieldDescriptor::kTypeToName[MAX_TYPE + 1] = {
89 "ERROR", // 0 is reserved for errors
90
91 "double", // TYPE_DOUBLE
92 "float", // TYPE_FLOAT
93 "int64", // TYPE_INT64
94 "uint64", // TYPE_UINT64
95 "int32", // TYPE_INT32
96 "fixed64", // TYPE_FIXED64
97 "fixed32", // TYPE_FIXED32
98 "bool", // TYPE_BOOL
99 "string", // TYPE_STRING
100 "group", // TYPE_GROUP
101 "message", // TYPE_MESSAGE
102 "bytes", // TYPE_BYTES
103 "uint32", // TYPE_UINT32
104 "enum", // TYPE_ENUM
105 "sfixed32", // TYPE_SFIXED32
106 "sfixed64", // TYPE_SFIXED64
107 "sint32", // TYPE_SINT32
108 "sint64", // TYPE_SINT64
109};
110
111const char * const FieldDescriptor::kLabelToName[MAX_LABEL + 1] = {
112 "ERROR", // 0 is reserved for errors
113
114 "optional", // LABEL_OPTIONAL
115 "required", // LABEL_REQUIRED
116 "repeated", // LABEL_REPEATED
117};
118
119#ifndef _MSC_VER // MSVC doesn't need these and won't even accept them.
120const int FieldDescriptor::kMaxNumber;
121const int FieldDescriptor::kFirstReservedNumber;
122const int FieldDescriptor::kLastReservedNumber;
123#endif
124
125namespace {
126
127const string kEmptyString;
128
kenton@google.com2d6daa72009-01-22 01:27:00 +0000129string ToCamelCase(const string& input) {
130 bool capitalize_next = false;
131 string result;
132 result.reserve(input.size());
133
134 for (int i = 0; i < input.size(); i++) {
135 if (input[i] == '_') {
136 capitalize_next = true;
137 } else if (capitalize_next) {
138 // Note: I distrust ctype.h due to locales.
139 if ('a' <= input[i] && input[i] <= 'z') {
140 result.push_back(input[i] - 'a' + 'A');
141 } else {
142 result.push_back(input[i]);
143 }
144 capitalize_next = false;
145 } else {
146 result.push_back(input[i]);
147 }
148 }
149
150 // Lower-case the first letter.
151 if (!result.empty() && 'A' <= result[0] && result[0] <= 'Z') {
152 result[0] = result[0] - 'A' + 'a';
153 }
154
155 return result;
156}
157
temporal40ee5512008-07-10 02:12:20 +0000158// A DescriptorPool contains a bunch of hash_maps to implement the
159// various Find*By*() methods. Since hashtable lookups are O(1), it's
160// most efficient to construct a fixed set of large hash_maps used by
161// all objects in the pool rather than construct one or more small
162// hash_maps for each object.
163//
164// The keys to these hash_maps are (parent, name) or (parent, number)
165// pairs. Unfortunately STL doesn't provide hash functions for pair<>,
166// so we must invent our own.
167//
168// TODO(kenton): Use StringPiece rather than const char* in keys? It would
169// be a lot cleaner but we'd just have to convert it back to const char*
170// for the open source release.
171
172typedef pair<const void*, const char*> PointerStringPair;
173
temporal40ee5512008-07-10 02:12:20 +0000174struct PointerStringPairEqual {
175 inline bool operator()(const PointerStringPair& a,
176 const PointerStringPair& b) const {
177 return a.first == b.first && strcmp(a.second, b.second) == 0;
178 }
179};
180
181template<typename PairType>
182struct PointerIntegerPairHash {
183 size_t operator()(const PairType& p) const {
184 // FIXME(kenton): What is the best way to compute this hash? I have
185 // no idea! This seems a bit better than an XOR.
186 return reinterpret_cast<intptr_t>(p.first) * ((1 << 16) - 1) + p.second;
187 }
188
189 // Used only by MSVC and platforms where hash_map is not available.
190 static const size_t bucket_size = 4;
191 static const size_t min_buckets = 8;
192 inline bool operator()(const PairType& a, const PairType& b) const {
193 return a.first < b.first ||
194 (a.first == b.first && a.second < b.second);
195 }
196};
197
198typedef pair<const Descriptor*, int> DescriptorIntPair;
199typedef pair<const EnumDescriptor*, int> EnumIntPair;
200
201struct PointerStringPairHash {
202 size_t operator()(const PointerStringPair& p) const {
203 // FIXME(kenton): What is the best way to compute this hash? I have
204 // no idea! This seems a bit better than an XOR.
205 hash<const char*> cstring_hash;
206 return reinterpret_cast<intptr_t>(p.first) * ((1 << 16) - 1) +
207 cstring_hash(p.second);
208 }
209
210 // Used only by MSVC and platforms where hash_map is not available.
211 static const size_t bucket_size = 4;
212 static const size_t min_buckets = 8;
213 inline bool operator()(const PointerStringPair& a,
214 const PointerStringPair& b) const {
215 if (a.first < b.first) return true;
216 if (a.first > b.first) return false;
217 return strcmp(a.second, b.second) < 0;
218 }
219};
220
221
222struct Symbol {
223 enum Type {
224 NULL_SYMBOL, MESSAGE, FIELD, ENUM, ENUM_VALUE, SERVICE, METHOD, PACKAGE
225 };
226 Type type;
227 union {
228 const Descriptor* descriptor;
229 const FieldDescriptor* field_descriptor;
230 const EnumDescriptor* enum_descriptor;
231 const EnumValueDescriptor* enum_value_descriptor;
232 const ServiceDescriptor* service_descriptor;
233 const MethodDescriptor* method_descriptor;
234 const FileDescriptor* package_file_descriptor;
235 };
236
237 inline Symbol() : type(NULL_SYMBOL) { descriptor = NULL; }
238 inline bool IsNull() const { return type == NULL_SYMBOL; }
kenton@google.comd37d46d2009-04-25 02:53:47 +0000239 inline bool IsType() const {
240 return type == MESSAGE || type == ENUM;
241 }
242 inline bool IsAggregate() const {
243 return type == MESSAGE || type == PACKAGE
244 || type == ENUM || type == SERVICE;
245 }
temporal40ee5512008-07-10 02:12:20 +0000246
247#define CONSTRUCTOR(TYPE, TYPE_CONSTANT, FIELD) \
248 inline explicit Symbol(const TYPE* value) { \
249 type = TYPE_CONSTANT; \
250 this->FIELD = value; \
251 }
252
253 CONSTRUCTOR(Descriptor , MESSAGE , descriptor )
254 CONSTRUCTOR(FieldDescriptor , FIELD , field_descriptor )
255 CONSTRUCTOR(EnumDescriptor , ENUM , enum_descriptor )
256 CONSTRUCTOR(EnumValueDescriptor, ENUM_VALUE, enum_value_descriptor )
257 CONSTRUCTOR(ServiceDescriptor , SERVICE , service_descriptor )
258 CONSTRUCTOR(MethodDescriptor , METHOD , method_descriptor )
259 CONSTRUCTOR(FileDescriptor , PACKAGE , package_file_descriptor)
260#undef CONSTRUCTOR
261
262 const FileDescriptor* GetFile() const {
263 switch (type) {
264 case NULL_SYMBOL: return NULL;
265 case MESSAGE : return descriptor ->file();
266 case FIELD : return field_descriptor ->file();
267 case ENUM : return enum_descriptor ->file();
268 case ENUM_VALUE : return enum_value_descriptor->type()->file();
269 case SERVICE : return service_descriptor ->file();
270 case METHOD : return method_descriptor ->service()->file();
271 case PACKAGE : return package_file_descriptor;
272 }
273 return NULL;
274 }
275};
276
277const Symbol kNullSymbol;
278
279typedef hash_map<const char*, Symbol,
kenton@google.comd37d46d2009-04-25 02:53:47 +0000280 hash<const char*>, streq>
temporal40ee5512008-07-10 02:12:20 +0000281 SymbolsByNameMap;
282typedef hash_map<PointerStringPair, Symbol,
283 PointerStringPairHash, PointerStringPairEqual>
284 SymbolsByParentMap;
285typedef hash_map<const char*, const FileDescriptor*,
kenton@google.comd37d46d2009-04-25 02:53:47 +0000286 hash<const char*>, streq>
temporal40ee5512008-07-10 02:12:20 +0000287 FilesByNameMap;
kenton@google.com2d6daa72009-01-22 01:27:00 +0000288typedef hash_map<PointerStringPair, const FieldDescriptor*,
289 PointerStringPairHash, PointerStringPairEqual>
290 FieldsByNameMap;
temporal40ee5512008-07-10 02:12:20 +0000291typedef hash_map<DescriptorIntPair, const FieldDescriptor*,
292 PointerIntegerPairHash<DescriptorIntPair> >
293 FieldsByNumberMap;
294typedef hash_map<EnumIntPair, const EnumValueDescriptor*,
295 PointerIntegerPairHash<EnumIntPair> >
296 EnumValuesByNumberMap;
kenton@google.comd37d46d2009-04-25 02:53:47 +0000297// This is a map rather than a hash_map, since we use it to iterate
298// through all the extensions that extend a given Descriptor, and an
299// ordered data structure that implements lower_bound is convenient
300// for that.
301typedef map<DescriptorIntPair, const FieldDescriptor*>
302 ExtensionsGroupedByDescriptorMap;
temporal40ee5512008-07-10 02:12:20 +0000303
304} // anonymous namespace
305
306// ===================================================================
307// DescriptorPool::Tables
308
309class DescriptorPool::Tables {
310 public:
311 Tables();
312 ~Tables();
313
314 // Checkpoint the state of the tables. Future calls to Rollback() will
315 // return the Tables to this state. This is used when building files, since
316 // some kinds of validation errors cannot be detected until the file's
317 // descriptors have already been added to the tables. BuildFile() calls
318 // Checkpoint() before it starts building and Rollback() if it encounters
319 // an error.
320 void Checkpoint();
321
322 // Roll back the Tables to the state of the last Checkpoint(), removing
323 // everything that was added after that point.
324 void Rollback();
325
326 // The stack of files which are currently being built. Used to detect
327 // cyclic dependencies when loading files from a DescriptorDatabase. Not
328 // used when fallback_database_ == NULL.
329 vector<string> pending_files_;
330
331 // A set of files which we have tried to load from the fallback database
332 // and encountered errors. We will not attempt to load them again.
333 // Not used when fallback_database_ == NULL.
334 hash_set<string> known_bad_files_;
335
kenton@google.comd37d46d2009-04-25 02:53:47 +0000336 // The set of descriptors for which we've already loaded the full
337 // set of extensions numbers from fallback_database_.
338 hash_set<const Descriptor*> extensions_loaded_from_db_;
339
temporal40ee5512008-07-10 02:12:20 +0000340 // -----------------------------------------------------------------
341 // Finding items.
342
kenton@google.comd37d46d2009-04-25 02:53:47 +0000343 // Find symbols. This returns a null Symbol (symbol.IsNull() is true)
kenton@google.com2d6daa72009-01-22 01:27:00 +0000344 // if not found.
temporal40ee5512008-07-10 02:12:20 +0000345 inline Symbol FindSymbol(const string& key) const;
temporal40ee5512008-07-10 02:12:20 +0000346
kenton@google.com2d6daa72009-01-22 01:27:00 +0000347 // This implements the body of DescriptorPool::Find*ByName(). It should
348 // really be a private method of DescriptorPool, but that would require
349 // declaring Symbol in descriptor.h, which would drag all kinds of other
350 // stuff into the header. Yay C++.
351 Symbol FindByNameHelper(
352 const DescriptorPool* pool, const string& name) const;
353
temporal40ee5512008-07-10 02:12:20 +0000354 // These return NULL if not found.
355 inline const FileDescriptor* FindFile(const string& key) const;
kenton@google.comd37d46d2009-04-25 02:53:47 +0000356 inline const FieldDescriptor* FindExtension(const Descriptor* extendee,
357 int number);
358 inline void FindAllExtensions(const Descriptor* extendee,
359 vector<const FieldDescriptor*>* out) const;
temporal40ee5512008-07-10 02:12:20 +0000360
361 // -----------------------------------------------------------------
362 // Adding items.
363
364 // These add items to the corresponding tables. They return false if
kenton@google.comd37d46d2009-04-25 02:53:47 +0000365 // the key already exists in the table. For AddSymbol(), the string passed
366 // in must be one that was constructed using AllocateString(), as it will
367 // be used as a key in the symbols_by_name_ map without copying.
368 bool AddSymbol(const string& full_name, Symbol symbol);
temporal40ee5512008-07-10 02:12:20 +0000369 bool AddFile(const FileDescriptor* file);
kenton@google.comd37d46d2009-04-25 02:53:47 +0000370 bool AddExtension(const FieldDescriptor* field);
temporal40ee5512008-07-10 02:12:20 +0000371
372 // -----------------------------------------------------------------
373 // Allocating memory.
374
375 // Allocate an object which will be reclaimed when the pool is
376 // destroyed. Note that the object's destructor will never be called,
377 // so its fields must be plain old data (primitive data types and
378 // pointers). All of the descriptor types are such objects.
379 template<typename Type> Type* Allocate();
380
381 // Allocate an array of objects which will be reclaimed when the
382 // pool in destroyed. Again, destructors are never called.
383 template<typename Type> Type* AllocateArray(int count);
384
385 // Allocate a string which will be destroyed when the pool is destroyed.
386 // The string is initialized to the given value for convenience.
387 string* AllocateString(const string& value);
388
kenton@google.comd37d46d2009-04-25 02:53:47 +0000389 // Allocate a protocol message object. Some older versions of GCC have
390 // trouble understanding explicit template instantiations in some cases, so
391 // in those cases we have to pass a dummy pointer of the right type as the
392 // parameter instead of specifying the type explicitly.
393 template<typename Type> Type* AllocateMessage(Type* dummy = NULL);
394
395 // Allocate a FileDescriptorTables object.
396 FileDescriptorTables* AllocateFileTables();
temporal40ee5512008-07-10 02:12:20 +0000397
398 private:
399 vector<string*> strings_; // All strings in the pool.
400 vector<Message*> messages_; // All messages in the pool.
kenton@google.comd37d46d2009-04-25 02:53:47 +0000401 vector<FileDescriptorTables*> file_tables_; // All file tables in the pool.
temporal40ee5512008-07-10 02:12:20 +0000402 vector<void*> allocations_; // All other memory allocated in the pool.
403
404 SymbolsByNameMap symbols_by_name_;
temporal40ee5512008-07-10 02:12:20 +0000405 FilesByNameMap files_by_name_;
kenton@google.comd37d46d2009-04-25 02:53:47 +0000406 ExtensionsGroupedByDescriptorMap extensions_;
temporal40ee5512008-07-10 02:12:20 +0000407
408 int strings_before_checkpoint_;
409 int messages_before_checkpoint_;
kenton@google.comd37d46d2009-04-25 02:53:47 +0000410 int file_tables_before_checkpoint_;
temporal40ee5512008-07-10 02:12:20 +0000411 int allocations_before_checkpoint_;
412 vector<const char* > symbols_after_checkpoint_;
temporal40ee5512008-07-10 02:12:20 +0000413 vector<const char* > files_after_checkpoint_;
kenton@google.comd37d46d2009-04-25 02:53:47 +0000414 vector<DescriptorIntPair> extensions_after_checkpoint_;
temporal40ee5512008-07-10 02:12:20 +0000415
416 // Allocate some bytes which will be reclaimed when the pool is
417 // destroyed.
418 void* AllocateBytes(int size);
419};
420
kenton@google.comd37d46d2009-04-25 02:53:47 +0000421// Contains tables specific to a particular file. These tables are not
422// modified once the file has been constructed, so they need not be
423// protected by a mutex. This makes operations that depend only on the
424// contents of a single file -- e.g. Descriptor::FindFieldByName() --
425// lock-free.
426//
427// For historical reasons, the definitions of the methods of
428// FileDescriptorTables and DescriptorPool::Tables are interleaved below.
429// These used to be a single class.
430class FileDescriptorTables {
431 public:
432 FileDescriptorTables();
433 ~FileDescriptorTables();
434
435 // Empty table, used with placeholder files.
436 static const FileDescriptorTables kEmpty;
437
438 // -----------------------------------------------------------------
439 // Finding items.
440
441 // Find symbols. These return a null Symbol (symbol.IsNull() is true)
442 // if not found.
443 inline Symbol FindNestedSymbol(const void* parent,
444 const string& name) const;
445 inline Symbol FindNestedSymbolOfType(const void* parent,
446 const string& name,
447 const Symbol::Type type) const;
448
449 // These return NULL if not found.
450 inline const FieldDescriptor* FindFieldByNumber(
451 const Descriptor* parent, int number) const;
452 inline const FieldDescriptor* FindFieldByLowercaseName(
453 const void* parent, const string& lowercase_name) const;
454 inline const FieldDescriptor* FindFieldByCamelcaseName(
455 const void* parent, const string& camelcase_name) const;
456 inline const EnumValueDescriptor* FindEnumValueByNumber(
457 const EnumDescriptor* parent, int number) const;
458
459 // -----------------------------------------------------------------
460 // Adding items.
461
462 // These add items to the corresponding tables. They return false if
463 // the key already exists in the table. For AddAliasUnderParent(), the
464 // string passed in must be one that was constructed using AllocateString(),
465 // as it will be used as a key in the symbols_by_parent_ map without copying.
466 bool AddAliasUnderParent(const void* parent, const string& name,
467 Symbol symbol);
468 bool AddFieldByNumber(const FieldDescriptor* field);
469 bool AddEnumValueByNumber(const EnumValueDescriptor* value);
470
471 // Adds the field to the lowercase_name and camelcase_name maps. Never
472 // fails because we allow duplicates; the first field by the name wins.
473 void AddFieldByStylizedNames(const FieldDescriptor* field);
474
475 private:
476 SymbolsByParentMap symbols_by_parent_;
477 FieldsByNameMap fields_by_lowercase_name_;
478 FieldsByNameMap fields_by_camelcase_name_;
479 FieldsByNumberMap fields_by_number_; // Not including extensions.
480 EnumValuesByNumberMap enum_values_by_number_;
481};
482
temporal40ee5512008-07-10 02:12:20 +0000483DescriptorPool::Tables::Tables()
484 : strings_before_checkpoint_(0),
485 messages_before_checkpoint_(0),
486 allocations_before_checkpoint_(0) {}
487
488DescriptorPool::Tables::~Tables() {
kenton@google.com24bf56f2008-09-24 20:31:01 +0000489 // Note that the deletion order is important, since the destructors of some
490 // messages may refer to objects in allocations_.
491 STLDeleteElements(&messages_);
temporal40ee5512008-07-10 02:12:20 +0000492 for (int i = 0; i < allocations_.size(); i++) {
493 operator delete(allocations_[i]);
494 }
495 STLDeleteElements(&strings_);
kenton@google.comd37d46d2009-04-25 02:53:47 +0000496 STLDeleteElements(&file_tables_);
temporal40ee5512008-07-10 02:12:20 +0000497}
498
kenton@google.comd37d46d2009-04-25 02:53:47 +0000499FileDescriptorTables::FileDescriptorTables() {}
500FileDescriptorTables::~FileDescriptorTables() {}
501
502const FileDescriptorTables FileDescriptorTables::kEmpty;
503
temporal40ee5512008-07-10 02:12:20 +0000504void DescriptorPool::Tables::Checkpoint() {
505 strings_before_checkpoint_ = strings_.size();
506 messages_before_checkpoint_ = messages_.size();
kenton@google.comd37d46d2009-04-25 02:53:47 +0000507 file_tables_before_checkpoint_ = file_tables_.size();
temporal40ee5512008-07-10 02:12:20 +0000508 allocations_before_checkpoint_ = allocations_.size();
509
510 symbols_after_checkpoint_.clear();
temporal40ee5512008-07-10 02:12:20 +0000511 files_after_checkpoint_.clear();
kenton@google.comd37d46d2009-04-25 02:53:47 +0000512 extensions_after_checkpoint_.clear();
temporal40ee5512008-07-10 02:12:20 +0000513}
514
515void DescriptorPool::Tables::Rollback() {
516 for (int i = 0; i < symbols_after_checkpoint_.size(); i++) {
517 symbols_by_name_.erase(symbols_after_checkpoint_[i]);
518 }
temporal40ee5512008-07-10 02:12:20 +0000519 for (int i = 0; i < files_after_checkpoint_.size(); i++) {
520 files_by_name_.erase(files_after_checkpoint_[i]);
521 }
kenton@google.comd37d46d2009-04-25 02:53:47 +0000522 for (int i = 0; i < extensions_after_checkpoint_.size(); i++) {
523 extensions_.erase(extensions_after_checkpoint_[i]);
temporal40ee5512008-07-10 02:12:20 +0000524 }
525
526 symbols_after_checkpoint_.clear();
temporal40ee5512008-07-10 02:12:20 +0000527 files_after_checkpoint_.clear();
kenton@google.comd37d46d2009-04-25 02:53:47 +0000528 extensions_after_checkpoint_.clear();
temporal40ee5512008-07-10 02:12:20 +0000529
530 STLDeleteContainerPointers(
531 strings_.begin() + strings_before_checkpoint_, strings_.end());
532 STLDeleteContainerPointers(
533 messages_.begin() + messages_before_checkpoint_, messages_.end());
kenton@google.comd37d46d2009-04-25 02:53:47 +0000534 STLDeleteContainerPointers(
535 file_tables_.begin() + file_tables_before_checkpoint_, file_tables_.end());
temporal40ee5512008-07-10 02:12:20 +0000536 for (int i = allocations_before_checkpoint_; i < allocations_.size(); i++) {
537 operator delete(allocations_[i]);
538 }
539
540 strings_.resize(strings_before_checkpoint_);
541 messages_.resize(messages_before_checkpoint_);
kenton@google.comd37d46d2009-04-25 02:53:47 +0000542 file_tables_.resize(file_tables_before_checkpoint_);
temporal40ee5512008-07-10 02:12:20 +0000543 allocations_.resize(allocations_before_checkpoint_);
544}
545
546// -------------------------------------------------------------------
547
548inline Symbol DescriptorPool::Tables::FindSymbol(const string& key) const {
549 const Symbol* result = FindOrNull(symbols_by_name_, key.c_str());
550 if (result == NULL) {
551 return kNullSymbol;
552 } else {
553 return *result;
554 }
555}
556
kenton@google.comd37d46d2009-04-25 02:53:47 +0000557inline Symbol FileDescriptorTables::FindNestedSymbol(
temporal40ee5512008-07-10 02:12:20 +0000558 const void* parent, const string& name) const {
559 const Symbol* result =
560 FindOrNull(symbols_by_parent_, PointerStringPair(parent, name.c_str()));
561 if (result == NULL) {
562 return kNullSymbol;
563 } else {
564 return *result;
565 }
566}
567
kenton@google.comd37d46d2009-04-25 02:53:47 +0000568inline Symbol FileDescriptorTables::FindNestedSymbolOfType(
temporal40ee5512008-07-10 02:12:20 +0000569 const void* parent, const string& name, const Symbol::Type type) const {
570 Symbol result = FindNestedSymbol(parent, name);
571 if (result.type != type) return kNullSymbol;
572 return result;
573}
574
kenton@google.com2d6daa72009-01-22 01:27:00 +0000575Symbol DescriptorPool::Tables::FindByNameHelper(
576 const DescriptorPool* pool, const string& name) const {
577 MutexLockMaybe lock(pool->mutex_);
578 Symbol result = FindSymbol(name);
579
580 if (result.IsNull() && pool->underlay_ != NULL) {
581 // Symbol not found; check the underlay.
582 result =
583 pool->underlay_->tables_->FindByNameHelper(pool->underlay_, name);
584 }
585
586 if (result.IsNull()) {
587 // Symbol still not found, so check fallback database.
588 if (pool->TryFindSymbolInFallbackDatabase(name)) {
589 result = FindSymbol(name);
590 }
591 }
592
593 return result;
594}
595
temporal40ee5512008-07-10 02:12:20 +0000596inline const FileDescriptor* DescriptorPool::Tables::FindFile(
597 const string& key) const {
598 return FindPtrOrNull(files_by_name_, key.c_str());
599}
600
kenton@google.comd37d46d2009-04-25 02:53:47 +0000601inline const FieldDescriptor* FileDescriptorTables::FindFieldByNumber(
temporal40ee5512008-07-10 02:12:20 +0000602 const Descriptor* parent, int number) const {
603 return FindPtrOrNull(fields_by_number_, make_pair(parent, number));
604}
605
kenton@google.comd37d46d2009-04-25 02:53:47 +0000606inline const FieldDescriptor* FileDescriptorTables::FindFieldByLowercaseName(
kenton@google.com2d6daa72009-01-22 01:27:00 +0000607 const void* parent, const string& lowercase_name) const {
608 return FindPtrOrNull(fields_by_lowercase_name_,
609 PointerStringPair(parent, lowercase_name.c_str()));
610}
611
kenton@google.comd37d46d2009-04-25 02:53:47 +0000612inline const FieldDescriptor* FileDescriptorTables::FindFieldByCamelcaseName(
kenton@google.com2d6daa72009-01-22 01:27:00 +0000613 const void* parent, const string& camelcase_name) const {
614 return FindPtrOrNull(fields_by_camelcase_name_,
615 PointerStringPair(parent, camelcase_name.c_str()));
616}
617
kenton@google.comd37d46d2009-04-25 02:53:47 +0000618inline const EnumValueDescriptor* FileDescriptorTables::FindEnumValueByNumber(
temporal40ee5512008-07-10 02:12:20 +0000619 const EnumDescriptor* parent, int number) const {
620 return FindPtrOrNull(enum_values_by_number_, make_pair(parent, number));
621}
622
kenton@google.comd37d46d2009-04-25 02:53:47 +0000623inline const FieldDescriptor* DescriptorPool::Tables::FindExtension(
624 const Descriptor* extendee, int number) {
625 return FindPtrOrNull(extensions_, make_pair(extendee, number));
626}
627
628inline void DescriptorPool::Tables::FindAllExtensions(
629 const Descriptor* extendee, vector<const FieldDescriptor*>* out) const {
630 ExtensionsGroupedByDescriptorMap::const_iterator it =
631 extensions_.lower_bound(make_pair(extendee, 0));
632 for (; it != extensions_.end() && it->first.first == extendee; ++it) {
633 out->push_back(it->second);
634 }
635}
636
temporal40ee5512008-07-10 02:12:20 +0000637// -------------------------------------------------------------------
638
639bool DescriptorPool::Tables::AddSymbol(
kenton@google.comd37d46d2009-04-25 02:53:47 +0000640 const string& full_name, Symbol symbol) {
temporal40ee5512008-07-10 02:12:20 +0000641 if (InsertIfNotPresent(&symbols_by_name_, full_name.c_str(), symbol)) {
642 symbols_after_checkpoint_.push_back(full_name.c_str());
temporal40ee5512008-07-10 02:12:20 +0000643 return true;
644 } else {
645 return false;
646 }
647}
648
kenton@google.comd37d46d2009-04-25 02:53:47 +0000649bool FileDescriptorTables::AddAliasUnderParent(
temporal40ee5512008-07-10 02:12:20 +0000650 const void* parent, const string& name, Symbol symbol) {
651 PointerStringPair by_parent_key(parent, name.c_str());
kenton@google.comd37d46d2009-04-25 02:53:47 +0000652 return InsertIfNotPresent(&symbols_by_parent_, by_parent_key, symbol);
temporal40ee5512008-07-10 02:12:20 +0000653}
654
655bool DescriptorPool::Tables::AddFile(const FileDescriptor* file) {
656 if (InsertIfNotPresent(&files_by_name_, file->name().c_str(), file)) {
657 files_after_checkpoint_.push_back(file->name().c_str());
658 return true;
659 } else {
660 return false;
661 }
662}
663
kenton@google.comd37d46d2009-04-25 02:53:47 +0000664void FileDescriptorTables::AddFieldByStylizedNames(
kenton@google.com2d6daa72009-01-22 01:27:00 +0000665 const FieldDescriptor* field) {
666 const void* parent;
667 if (field->is_extension()) {
668 if (field->extension_scope() == NULL) {
669 parent = field->file();
670 } else {
671 parent = field->extension_scope();
672 }
673 } else {
674 parent = field->containing_type();
675 }
676
677 PointerStringPair lowercase_key(parent, field->lowercase_name().c_str());
kenton@google.comd37d46d2009-04-25 02:53:47 +0000678 InsertIfNotPresent(&fields_by_lowercase_name_, lowercase_key, field);
kenton@google.com2d6daa72009-01-22 01:27:00 +0000679
680 PointerStringPair camelcase_key(parent, field->camelcase_name().c_str());
kenton@google.comd37d46d2009-04-25 02:53:47 +0000681 InsertIfNotPresent(&fields_by_camelcase_name_, camelcase_key, field);
kenton@google.com2d6daa72009-01-22 01:27:00 +0000682}
683
kenton@google.comd37d46d2009-04-25 02:53:47 +0000684bool FileDescriptorTables::AddFieldByNumber(const FieldDescriptor* field) {
temporal40ee5512008-07-10 02:12:20 +0000685 DescriptorIntPair key(field->containing_type(), field->number());
kenton@google.comd37d46d2009-04-25 02:53:47 +0000686 return InsertIfNotPresent(&fields_by_number_, key, field);
temporal40ee5512008-07-10 02:12:20 +0000687}
688
kenton@google.comd37d46d2009-04-25 02:53:47 +0000689bool FileDescriptorTables::AddEnumValueByNumber(
temporal40ee5512008-07-10 02:12:20 +0000690 const EnumValueDescriptor* value) {
691 EnumIntPair key(value->type(), value->number());
kenton@google.comd37d46d2009-04-25 02:53:47 +0000692 return InsertIfNotPresent(&enum_values_by_number_, key, value);
693}
694
695bool DescriptorPool::Tables::AddExtension(const FieldDescriptor* field) {
696 DescriptorIntPair key(field->containing_type(), field->number());
697 if (InsertIfNotPresent(&extensions_, key, field)) {
698 extensions_after_checkpoint_.push_back(key);
temporal40ee5512008-07-10 02:12:20 +0000699 return true;
700 } else {
701 return false;
702 }
703}
704
705// -------------------------------------------------------------------
706
707template<typename Type>
708Type* DescriptorPool::Tables::Allocate() {
709 return reinterpret_cast<Type*>(AllocateBytes(sizeof(Type)));
710}
711
712template<typename Type>
713Type* DescriptorPool::Tables::AllocateArray(int count) {
714 return reinterpret_cast<Type*>(AllocateBytes(sizeof(Type) * count));
715}
716
717string* DescriptorPool::Tables::AllocateString(const string& value) {
718 string* result = new string(value);
719 strings_.push_back(result);
720 return result;
721}
722
723template<typename Type>
kenton@google.comd37d46d2009-04-25 02:53:47 +0000724Type* DescriptorPool::Tables::AllocateMessage(Type* dummy) {
temporal40ee5512008-07-10 02:12:20 +0000725 Type* result = new Type;
726 messages_.push_back(result);
727 return result;
728}
729
kenton@google.comd37d46d2009-04-25 02:53:47 +0000730FileDescriptorTables* DescriptorPool::Tables::AllocateFileTables() {
731 FileDescriptorTables* result = new FileDescriptorTables;
732 file_tables_.push_back(result);
733 return result;
734}
735
temporal40ee5512008-07-10 02:12:20 +0000736void* DescriptorPool::Tables::AllocateBytes(int size) {
737 // TODO(kenton): Would it be worthwhile to implement this in some more
738 // sophisticated way? Probably not for the open source release, but for
739 // internal use we could easily plug in one of our existing memory pool
740 // allocators...
741 if (size == 0) return NULL;
742
743 void* result = operator new(size);
744 allocations_.push_back(result);
745 return result;
746}
747
748// ===================================================================
749// DescriptorPool
750
751DescriptorPool::ErrorCollector::~ErrorCollector() {}
752
753DescriptorPool::DescriptorPool()
754 : mutex_(NULL),
755 fallback_database_(NULL),
756 default_error_collector_(NULL),
757 underlay_(NULL),
758 tables_(new Tables),
759 enforce_dependencies_(true),
kenton@google.comd37d46d2009-04-25 02:53:47 +0000760 allow_unknown_(false) {}
temporal40ee5512008-07-10 02:12:20 +0000761
762DescriptorPool::DescriptorPool(DescriptorDatabase* fallback_database,
763 ErrorCollector* error_collector)
764 : mutex_(new Mutex),
765 fallback_database_(fallback_database),
766 default_error_collector_(error_collector),
767 underlay_(NULL),
768 tables_(new Tables),
769 enforce_dependencies_(true),
kenton@google.comd37d46d2009-04-25 02:53:47 +0000770 allow_unknown_(false) {
temporal40ee5512008-07-10 02:12:20 +0000771}
772
773DescriptorPool::DescriptorPool(const DescriptorPool* underlay)
774 : mutex_(NULL),
775 fallback_database_(NULL),
776 default_error_collector_(NULL),
777 underlay_(underlay),
778 tables_(new Tables),
kenton@google.comd37d46d2009-04-25 02:53:47 +0000779 enforce_dependencies_(true),
780 allow_unknown_(false) {}
temporal40ee5512008-07-10 02:12:20 +0000781
782DescriptorPool::~DescriptorPool() {
783 if (mutex_ != NULL) delete mutex_;
784}
785
786// DescriptorPool::BuildFile() defined later.
787// DescriptorPool::BuildFileCollectingErrors() defined later.
temporal40ee5512008-07-10 02:12:20 +0000788
789void DescriptorPool::InternalDontEnforceDependencies() {
790 enforce_dependencies_ = false;
791}
792
kenton@google.comd37d46d2009-04-25 02:53:47 +0000793bool DescriptorPool::InternalIsFileLoaded(const string& filename) const {
794 MutexLockMaybe lock(mutex_);
795 return tables_->FindFile(filename) != NULL;
796}
797
798// generated_pool ====================================================
799
800namespace {
801
kenton@google.comfccb1462009-12-18 02:11:36 +0000802
kenton@google.comd37d46d2009-04-25 02:53:47 +0000803EncodedDescriptorDatabase* generated_database_ = NULL;
804DescriptorPool* generated_pool_ = NULL;
temporalbdbb8632009-12-18 08:21:00 +0000805GOOGLE_PROTOBUF_DECLARE_ONCE(generated_pool_init_);
kenton@google.comd37d46d2009-04-25 02:53:47 +0000806
kenton@google.com63e646b2009-05-06 19:27:03 +0000807void DeleteGeneratedPool() {
808 delete generated_database_;
809 generated_database_ = NULL;
810 delete generated_pool_;
811 generated_pool_ = NULL;
kenton@google.comd37d46d2009-04-25 02:53:47 +0000812}
813
kenton@google.com63e646b2009-05-06 19:27:03 +0000814void InitGeneratedPool() {
815 generated_database_ = new EncodedDescriptorDatabase;
816 generated_pool_ = new DescriptorPool(generated_database_);
kenton@google.comfccb1462009-12-18 02:11:36 +0000817
kenton@google.com63e646b2009-05-06 19:27:03 +0000818 internal::OnShutdown(&DeleteGeneratedPool);
819}
820
821inline void InitGeneratedPoolOnce() {
kenton@google.com80b1d622009-07-29 01:13:20 +0000822 ::google::protobuf::GoogleOnceInit(&generated_pool_init_, &InitGeneratedPool);
kenton@google.com63e646b2009-05-06 19:27:03 +0000823}
kenton@google.comd37d46d2009-04-25 02:53:47 +0000824
825} // anonymous namespace
826
827const DescriptorPool* DescriptorPool::generated_pool() {
kenton@google.com63e646b2009-05-06 19:27:03 +0000828 InitGeneratedPoolOnce();
kenton@google.comd37d46d2009-04-25 02:53:47 +0000829 return generated_pool_;
830}
831
832DescriptorPool* DescriptorPool::internal_generated_pool() {
kenton@google.com63e646b2009-05-06 19:27:03 +0000833 InitGeneratedPoolOnce();
kenton@google.comd37d46d2009-04-25 02:53:47 +0000834 return generated_pool_;
835}
836
837void DescriptorPool::InternalAddGeneratedFile(
838 const void* encoded_file_descriptor, int size) {
839 // So, this function is called in the process of initializing the
840 // descriptors for generated proto classes. Each generated .pb.cc file
841 // has an internal procedure called AddDescriptors() which is called at
842 // process startup, and that function calls this one in order to register
843 // the raw bytes of the FileDescriptorProto representing the file.
844 //
845 // We do not actually construct the descriptor objects right away. We just
846 // hang on to the bytes until they are actually needed. We actually construct
847 // the descriptor the first time one of the following things happens:
848 // * Someone calls a method like descriptor(), GetDescriptor(), or
849 // GetReflection() on the generated types, which requires returning the
850 // descriptor or an object based on it.
851 // * Someone looks up the descriptor in DescriptorPool::generated_pool().
852 //
853 // Once one of these happens, the DescriptorPool actually parses the
854 // FileDescriptorProto and generates a FileDescriptor (and all its children)
855 // based on it.
856 //
857 // Note that FileDescriptorProto is itself a generated protocol message.
858 // Therefore, when we parse one, we have to be very careful to avoid using
859 // any descriptor-based operations, since this might cause infinite recursion
860 // or deadlock.
kenton@google.com63e646b2009-05-06 19:27:03 +0000861 InitGeneratedPoolOnce();
kenton@google.comd37d46d2009-04-25 02:53:47 +0000862 GOOGLE_CHECK(generated_database_->Add(encoded_file_descriptor, size));
863}
864
865
temporal40ee5512008-07-10 02:12:20 +0000866// Find*By* methods ==================================================
867
868// TODO(kenton): There's a lot of repeated code here, but I'm not sure if
869// there's any good way to factor it out. Think about this some time when
870// there's nothing more important to do (read: never).
871
872const FileDescriptor* DescriptorPool::FindFileByName(const string& name) const {
873 MutexLockMaybe lock(mutex_);
874 const FileDescriptor* result = tables_->FindFile(name);
875 if (result != NULL) return result;
876 if (underlay_ != NULL) {
877 const FileDescriptor* result = underlay_->FindFileByName(name);
878 if (result != NULL) return result;
879 }
880 if (TryFindFileInFallbackDatabase(name)) {
881 const FileDescriptor* result = tables_->FindFile(name);
882 if (result != NULL) return result;
883 }
884 return NULL;
885}
886
887const FileDescriptor* DescriptorPool::FindFileContainingSymbol(
888 const string& symbol_name) const {
889 MutexLockMaybe lock(mutex_);
890 Symbol result = tables_->FindSymbol(symbol_name);
891 if (!result.IsNull()) return result.GetFile();
892 if (underlay_ != NULL) {
893 const FileDescriptor* result =
894 underlay_->FindFileContainingSymbol(symbol_name);
895 if (result != NULL) return result;
896 }
897 if (TryFindSymbolInFallbackDatabase(symbol_name)) {
898 Symbol result = tables_->FindSymbol(symbol_name);
899 if (!result.IsNull()) return result.GetFile();
900 }
901 return NULL;
902}
903
904const Descriptor* DescriptorPool::FindMessageTypeByName(
905 const string& name) const {
kenton@google.com2d6daa72009-01-22 01:27:00 +0000906 Symbol result = tables_->FindByNameHelper(this, name);
907 return (result.type == Symbol::MESSAGE) ? result.descriptor : NULL;
temporal40ee5512008-07-10 02:12:20 +0000908}
909
910const FieldDescriptor* DescriptorPool::FindFieldByName(
911 const string& name) const {
kenton@google.com2d6daa72009-01-22 01:27:00 +0000912 Symbol result = tables_->FindByNameHelper(this, name);
913 if (result.type == Symbol::FIELD &&
914 !result.field_descriptor->is_extension()) {
temporal40ee5512008-07-10 02:12:20 +0000915 return result.field_descriptor;
kenton@google.com2d6daa72009-01-22 01:27:00 +0000916 } else {
917 return NULL;
temporal40ee5512008-07-10 02:12:20 +0000918 }
temporal40ee5512008-07-10 02:12:20 +0000919}
920
921const FieldDescriptor* DescriptorPool::FindExtensionByName(
922 const string& name) const {
kenton@google.com2d6daa72009-01-22 01:27:00 +0000923 Symbol result = tables_->FindByNameHelper(this, name);
924 if (result.type == Symbol::FIELD &&
925 result.field_descriptor->is_extension()) {
temporal40ee5512008-07-10 02:12:20 +0000926 return result.field_descriptor;
kenton@google.com2d6daa72009-01-22 01:27:00 +0000927 } else {
928 return NULL;
temporal40ee5512008-07-10 02:12:20 +0000929 }
temporal40ee5512008-07-10 02:12:20 +0000930}
931
932const EnumDescriptor* DescriptorPool::FindEnumTypeByName(
933 const string& name) const {
kenton@google.com2d6daa72009-01-22 01:27:00 +0000934 Symbol result = tables_->FindByNameHelper(this, name);
935 return (result.type == Symbol::ENUM) ? result.enum_descriptor : NULL;
temporal40ee5512008-07-10 02:12:20 +0000936}
937
938const EnumValueDescriptor* DescriptorPool::FindEnumValueByName(
939 const string& name) const {
kenton@google.com2d6daa72009-01-22 01:27:00 +0000940 Symbol result = tables_->FindByNameHelper(this, name);
941 return (result.type == Symbol::ENUM_VALUE) ?
942 result.enum_value_descriptor : NULL;
temporal40ee5512008-07-10 02:12:20 +0000943}
944
945const ServiceDescriptor* DescriptorPool::FindServiceByName(
946 const string& name) const {
kenton@google.com2d6daa72009-01-22 01:27:00 +0000947 Symbol result = tables_->FindByNameHelper(this, name);
948 return (result.type == Symbol::SERVICE) ? result.service_descriptor : NULL;
temporal40ee5512008-07-10 02:12:20 +0000949}
950
951const MethodDescriptor* DescriptorPool::FindMethodByName(
952 const string& name) const {
kenton@google.com2d6daa72009-01-22 01:27:00 +0000953 Symbol result = tables_->FindByNameHelper(this, name);
954 return (result.type == Symbol::METHOD) ? result.method_descriptor : NULL;
temporal40ee5512008-07-10 02:12:20 +0000955}
956
957const FieldDescriptor* DescriptorPool::FindExtensionByNumber(
958 const Descriptor* extendee, int number) const {
959 MutexLockMaybe lock(mutex_);
kenton@google.comd37d46d2009-04-25 02:53:47 +0000960 const FieldDescriptor* result = tables_->FindExtension(extendee, number);
961 if (result != NULL) {
temporal40ee5512008-07-10 02:12:20 +0000962 return result;
963 }
964 if (underlay_ != NULL) {
965 const FieldDescriptor* result =
966 underlay_->FindExtensionByNumber(extendee, number);
967 if (result != NULL) return result;
968 }
969 if (TryFindExtensionInFallbackDatabase(extendee, number)) {
kenton@google.comd37d46d2009-04-25 02:53:47 +0000970 const FieldDescriptor* result = tables_->FindExtension(extendee, number);
971 if (result != NULL) {
temporal40ee5512008-07-10 02:12:20 +0000972 return result;
973 }
974 }
975 return NULL;
976}
977
kenton@google.comd37d46d2009-04-25 02:53:47 +0000978void DescriptorPool::FindAllExtensions(
979 const Descriptor* extendee, vector<const FieldDescriptor*>* out) const {
980 MutexLockMaybe lock(mutex_);
981
982 // Initialize tables_->extensions_ from the fallback database first
983 // (but do this only once per descriptor).
984 if (fallback_database_ != NULL &&
985 tables_->extensions_loaded_from_db_.count(extendee) == 0) {
986 vector<int> numbers;
987 if (fallback_database_->FindAllExtensionNumbers(extendee->full_name(),
988 &numbers)) {
989 for (int i = 0; i < numbers.size(); ++i) {
990 int number = numbers[i];
991 if (tables_->FindExtension(extendee, number) == NULL) {
992 TryFindExtensionInFallbackDatabase(extendee, number);
993 }
994 }
995 tables_->extensions_loaded_from_db_.insert(extendee);
996 }
997 }
998
999 tables_->FindAllExtensions(extendee, out);
1000 if (underlay_ != NULL) {
1001 underlay_->FindAllExtensions(extendee, out);
1002 }
1003}
1004
temporal40ee5512008-07-10 02:12:20 +00001005// -------------------------------------------------------------------
1006
1007const FieldDescriptor*
1008Descriptor::FindFieldByNumber(int key) const {
temporal40ee5512008-07-10 02:12:20 +00001009 const FieldDescriptor* result =
kenton@google.comd37d46d2009-04-25 02:53:47 +00001010 file()->tables_->FindFieldByNumber(this, key);
temporal40ee5512008-07-10 02:12:20 +00001011 if (result == NULL || result->is_extension()) {
1012 return NULL;
1013 } else {
1014 return result;
1015 }
1016}
1017
1018const FieldDescriptor*
kenton@google.com2d6daa72009-01-22 01:27:00 +00001019Descriptor::FindFieldByLowercaseName(const string& key) const {
kenton@google.com2d6daa72009-01-22 01:27:00 +00001020 const FieldDescriptor* result =
kenton@google.comd37d46d2009-04-25 02:53:47 +00001021 file()->tables_->FindFieldByLowercaseName(this, key);
kenton@google.com2d6daa72009-01-22 01:27:00 +00001022 if (result == NULL || result->is_extension()) {
1023 return NULL;
1024 } else {
1025 return result;
1026 }
1027}
1028
1029const FieldDescriptor*
1030Descriptor::FindFieldByCamelcaseName(const string& key) const {
kenton@google.com2d6daa72009-01-22 01:27:00 +00001031 const FieldDescriptor* result =
kenton@google.comd37d46d2009-04-25 02:53:47 +00001032 file()->tables_->FindFieldByCamelcaseName(this, key);
kenton@google.com2d6daa72009-01-22 01:27:00 +00001033 if (result == NULL || result->is_extension()) {
1034 return NULL;
1035 } else {
1036 return result;
1037 }
1038}
1039
1040const FieldDescriptor*
temporal40ee5512008-07-10 02:12:20 +00001041Descriptor::FindFieldByName(const string& key) const {
temporal40ee5512008-07-10 02:12:20 +00001042 Symbol result =
kenton@google.comd37d46d2009-04-25 02:53:47 +00001043 file()->tables_->FindNestedSymbolOfType(this, key, Symbol::FIELD);
temporal40ee5512008-07-10 02:12:20 +00001044 if (!result.IsNull() && !result.field_descriptor->is_extension()) {
1045 return result.field_descriptor;
1046 } else {
1047 return NULL;
1048 }
1049}
1050
1051const FieldDescriptor*
1052Descriptor::FindExtensionByName(const string& key) const {
temporal40ee5512008-07-10 02:12:20 +00001053 Symbol result =
kenton@google.comd37d46d2009-04-25 02:53:47 +00001054 file()->tables_->FindNestedSymbolOfType(this, key, Symbol::FIELD);
temporal40ee5512008-07-10 02:12:20 +00001055 if (!result.IsNull() && result.field_descriptor->is_extension()) {
1056 return result.field_descriptor;
1057 } else {
1058 return NULL;
1059 }
1060}
1061
kenton@google.com2d6daa72009-01-22 01:27:00 +00001062const FieldDescriptor*
1063Descriptor::FindExtensionByLowercaseName(const string& key) const {
kenton@google.com2d6daa72009-01-22 01:27:00 +00001064 const FieldDescriptor* result =
kenton@google.comd37d46d2009-04-25 02:53:47 +00001065 file()->tables_->FindFieldByLowercaseName(this, key);
kenton@google.com2d6daa72009-01-22 01:27:00 +00001066 if (result == NULL || !result->is_extension()) {
1067 return NULL;
1068 } else {
1069 return result;
1070 }
1071}
1072
1073const FieldDescriptor*
1074Descriptor::FindExtensionByCamelcaseName(const string& key) const {
kenton@google.com2d6daa72009-01-22 01:27:00 +00001075 const FieldDescriptor* result =
kenton@google.comd37d46d2009-04-25 02:53:47 +00001076 file()->tables_->FindFieldByCamelcaseName(this, key);
kenton@google.com2d6daa72009-01-22 01:27:00 +00001077 if (result == NULL || !result->is_extension()) {
1078 return NULL;
1079 } else {
1080 return result;
1081 }
1082}
1083
temporal40ee5512008-07-10 02:12:20 +00001084const Descriptor*
1085Descriptor::FindNestedTypeByName(const string& key) const {
temporal40ee5512008-07-10 02:12:20 +00001086 Symbol result =
kenton@google.comd37d46d2009-04-25 02:53:47 +00001087 file()->tables_->FindNestedSymbolOfType(this, key, Symbol::MESSAGE);
temporal40ee5512008-07-10 02:12:20 +00001088 if (!result.IsNull()) {
1089 return result.descriptor;
1090 } else {
1091 return NULL;
1092 }
1093}
1094
1095const EnumDescriptor*
1096Descriptor::FindEnumTypeByName(const string& key) const {
temporal40ee5512008-07-10 02:12:20 +00001097 Symbol result =
kenton@google.comd37d46d2009-04-25 02:53:47 +00001098 file()->tables_->FindNestedSymbolOfType(this, key, Symbol::ENUM);
temporal40ee5512008-07-10 02:12:20 +00001099 if (!result.IsNull()) {
1100 return result.enum_descriptor;
1101 } else {
1102 return NULL;
1103 }
1104}
1105
1106const EnumValueDescriptor*
1107Descriptor::FindEnumValueByName(const string& key) const {
temporal40ee5512008-07-10 02:12:20 +00001108 Symbol result =
kenton@google.comd37d46d2009-04-25 02:53:47 +00001109 file()->tables_->FindNestedSymbolOfType(this, key, Symbol::ENUM_VALUE);
temporal40ee5512008-07-10 02:12:20 +00001110 if (!result.IsNull()) {
1111 return result.enum_value_descriptor;
1112 } else {
1113 return NULL;
1114 }
1115}
1116
1117const EnumValueDescriptor*
1118EnumDescriptor::FindValueByName(const string& key) const {
temporal40ee5512008-07-10 02:12:20 +00001119 Symbol result =
kenton@google.comd37d46d2009-04-25 02:53:47 +00001120 file()->tables_->FindNestedSymbolOfType(this, key, Symbol::ENUM_VALUE);
temporal40ee5512008-07-10 02:12:20 +00001121 if (!result.IsNull()) {
1122 return result.enum_value_descriptor;
1123 } else {
1124 return NULL;
1125 }
1126}
1127
1128const EnumValueDescriptor*
1129EnumDescriptor::FindValueByNumber(int key) const {
kenton@google.comd37d46d2009-04-25 02:53:47 +00001130 return file()->tables_->FindEnumValueByNumber(this, key);
temporal40ee5512008-07-10 02:12:20 +00001131}
1132
1133const MethodDescriptor*
1134ServiceDescriptor::FindMethodByName(const string& key) const {
temporal40ee5512008-07-10 02:12:20 +00001135 Symbol result =
kenton@google.comd37d46d2009-04-25 02:53:47 +00001136 file()->tables_->FindNestedSymbolOfType(this, key, Symbol::METHOD);
temporal40ee5512008-07-10 02:12:20 +00001137 if (!result.IsNull()) {
1138 return result.method_descriptor;
1139 } else {
1140 return NULL;
1141 }
1142}
1143
1144const Descriptor*
1145FileDescriptor::FindMessageTypeByName(const string& key) const {
kenton@google.comd37d46d2009-04-25 02:53:47 +00001146 Symbol result = tables_->FindNestedSymbolOfType(this, key, Symbol::MESSAGE);
temporal40ee5512008-07-10 02:12:20 +00001147 if (!result.IsNull()) {
1148 return result.descriptor;
1149 } else {
1150 return NULL;
1151 }
1152}
1153
1154const EnumDescriptor*
1155FileDescriptor::FindEnumTypeByName(const string& key) const {
kenton@google.comd37d46d2009-04-25 02:53:47 +00001156 Symbol result = tables_->FindNestedSymbolOfType(this, key, Symbol::ENUM);
temporal40ee5512008-07-10 02:12:20 +00001157 if (!result.IsNull()) {
1158 return result.enum_descriptor;
1159 } else {
1160 return NULL;
1161 }
1162}
1163
1164const EnumValueDescriptor*
1165FileDescriptor::FindEnumValueByName(const string& key) const {
temporal40ee5512008-07-10 02:12:20 +00001166 Symbol result =
kenton@google.comd37d46d2009-04-25 02:53:47 +00001167 tables_->FindNestedSymbolOfType(this, key, Symbol::ENUM_VALUE);
temporal40ee5512008-07-10 02:12:20 +00001168 if (!result.IsNull()) {
1169 return result.enum_value_descriptor;
1170 } else {
1171 return NULL;
1172 }
1173}
1174
1175const ServiceDescriptor*
1176FileDescriptor::FindServiceByName(const string& key) const {
kenton@google.comd37d46d2009-04-25 02:53:47 +00001177 Symbol result = tables_->FindNestedSymbolOfType(this, key, Symbol::SERVICE);
temporal40ee5512008-07-10 02:12:20 +00001178 if (!result.IsNull()) {
1179 return result.service_descriptor;
1180 } else {
1181 return NULL;
1182 }
1183}
1184
1185const FieldDescriptor*
1186FileDescriptor::FindExtensionByName(const string& key) const {
kenton@google.comd37d46d2009-04-25 02:53:47 +00001187 Symbol result = tables_->FindNestedSymbolOfType(this, key, Symbol::FIELD);
temporal40ee5512008-07-10 02:12:20 +00001188 if (!result.IsNull() && result.field_descriptor->is_extension()) {
1189 return result.field_descriptor;
1190 } else {
1191 return NULL;
1192 }
1193}
1194
kenton@google.com2d6daa72009-01-22 01:27:00 +00001195const FieldDescriptor*
1196FileDescriptor::FindExtensionByLowercaseName(const string& key) const {
kenton@google.comd37d46d2009-04-25 02:53:47 +00001197 const FieldDescriptor* result = tables_->FindFieldByLowercaseName(this, key);
kenton@google.com2d6daa72009-01-22 01:27:00 +00001198 if (result == NULL || !result->is_extension()) {
1199 return NULL;
1200 } else {
1201 return result;
1202 }
1203}
1204
1205const FieldDescriptor*
1206FileDescriptor::FindExtensionByCamelcaseName(const string& key) const {
kenton@google.comd37d46d2009-04-25 02:53:47 +00001207 const FieldDescriptor* result = tables_->FindFieldByCamelcaseName(this, key);
kenton@google.com2d6daa72009-01-22 01:27:00 +00001208 if (result == NULL || !result->is_extension()) {
1209 return NULL;
1210 } else {
1211 return result;
1212 }
1213}
1214
temporal40ee5512008-07-10 02:12:20 +00001215bool Descriptor::IsExtensionNumber(int number) const {
1216 // Linear search should be fine because we don't expect a message to have
1217 // more than a couple extension ranges.
1218 for (int i = 0; i < extension_range_count(); i++) {
1219 if (number >= extension_range(i)->start &&
1220 number < extension_range(i)->end) {
1221 return true;
1222 }
1223 }
1224 return false;
1225}
1226
1227// -------------------------------------------------------------------
1228
1229bool DescriptorPool::TryFindFileInFallbackDatabase(const string& name) const {
1230 if (fallback_database_ == NULL) return false;
1231
1232 if (tables_->known_bad_files_.count(name) > 0) return false;
1233
1234 FileDescriptorProto file_proto;
1235 if (!fallback_database_->FindFileByName(name, &file_proto) ||
1236 BuildFileFromDatabase(file_proto) == NULL) {
1237 tables_->known_bad_files_.insert(name);
1238 return false;
1239 }
1240
1241 return true;
1242}
1243
1244bool DescriptorPool::TryFindSymbolInFallbackDatabase(const string& name) const {
1245 if (fallback_database_ == NULL) return false;
1246
1247 FileDescriptorProto file_proto;
1248 if (!fallback_database_->FindFileContainingSymbol(name, &file_proto)) {
1249 return false;
1250 }
1251
1252 if (tables_->FindFile(file_proto.name()) != NULL) {
1253 // We've already loaded this file, and it apparently doesn't contain the
1254 // symbol we're looking for. Some DescriptorDatabases return false
1255 // positives.
1256 return false;
1257 }
1258
1259 if (BuildFileFromDatabase(file_proto) == NULL) {
1260 return false;
1261 }
1262
1263 return true;
1264}
1265
1266bool DescriptorPool::TryFindExtensionInFallbackDatabase(
1267 const Descriptor* containing_type, int field_number) const {
1268 if (fallback_database_ == NULL) return false;
1269
1270 FileDescriptorProto file_proto;
1271 if (!fallback_database_->FindFileContainingExtension(
1272 containing_type->full_name(), field_number, &file_proto)) {
1273 return false;
1274 }
1275
1276 if (tables_->FindFile(file_proto.name()) != NULL) {
1277 // We've already loaded this file, and it apparently doesn't contain the
1278 // extension we're looking for. Some DescriptorDatabases return false
1279 // positives.
1280 return false;
1281 }
1282
1283 if (BuildFileFromDatabase(file_proto) == NULL) {
1284 return false;
1285 }
1286
1287 return true;
1288}
1289
1290// ===================================================================
1291
1292string FieldDescriptor::DefaultValueAsString(bool quote_string_type) const {
1293 GOOGLE_CHECK(has_default_value()) << "No default value";
1294 switch (cpp_type()) {
1295 case CPPTYPE_INT32:
1296 return SimpleItoa(default_value_int32());
1297 break;
1298 case CPPTYPE_INT64:
1299 return SimpleItoa(default_value_int64());
1300 break;
1301 case CPPTYPE_UINT32:
1302 return SimpleItoa(default_value_uint32());
1303 break;
1304 case CPPTYPE_UINT64:
1305 return SimpleItoa(default_value_uint64());
1306 break;
1307 case CPPTYPE_FLOAT:
1308 return SimpleFtoa(default_value_float());
1309 break;
1310 case CPPTYPE_DOUBLE:
1311 return SimpleDtoa(default_value_double());
1312 break;
1313 case CPPTYPE_BOOL:
1314 return default_value_bool() ? "true" : "false";
1315 break;
1316 case CPPTYPE_STRING:
1317 if (quote_string_type) {
1318 return "\"" + CEscape(default_value_string()) + "\"";
1319 } else {
1320 if (type() == TYPE_BYTES) {
1321 return CEscape(default_value_string());
1322 } else {
1323 return default_value_string();
1324 }
1325 }
1326 break;
1327 case CPPTYPE_ENUM:
1328 return default_value_enum()->name();
1329 break;
1330 case CPPTYPE_MESSAGE:
1331 GOOGLE_LOG(DFATAL) << "Messages can't have default values!";
1332 break;
1333 }
1334 GOOGLE_LOG(FATAL) << "Can't get here: failed to get default value as string";
1335 return "";
1336}
1337
1338// CopyTo methods ====================================================
1339
1340void FileDescriptor::CopyTo(FileDescriptorProto* proto) const {
1341 proto->set_name(name());
1342 if (!package().empty()) proto->set_package(package());
1343
1344 for (int i = 0; i < dependency_count(); i++) {
1345 proto->add_dependency(dependency(i)->name());
1346 }
1347
1348 for (int i = 0; i < message_type_count(); i++) {
1349 message_type(i)->CopyTo(proto->add_message_type());
1350 }
1351 for (int i = 0; i < enum_type_count(); i++) {
1352 enum_type(i)->CopyTo(proto->add_enum_type());
1353 }
1354 for (int i = 0; i < service_count(); i++) {
1355 service(i)->CopyTo(proto->add_service());
1356 }
1357 for (int i = 0; i < extension_count(); i++) {
1358 extension(i)->CopyTo(proto->add_extension());
1359 }
1360
1361 if (&options() != &FileOptions::default_instance()) {
1362 proto->mutable_options()->CopyFrom(options());
1363 }
1364}
1365
1366void Descriptor::CopyTo(DescriptorProto* proto) const {
1367 proto->set_name(name());
1368
1369 for (int i = 0; i < field_count(); i++) {
1370 field(i)->CopyTo(proto->add_field());
1371 }
1372 for (int i = 0; i < nested_type_count(); i++) {
1373 nested_type(i)->CopyTo(proto->add_nested_type());
1374 }
1375 for (int i = 0; i < enum_type_count(); i++) {
1376 enum_type(i)->CopyTo(proto->add_enum_type());
1377 }
1378 for (int i = 0; i < extension_range_count(); i++) {
1379 DescriptorProto::ExtensionRange* range = proto->add_extension_range();
1380 range->set_start(extension_range(i)->start);
1381 range->set_end(extension_range(i)->end);
1382 }
1383 for (int i = 0; i < extension_count(); i++) {
1384 extension(i)->CopyTo(proto->add_extension());
1385 }
1386
1387 if (&options() != &MessageOptions::default_instance()) {
1388 proto->mutable_options()->CopyFrom(options());
1389 }
1390}
1391
1392void FieldDescriptor::CopyTo(FieldDescriptorProto* proto) const {
1393 proto->set_name(name());
1394 proto->set_number(number());
kenton@google.coma2a32c22008-11-14 17:29:32 +00001395
1396 // Some compilers do not allow static_cast directly between two enum types,
1397 // so we must cast to int first.
1398 proto->set_label(static_cast<FieldDescriptorProto::Label>(
1399 implicit_cast<int>(label())));
1400 proto->set_type(static_cast<FieldDescriptorProto::Type>(
1401 implicit_cast<int>(type())));
temporal40ee5512008-07-10 02:12:20 +00001402
1403 if (is_extension()) {
kenton@google.comd37d46d2009-04-25 02:53:47 +00001404 if (!containing_type()->is_unqualified_placeholder_) {
1405 proto->set_extendee(".");
1406 }
temporal40ee5512008-07-10 02:12:20 +00001407 proto->mutable_extendee()->append(containing_type()->full_name());
1408 }
1409
1410 if (cpp_type() == CPPTYPE_MESSAGE) {
kenton@google.comd37d46d2009-04-25 02:53:47 +00001411 if (message_type()->is_placeholder_) {
1412 // We don't actually know if the type is a message type. It could be
1413 // an enum.
1414 proto->clear_type();
1415 }
1416
1417 if (!message_type()->is_unqualified_placeholder_) {
1418 proto->set_type_name(".");
1419 }
temporal40ee5512008-07-10 02:12:20 +00001420 proto->mutable_type_name()->append(message_type()->full_name());
1421 } else if (cpp_type() == CPPTYPE_ENUM) {
kenton@google.comd37d46d2009-04-25 02:53:47 +00001422 if (!enum_type()->is_unqualified_placeholder_) {
1423 proto->set_type_name(".");
1424 }
temporal40ee5512008-07-10 02:12:20 +00001425 proto->mutable_type_name()->append(enum_type()->full_name());
1426 }
1427
1428 if (has_default_value()) {
1429 proto->set_default_value(DefaultValueAsString(false));
1430 }
1431
1432 if (&options() != &FieldOptions::default_instance()) {
1433 proto->mutable_options()->CopyFrom(options());
1434 }
1435}
1436
1437void EnumDescriptor::CopyTo(EnumDescriptorProto* proto) const {
1438 proto->set_name(name());
1439
1440 for (int i = 0; i < value_count(); i++) {
1441 value(i)->CopyTo(proto->add_value());
1442 }
1443
1444 if (&options() != &EnumOptions::default_instance()) {
1445 proto->mutable_options()->CopyFrom(options());
1446 }
1447}
1448
1449void EnumValueDescriptor::CopyTo(EnumValueDescriptorProto* proto) const {
1450 proto->set_name(name());
1451 proto->set_number(number());
1452
1453 if (&options() != &EnumValueOptions::default_instance()) {
1454 proto->mutable_options()->CopyFrom(options());
1455 }
1456}
1457
1458void ServiceDescriptor::CopyTo(ServiceDescriptorProto* proto) const {
1459 proto->set_name(name());
1460
1461 for (int i = 0; i < method_count(); i++) {
1462 method(i)->CopyTo(proto->add_method());
1463 }
1464
1465 if (&options() != &ServiceOptions::default_instance()) {
1466 proto->mutable_options()->CopyFrom(options());
1467 }
1468}
1469
1470void MethodDescriptor::CopyTo(MethodDescriptorProto* proto) const {
1471 proto->set_name(name());
1472
kenton@google.comd37d46d2009-04-25 02:53:47 +00001473 if (!input_type()->is_unqualified_placeholder_) {
1474 proto->set_input_type(".");
1475 }
temporal40ee5512008-07-10 02:12:20 +00001476 proto->mutable_input_type()->append(input_type()->full_name());
kenton@google.comd37d46d2009-04-25 02:53:47 +00001477
1478 if (!output_type()->is_unqualified_placeholder_) {
1479 proto->set_output_type(".");
1480 }
temporal40ee5512008-07-10 02:12:20 +00001481 proto->mutable_output_type()->append(output_type()->full_name());
1482
1483 if (&options() != &MethodOptions::default_instance()) {
1484 proto->mutable_options()->CopyFrom(options());
1485 }
1486}
1487
1488// DebugString methods ===============================================
1489
1490namespace {
1491
1492// Used by each of the option formatters.
1493bool RetrieveOptions(const Message &options, vector<string> *option_entries) {
1494 option_entries->clear();
temporal779f61c2008-08-13 03:15:00 +00001495 const Reflection* reflection = options.GetReflection();
temporal40ee5512008-07-10 02:12:20 +00001496 vector<const FieldDescriptor*> fields;
temporal779f61c2008-08-13 03:15:00 +00001497 reflection->ListFields(options, &fields);
temporal40ee5512008-07-10 02:12:20 +00001498 for (int i = 0; i < fields.size(); i++) {
1499 // Doesn't make sense to have message type fields here
1500 if (fields[i]->cpp_type() == FieldDescriptor::CPPTYPE_MESSAGE) {
1501 continue;
1502 }
1503 int count = 1;
1504 bool repeated = false;
1505 if (fields[i]->is_repeated()) {
temporal779f61c2008-08-13 03:15:00 +00001506 count = reflection->FieldSize(options, fields[i]);
temporal40ee5512008-07-10 02:12:20 +00001507 repeated = true;
1508 }
1509 for (int j = 0; j < count; j++) {
1510 string fieldval;
1511 TextFormat::PrintFieldValueToString(options, fields[i],
1512 repeated ? count : -1, &fieldval);
1513 option_entries->push_back(fields[i]->name() + " = " + fieldval);
1514 }
1515 }
1516 return !option_entries->empty();
1517}
1518
1519// Formats options that all appear together in brackets. Does not include
1520// brackets.
1521bool FormatBracketedOptions(const Message &options, string *output) {
1522 vector<string> all_options;
1523 if (RetrieveOptions(options, &all_options)) {
1524 output->append(JoinStrings(all_options, ", "));
1525 }
1526 return !all_options.empty();
1527}
1528
1529// Formats options one per line
1530bool FormatLineOptions(int depth, const Message &options, string *output) {
1531 string prefix(depth * 2, ' ');
1532 vector<string> all_options;
1533 if (RetrieveOptions(options, &all_options)) {
1534 for (int i = 0; i < all_options.size(); i++) {
1535 strings::SubstituteAndAppend(output, "$0option $1;\n",
1536 prefix, all_options[i]);
1537 }
1538 }
1539 return !all_options.empty();
1540}
1541
1542} // anonymous namespace
1543
1544string FileDescriptor::DebugString() const {
1545 string contents = "syntax = \"proto2\";\n\n";
1546
1547 for (int i = 0; i < dependency_count(); i++) {
1548 strings::SubstituteAndAppend(&contents, "import \"$0\";\n",
1549 dependency(i)->name());
1550 }
1551
1552 if (!package().empty()) {
1553 strings::SubstituteAndAppend(&contents, "package $0;\n\n", package());
1554 }
1555
1556 if (FormatLineOptions(0, options(), &contents)) {
1557 contents.append("\n"); // add some space if we had options
1558 }
1559
1560 for (int i = 0; i < enum_type_count(); i++) {
1561 enum_type(i)->DebugString(0, &contents);
1562 contents.append("\n");
1563 }
1564
1565 // Find all the 'group' type extensions; we will not output their nested
1566 // definitions (those will be done with their group field descriptor).
1567 set<const Descriptor*> groups;
1568 for (int i = 0; i < extension_count(); i++) {
1569 if (extension(i)->type() == FieldDescriptor::TYPE_GROUP) {
1570 groups.insert(extension(i)->message_type());
1571 }
1572 }
1573
1574 for (int i = 0; i < message_type_count(); i++) {
1575 if (groups.count(message_type(i)) == 0) {
1576 strings::SubstituteAndAppend(&contents, "message $0",
1577 message_type(i)->name());
1578 message_type(i)->DebugString(0, &contents);
1579 contents.append("\n");
1580 }
1581 }
1582
1583 for (int i = 0; i < service_count(); i++) {
1584 service(i)->DebugString(&contents);
1585 contents.append("\n");
1586 }
1587
1588 const Descriptor* containing_type = NULL;
1589 for (int i = 0; i < extension_count(); i++) {
1590 if (extension(i)->containing_type() != containing_type) {
1591 if (i > 0) contents.append("}\n\n");
1592 containing_type = extension(i)->containing_type();
1593 strings::SubstituteAndAppend(&contents, "extend .$0 {\n",
1594 containing_type->full_name());
1595 }
1596 extension(i)->DebugString(1, &contents);
1597 }
1598 if (extension_count() > 0) contents.append("}\n\n");
1599
1600 return contents;
1601}
1602
1603string Descriptor::DebugString() const {
1604 string contents;
1605 strings::SubstituteAndAppend(&contents, "message $0", name());
1606 DebugString(0, &contents);
1607 return contents;
1608}
1609
1610void Descriptor::DebugString(int depth, string *contents) const {
1611 string prefix(depth * 2, ' ');
1612 ++depth;
1613 contents->append(" {\n");
1614
1615 FormatLineOptions(depth, options(), contents);
1616
1617 // Find all the 'group' types for fields and extensions; we will not output
1618 // their nested definitions (those will be done with their group field
1619 // descriptor).
1620 set<const Descriptor*> groups;
1621 for (int i = 0; i < field_count(); i++) {
1622 if (field(i)->type() == FieldDescriptor::TYPE_GROUP) {
1623 groups.insert(field(i)->message_type());
1624 }
1625 }
1626 for (int i = 0; i < extension_count(); i++) {
1627 if (extension(i)->type() == FieldDescriptor::TYPE_GROUP) {
1628 groups.insert(extension(i)->message_type());
1629 }
1630 }
1631
1632 for (int i = 0; i < nested_type_count(); i++) {
1633 if (groups.count(nested_type(i)) == 0) {
1634 strings::SubstituteAndAppend(contents, "$0 message $1",
1635 prefix, nested_type(i)->name());
1636 nested_type(i)->DebugString(depth, contents);
1637 }
1638 }
1639 for (int i = 0; i < enum_type_count(); i++) {
1640 enum_type(i)->DebugString(depth, contents);
1641 }
1642 for (int i = 0; i < field_count(); i++) {
1643 field(i)->DebugString(depth, contents);
1644 }
1645
1646 for (int i = 0; i < extension_range_count(); i++) {
1647 strings::SubstituteAndAppend(contents, "$0 extensions $1 to $2;\n",
1648 prefix,
1649 extension_range(i)->start,
1650 extension_range(i)->end - 1);
1651 }
1652
1653 // Group extensions by what they extend, so they can be printed out together.
1654 const Descriptor* containing_type = NULL;
1655 for (int i = 0; i < extension_count(); i++) {
1656 if (extension(i)->containing_type() != containing_type) {
1657 if (i > 0) strings::SubstituteAndAppend(contents, "$0 }\n", prefix);
1658 containing_type = extension(i)->containing_type();
1659 strings::SubstituteAndAppend(contents, "$0 extend .$1 {\n",
1660 prefix, containing_type->full_name());
1661 }
1662 extension(i)->DebugString(depth + 1, contents);
1663 }
1664 if (extension_count() > 0)
1665 strings::SubstituteAndAppend(contents, "$0 }\n", prefix);
1666
1667 strings::SubstituteAndAppend(contents, "$0}\n", prefix);
1668}
1669
1670string FieldDescriptor::DebugString() const {
1671 string contents;
1672 int depth = 0;
1673 if (is_extension()) {
1674 strings::SubstituteAndAppend(&contents, "extend .$0 {\n",
1675 containing_type()->full_name());
1676 depth = 1;
1677 }
1678 DebugString(depth, &contents);
1679 if (is_extension()) {
1680 contents.append("}\n");
1681 }
1682 return contents;
1683}
1684
1685void FieldDescriptor::DebugString(int depth, string *contents) const {
1686 string prefix(depth * 2, ' ');
1687 string field_type;
1688 switch (type()) {
1689 case TYPE_MESSAGE:
1690 field_type = "." + message_type()->full_name();
1691 break;
1692 case TYPE_ENUM:
1693 field_type = "." + enum_type()->full_name();
1694 break;
1695 default:
1696 field_type = kTypeToName[type()];
1697 }
1698
1699 strings::SubstituteAndAppend(contents, "$0$1 $2 $3 = $4",
1700 prefix,
1701 kLabelToName[label()],
1702 field_type,
1703 type() == TYPE_GROUP ? message_type()->name() :
1704 name(),
1705 number());
1706
1707 bool bracketed = false;
1708 if (has_default_value()) {
1709 bracketed = true;
1710 strings::SubstituteAndAppend(contents, " [default = $0",
1711 DefaultValueAsString(true));
1712 }
1713
1714 string formatted_options;
1715 if (FormatBracketedOptions(options(), &formatted_options)) {
1716 contents->append(bracketed ? ", " : " [");
1717 bracketed = true;
1718 contents->append(formatted_options);
1719 }
1720
1721 if (bracketed) {
1722 contents->append("]");
1723 }
1724
1725 if (type() == TYPE_GROUP) {
1726 message_type()->DebugString(depth, contents);
1727 } else {
1728 contents->append(";\n");
1729 }
1730}
1731
1732string EnumDescriptor::DebugString() const {
1733 string contents;
1734 DebugString(0, &contents);
1735 return contents;
1736}
1737
1738void EnumDescriptor::DebugString(int depth, string *contents) const {
1739 string prefix(depth * 2, ' ');
1740 ++depth;
1741 strings::SubstituteAndAppend(contents, "$0enum $1 {\n",
1742 prefix, name());
1743
1744 FormatLineOptions(depth, options(), contents);
1745
1746 for (int i = 0; i < value_count(); i++) {
1747 value(i)->DebugString(depth, contents);
1748 }
1749 strings::SubstituteAndAppend(contents, "$0}\n", prefix);
1750}
1751
1752string EnumValueDescriptor::DebugString() const {
1753 string contents;
1754 DebugString(0, &contents);
1755 return contents;
1756}
1757
1758void EnumValueDescriptor::DebugString(int depth, string *contents) const {
1759 string prefix(depth * 2, ' ');
1760 strings::SubstituteAndAppend(contents, "$0$1 = $2",
1761 prefix, name(), number());
1762
1763 string formatted_options;
1764 if (FormatBracketedOptions(options(), &formatted_options)) {
1765 strings::SubstituteAndAppend(contents, " [$0]", formatted_options);
1766 }
1767 contents->append(";\n");
1768}
1769
1770string ServiceDescriptor::DebugString() const {
1771 string contents;
1772 DebugString(&contents);
1773 return contents;
1774}
1775
1776void ServiceDescriptor::DebugString(string *contents) const {
1777 strings::SubstituteAndAppend(contents, "service $0 {\n", name());
1778
1779 FormatLineOptions(1, options(), contents);
1780
1781 for (int i = 0; i < method_count(); i++) {
1782 method(i)->DebugString(1, contents);
1783 }
1784
1785 contents->append("}\n");
1786}
1787
1788string MethodDescriptor::DebugString() const {
1789 string contents;
1790 DebugString(0, &contents);
1791 return contents;
1792}
1793
1794void MethodDescriptor::DebugString(int depth, string *contents) const {
1795 string prefix(depth * 2, ' ');
1796 ++depth;
1797 strings::SubstituteAndAppend(contents, "$0rpc $1(.$2) returns (.$3)",
1798 prefix, name(),
1799 input_type()->full_name(),
1800 output_type()->full_name());
1801
1802 string formatted_options;
1803 if (FormatLineOptions(depth, options(), &formatted_options)) {
1804 strings::SubstituteAndAppend(contents, " {\n$0$1}\n",
1805 formatted_options, prefix);
1806 } else {
1807 contents->append(";\n");
1808 }
1809}
1810// ===================================================================
1811
kenton@google.com24bf56f2008-09-24 20:31:01 +00001812namespace {
1813
1814// Represents an options message to interpret. Extension names in the option
1815// name are respolved relative to name_scope. element_name and orig_opt are
1816// used only for error reporting (since the parser records locations against
1817// pointers in the original options, not the mutable copy). The Message must be
1818// one of the Options messages in descriptor.proto.
1819struct OptionsToInterpret {
1820 OptionsToInterpret(const string& ns,
1821 const string& el,
1822 const Message* orig_opt,
1823 Message* opt)
1824 : name_scope(ns),
1825 element_name(el),
1826 original_options(orig_opt),
1827 options(opt) {
1828 }
1829 string name_scope;
1830 string element_name;
1831 const Message* original_options;
1832 Message* options;
1833};
1834
1835} // namespace
1836
temporal40ee5512008-07-10 02:12:20 +00001837class DescriptorBuilder {
1838 public:
1839 DescriptorBuilder(const DescriptorPool* pool,
1840 DescriptorPool::Tables* tables,
1841 DescriptorPool::ErrorCollector* error_collector);
1842 ~DescriptorBuilder();
1843
kenton@google.comd37d46d2009-04-25 02:53:47 +00001844 const FileDescriptor* BuildFile(const FileDescriptorProto& proto);
temporal40ee5512008-07-10 02:12:20 +00001845
1846 private:
kenton@google.com24bf56f2008-09-24 20:31:01 +00001847 friend class OptionInterpreter;
1848
temporal40ee5512008-07-10 02:12:20 +00001849 const DescriptorPool* pool_;
1850 DescriptorPool::Tables* tables_; // for convenience
1851 DescriptorPool::ErrorCollector* error_collector_;
kenton@google.com24bf56f2008-09-24 20:31:01 +00001852
1853 // As we build descriptors we store copies of the options messages in
1854 // them. We put pointers to those copies in this vector, as we build, so we
1855 // can later (after cross-linking) interpret those options.
1856 vector<OptionsToInterpret> options_to_interpret_;
1857
temporal40ee5512008-07-10 02:12:20 +00001858 bool had_errors_;
1859 string filename_;
1860 FileDescriptor* file_;
kenton@google.comd37d46d2009-04-25 02:53:47 +00001861 FileDescriptorTables* file_tables_;
temporal40ee5512008-07-10 02:12:20 +00001862
1863 // If LookupSymbol() finds a symbol that is in a file which is not a declared
1864 // dependency of this file, it will fail, but will set
1865 // possible_undeclared_dependency_ to point at that file. This is only used
1866 // by AddNotDefinedError() to report a more useful error message.
kenton@google.com80b1d622009-07-29 01:13:20 +00001867 // possible_undeclared_dependency_name_ is the name of the symbol that was
1868 // actually found in possible_undeclared_dependency_, which may be a parent
1869 // of the symbol actually looked for.
temporal40ee5512008-07-10 02:12:20 +00001870 const FileDescriptor* possible_undeclared_dependency_;
kenton@google.com80b1d622009-07-29 01:13:20 +00001871 string possible_undeclared_dependency_name_;
temporal40ee5512008-07-10 02:12:20 +00001872
1873 void AddError(const string& element_name,
1874 const Message& descriptor,
1875 DescriptorPool::ErrorCollector::ErrorLocation location,
1876 const string& error);
1877
1878 // Adds an error indicating that undefined_symbol was not defined. Must
1879 // only be called after LookupSymbol() fails.
1880 void AddNotDefinedError(
1881 const string& element_name,
1882 const Message& descriptor,
1883 DescriptorPool::ErrorCollector::ErrorLocation location,
1884 const string& undefined_symbol);
1885
1886 // Silly helper which determines if the given file is in the given package.
1887 // I.e., either file->package() == package_name or file->package() is a
1888 // nested package within package_name.
1889 bool IsInPackage(const FileDescriptor* file, const string& package_name);
1890
1891 // Like tables_->FindSymbol(), but additionally:
1892 // - Search the pool's underlay if not found in tables_.
1893 // - Insure that the resulting Symbol is from one of the file's declared
1894 // dependencies.
1895 Symbol FindSymbol(const string& name);
1896
kenton@google.com26bd9ee2008-11-21 00:06:27 +00001897 // Like FindSymbol() but does not require that the symbol is in one of the
1898 // file's declared dependencies.
1899 Symbol FindSymbolNotEnforcingDeps(const string& name);
1900
temporal40ee5512008-07-10 02:12:20 +00001901 // Like FindSymbol(), but looks up the name relative to some other symbol
kenton@google.com24bf56f2008-09-24 20:31:01 +00001902 // name. This first searches siblings of relative_to, then siblings of its
temporal40ee5512008-07-10 02:12:20 +00001903 // parents, etc. For example, LookupSymbol("foo.bar", "baz.qux.corge") makes
1904 // the following calls, returning the first non-null result:
1905 // FindSymbol("baz.qux.foo.bar"), FindSymbol("baz.foo.bar"),
kenton@google.comd37d46d2009-04-25 02:53:47 +00001906 // FindSymbol("foo.bar"). If AllowUnknownDependencies() has been called
1907 // on the DescriptorPool, this will generate a placeholder type if
1908 // the name is not found (unless the name itself is malformed). The
1909 // placeholder_type parameter indicates what kind of placeholder should be
1910 // constructed in this case. The resolve_mode parameter determines whether
1911 // any symbol is returned, or only symbols that are types. Note, however,
1912 // that LookupSymbol may still return a non-type symbol in LOOKUP_TYPES mode,
1913 // if it believes that's all it could refer to. The caller should always
1914 // check that it receives the type of symbol it was expecting.
1915 enum PlaceholderType {
1916 PLACEHOLDER_MESSAGE,
1917 PLACEHOLDER_ENUM,
1918 PLACEHOLDER_EXTENDABLE_MESSAGE
1919 };
1920 enum ResolveMode {
1921 LOOKUP_ALL, LOOKUP_TYPES
1922 };
1923 Symbol LookupSymbol(const string& name, const string& relative_to,
1924 PlaceholderType placeholder_type = PLACEHOLDER_MESSAGE,
1925 ResolveMode resolve_mode = LOOKUP_ALL);
1926
1927 // Like LookupSymbol() but will not return a placeholder even if
1928 // AllowUnknownDependencies() has been used.
1929 Symbol LookupSymbolNoPlaceholder(const string& name,
1930 const string& relative_to,
1931 ResolveMode resolve_mode = LOOKUP_ALL);
1932
1933 // Creates a placeholder type suitable for return from LookupSymbol(). May
1934 // return kNullSymbol if the name is not a valid type name.
1935 Symbol NewPlaceholder(const string& name, PlaceholderType placeholder_type);
1936
1937 // Creates a placeholder file. Never returns NULL. This is used when an
1938 // import is not found and AllowUnknownDependencies() is enabled.
1939 const FileDescriptor* NewPlaceholderFile(const string& name);
temporal40ee5512008-07-10 02:12:20 +00001940
temporalf2063512008-07-23 01:19:07 +00001941 // Calls tables_->AddSymbol() and records an error if it fails. Returns
1942 // true if successful or false if failed, though most callers can ignore
1943 // the return value since an error has already been recorded.
1944 bool AddSymbol(const string& full_name,
temporal40ee5512008-07-10 02:12:20 +00001945 const void* parent, const string& name,
1946 const Message& proto, Symbol symbol);
1947
1948 // Like AddSymbol(), but succeeds if the symbol is already defined as long
1949 // as the existing definition is also a package (because it's OK to define
1950 // the same package in two different files). Also adds all parents of the
1951 // packgae to the symbol table (e.g. AddPackage("foo.bar", ...) will add
1952 // "foo.bar" and "foo" to the table).
1953 void AddPackage(const string& name, const Message& proto,
1954 const FileDescriptor* file);
1955
1956 // Checks that the symbol name contains only alphanumeric characters and
1957 // underscores. Records an error otherwise.
1958 void ValidateSymbolName(const string& name, const string& full_name,
1959 const Message& proto);
1960
kenton@google.comd37d46d2009-04-25 02:53:47 +00001961 // Like ValidateSymbolName(), but the name is allowed to contain periods and
1962 // an error is indicated by returning false (not recording the error).
1963 bool ValidateQualifiedName(const string& name);
1964
temporal40ee5512008-07-10 02:12:20 +00001965 // Used by BUILD_ARRAY macro (below) to avoid having to have the type
1966 // specified as a macro parameter.
1967 template <typename Type>
1968 inline void AllocateArray(int size, Type** output) {
1969 *output = tables_->AllocateArray<Type>(size);
1970 }
1971
kenton@google.com24bf56f2008-09-24 20:31:01 +00001972 // Allocates a copy of orig_options in tables_ and stores it in the
1973 // descriptor. Remembers its uninterpreted options, to be interpreted
1974 // later. DescriptorT must be one of the Descriptor messages from
1975 // descriptor.proto.
1976 template<class DescriptorT> void AllocateOptions(
1977 const typename DescriptorT::OptionsType& orig_options,
1978 DescriptorT* descriptor);
kenton@google.comde754372008-11-06 21:36:28 +00001979 // Specialization for FileOptions.
1980 void AllocateOptions(const FileOptions& orig_options,
1981 FileDescriptor* descriptor);
kenton@google.com24bf56f2008-09-24 20:31:01 +00001982
1983 // Implementation for AllocateOptions(). Don't call this directly.
1984 template<class DescriptorT> void AllocateOptionsImpl(
1985 const string& name_scope,
1986 const string& element_name,
1987 const typename DescriptorT::OptionsType& orig_options,
1988 DescriptorT* descriptor);
1989
temporal40ee5512008-07-10 02:12:20 +00001990 // These methods all have the same signature for the sake of the BUILD_ARRAY
1991 // macro, below.
1992 void BuildMessage(const DescriptorProto& proto,
1993 const Descriptor* parent,
1994 Descriptor* result);
1995 void BuildFieldOrExtension(const FieldDescriptorProto& proto,
1996 const Descriptor* parent,
1997 FieldDescriptor* result,
1998 bool is_extension);
1999 void BuildField(const FieldDescriptorProto& proto,
2000 const Descriptor* parent,
2001 FieldDescriptor* result) {
2002 BuildFieldOrExtension(proto, parent, result, false);
2003 }
2004 void BuildExtension(const FieldDescriptorProto& proto,
2005 const Descriptor* parent,
2006 FieldDescriptor* result) {
2007 BuildFieldOrExtension(proto, parent, result, true);
2008 }
2009 void BuildExtensionRange(const DescriptorProto::ExtensionRange& proto,
2010 const Descriptor* parent,
2011 Descriptor::ExtensionRange* result);
2012 void BuildEnum(const EnumDescriptorProto& proto,
2013 const Descriptor* parent,
2014 EnumDescriptor* result);
2015 void BuildEnumValue(const EnumValueDescriptorProto& proto,
2016 const EnumDescriptor* parent,
2017 EnumValueDescriptor* result);
2018 void BuildService(const ServiceDescriptorProto& proto,
2019 const void* dummy,
2020 ServiceDescriptor* result);
2021 void BuildMethod(const MethodDescriptorProto& proto,
2022 const ServiceDescriptor* parent,
2023 MethodDescriptor* result);
2024
kenton@google.com24bf56f2008-09-24 20:31:01 +00002025 // Must be run only after building.
2026 //
2027 // NOTE: Options will not be available during cross-linking, as they
2028 // have not yet been interpreted. Defer any handling of options to the
2029 // Validate*Options methods.
temporal40ee5512008-07-10 02:12:20 +00002030 void CrossLinkFile(FileDescriptor* file, const FileDescriptorProto& proto);
2031 void CrossLinkMessage(Descriptor* message, const DescriptorProto& proto);
2032 void CrossLinkField(FieldDescriptor* field,
2033 const FieldDescriptorProto& proto);
kenton@google.com24bf56f2008-09-24 20:31:01 +00002034 void CrossLinkEnum(EnumDescriptor* enum_type,
2035 const EnumDescriptorProto& proto);
2036 void CrossLinkEnumValue(EnumValueDescriptor* enum_value,
2037 const EnumValueDescriptorProto& proto);
temporal40ee5512008-07-10 02:12:20 +00002038 void CrossLinkService(ServiceDescriptor* service,
2039 const ServiceDescriptorProto& proto);
2040 void CrossLinkMethod(MethodDescriptor* method,
2041 const MethodDescriptorProto& proto);
kenton@google.com24bf56f2008-09-24 20:31:01 +00002042
2043 // Must be run only after cross-linking.
2044 void InterpretOptions();
2045
2046 // A helper class for interpreting options.
2047 class OptionInterpreter {
2048 public:
2049 // Creates an interpreter that operates in the context of the pool of the
2050 // specified builder, which must not be NULL. We don't take ownership of the
2051 // builder.
2052 explicit OptionInterpreter(DescriptorBuilder* builder);
2053
2054 ~OptionInterpreter();
2055
2056 // Interprets the uninterpreted options in the specified Options message.
2057 // On error, calls AddError() on the underlying builder and returns false.
2058 // Otherwise returns true.
2059 bool InterpretOptions(OptionsToInterpret* options_to_interpret);
2060
liujisi@google.com33165fe2010-11-02 13:14:58 +00002061 class AggregateOptionFinder;
2062
kenton@google.com24bf56f2008-09-24 20:31:01 +00002063 private:
2064 // Interprets uninterpreted_option_ on the specified message, which
2065 // must be the mutable copy of the original options message to which
2066 // uninterpreted_option_ belongs.
2067 bool InterpretSingleOption(Message* options);
2068
kenton@google.comd37d46d2009-04-25 02:53:47 +00002069 // Adds the uninterpreted_option to the given options message verbatim.
2070 // Used when AllowUnknownDependencies() is in effect and we can't find
2071 // the option's definition.
2072 void AddWithoutInterpreting(const UninterpretedOption& uninterpreted_option,
2073 Message* options);
2074
kenton@google.com24bf56f2008-09-24 20:31:01 +00002075 // A recursive helper function that drills into the intermediate fields
kenton@google.com80b1d622009-07-29 01:13:20 +00002076 // in unknown_fields to check if field innermost_field is set on the
kenton@google.com24bf56f2008-09-24 20:31:01 +00002077 // innermost message. Returns false and sets an error if so.
2078 bool ExamineIfOptionIsSet(
2079 vector<const FieldDescriptor*>::const_iterator intermediate_fields_iter,
2080 vector<const FieldDescriptor*>::const_iterator intermediate_fields_end,
2081 const FieldDescriptor* innermost_field, const string& debug_msg_name,
2082 const UnknownFieldSet& unknown_fields);
2083
2084 // Validates the value for the option field of the currently interpreted
2085 // option and then sets it on the unknown_field.
2086 bool SetOptionValue(const FieldDescriptor* option_field,
kenton@google.comd37d46d2009-04-25 02:53:47 +00002087 UnknownFieldSet* unknown_fields);
kenton@google.com24bf56f2008-09-24 20:31:01 +00002088
liujisi@google.com33165fe2010-11-02 13:14:58 +00002089 // Parses an aggregate value for a CPPTYPE_MESSAGE option and
2090 // saves it into *unknown_fields.
2091 bool SetAggregateOption(const FieldDescriptor* option_field,
2092 UnknownFieldSet* unknown_fields);
2093
kenton@google.com24bf56f2008-09-24 20:31:01 +00002094 // Convenience functions to set an int field the right way, depending on
2095 // its wire type (a single int CppType can represent multiple wire types).
kenton@google.comd37d46d2009-04-25 02:53:47 +00002096 void SetInt32(int number, int32 value, FieldDescriptor::Type type,
2097 UnknownFieldSet* unknown_fields);
2098 void SetInt64(int number, int64 value, FieldDescriptor::Type type,
2099 UnknownFieldSet* unknown_fields);
2100 void SetUInt32(int number, uint32 value, FieldDescriptor::Type type,
2101 UnknownFieldSet* unknown_fields);
2102 void SetUInt64(int number, uint64 value, FieldDescriptor::Type type,
2103 UnknownFieldSet* unknown_fields);
kenton@google.com24bf56f2008-09-24 20:31:01 +00002104
2105 // A helper function that adds an error at the specified location of the
2106 // option we're currently interpreting, and returns false.
2107 bool AddOptionError(DescriptorPool::ErrorCollector::ErrorLocation location,
2108 const string& msg) {
2109 builder_->AddError(options_to_interpret_->element_name,
2110 *uninterpreted_option_, location, msg);
2111 return false;
2112 }
2113
2114 // A helper function that adds an error at the location of the option name
2115 // and returns false.
2116 bool AddNameError(const string& msg) {
2117 return AddOptionError(DescriptorPool::ErrorCollector::OPTION_NAME, msg);
2118 }
2119
2120 // A helper function that adds an error at the location of the option name
2121 // and returns false.
2122 bool AddValueError(const string& msg) {
2123 return AddOptionError(DescriptorPool::ErrorCollector::OPTION_VALUE, msg);
2124 }
2125
2126 // We interpret against this builder's pool. Is never NULL. We don't own
2127 // this pointer.
2128 DescriptorBuilder* builder_;
2129
2130 // The options we're currently interpreting, or NULL if we're not in a call
2131 // to InterpretOptions.
2132 const OptionsToInterpret* options_to_interpret_;
2133
2134 // The option we're currently interpreting within options_to_interpret_, or
2135 // NULL if we're not in a call to InterpretOptions(). This points to a
2136 // submessage of the original option, not the mutable copy. Therefore we
2137 // can use it to find locations recorded by the parser.
2138 const UninterpretedOption* uninterpreted_option_;
2139
liujisi@google.com33165fe2010-11-02 13:14:58 +00002140 // Factory used to create the dynamic messages we need to parse
2141 // any aggregate option values we encounter.
2142 DynamicMessageFactory dynamic_factory_;
2143
kenton@google.com24bf56f2008-09-24 20:31:01 +00002144 GOOGLE_DISALLOW_EVIL_CONSTRUCTORS(OptionInterpreter);
2145 };
2146
kenton@google.comd37d46d2009-04-25 02:53:47 +00002147 // Work-around for broken compilers: According to the C++ standard,
2148 // OptionInterpreter should have access to the private members of any class
2149 // which has declared DescriptorBuilder as a friend. Unfortunately some old
2150 // versions of GCC and other compilers do not implement this correctly. So,
2151 // we have to have these intermediate methods to provide access. We also
2152 // redundantly declare OptionInterpreter a friend just to make things extra
2153 // clear for these bad compilers.
2154 friend class OptionInterpreter;
liujisi@google.com33165fe2010-11-02 13:14:58 +00002155 friend class OptionInterpreter::AggregateOptionFinder;
2156
kenton@google.comd37d46d2009-04-25 02:53:47 +00002157 static inline bool get_allow_unknown(const DescriptorPool* pool) {
2158 return pool->allow_unknown_;
2159 }
2160 static inline bool get_is_placeholder(const Descriptor* descriptor) {
2161 return descriptor->is_placeholder_;
2162 }
2163
kenton@google.com24bf56f2008-09-24 20:31:01 +00002164 // Must be run only after options have been interpreted.
2165 //
2166 // NOTE: Validation code must only reference the options in the mutable
2167 // descriptors, which are the ones that have been interpreted. The const
2168 // proto references are passed in only so they can be provided to calls to
2169 // AddError(). Do not look at their options, which have not been interpreted.
2170 void ValidateFileOptions(FileDescriptor* file,
2171 const FileDescriptorProto& proto);
2172 void ValidateMessageOptions(Descriptor* message,
2173 const DescriptorProto& proto);
2174 void ValidateFieldOptions(FieldDescriptor* field,
2175 const FieldDescriptorProto& proto);
2176 void ValidateEnumOptions(EnumDescriptor* enm,
2177 const EnumDescriptorProto& proto);
2178 void ValidateEnumValueOptions(EnumValueDescriptor* enum_value,
2179 const EnumValueDescriptorProto& proto);
2180 void ValidateServiceOptions(ServiceDescriptor* service,
2181 const ServiceDescriptorProto& proto);
2182 void ValidateMethodOptions(MethodDescriptor* method,
2183 const MethodDescriptorProto& proto);
2184
2185 void ValidateMapKey(FieldDescriptor* field,
2186 const FieldDescriptorProto& proto);
temporal40ee5512008-07-10 02:12:20 +00002187};
2188
2189const FileDescriptor* DescriptorPool::BuildFile(
2190 const FileDescriptorProto& proto) {
2191 GOOGLE_CHECK(fallback_database_ == NULL)
2192 << "Cannot call BuildFile on a DescriptorPool that uses a "
2193 "DescriptorDatabase. You must instead find a way to get your file "
2194 "into the underlying database.";
2195 GOOGLE_CHECK(mutex_ == NULL); // Implied by the above GOOGLE_CHECK.
kenton@google.comd37d46d2009-04-25 02:53:47 +00002196 return DescriptorBuilder(this, tables_.get(), NULL).BuildFile(proto);
temporal40ee5512008-07-10 02:12:20 +00002197}
2198
2199const FileDescriptor* DescriptorPool::BuildFileCollectingErrors(
2200 const FileDescriptorProto& proto,
2201 ErrorCollector* error_collector) {
2202 GOOGLE_CHECK(fallback_database_ == NULL)
2203 << "Cannot call BuildFile on a DescriptorPool that uses a "
2204 "DescriptorDatabase. You must instead find a way to get your file "
2205 "into the underlying database.";
2206 GOOGLE_CHECK(mutex_ == NULL); // Implied by the above GOOGLE_CHECK.
2207 return DescriptorBuilder(this, tables_.get(),
kenton@google.comd37d46d2009-04-25 02:53:47 +00002208 error_collector).BuildFile(proto);
temporal40ee5512008-07-10 02:12:20 +00002209}
2210
2211const FileDescriptor* DescriptorPool::BuildFileFromDatabase(
2212 const FileDescriptorProto& proto) const {
2213 mutex_->AssertHeld();
2214 return DescriptorBuilder(this, tables_.get(),
kenton@google.comd37d46d2009-04-25 02:53:47 +00002215 default_error_collector_).BuildFile(proto);
temporal40ee5512008-07-10 02:12:20 +00002216}
2217
2218DescriptorBuilder::DescriptorBuilder(
2219 const DescriptorPool* pool,
2220 DescriptorPool::Tables* tables,
2221 DescriptorPool::ErrorCollector* error_collector)
2222 : pool_(pool),
2223 tables_(tables),
2224 error_collector_(error_collector),
2225 had_errors_(false),
2226 possible_undeclared_dependency_(NULL) {}
2227
2228DescriptorBuilder::~DescriptorBuilder() {}
2229
2230void DescriptorBuilder::AddError(
2231 const string& element_name,
2232 const Message& descriptor,
2233 DescriptorPool::ErrorCollector::ErrorLocation location,
2234 const string& error) {
2235 if (error_collector_ == NULL) {
2236 if (!had_errors_) {
2237 GOOGLE_LOG(ERROR) << "Invalid proto descriptor for file \"" << filename_
2238 << "\":";
2239 }
2240 GOOGLE_LOG(ERROR) << " " << element_name << ": " << error;
2241 } else {
2242 error_collector_->AddError(filename_, element_name,
2243 &descriptor, location, error);
2244 }
2245 had_errors_ = true;
2246}
2247
2248void DescriptorBuilder::AddNotDefinedError(
2249 const string& element_name,
2250 const Message& descriptor,
2251 DescriptorPool::ErrorCollector::ErrorLocation location,
2252 const string& undefined_symbol) {
2253 if (possible_undeclared_dependency_ == NULL) {
2254 AddError(element_name, descriptor, location,
2255 "\"" + undefined_symbol + "\" is not defined.");
2256 } else {
2257 AddError(element_name, descriptor, location,
kenton@google.com80b1d622009-07-29 01:13:20 +00002258 "\"" + possible_undeclared_dependency_name_ +
2259 "\" seems to be defined in \"" +
2260 possible_undeclared_dependency_->name() + "\", which is not "
temporal40ee5512008-07-10 02:12:20 +00002261 "imported by \"" + filename_ + "\". To use it here, please "
2262 "add the necessary import.");
2263 }
2264}
2265
2266bool DescriptorBuilder::IsInPackage(const FileDescriptor* file,
2267 const string& package_name) {
2268 return HasPrefixString(file->package(), package_name) &&
2269 (file->package().size() == package_name.size() ||
2270 file->package()[package_name.size()] == '.');
2271}
2272
kenton@google.com26bd9ee2008-11-21 00:06:27 +00002273Symbol DescriptorBuilder::FindSymbolNotEnforcingDeps(const string& name) {
temporal40ee5512008-07-10 02:12:20 +00002274 Symbol result;
2275
2276 // We need to search our pool and all its underlays.
2277 const DescriptorPool* pool = pool_;
2278 while (true) {
2279 // If we are looking at an underlay, we must lock its mutex_, since we are
2280 // accessing the underlay's tables_ dircetly.
2281 MutexLockMaybe lock((pool == pool_) ? NULL : pool->mutex_);
2282
2283 // Note that we don't have to check fallback_database_ here because the
2284 // symbol has to be in one of its file's direct dependencies, and we have
2285 // already loaded those by the time we get here.
2286 result = pool->tables_->FindSymbol(name);
2287 if (!result.IsNull()) break;
2288 if (pool->underlay_ == NULL) return kNullSymbol;
2289 pool = pool->underlay_;
2290 }
2291
kenton@google.com26bd9ee2008-11-21 00:06:27 +00002292 return result;
2293}
2294
2295Symbol DescriptorBuilder::FindSymbol(const string& name) {
2296 Symbol result = FindSymbolNotEnforcingDeps(name);
2297
temporal40ee5512008-07-10 02:12:20 +00002298 if (!pool_->enforce_dependencies_) {
2299 // Hack for CompilerUpgrader.
2300 return result;
2301 }
2302
2303 // Only find symbols which were defined in this file or one of its
2304 // dependencies.
2305 const FileDescriptor* file = result.GetFile();
2306 if (file == file_) return result;
2307 for (int i = 0; i < file_->dependency_count(); i++) {
2308 if (file == file_->dependency(i)) return result;
2309 }
2310
2311 if (result.type == Symbol::PACKAGE) {
2312 // Arg, this is overcomplicated. The symbol is a package name. It could
2313 // be that the package was defined in multiple files. result.GetFile()
2314 // returns the first file we saw that used this package. We've determined
2315 // that that file is not a direct dependency of the file we are currently
2316 // building, but it could be that some other file which *is* a direct
2317 // dependency also defines the same package. We can't really rule out this
2318 // symbol unless none of the dependencies define it.
2319 if (IsInPackage(file_, name)) return result;
2320 for (int i = 0; i < file_->dependency_count(); i++) {
kenton@google.com80b1d622009-07-29 01:13:20 +00002321 // Note: A dependency may be NULL if it was not found or had errors.
2322 if (file_->dependency(i) != NULL &&
2323 IsInPackage(file_->dependency(i), name)) {
2324 return result;
2325 }
temporal40ee5512008-07-10 02:12:20 +00002326 }
2327 }
2328
2329 possible_undeclared_dependency_ = file;
kenton@google.com80b1d622009-07-29 01:13:20 +00002330 possible_undeclared_dependency_name_ = name;
temporal40ee5512008-07-10 02:12:20 +00002331 return kNullSymbol;
2332}
2333
kenton@google.comd37d46d2009-04-25 02:53:47 +00002334Symbol DescriptorBuilder::LookupSymbolNoPlaceholder(
2335 const string& name, const string& relative_to, ResolveMode resolve_mode) {
temporal40ee5512008-07-10 02:12:20 +00002336 possible_undeclared_dependency_ = NULL;
2337
2338 if (name.size() > 0 && name[0] == '.') {
2339 // Fully-qualified name.
2340 return FindSymbol(name.substr(1));
2341 }
2342
2343 // If name is something like "Foo.Bar.baz", and symbols named "Foo" are
2344 // defined in multiple parent scopes, we only want to find "Bar.baz" in the
2345 // innermost one. E.g., the following should produce an error:
2346 // message Bar { message Baz {} }
2347 // message Foo {
2348 // message Bar {
2349 // }
2350 // optional Bar.Baz baz = 1;
2351 // }
2352 // So, we look for just "Foo" first, then look for "Bar.baz" within it if
2353 // found.
2354 int name_dot_pos = name.find_first_of('.');
2355 string first_part_of_name;
2356 if (name_dot_pos == string::npos) {
2357 first_part_of_name = name;
2358 } else {
2359 first_part_of_name = name.substr(0, name_dot_pos);
2360 }
2361
2362 string scope_to_try(relative_to);
2363
2364 while (true) {
2365 // Chop off the last component of the scope.
2366 string::size_type dot_pos = scope_to_try.find_last_of('.');
2367 if (dot_pos == string::npos) {
2368 return FindSymbol(name);
2369 } else {
2370 scope_to_try.erase(dot_pos);
2371 }
2372
2373 // Append ".first_part_of_name" and try to find.
2374 string::size_type old_size = scope_to_try.size();
2375 scope_to_try.append(1, '.');
2376 scope_to_try.append(first_part_of_name);
2377 Symbol result = FindSymbol(scope_to_try);
2378 if (!result.IsNull()) {
2379 if (first_part_of_name.size() < name.size()) {
2380 // name is a compound symbol, of which we only found the first part.
2381 // Now try to look up the rest of it.
kenton@google.comd37d46d2009-04-25 02:53:47 +00002382 if (result.IsAggregate()) {
2383 scope_to_try.append(name, first_part_of_name.size(),
2384 name.size() - first_part_of_name.size());
2385 return FindSymbol(scope_to_try);
2386 } else {
2387 // We found a symbol but it's not an aggregate. Continue the loop.
2388 }
2389 } else {
2390 if (resolve_mode == LOOKUP_TYPES && !result.IsType()) {
2391 // We found a symbol but it's not a type. Continue the loop.
2392 } else {
2393 return result;
2394 }
temporal40ee5512008-07-10 02:12:20 +00002395 }
temporal40ee5512008-07-10 02:12:20 +00002396 }
2397
2398 // Not found. Remove the name so we can try again.
2399 scope_to_try.erase(old_size);
2400 }
2401}
2402
kenton@google.comd37d46d2009-04-25 02:53:47 +00002403Symbol DescriptorBuilder::LookupSymbol(
2404 const string& name, const string& relative_to,
2405 PlaceholderType placeholder_type, ResolveMode resolve_mode) {
2406 Symbol result = LookupSymbolNoPlaceholder(
2407 name, relative_to, resolve_mode);
2408 if (result.IsNull() && pool_->allow_unknown_) {
2409 // Not found, but AllowUnknownDependencies() is enabled. Return a
2410 // placeholder instead.
2411 result = NewPlaceholder(name, placeholder_type);
2412 }
2413 return result;
2414}
2415
2416Symbol DescriptorBuilder::NewPlaceholder(const string& name,
2417 PlaceholderType placeholder_type) {
2418 // Compute names.
2419 const string* placeholder_full_name;
2420 const string* placeholder_name;
2421 const string* placeholder_package;
2422
2423 if (!ValidateQualifiedName(name)) return kNullSymbol;
2424 if (name[0] == '.') {
2425 // Fully-qualified.
2426 placeholder_full_name = tables_->AllocateString(name.substr(1));
2427 } else {
2428 placeholder_full_name = tables_->AllocateString(name);
2429 }
2430
2431 string::size_type dotpos = placeholder_full_name->find_last_of('.');
2432 if (dotpos != string::npos) {
2433 placeholder_package = tables_->AllocateString(
2434 placeholder_full_name->substr(0, dotpos));
2435 placeholder_name = tables_->AllocateString(
2436 placeholder_full_name->substr(dotpos + 1));
2437 } else {
2438 placeholder_package = &kEmptyString;
2439 placeholder_name = placeholder_full_name;
2440 }
2441
2442 // Create the placeholders.
2443 FileDescriptor* placeholder_file = tables_->Allocate<FileDescriptor>();
2444 memset(placeholder_file, 0, sizeof(*placeholder_file));
2445
2446 placeholder_file->name_ =
2447 tables_->AllocateString(*placeholder_full_name + ".placeholder.proto");
2448 placeholder_file->package_ = placeholder_package;
2449 placeholder_file->pool_ = pool_;
2450 placeholder_file->options_ = &FileOptions::default_instance();
2451 placeholder_file->tables_ = &FileDescriptorTables::kEmpty;
2452 // All other fields are zero or NULL.
2453
2454 if (placeholder_type == PLACEHOLDER_ENUM) {
2455 placeholder_file->enum_type_count_ = 1;
2456 placeholder_file->enum_types_ =
2457 tables_->AllocateArray<EnumDescriptor>(1);
2458
2459 EnumDescriptor* placeholder_enum = &placeholder_file->enum_types_[0];
2460 memset(placeholder_enum, 0, sizeof(*placeholder_enum));
2461
2462 placeholder_enum->full_name_ = placeholder_full_name;
2463 placeholder_enum->name_ = placeholder_name;
2464 placeholder_enum->file_ = placeholder_file;
2465 placeholder_enum->options_ = &EnumOptions::default_instance();
2466 placeholder_enum->is_placeholder_ = true;
2467 placeholder_enum->is_unqualified_placeholder_ = (name[0] != '.');
2468
2469 // Enums must have at least one value.
2470 placeholder_enum->value_count_ = 1;
2471 placeholder_enum->values_ = tables_->AllocateArray<EnumValueDescriptor>(1);
2472
2473 EnumValueDescriptor* placeholder_value = &placeholder_enum->values_[0];
2474 memset(placeholder_value, 0, sizeof(*placeholder_value));
2475
2476 placeholder_value->name_ = tables_->AllocateString("PLACEHOLDER_VALUE");
2477 // Note that enum value names are siblings of their type, not children.
2478 placeholder_value->full_name_ =
2479 placeholder_package->empty() ? placeholder_value->name_ :
2480 tables_->AllocateString(*placeholder_package + ".PLACEHOLDER_VALUE");
2481
2482 placeholder_value->number_ = 0;
2483 placeholder_value->type_ = placeholder_enum;
2484 placeholder_value->options_ = &EnumValueOptions::default_instance();
2485
2486 return Symbol(placeholder_enum);
2487 } else {
2488 placeholder_file->message_type_count_ = 1;
2489 placeholder_file->message_types_ =
2490 tables_->AllocateArray<Descriptor>(1);
2491
2492 Descriptor* placeholder_message = &placeholder_file->message_types_[0];
2493 memset(placeholder_message, 0, sizeof(*placeholder_message));
2494
2495 placeholder_message->full_name_ = placeholder_full_name;
2496 placeholder_message->name_ = placeholder_name;
2497 placeholder_message->file_ = placeholder_file;
2498 placeholder_message->options_ = &MessageOptions::default_instance();
2499 placeholder_message->is_placeholder_ = true;
2500 placeholder_message->is_unqualified_placeholder_ = (name[0] != '.');
2501
2502 if (placeholder_type == PLACEHOLDER_EXTENDABLE_MESSAGE) {
2503 placeholder_message->extension_range_count_ = 1;
2504 placeholder_message->extension_ranges_ =
2505 tables_->AllocateArray<Descriptor::ExtensionRange>(1);
2506 placeholder_message->extension_ranges_->start = 1;
2507 // kMaxNumber + 1 because ExtensionRange::end is exclusive.
2508 placeholder_message->extension_ranges_->end =
2509 FieldDescriptor::kMaxNumber + 1;
2510 }
2511
2512 return Symbol(placeholder_message);
2513 }
2514}
2515
2516const FileDescriptor* DescriptorBuilder::NewPlaceholderFile(
2517 const string& name) {
2518 FileDescriptor* placeholder = tables_->Allocate<FileDescriptor>();
2519 memset(placeholder, 0, sizeof(*placeholder));
2520
2521 placeholder->name_ = tables_->AllocateString(name);
2522 placeholder->package_ = &kEmptyString;
2523 placeholder->pool_ = pool_;
2524 placeholder->options_ = &FileOptions::default_instance();
2525 placeholder->tables_ = &FileDescriptorTables::kEmpty;
2526 // All other fields are zero or NULL.
2527
2528 return placeholder;
2529}
2530
temporalf2063512008-07-23 01:19:07 +00002531bool DescriptorBuilder::AddSymbol(
temporal40ee5512008-07-10 02:12:20 +00002532 const string& full_name, const void* parent, const string& name,
2533 const Message& proto, Symbol symbol) {
2534 // If the caller passed NULL for the parent, the symbol is at file scope.
2535 // Use its file as the parent instead.
2536 if (parent == NULL) parent = file_;
2537
kenton@google.comd37d46d2009-04-25 02:53:47 +00002538 if (tables_->AddSymbol(full_name, symbol)) {
2539 if (!file_tables_->AddAliasUnderParent(parent, name, symbol)) {
2540 GOOGLE_LOG(DFATAL) << "\"" << full_name << "\" not previously defined in "
2541 "symbols_by_name_, but was defined in symbols_by_parent_; "
2542 "this shouldn't be possible.";
2543 return false;
2544 }
temporalf2063512008-07-23 01:19:07 +00002545 return true;
2546 } else {
temporal40ee5512008-07-10 02:12:20 +00002547 const FileDescriptor* other_file = tables_->FindSymbol(full_name).GetFile();
2548 if (other_file == file_) {
2549 string::size_type dot_pos = full_name.find_last_of('.');
2550 if (dot_pos == string::npos) {
2551 AddError(full_name, proto, DescriptorPool::ErrorCollector::NAME,
2552 "\"" + full_name + "\" is already defined.");
2553 } else {
2554 AddError(full_name, proto, DescriptorPool::ErrorCollector::NAME,
2555 "\"" + full_name.substr(dot_pos + 1) +
2556 "\" is already defined in \"" +
2557 full_name.substr(0, dot_pos) + "\".");
2558 }
2559 } else {
2560 // Symbol seems to have been defined in a different file.
2561 AddError(full_name, proto, DescriptorPool::ErrorCollector::NAME,
2562 "\"" + full_name + "\" is already defined in file \"" +
2563 other_file->name() + "\".");
2564 }
temporalf2063512008-07-23 01:19:07 +00002565 return false;
temporal40ee5512008-07-10 02:12:20 +00002566 }
2567}
2568
2569void DescriptorBuilder::AddPackage(
2570 const string& name, const Message& proto, const FileDescriptor* file) {
kenton@google.comd37d46d2009-04-25 02:53:47 +00002571 if (tables_->AddSymbol(name, Symbol(file))) {
temporal40ee5512008-07-10 02:12:20 +00002572 // Success. Also add parent package, if any.
2573 string::size_type dot_pos = name.find_last_of('.');
2574 if (dot_pos == string::npos) {
2575 // No parents.
2576 ValidateSymbolName(name, name, proto);
2577 } else {
2578 // Has parent.
2579 string* parent_name = tables_->AllocateString(name.substr(0, dot_pos));
2580 AddPackage(*parent_name, proto, file);
2581 ValidateSymbolName(name.substr(dot_pos + 1), name, proto);
2582 }
2583 } else {
2584 Symbol existing_symbol = tables_->FindSymbol(name);
2585 // It's OK to redefine a package.
2586 if (existing_symbol.type != Symbol::PACKAGE) {
2587 // Symbol seems to have been defined in a different file.
2588 AddError(name, proto, DescriptorPool::ErrorCollector::NAME,
2589 "\"" + name + "\" is already defined (as something other than "
2590 "a package) in file \"" + existing_symbol.GetFile()->name() +
2591 "\".");
2592 }
2593 }
2594}
2595
2596void DescriptorBuilder::ValidateSymbolName(
2597 const string& name, const string& full_name, const Message& proto) {
2598 if (name.empty()) {
2599 AddError(full_name, proto, DescriptorPool::ErrorCollector::NAME,
2600 "Missing name.");
2601 } else {
2602 for (int i = 0; i < name.size(); i++) {
2603 // I don't trust isalnum() due to locales. :(
2604 if ((name[i] < 'a' || 'z' < name[i]) &&
2605 (name[i] < 'A' || 'Z' < name[i]) &&
2606 (name[i] < '0' || '9' < name[i]) &&
2607 (name[i] != '_')) {
2608 AddError(full_name, proto, DescriptorPool::ErrorCollector::NAME,
2609 "\"" + name + "\" is not a valid identifier.");
2610 }
2611 }
2612 }
2613}
2614
kenton@google.comd37d46d2009-04-25 02:53:47 +00002615bool DescriptorBuilder::ValidateQualifiedName(const string& name) {
2616 bool last_was_period = false;
2617
2618 for (int i = 0; i < name.size(); i++) {
2619 // I don't trust isalnum() due to locales. :(
2620 if (('a' <= name[i] && name[i] <= 'z') ||
2621 ('A' <= name[i] && name[i] <= 'Z') ||
2622 ('0' <= name[i] && name[i] <= '9') ||
2623 (name[i] == '_')) {
2624 last_was_period = false;
2625 } else if (name[i] == '.') {
2626 if (last_was_period) return false;
2627 last_was_period = true;
2628 } else {
2629 return false;
2630 }
2631 }
2632
2633 return !name.empty() && !last_was_period;
2634}
2635
temporal40ee5512008-07-10 02:12:20 +00002636// -------------------------------------------------------------------
2637
kenton@google.com24bf56f2008-09-24 20:31:01 +00002638// This generic implementation is good for all descriptors except
2639// FileDescriptor.
2640template<class DescriptorT> void DescriptorBuilder::AllocateOptions(
2641 const typename DescriptorT::OptionsType& orig_options,
2642 DescriptorT* descriptor) {
2643 AllocateOptionsImpl(descriptor->full_name(), descriptor->full_name(),
2644 orig_options, descriptor);
2645}
2646
2647// We specialize for FileDescriptor.
kenton@google.comde754372008-11-06 21:36:28 +00002648void DescriptorBuilder::AllocateOptions(const FileOptions& orig_options,
2649 FileDescriptor* descriptor) {
kenton@google.com24bf56f2008-09-24 20:31:01 +00002650 // We add the dummy token so that LookupSymbol does the right thing.
2651 AllocateOptionsImpl(descriptor->package() + ".dummy", descriptor->name(),
2652 orig_options, descriptor);
2653}
2654
2655template<class DescriptorT> void DescriptorBuilder::AllocateOptionsImpl(
2656 const string& name_scope,
2657 const string& element_name,
2658 const typename DescriptorT::OptionsType& orig_options,
2659 DescriptorT* descriptor) {
kenton@google.comd37d46d2009-04-25 02:53:47 +00002660 // We need to use a dummy pointer to work around a bug in older versions of
2661 // GCC. Otherwise, the following two lines could be replaced with:
2662 // typename DescriptorT::OptionsType* options =
2663 // tables_->AllocateMessage<typename DescriptorT::OptionsType>();
2664 typename DescriptorT::OptionsType* const dummy = NULL;
2665 typename DescriptorT::OptionsType* options = tables_->AllocateMessage(dummy);
kenton@google.com24bf56f2008-09-24 20:31:01 +00002666 options->CopyFrom(orig_options);
kenton@google.com24bf56f2008-09-24 20:31:01 +00002667 descriptor->options_ = options;
kenton@google.comd37d46d2009-04-25 02:53:47 +00002668
2669 // Don't add to options_to_interpret_ unless there were uninterpreted
2670 // options. This not only avoids unnecessary work, but prevents a
2671 // bootstrapping problem when building descriptors for descriptor.proto.
2672 // descriptor.proto does not contain any uninterpreted options, but
2673 // attempting to interpret options anyway will cause
2674 // OptionsType::GetDescriptor() to be called which may then deadlock since
2675 // we're still trying to build it.
2676 if (options->uninterpreted_option_size() > 0) {
2677 options_to_interpret_.push_back(
2678 OptionsToInterpret(name_scope, element_name, &orig_options, options));
2679 }
kenton@google.com24bf56f2008-09-24 20:31:01 +00002680}
2681
2682
temporal40ee5512008-07-10 02:12:20 +00002683// A common pattern: We want to convert a repeated field in the descriptor
2684// to an array of values, calling some method to build each value.
2685#define BUILD_ARRAY(INPUT, OUTPUT, NAME, METHOD, PARENT) \
2686 OUTPUT->NAME##_count_ = INPUT.NAME##_size(); \
2687 AllocateArray(INPUT.NAME##_size(), &OUTPUT->NAME##s_); \
2688 for (int i = 0; i < INPUT.NAME##_size(); i++) { \
2689 METHOD(INPUT.NAME(i), PARENT, OUTPUT->NAME##s_ + i); \
2690 }
2691
2692const FileDescriptor* DescriptorBuilder::BuildFile(
kenton@google.comd37d46d2009-04-25 02:53:47 +00002693 const FileDescriptorProto& proto) {
temporal40ee5512008-07-10 02:12:20 +00002694 filename_ = proto.name();
2695
kenton@google.comd37d46d2009-04-25 02:53:47 +00002696 // Check if the file already exists and is identical to the one being built.
2697 // Note: This only works if the input is canonical -- that is, it
2698 // fully-qualifies all type names, has no UninterpretedOptions, etc.
2699 // This is fine, because this idempotency "feature" really only exists to
2700 // accomodate one hack in the proto1->proto2 migration layer.
2701 const FileDescriptor* existing_file = tables_->FindFile(filename_);
2702 if (existing_file != NULL) {
2703 // File already in pool. Compare the existing one to the input.
2704 FileDescriptorProto existing_proto;
2705 existing_file->CopyTo(&existing_proto);
2706 if (existing_proto.SerializeAsString() == proto.SerializeAsString()) {
2707 // They're identical. Return the existing descriptor.
2708 return existing_file;
2709 }
2710
2711 // Not a match. The error will be detected and handled later.
2712 }
2713
temporal40ee5512008-07-10 02:12:20 +00002714 // Check to see if this file is already on the pending files list.
2715 // TODO(kenton): Allow recursive imports? It may not work with some
2716 // (most?) programming languages. E.g., in C++, a forward declaration
2717 // of a type is not sufficient to allow it to be used even in a
2718 // generated header file due to inlining. This could perhaps be
2719 // worked around using tricks involving inserting #include statements
2720 // mid-file, but that's pretty ugly, and I'm pretty sure there are
2721 // some languages out there that do not allow recursive dependencies
2722 // at all.
2723 for (int i = 0; i < tables_->pending_files_.size(); i++) {
2724 if (tables_->pending_files_[i] == proto.name()) {
2725 string error_message("File recursively imports itself: ");
2726 for (; i < tables_->pending_files_.size(); i++) {
2727 error_message.append(tables_->pending_files_[i]);
2728 error_message.append(" -> ");
2729 }
2730 error_message.append(proto.name());
2731
2732 AddError(proto.name(), proto, DescriptorPool::ErrorCollector::OTHER,
2733 error_message);
2734 return NULL;
2735 }
2736 }
2737
2738 // If we have a fallback_database_, attempt to load all dependencies now,
2739 // before checkpointing tables_. This avoids confusion with recursive
2740 // checkpoints.
2741 if (pool_->fallback_database_ != NULL) {
2742 tables_->pending_files_.push_back(proto.name());
2743 for (int i = 0; i < proto.dependency_size(); i++) {
2744 if (tables_->FindFile(proto.dependency(i)) == NULL &&
2745 (pool_->underlay_ == NULL ||
2746 pool_->underlay_->FindFileByName(proto.dependency(i)) == NULL)) {
2747 // We don't care what this returns since we'll find out below anyway.
2748 pool_->TryFindFileInFallbackDatabase(proto.dependency(i));
2749 }
2750 }
2751 tables_->pending_files_.pop_back();
2752 }
2753
2754 // Checkpoint the tables so that we can roll back if something goes wrong.
2755 tables_->Checkpoint();
2756
2757 FileDescriptor* result = tables_->Allocate<FileDescriptor>();
2758 file_ = result;
2759
kenton@google.comd37d46d2009-04-25 02:53:47 +00002760 file_tables_ = tables_->AllocateFileTables();
2761 file_->tables_ = file_tables_;
2762
temporal40ee5512008-07-10 02:12:20 +00002763 if (!proto.has_name()) {
2764 AddError("", proto, DescriptorPool::ErrorCollector::OTHER,
2765 "Missing field: FileDescriptorProto.name.");
2766 }
2767
2768 result->name_ = tables_->AllocateString(proto.name());
temporal6fdb0962008-07-25 04:38:05 +00002769 if (proto.has_package()) {
2770 result->package_ = tables_->AllocateString(proto.package());
2771 } else {
2772 // We cannot rely on proto.package() returning a valid string if
2773 // proto.has_package() is false, because we might be running at static
2774 // initialization time, in which case default values have not yet been
2775 // initialized.
2776 result->package_ = tables_->AllocateString("");
2777 }
temporal40ee5512008-07-10 02:12:20 +00002778 result->pool_ = pool_;
2779
2780 // Add to tables.
2781 if (!tables_->AddFile(result)) {
2782 AddError(proto.name(), proto, DescriptorPool::ErrorCollector::OTHER,
2783 "A file with this name is already in the pool.");
2784 // Bail out early so that if this is actually the exact same file, we
2785 // don't end up reporting that every single symbol is already defined.
2786 tables_->Rollback();
2787 return NULL;
2788 }
2789 if (!result->package().empty()) {
2790 AddPackage(result->package(), proto, result);
2791 }
2792
2793 // Make sure all dependencies are loaded.
2794 set<string> seen_dependencies;
2795 result->dependency_count_ = proto.dependency_size();
2796 result->dependencies_ =
2797 tables_->AllocateArray<const FileDescriptor*>(proto.dependency_size());
2798 for (int i = 0; i < proto.dependency_size(); i++) {
2799 if (!seen_dependencies.insert(proto.dependency(i)).second) {
2800 AddError(proto.name(), proto,
2801 DescriptorPool::ErrorCollector::OTHER,
2802 "Import \"" + proto.dependency(i) + "\" was listed twice.");
2803 }
2804
2805 const FileDescriptor* dependency = tables_->FindFile(proto.dependency(i));
2806 if (dependency == NULL && pool_->underlay_ != NULL) {
2807 dependency = pool_->underlay_->FindFileByName(proto.dependency(i));
2808 }
2809
2810 if (dependency == NULL) {
kenton@google.comd37d46d2009-04-25 02:53:47 +00002811 if (pool_->allow_unknown_) {
2812 dependency = NewPlaceholderFile(proto.dependency(i));
temporal40ee5512008-07-10 02:12:20 +00002813 } else {
kenton@google.comd37d46d2009-04-25 02:53:47 +00002814 string message;
2815 if (pool_->fallback_database_ == NULL) {
2816 message = "Import \"" + proto.dependency(i) +
2817 "\" has not been loaded.";
2818 } else {
2819 message = "Import \"" + proto.dependency(i) +
2820 "\" was not found or had errors.";
2821 }
2822 AddError(proto.name(), proto,
2823 DescriptorPool::ErrorCollector::OTHER,
2824 message);
temporal40ee5512008-07-10 02:12:20 +00002825 }
temporal40ee5512008-07-10 02:12:20 +00002826 }
2827
2828 result->dependencies_[i] = dependency;
2829 }
2830
2831 // Convert children.
2832 BUILD_ARRAY(proto, result, message_type, BuildMessage , NULL);
2833 BUILD_ARRAY(proto, result, enum_type , BuildEnum , NULL);
2834 BUILD_ARRAY(proto, result, service , BuildService , NULL);
2835 BUILD_ARRAY(proto, result, extension , BuildExtension, NULL);
2836
2837 // Copy options.
2838 if (!proto.has_options()) {
kenton@google.com24bf56f2008-09-24 20:31:01 +00002839 result->options_ = NULL; // Will set to default_instance later.
temporal40ee5512008-07-10 02:12:20 +00002840 } else {
kenton@google.com24bf56f2008-09-24 20:31:01 +00002841 AllocateOptions(proto.options(), result);
2842 }
2843
2844 // Note that the following steps must occur in exactly the specified order.
2845
temporal40ee5512008-07-10 02:12:20 +00002846 // Cross-link.
2847 CrossLinkFile(result, proto);
2848
kenton@google.com24bf56f2008-09-24 20:31:01 +00002849 // Interpret any remaining uninterpreted options gathered into
2850 // options_to_interpret_ during descriptor building. Cross-linking has made
2851 // extension options known, so all interpretations should now succeed.
2852 if (!had_errors_) {
2853 OptionInterpreter option_interpreter(this);
2854 for (vector<OptionsToInterpret>::iterator iter =
2855 options_to_interpret_.begin();
2856 iter != options_to_interpret_.end(); ++iter) {
2857 option_interpreter.InterpretOptions(&(*iter));
2858 }
2859 options_to_interpret_.clear();
2860 }
2861
2862 // Validate options.
2863 if (!had_errors_) {
2864 ValidateFileOptions(result, proto);
2865 }
2866
temporal40ee5512008-07-10 02:12:20 +00002867 if (had_errors_) {
2868 tables_->Rollback();
2869 return NULL;
2870 } else {
2871 tables_->Checkpoint();
2872 return result;
2873 }
2874}
2875
2876void DescriptorBuilder::BuildMessage(const DescriptorProto& proto,
2877 const Descriptor* parent,
2878 Descriptor* result) {
2879 const string& scope = (parent == NULL) ?
2880 file_->package() : parent->full_name();
2881 string* full_name = tables_->AllocateString(scope);
2882 if (!full_name->empty()) full_name->append(1, '.');
2883 full_name->append(proto.name());
2884
2885 ValidateSymbolName(proto.name(), *full_name, proto);
2886
2887 result->name_ = tables_->AllocateString(proto.name());
2888 result->full_name_ = full_name;
2889 result->file_ = file_;
2890 result->containing_type_ = parent;
kenton@google.comd37d46d2009-04-25 02:53:47 +00002891 result->is_placeholder_ = false;
2892 result->is_unqualified_placeholder_ = false;
temporal40ee5512008-07-10 02:12:20 +00002893
2894 BUILD_ARRAY(proto, result, field , BuildField , result);
2895 BUILD_ARRAY(proto, result, nested_type , BuildMessage , result);
2896 BUILD_ARRAY(proto, result, enum_type , BuildEnum , result);
2897 BUILD_ARRAY(proto, result, extension_range, BuildExtensionRange, result);
2898 BUILD_ARRAY(proto, result, extension , BuildExtension , result);
2899
2900 // Copy options.
2901 if (!proto.has_options()) {
kenton@google.com24bf56f2008-09-24 20:31:01 +00002902 result->options_ = NULL; // Will set to default_instance later.
temporal40ee5512008-07-10 02:12:20 +00002903 } else {
kenton@google.com24bf56f2008-09-24 20:31:01 +00002904 AllocateOptions(proto.options(), result);
temporal40ee5512008-07-10 02:12:20 +00002905 }
2906
2907 AddSymbol(result->full_name(), parent, result->name(),
2908 proto, Symbol(result));
2909
2910 // Check that no fields have numbers in extension ranges.
2911 for (int i = 0; i < result->field_count(); i++) {
2912 const FieldDescriptor* field = result->field(i);
2913 for (int j = 0; j < result->extension_range_count(); j++) {
2914 const Descriptor::ExtensionRange* range = result->extension_range(j);
2915 if (range->start <= field->number() && field->number() < range->end) {
2916 AddError(field->full_name(), proto.extension_range(j),
2917 DescriptorPool::ErrorCollector::NUMBER,
2918 strings::Substitute(
2919 "Extension range $0 to $1 includes field \"$2\" ($3).",
2920 range->start, range->end - 1,
2921 field->name(), field->number()));
2922 }
2923 }
2924 }
2925
2926 // Check that extension ranges don't overlap.
2927 for (int i = 0; i < result->extension_range_count(); i++) {
2928 const Descriptor::ExtensionRange* range1 = result->extension_range(i);
2929 for (int j = i + 1; j < result->extension_range_count(); j++) {
2930 const Descriptor::ExtensionRange* range2 = result->extension_range(j);
2931 if (range1->end > range2->start && range2->end > range1->start) {
2932 AddError(result->full_name(), proto.extension_range(j),
2933 DescriptorPool::ErrorCollector::NUMBER,
2934 strings::Substitute("Extension range $0 to $1 overlaps with "
2935 "already-defined range $2 to $3.",
2936 range2->start, range2->end - 1,
2937 range1->start, range1->end - 1));
2938 }
2939 }
2940 }
2941}
2942
2943void DescriptorBuilder::BuildFieldOrExtension(const FieldDescriptorProto& proto,
2944 const Descriptor* parent,
2945 FieldDescriptor* result,
2946 bool is_extension) {
2947 const string& scope = (parent == NULL) ?
2948 file_->package() : parent->full_name();
2949 string* full_name = tables_->AllocateString(scope);
2950 if (!full_name->empty()) full_name->append(1, '.');
2951 full_name->append(proto.name());
2952
2953 ValidateSymbolName(proto.name(), *full_name, proto);
2954
2955 result->name_ = tables_->AllocateString(proto.name());
2956 result->full_name_ = full_name;
2957 result->file_ = file_;
2958 result->number_ = proto.number();
temporal40ee5512008-07-10 02:12:20 +00002959 result->is_extension_ = is_extension;
2960
kenton@google.com2d6daa72009-01-22 01:27:00 +00002961 // If .proto files follow the style guide then the name should already be
2962 // lower-cased. If that's the case we can just reuse the string we already
2963 // allocated rather than allocate a new one.
2964 string lowercase_name(proto.name());
2965 LowerString(&lowercase_name);
2966 if (lowercase_name == proto.name()) {
2967 result->lowercase_name_ = result->name_;
2968 } else {
2969 result->lowercase_name_ = tables_->AllocateString(lowercase_name);
2970 }
2971
2972 // Don't bother with the above optimization for camel-case names since
2973 // .proto files that follow the guide shouldn't be using names in this
2974 // format, so the optimization wouldn't help much.
2975 result->camelcase_name_ = tables_->AllocateString(ToCamelCase(proto.name()));
2976
kenton@google.coma2a32c22008-11-14 17:29:32 +00002977 // Some compilers do not allow static_cast directly between two enum types,
2978 // so we must cast to int first.
2979 result->type_ = static_cast<FieldDescriptor::Type>(
2980 implicit_cast<int>(proto.type()));
2981 result->label_ = static_cast<FieldDescriptor::Label>(
2982 implicit_cast<int>(proto.label()));
2983
temporal40ee5512008-07-10 02:12:20 +00002984 // Some of these may be filled in when cross-linking.
2985 result->containing_type_ = NULL;
2986 result->extension_scope_ = NULL;
2987 result->experimental_map_key_ = NULL;
2988 result->message_type_ = NULL;
2989 result->enum_type_ = NULL;
2990
2991 result->has_default_value_ = proto.has_default_value();
kenton@google.comd37d46d2009-04-25 02:53:47 +00002992 if (proto.has_default_value() && result->is_repeated()) {
2993 AddError(result->full_name(), proto,
2994 DescriptorPool::ErrorCollector::DEFAULT_VALUE,
2995 "Repeated fields can't have default values.");
2996 }
2997
temporal40ee5512008-07-10 02:12:20 +00002998 if (proto.has_type()) {
2999 if (proto.has_default_value()) {
3000 char* end_pos = NULL;
3001 switch (result->cpp_type()) {
3002 case FieldDescriptor::CPPTYPE_INT32:
3003 result->default_value_int32_ =
3004 strtol(proto.default_value().c_str(), &end_pos, 0);
3005 break;
3006 case FieldDescriptor::CPPTYPE_INT64:
3007 result->default_value_int64_ =
3008 strto64(proto.default_value().c_str(), &end_pos, 0);
3009 break;
3010 case FieldDescriptor::CPPTYPE_UINT32:
3011 result->default_value_uint32_ =
3012 strtoul(proto.default_value().c_str(), &end_pos, 0);
3013 break;
3014 case FieldDescriptor::CPPTYPE_UINT64:
3015 result->default_value_uint64_ =
3016 strtou64(proto.default_value().c_str(), &end_pos, 0);
3017 break;
3018 case FieldDescriptor::CPPTYPE_FLOAT:
kenton@google.com684d45b2009-12-19 04:50:00 +00003019 if (proto.default_value() == "inf") {
3020 result->default_value_float_ = numeric_limits<float>::infinity();
3021 } else if (proto.default_value() == "-inf") {
3022 result->default_value_float_ = -numeric_limits<float>::infinity();
3023 } else if (proto.default_value() == "nan") {
3024 result->default_value_float_ = numeric_limits<float>::quiet_NaN();
3025 } else {
3026 result->default_value_float_ =
3027 NoLocaleStrtod(proto.default_value().c_str(), &end_pos);
3028 }
temporal40ee5512008-07-10 02:12:20 +00003029 break;
3030 case FieldDescriptor::CPPTYPE_DOUBLE:
kenton@google.com684d45b2009-12-19 04:50:00 +00003031 if (proto.default_value() == "inf") {
3032 result->default_value_double_ = numeric_limits<double>::infinity();
3033 } else if (proto.default_value() == "-inf") {
3034 result->default_value_double_ = -numeric_limits<double>::infinity();
3035 } else if (proto.default_value() == "nan") {
3036 result->default_value_double_ = numeric_limits<double>::quiet_NaN();
3037 } else {
3038 result->default_value_double_ =
3039 NoLocaleStrtod(proto.default_value().c_str(), &end_pos);
3040 }
temporal40ee5512008-07-10 02:12:20 +00003041 break;
3042 case FieldDescriptor::CPPTYPE_BOOL:
3043 if (proto.default_value() == "true") {
3044 result->default_value_bool_ = true;
3045 } else if (proto.default_value() == "false") {
3046 result->default_value_bool_ = false;
3047 } else {
3048 AddError(result->full_name(), proto,
3049 DescriptorPool::ErrorCollector::DEFAULT_VALUE,
3050 "Boolean default must be true or false.");
3051 }
3052 break;
3053 case FieldDescriptor::CPPTYPE_ENUM:
3054 // This will be filled in when cross-linking.
3055 result->default_value_enum_ = NULL;
3056 break;
3057 case FieldDescriptor::CPPTYPE_STRING:
3058 if (result->type() == FieldDescriptor::TYPE_BYTES) {
3059 result->default_value_string_ = tables_->AllocateString(
3060 UnescapeCEscapeString(proto.default_value()));
3061 } else {
3062 result->default_value_string_ =
3063 tables_->AllocateString(proto.default_value());
3064 }
3065 break;
3066 case FieldDescriptor::CPPTYPE_MESSAGE:
3067 AddError(result->full_name(), proto,
3068 DescriptorPool::ErrorCollector::DEFAULT_VALUE,
3069 "Messages can't have default values.");
3070 result->has_default_value_ = false;
3071 break;
3072 }
3073
3074 if (end_pos != NULL) {
3075 // end_pos is only set non-NULL by the parsers for numeric types, above.
3076 // This checks that the default was non-empty and had no extra junk
3077 // after the end of the number.
3078 if (proto.default_value().empty() || *end_pos != '\0') {
3079 AddError(result->full_name(), proto,
3080 DescriptorPool::ErrorCollector::DEFAULT_VALUE,
3081 "Couldn't parse default value.");
3082 }
3083 }
3084 } else {
3085 // No explicit default value
3086 switch (result->cpp_type()) {
3087 case FieldDescriptor::CPPTYPE_INT32:
3088 result->default_value_int32_ = 0;
3089 break;
3090 case FieldDescriptor::CPPTYPE_INT64:
3091 result->default_value_int64_ = 0;
3092 break;
3093 case FieldDescriptor::CPPTYPE_UINT32:
3094 result->default_value_uint32_ = 0;
3095 break;
3096 case FieldDescriptor::CPPTYPE_UINT64:
3097 result->default_value_uint64_ = 0;
3098 break;
3099 case FieldDescriptor::CPPTYPE_FLOAT:
3100 result->default_value_float_ = 0.0f;
3101 break;
3102 case FieldDescriptor::CPPTYPE_DOUBLE:
3103 result->default_value_double_ = 0.0;
3104 break;
3105 case FieldDescriptor::CPPTYPE_BOOL:
3106 result->default_value_bool_ = false;
3107 break;
3108 case FieldDescriptor::CPPTYPE_ENUM:
3109 // This will be filled in when cross-linking.
3110 result->default_value_enum_ = NULL;
3111 break;
3112 case FieldDescriptor::CPPTYPE_STRING:
3113 result->default_value_string_ = &kEmptyString;
3114 break;
3115 case FieldDescriptor::CPPTYPE_MESSAGE:
3116 break;
3117 }
3118 }
3119 }
3120
3121 if (result->number() <= 0) {
3122 AddError(result->full_name(), proto, DescriptorPool::ErrorCollector::NUMBER,
3123 "Field numbers must be positive integers.");
3124 } else if (result->number() > FieldDescriptor::kMaxNumber) {
3125 AddError(result->full_name(), proto, DescriptorPool::ErrorCollector::NUMBER,
3126 strings::Substitute("Field numbers cannot be greater than $0.",
3127 FieldDescriptor::kMaxNumber));
3128 } else if (result->number() >= FieldDescriptor::kFirstReservedNumber &&
3129 result->number() <= FieldDescriptor::kLastReservedNumber) {
3130 AddError(result->full_name(), proto, DescriptorPool::ErrorCollector::NUMBER,
3131 strings::Substitute(
3132 "Field numbers $0 through $1 are reserved for the protocol "
3133 "buffer library implementation.",
3134 FieldDescriptor::kFirstReservedNumber,
3135 FieldDescriptor::kLastReservedNumber));
3136 }
3137
3138 if (is_extension) {
3139 if (!proto.has_extendee()) {
3140 AddError(result->full_name(), proto,
3141 DescriptorPool::ErrorCollector::EXTENDEE,
3142 "FieldDescriptorProto.extendee not set for extension field.");
3143 }
3144
3145 result->extension_scope_ = parent;
3146 } else {
3147 if (proto.has_extendee()) {
3148 AddError(result->full_name(), proto,
3149 DescriptorPool::ErrorCollector::EXTENDEE,
3150 "FieldDescriptorProto.extendee set for non-extension field.");
3151 }
3152
3153 result->containing_type_ = parent;
3154 }
3155
3156 // Copy options.
3157 if (!proto.has_options()) {
kenton@google.com24bf56f2008-09-24 20:31:01 +00003158 result->options_ = NULL; // Will set to default_instance later.
temporal40ee5512008-07-10 02:12:20 +00003159 } else {
kenton@google.com24bf56f2008-09-24 20:31:01 +00003160 AllocateOptions(proto.options(), result);
temporal40ee5512008-07-10 02:12:20 +00003161 }
3162
3163 AddSymbol(result->full_name(), parent, result->name(),
3164 proto, Symbol(result));
3165}
3166
3167void DescriptorBuilder::BuildExtensionRange(
3168 const DescriptorProto::ExtensionRange& proto,
3169 const Descriptor* parent,
3170 Descriptor::ExtensionRange* result) {
3171 result->start = proto.start();
3172 result->end = proto.end();
3173 if (result->start <= 0) {
3174 AddError(parent->full_name(), proto,
3175 DescriptorPool::ErrorCollector::NUMBER,
3176 "Extension numbers must be positive integers.");
3177 }
3178
3179 if (result->end > FieldDescriptor::kMaxNumber + 1) {
3180 AddError(parent->full_name(), proto,
3181 DescriptorPool::ErrorCollector::NUMBER,
3182 strings::Substitute("Extension numbers cannot be greater than $0.",
3183 FieldDescriptor::kMaxNumber));
3184 }
3185
3186 if (result->start >= result->end) {
3187 AddError(parent->full_name(), proto,
3188 DescriptorPool::ErrorCollector::NUMBER,
3189 "Extension range end number must be greater than start number.");
3190 }
3191}
3192
3193void DescriptorBuilder::BuildEnum(const EnumDescriptorProto& proto,
3194 const Descriptor* parent,
3195 EnumDescriptor* result) {
3196 const string& scope = (parent == NULL) ?
3197 file_->package() : parent->full_name();
3198 string* full_name = tables_->AllocateString(scope);
3199 if (!full_name->empty()) full_name->append(1, '.');
3200 full_name->append(proto.name());
3201
3202 ValidateSymbolName(proto.name(), *full_name, proto);
3203
3204 result->name_ = tables_->AllocateString(proto.name());
3205 result->full_name_ = full_name;
3206 result->file_ = file_;
3207 result->containing_type_ = parent;
kenton@google.comd37d46d2009-04-25 02:53:47 +00003208 result->is_placeholder_ = false;
3209 result->is_unqualified_placeholder_ = false;
temporal40ee5512008-07-10 02:12:20 +00003210
3211 if (proto.value_size() == 0) {
3212 // We cannot allow enums with no values because this would mean there
3213 // would be no valid default value for fields of this type.
3214 AddError(result->full_name(), proto,
3215 DescriptorPool::ErrorCollector::NAME,
3216 "Enums must contain at least one value.");
3217 }
3218
3219 BUILD_ARRAY(proto, result, value, BuildEnumValue, result);
3220
3221 // Copy options.
3222 if (!proto.has_options()) {
kenton@google.com24bf56f2008-09-24 20:31:01 +00003223 result->options_ = NULL; // Will set to default_instance later.
temporal40ee5512008-07-10 02:12:20 +00003224 } else {
kenton@google.com24bf56f2008-09-24 20:31:01 +00003225 AllocateOptions(proto.options(), result);
temporal40ee5512008-07-10 02:12:20 +00003226 }
3227
3228 AddSymbol(result->full_name(), parent, result->name(),
3229 proto, Symbol(result));
3230}
3231
3232void DescriptorBuilder::BuildEnumValue(const EnumValueDescriptorProto& proto,
3233 const EnumDescriptor* parent,
3234 EnumValueDescriptor* result) {
3235 result->name_ = tables_->AllocateString(proto.name());
3236 result->number_ = proto.number();
3237 result->type_ = parent;
3238
3239 // Note: full_name for enum values is a sibling to the parent's name, not a
3240 // child of it.
3241 string* full_name = tables_->AllocateString(*parent->full_name_);
3242 full_name->resize(full_name->size() - parent->name_->size());
3243 full_name->append(*result->name_);
3244 result->full_name_ = full_name;
3245
3246 ValidateSymbolName(proto.name(), *full_name, proto);
3247
3248 // Copy options.
3249 if (!proto.has_options()) {
kenton@google.com24bf56f2008-09-24 20:31:01 +00003250 result->options_ = NULL; // Will set to default_instance later.
temporal40ee5512008-07-10 02:12:20 +00003251 } else {
kenton@google.com24bf56f2008-09-24 20:31:01 +00003252 AllocateOptions(proto.options(), result);
temporal40ee5512008-07-10 02:12:20 +00003253 }
3254
3255 // Again, enum values are weird because we makes them appear as siblings
3256 // of the enum type instead of children of it. So, we use
3257 // parent->containing_type() as the value's parent.
temporalf2063512008-07-23 01:19:07 +00003258 bool added_to_outer_scope =
3259 AddSymbol(result->full_name(), parent->containing_type(), result->name(),
3260 proto, Symbol(result));
temporal40ee5512008-07-10 02:12:20 +00003261
3262 // However, we also want to be able to search for values within a single
3263 // enum type, so we add it as a child of the enum type itself, too.
3264 // Note: This could fail, but if it does, the error has already been
3265 // reported by the above AddSymbol() call, so we ignore the return code.
temporalf2063512008-07-23 01:19:07 +00003266 bool added_to_inner_scope =
kenton@google.comd37d46d2009-04-25 02:53:47 +00003267 file_tables_->AddAliasUnderParent(parent, result->name(), Symbol(result));
temporalf2063512008-07-23 01:19:07 +00003268
3269 if (added_to_inner_scope && !added_to_outer_scope) {
3270 // This value did not conflict with any values defined in the same enum,
3271 // but it did conflict with some other symbol defined in the enum type's
3272 // scope. Let's print an additional error to explain this.
3273 string outer_scope;
3274 if (parent->containing_type() == NULL) {
3275 outer_scope = file_->package();
3276 } else {
3277 outer_scope = parent->containing_type()->full_name();
3278 }
3279
3280 if (outer_scope.empty()) {
3281 outer_scope = "the global scope";
3282 } else {
3283 outer_scope = "\"" + outer_scope + "\"";
3284 }
3285
3286 AddError(result->full_name(), proto,
3287 DescriptorPool::ErrorCollector::NAME,
3288 "Note that enum values use C++ scoping rules, meaning that "
3289 "enum values are siblings of their type, not children of it. "
3290 "Therefore, \"" + result->name() + "\" must be unique within "
3291 + outer_scope + ", not just within \"" + parent->name() + "\".");
3292 }
temporal40ee5512008-07-10 02:12:20 +00003293
3294 // An enum is allowed to define two numbers that refer to the same value.
3295 // FindValueByNumber() should return the first such value, so we simply
3296 // ignore AddEnumValueByNumber()'s return code.
kenton@google.comd37d46d2009-04-25 02:53:47 +00003297 file_tables_->AddEnumValueByNumber(result);
temporal40ee5512008-07-10 02:12:20 +00003298}
3299
3300void DescriptorBuilder::BuildService(const ServiceDescriptorProto& proto,
3301 const void* dummy,
3302 ServiceDescriptor* result) {
3303 string* full_name = tables_->AllocateString(file_->package());
3304 if (!full_name->empty()) full_name->append(1, '.');
3305 full_name->append(proto.name());
3306
3307 ValidateSymbolName(proto.name(), *full_name, proto);
3308
3309 result->name_ = tables_->AllocateString(proto.name());
3310 result->full_name_ = full_name;
3311 result->file_ = file_;
3312
3313 BUILD_ARRAY(proto, result, method, BuildMethod, result);
3314
3315 // Copy options.
3316 if (!proto.has_options()) {
kenton@google.com24bf56f2008-09-24 20:31:01 +00003317 result->options_ = NULL; // Will set to default_instance later.
temporal40ee5512008-07-10 02:12:20 +00003318 } else {
kenton@google.com24bf56f2008-09-24 20:31:01 +00003319 AllocateOptions(proto.options(), result);
temporal40ee5512008-07-10 02:12:20 +00003320 }
3321
3322 AddSymbol(result->full_name(), NULL, result->name(),
3323 proto, Symbol(result));
3324}
3325
3326void DescriptorBuilder::BuildMethod(const MethodDescriptorProto& proto,
3327 const ServiceDescriptor* parent,
3328 MethodDescriptor* result) {
3329 result->name_ = tables_->AllocateString(proto.name());
3330 result->service_ = parent;
3331
3332 string* full_name = tables_->AllocateString(parent->full_name());
3333 full_name->append(1, '.');
3334 full_name->append(*result->name_);
3335 result->full_name_ = full_name;
3336
3337 ValidateSymbolName(proto.name(), *full_name, proto);
3338
3339 // These will be filled in when cross-linking.
3340 result->input_type_ = NULL;
3341 result->output_type_ = NULL;
3342
3343 // Copy options.
3344 if (!proto.has_options()) {
kenton@google.com24bf56f2008-09-24 20:31:01 +00003345 result->options_ = NULL; // Will set to default_instance later.
temporal40ee5512008-07-10 02:12:20 +00003346 } else {
kenton@google.com24bf56f2008-09-24 20:31:01 +00003347 AllocateOptions(proto.options(), result);
temporal40ee5512008-07-10 02:12:20 +00003348 }
3349
3350 AddSymbol(result->full_name(), parent, result->name(),
3351 proto, Symbol(result));
3352}
3353
3354#undef BUILD_ARRAY
3355
3356// -------------------------------------------------------------------
3357
3358void DescriptorBuilder::CrossLinkFile(
3359 FileDescriptor* file, const FileDescriptorProto& proto) {
kenton@google.com24bf56f2008-09-24 20:31:01 +00003360 if (file->options_ == NULL) {
3361 file->options_ = &FileOptions::default_instance();
3362 }
3363
temporal40ee5512008-07-10 02:12:20 +00003364 for (int i = 0; i < file->message_type_count(); i++) {
3365 CrossLinkMessage(&file->message_types_[i], proto.message_type(i));
3366 }
3367
3368 for (int i = 0; i < file->extension_count(); i++) {
3369 CrossLinkField(&file->extensions_[i], proto.extension(i));
3370 }
3371
kenton@google.com24bf56f2008-09-24 20:31:01 +00003372 for (int i = 0; i < file->enum_type_count(); i++) {
3373 CrossLinkEnum(&file->enum_types_[i], proto.enum_type(i));
3374 }
3375
temporal40ee5512008-07-10 02:12:20 +00003376 for (int i = 0; i < file->service_count(); i++) {
3377 CrossLinkService(&file->services_[i], proto.service(i));
3378 }
3379}
3380
3381void DescriptorBuilder::CrossLinkMessage(
3382 Descriptor* message, const DescriptorProto& proto) {
kenton@google.com24bf56f2008-09-24 20:31:01 +00003383 if (message->options_ == NULL) {
3384 message->options_ = &MessageOptions::default_instance();
3385 }
3386
temporal40ee5512008-07-10 02:12:20 +00003387 for (int i = 0; i < message->nested_type_count(); i++) {
3388 CrossLinkMessage(&message->nested_types_[i], proto.nested_type(i));
3389 }
3390
kenton@google.com24bf56f2008-09-24 20:31:01 +00003391 for (int i = 0; i < message->enum_type_count(); i++) {
3392 CrossLinkEnum(&message->enum_types_[i], proto.enum_type(i));
3393 }
3394
temporal40ee5512008-07-10 02:12:20 +00003395 for (int i = 0; i < message->field_count(); i++) {
3396 CrossLinkField(&message->fields_[i], proto.field(i));
3397 }
3398
3399 for (int i = 0; i < message->extension_count(); i++) {
3400 CrossLinkField(&message->extensions_[i], proto.extension(i));
3401 }
3402}
3403
3404void DescriptorBuilder::CrossLinkField(
3405 FieldDescriptor* field, const FieldDescriptorProto& proto) {
kenton@google.com24bf56f2008-09-24 20:31:01 +00003406 if (field->options_ == NULL) {
3407 field->options_ = &FieldOptions::default_instance();
3408 }
3409
temporal40ee5512008-07-10 02:12:20 +00003410 if (proto.has_extendee()) {
kenton@google.comd37d46d2009-04-25 02:53:47 +00003411 Symbol extendee = LookupSymbol(proto.extendee(), field->full_name(),
3412 PLACEHOLDER_EXTENDABLE_MESSAGE);
temporal40ee5512008-07-10 02:12:20 +00003413 if (extendee.IsNull()) {
3414 AddNotDefinedError(field->full_name(), proto,
3415 DescriptorPool::ErrorCollector::EXTENDEE,
3416 proto.extendee());
3417 return;
3418 } else if (extendee.type != Symbol::MESSAGE) {
3419 AddError(field->full_name(), proto,
3420 DescriptorPool::ErrorCollector::EXTENDEE,
3421 "\"" + proto.extendee() + "\" is not a message type.");
3422 return;
3423 }
3424 field->containing_type_ = extendee.descriptor;
3425
3426 if (!field->containing_type()->IsExtensionNumber(field->number())) {
3427 AddError(field->full_name(), proto,
3428 DescriptorPool::ErrorCollector::NUMBER,
3429 strings::Substitute("\"$0\" does not declare $1 as an "
3430 "extension number.",
3431 field->containing_type()->full_name(),
3432 field->number()));
3433 }
3434 }
3435
3436 if (proto.has_type_name()) {
kenton@google.comd37d46d2009-04-25 02:53:47 +00003437 // Assume we are expecting a message type unless the proto contains some
3438 // evidence that it expects an enum type. This only makes a difference if
3439 // we end up creating a placeholder.
3440 bool expecting_enum = (proto.type() == FieldDescriptorProto::TYPE_ENUM) ||
3441 proto.has_default_value();
3442
3443 Symbol type =
3444 LookupSymbol(proto.type_name(), field->full_name(),
3445 expecting_enum ? PLACEHOLDER_ENUM : PLACEHOLDER_MESSAGE,
3446 LOOKUP_TYPES);
3447
temporal40ee5512008-07-10 02:12:20 +00003448 if (type.IsNull()) {
3449 AddNotDefinedError(field->full_name(), proto,
3450 DescriptorPool::ErrorCollector::TYPE,
3451 proto.type_name());
3452 return;
3453 }
3454
3455 if (!proto.has_type()) {
3456 // Choose field type based on symbol.
3457 if (type.type == Symbol::MESSAGE) {
3458 field->type_ = FieldDescriptor::TYPE_MESSAGE;
3459 } else if (type.type == Symbol::ENUM) {
3460 field->type_ = FieldDescriptor::TYPE_ENUM;
3461 } else {
3462 AddError(field->full_name(), proto,
3463 DescriptorPool::ErrorCollector::TYPE,
3464 "\"" + proto.type_name() + "\" is not a type.");
3465 return;
3466 }
3467 }
3468
3469 if (field->cpp_type() == FieldDescriptor::CPPTYPE_MESSAGE) {
3470 if (type.type != Symbol::MESSAGE) {
3471 AddError(field->full_name(), proto,
3472 DescriptorPool::ErrorCollector::TYPE,
3473 "\"" + proto.type_name() + "\" is not a message type.");
3474 return;
3475 }
3476 field->message_type_ = type.descriptor;
3477
3478 if (field->has_default_value()) {
3479 AddError(field->full_name(), proto,
3480 DescriptorPool::ErrorCollector::DEFAULT_VALUE,
3481 "Messages can't have default values.");
3482 }
3483 } else if (field->cpp_type() == FieldDescriptor::CPPTYPE_ENUM) {
3484 if (type.type != Symbol::ENUM) {
3485 AddError(field->full_name(), proto,
3486 DescriptorPool::ErrorCollector::TYPE,
3487 "\"" + proto.type_name() + "\" is not an enum type.");
3488 return;
3489 }
3490 field->enum_type_ = type.enum_descriptor;
3491
kenton@google.comd37d46d2009-04-25 02:53:47 +00003492 if (field->enum_type()->is_placeholder_) {
3493 // We can't look up default values for placeholder types. We'll have
3494 // to just drop them.
3495 field->has_default_value_ = false;
3496 }
3497
temporal40ee5512008-07-10 02:12:20 +00003498 if (field->has_default_value()) {
3499 // We can't just use field->enum_type()->FindValueByName() here
3500 // because that locks the pool's mutex, which we have already locked
3501 // at this point.
3502 Symbol default_value =
kenton@google.comd37d46d2009-04-25 02:53:47 +00003503 LookupSymbolNoPlaceholder(proto.default_value(),
3504 field->enum_type()->full_name());
temporal40ee5512008-07-10 02:12:20 +00003505
3506 if (default_value.type == Symbol::ENUM_VALUE &&
3507 default_value.enum_value_descriptor->type() == field->enum_type()) {
3508 field->default_value_enum_ = default_value.enum_value_descriptor;
3509 } else {
3510 AddError(field->full_name(), proto,
3511 DescriptorPool::ErrorCollector::DEFAULT_VALUE,
3512 "Enum type \"" + field->enum_type()->full_name() +
3513 "\" has no value named \"" + proto.default_value() + "\".");
3514 }
3515 } else if (field->enum_type()->value_count() > 0) {
3516 // All enums must have at least one value, or we would have reported
3517 // an error elsewhere. We use the first defined value as the default
3518 // if a default is not explicitly defined.
3519 field->default_value_enum_ = field->enum_type()->value(0);
3520 }
3521 } else {
3522 AddError(field->full_name(), proto, DescriptorPool::ErrorCollector::TYPE,
3523 "Field with primitive type has type_name.");
3524 }
3525 } else {
3526 if (field->cpp_type() == FieldDescriptor::CPPTYPE_MESSAGE ||
3527 field->cpp_type() == FieldDescriptor::CPPTYPE_ENUM) {
3528 AddError(field->full_name(), proto, DescriptorPool::ErrorCollector::TYPE,
3529 "Field with message or enum type missing type_name.");
3530 }
3531 }
3532
temporal40ee5512008-07-10 02:12:20 +00003533 // Add the field to the fields-by-number table.
3534 // Note: We have to do this *after* cross-linking because extensions do not
3535 // know their containing type until now.
kenton@google.comd37d46d2009-04-25 02:53:47 +00003536 if (!file_tables_->AddFieldByNumber(field)) {
temporal40ee5512008-07-10 02:12:20 +00003537 const FieldDescriptor* conflicting_field =
kenton@google.comd37d46d2009-04-25 02:53:47 +00003538 file_tables_->FindFieldByNumber(field->containing_type(),
3539 field->number());
temporal40ee5512008-07-10 02:12:20 +00003540 if (field->is_extension()) {
3541 AddError(field->full_name(), proto,
3542 DescriptorPool::ErrorCollector::NUMBER,
3543 strings::Substitute("Extension number $0 has already been used "
3544 "in \"$1\" by extension \"$2\".",
3545 field->number(),
3546 field->containing_type()->full_name(),
3547 conflicting_field->full_name()));
3548 } else {
3549 AddError(field->full_name(), proto,
3550 DescriptorPool::ErrorCollector::NUMBER,
3551 strings::Substitute("Field number $0 has already been used in "
3552 "\"$1\" by field \"$2\".",
3553 field->number(),
3554 field->containing_type()->full_name(),
3555 conflicting_field->name()));
3556 }
3557 }
kenton@google.com2d6daa72009-01-22 01:27:00 +00003558
kenton@google.comd37d46d2009-04-25 02:53:47 +00003559 if (field->is_extension()) {
3560 // No need for error checking: if the extension number collided,
3561 // we've already been informed of it by the if() above.
3562 tables_->AddExtension(field);
3563 }
3564
kenton@google.com2d6daa72009-01-22 01:27:00 +00003565 // Add the field to the lowercase-name and camelcase-name tables.
kenton@google.comd37d46d2009-04-25 02:53:47 +00003566 file_tables_->AddFieldByStylizedNames(field);
kenton@google.com24bf56f2008-09-24 20:31:01 +00003567}
temporal40ee5512008-07-10 02:12:20 +00003568
kenton@google.com24bf56f2008-09-24 20:31:01 +00003569void DescriptorBuilder::CrossLinkEnum(
3570 EnumDescriptor* enum_type, const EnumDescriptorProto& proto) {
3571 if (enum_type->options_ == NULL) {
3572 enum_type->options_ = &EnumOptions::default_instance();
3573 }
3574
3575 for (int i = 0; i < enum_type->value_count(); i++) {
3576 CrossLinkEnumValue(&enum_type->values_[i], proto.value(i));
3577 }
3578}
3579
3580void DescriptorBuilder::CrossLinkEnumValue(
3581 EnumValueDescriptor* enum_value, const EnumValueDescriptorProto& proto) {
3582 if (enum_value->options_ == NULL) {
3583 enum_value->options_ = &EnumValueOptions::default_instance();
temporal40ee5512008-07-10 02:12:20 +00003584 }
3585}
3586
3587void DescriptorBuilder::CrossLinkService(
3588 ServiceDescriptor* service, const ServiceDescriptorProto& proto) {
kenton@google.com24bf56f2008-09-24 20:31:01 +00003589 if (service->options_ == NULL) {
3590 service->options_ = &ServiceOptions::default_instance();
3591 }
3592
temporal40ee5512008-07-10 02:12:20 +00003593 for (int i = 0; i < service->method_count(); i++) {
3594 CrossLinkMethod(&service->methods_[i], proto.method(i));
3595 }
3596}
3597
3598void DescriptorBuilder::CrossLinkMethod(
3599 MethodDescriptor* method, const MethodDescriptorProto& proto) {
kenton@google.com24bf56f2008-09-24 20:31:01 +00003600 if (method->options_ == NULL) {
3601 method->options_ = &MethodOptions::default_instance();
3602 }
3603
temporal40ee5512008-07-10 02:12:20 +00003604 Symbol input_type = LookupSymbol(proto.input_type(), method->full_name());
3605 if (input_type.IsNull()) {
3606 AddNotDefinedError(method->full_name(), proto,
3607 DescriptorPool::ErrorCollector::INPUT_TYPE,
3608 proto.input_type());
3609 } else if (input_type.type != Symbol::MESSAGE) {
3610 AddError(method->full_name(), proto,
3611 DescriptorPool::ErrorCollector::INPUT_TYPE,
3612 "\"" + proto.input_type() + "\" is not a message type.");
3613 } else {
3614 method->input_type_ = input_type.descriptor;
3615 }
3616
3617 Symbol output_type = LookupSymbol(proto.output_type(), method->full_name());
3618 if (output_type.IsNull()) {
3619 AddNotDefinedError(method->full_name(), proto,
3620 DescriptorPool::ErrorCollector::OUTPUT_TYPE,
3621 proto.output_type());
3622 } else if (output_type.type != Symbol::MESSAGE) {
3623 AddError(method->full_name(), proto,
3624 DescriptorPool::ErrorCollector::OUTPUT_TYPE,
3625 "\"" + proto.output_type() + "\" is not a message type.");
3626 } else {
3627 method->output_type_ = output_type.descriptor;
3628 }
3629}
3630
kenton@google.com24bf56f2008-09-24 20:31:01 +00003631// -------------------------------------------------------------------
3632
3633#define VALIDATE_OPTIONS_FROM_ARRAY(descriptor, array_name, type) \
3634 for (int i = 0; i < descriptor->array_name##_count(); ++i) { \
3635 Validate##type##Options(descriptor->array_name##s_ + i, \
3636 proto.array_name(i)); \
3637 }
3638
kenton@google.com80b1d622009-07-29 01:13:20 +00003639// Determine if the file uses optimize_for = LITE_RUNTIME, being careful to
3640// avoid problems that exist at init time.
3641static bool IsLite(const FileDescriptor* file) {
3642 // TODO(kenton): I don't even remember how many of these conditions are
3643 // actually possible. I'm just being super-safe.
3644 return file != NULL &&
3645 &file->options() != NULL &&
3646 &file->options() != &FileOptions::default_instance() &&
3647 file->options().optimize_for() == FileOptions::LITE_RUNTIME;
3648}
3649
kenton@google.com24bf56f2008-09-24 20:31:01 +00003650void DescriptorBuilder::ValidateFileOptions(FileDescriptor* file,
3651 const FileDescriptorProto& proto) {
3652 VALIDATE_OPTIONS_FROM_ARRAY(file, message_type, Message);
3653 VALIDATE_OPTIONS_FROM_ARRAY(file, enum_type, Enum);
3654 VALIDATE_OPTIONS_FROM_ARRAY(file, service, Service);
3655 VALIDATE_OPTIONS_FROM_ARRAY(file, extension, Field);
kenton@google.com80b1d622009-07-29 01:13:20 +00003656
3657 // Lite files can only be imported by other Lite files.
3658 if (!IsLite(file)) {
3659 for (int i = 0; i < file->dependency_count(); i++) {
3660 if (IsLite(file->dependency(i))) {
3661 AddError(
3662 file->name(), proto,
3663 DescriptorPool::ErrorCollector::OTHER,
3664 "Files that do not use optimize_for = LITE_RUNTIME cannot import "
3665 "files which do use this option. This file is not lite, but it "
3666 "imports \"" + file->dependency(i)->name() + "\" which is.");
3667 break;
3668 }
3669 }
3670 }
kenton@google.com24bf56f2008-09-24 20:31:01 +00003671}
3672
3673void DescriptorBuilder::ValidateMessageOptions(Descriptor* message,
3674 const DescriptorProto& proto) {
3675 VALIDATE_OPTIONS_FROM_ARRAY(message, field, Field);
3676 VALIDATE_OPTIONS_FROM_ARRAY(message, nested_type, Message);
3677 VALIDATE_OPTIONS_FROM_ARRAY(message, enum_type, Enum);
3678 VALIDATE_OPTIONS_FROM_ARRAY(message, extension, Field);
3679}
3680
3681void DescriptorBuilder::ValidateFieldOptions(FieldDescriptor* field,
temporal40ee5512008-07-10 02:12:20 +00003682 const FieldDescriptorProto& proto) {
kenton@google.com24bf56f2008-09-24 20:31:01 +00003683 if (field->options().has_experimental_map_key()) {
3684 ValidateMapKey(field, proto);
3685 }
3686
kenton@google.com2d6daa72009-01-22 01:27:00 +00003687 // Only repeated primitive fields may be packed.
kenton@google.comfccb1462009-12-18 02:11:36 +00003688 if (field->options().packed() && !field->is_packable()) {
3689 AddError(
3690 field->full_name(), proto,
3691 DescriptorPool::ErrorCollector::TYPE,
3692 "[packed = true] can only be specified for repeated primitive fields.");
kenton@google.com2d6daa72009-01-22 01:27:00 +00003693 }
3694
kenton@google.com24bf56f2008-09-24 20:31:01 +00003695 // Note: Default instance may not yet be initialized here, so we have to
3696 // avoid reading from it.
3697 if (field->containing_type_ != NULL &&
3698 &field->containing_type()->options() !=
3699 &MessageOptions::default_instance() &&
3700 field->containing_type()->options().message_set_wire_format()) {
3701 if (field->is_extension()) {
3702 if (!field->is_optional() ||
3703 field->type() != FieldDescriptor::TYPE_MESSAGE) {
3704 AddError(field->full_name(), proto,
3705 DescriptorPool::ErrorCollector::TYPE,
3706 "Extensions of MessageSets must be optional messages.");
3707 }
3708 } else {
3709 AddError(field->full_name(), proto,
3710 DescriptorPool::ErrorCollector::NAME,
3711 "MessageSets cannot have fields, only extensions.");
3712 }
3713 }
kenton@google.com80b1d622009-07-29 01:13:20 +00003714
3715 // Lite extensions can only be of Lite types.
3716 if (IsLite(field->file()) &&
3717 field->containing_type_ != NULL &&
3718 !IsLite(field->containing_type()->file())) {
3719 AddError(field->full_name(), proto,
3720 DescriptorPool::ErrorCollector::EXTENDEE,
3721 "Extensions to non-lite types can only be declared in non-lite "
3722 "files. Note that you cannot extend a non-lite type to contain "
3723 "a lite type, but the reverse is allowed.");
3724 }
kenton@google.com24bf56f2008-09-24 20:31:01 +00003725}
3726
3727void DescriptorBuilder::ValidateEnumOptions(EnumDescriptor* enm,
3728 const EnumDescriptorProto& proto) {
3729 VALIDATE_OPTIONS_FROM_ARRAY(enm, value, EnumValue);
3730}
3731
3732void DescriptorBuilder::ValidateEnumValueOptions(
3733 EnumValueDescriptor* enum_value, const EnumValueDescriptorProto& proto) {
3734 // Nothing to do so far.
3735}
3736void DescriptorBuilder::ValidateServiceOptions(ServiceDescriptor* service,
3737 const ServiceDescriptorProto& proto) {
kenton@google.comd09ab852010-04-19 19:15:12 +00003738 if (IsLite(service->file()) &&
3739 (service->file()->options().cc_generic_services() ||
3740 service->file()->options().java_generic_services())) {
kenton@google.com80b1d622009-07-29 01:13:20 +00003741 AddError(service->full_name(), proto,
3742 DescriptorPool::ErrorCollector::NAME,
kenton@google.comd09ab852010-04-19 19:15:12 +00003743 "Files with optimize_for = LITE_RUNTIME cannot define services "
3744 "unless you set both options cc_generic_services and "
3745 "java_generic_sevices to false.");
kenton@google.com80b1d622009-07-29 01:13:20 +00003746 }
3747
kenton@google.com24bf56f2008-09-24 20:31:01 +00003748 VALIDATE_OPTIONS_FROM_ARRAY(service, method, Method);
3749}
3750
3751void DescriptorBuilder::ValidateMethodOptions(MethodDescriptor* method,
3752 const MethodDescriptorProto& proto) {
3753 // Nothing to do so far.
3754}
3755
3756void DescriptorBuilder::ValidateMapKey(FieldDescriptor* field,
3757 const FieldDescriptorProto& proto) {
temporal40ee5512008-07-10 02:12:20 +00003758 if (!field->is_repeated()) {
3759 AddError(field->full_name(), proto, DescriptorPool::ErrorCollector::TYPE,
3760 "map type is only allowed for repeated fields.");
3761 return;
3762 }
3763
3764 if (field->cpp_type() != FieldDescriptor::CPPTYPE_MESSAGE) {
3765 AddError(field->full_name(), proto, DescriptorPool::ErrorCollector::TYPE,
3766 "map type is only allowed for fields with a message type.");
3767 return;
3768 }
3769
3770 const Descriptor* item_type = field->message_type();
3771 if (item_type == NULL) {
3772 AddError(field->full_name(), proto, DescriptorPool::ErrorCollector::TYPE,
3773 "Could not find field type.");
3774 return;
3775 }
3776
3777 // Find the field in item_type named by "experimental_map_key"
kenton@google.com24bf56f2008-09-24 20:31:01 +00003778 const string& key_name = field->options().experimental_map_key();
temporal40ee5512008-07-10 02:12:20 +00003779 const Symbol key_symbol = LookupSymbol(
3780 key_name,
3781 // We append ".key_name" to the containing type's name since
3782 // LookupSymbol() searches for peers of the supplied name, not
3783 // children of the supplied name.
3784 item_type->full_name() + "." + key_name);
3785
3786 if (key_symbol.IsNull() || key_symbol.field_descriptor->is_extension()) {
3787 AddError(field->full_name(), proto, DescriptorPool::ErrorCollector::TYPE,
3788 "Could not find field named \"" + key_name + "\" in type \"" +
3789 item_type->full_name() + "\".");
3790 return;
3791 }
3792 const FieldDescriptor* key_field = key_symbol.field_descriptor;
3793
3794 if (key_field->is_repeated()) {
3795 AddError(field->full_name(), proto, DescriptorPool::ErrorCollector::TYPE,
3796 "map_key must not name a repeated field.");
3797 return;
3798 }
3799
3800 if (key_field->cpp_type() == FieldDescriptor::CPPTYPE_MESSAGE) {
3801 AddError(field->full_name(), proto, DescriptorPool::ErrorCollector::TYPE,
3802 "map key must name a scalar or string field.");
3803 return;
3804 }
3805
3806 field->experimental_map_key_ = key_field;
3807}
3808
kenton@google.com24bf56f2008-09-24 20:31:01 +00003809#undef VALIDATE_OPTIONS_FROM_ARRAY
3810
3811// -------------------------------------------------------------------
3812
3813DescriptorBuilder::OptionInterpreter::OptionInterpreter(
3814 DescriptorBuilder* builder) : builder_(builder) {
3815 GOOGLE_CHECK(builder_);
3816}
3817
3818DescriptorBuilder::OptionInterpreter::~OptionInterpreter() {
3819}
3820
3821bool DescriptorBuilder::OptionInterpreter::InterpretOptions(
3822 OptionsToInterpret* options_to_interpret) {
3823 // Note that these may be in different pools, so we can't use the same
3824 // descriptor and reflection objects on both.
3825 Message* options = options_to_interpret->options;
3826 const Message* original_options = options_to_interpret->original_options;
3827
3828 bool failed = false;
3829 options_to_interpret_ = options_to_interpret;
3830
kenton@google.comd37d46d2009-04-25 02:53:47 +00003831 // Find the uninterpreted_option field in the mutable copy of the options
3832 // and clear them, since we're about to interpret them.
3833 const FieldDescriptor* uninterpreted_options_field =
3834 options->GetDescriptor()->FindFieldByName("uninterpreted_option");
3835 GOOGLE_CHECK(uninterpreted_options_field != NULL)
3836 << "No field named \"uninterpreted_option\" in the Options proto.";
3837 options->GetReflection()->ClearField(options, uninterpreted_options_field);
3838
kenton@google.com24bf56f2008-09-24 20:31:01 +00003839 // Find the uninterpreted_option field in the original options.
3840 const FieldDescriptor* original_uninterpreted_options_field =
3841 original_options->GetDescriptor()->
3842 FindFieldByName("uninterpreted_option");
3843 GOOGLE_CHECK(original_uninterpreted_options_field != NULL)
3844 << "No field named \"uninterpreted_option\" in the Options proto.";
3845
3846 const int num_uninterpreted_options = original_options->GetReflection()->
3847 FieldSize(*original_options, original_uninterpreted_options_field);
3848 for (int i = 0; i < num_uninterpreted_options; ++i) {
kenton@google.com80b1d622009-07-29 01:13:20 +00003849 uninterpreted_option_ = down_cast<const UninterpretedOption*>(
kenton@google.com24bf56f2008-09-24 20:31:01 +00003850 &original_options->GetReflection()->GetRepeatedMessage(
3851 *original_options, original_uninterpreted_options_field, i));
3852 if (!InterpretSingleOption(options)) {
3853 // Error already added by InterpretSingleOption().
3854 failed = true;
3855 break;
3856 }
3857 }
3858 // Reset these, so we don't have any dangling pointers.
3859 uninterpreted_option_ = NULL;
3860 options_to_interpret_ = NULL;
3861
3862 if (!failed) {
kenton@google.com24bf56f2008-09-24 20:31:01 +00003863 // InterpretSingleOption() added the interpreted options in the
3864 // UnknownFieldSet, in case the option isn't yet known to us. Now we
3865 // serialize the options message and deserialize it back. That way, any
3866 // option fields that we do happen to know about will get moved from the
3867 // UnknownFieldSet into the real fields, and thus be available right away.
3868 // If they are not known, that's OK too. They will get reparsed into the
3869 // UnknownFieldSet and wait there until the message is parsed by something
3870 // that does know about the options.
3871 string buf;
3872 options->AppendToString(&buf);
3873 GOOGLE_CHECK(options->ParseFromString(buf))
3874 << "Protocol message serialized itself in invalid fashion.";
3875 }
3876
3877 return !failed;
3878}
3879
3880bool DescriptorBuilder::OptionInterpreter::InterpretSingleOption(
3881 Message* options) {
3882 // First do some basic validation.
3883 if (uninterpreted_option_->name_size() == 0) {
3884 // This should never happen unless the parser has gone seriously awry or
3885 // someone has manually created the uninterpreted option badly.
3886 return AddNameError("Option must have a name.");
3887 }
3888 if (uninterpreted_option_->name(0).name_part() == "uninterpreted_option") {
3889 return AddNameError("Option must not use reserved name "
3890 "\"uninterpreted_option\".");
3891 }
3892
3893 const Descriptor* options_descriptor = NULL;
3894 // Get the options message's descriptor from the builder's pool, so that we
3895 // get the version that knows about any extension options declared in the
3896 // file we're currently building. The descriptor should be there as long as
3897 // the file we're building imported "google/protobuf/descriptors.proto".
3898
3899 // Note that we use DescriptorBuilder::FindSymbol(), not
3900 // DescriptorPool::FindMessageTypeByName() because we're already holding the
3901 // pool's mutex, and the latter method locks it again.
kenton@google.com26bd9ee2008-11-21 00:06:27 +00003902 Symbol symbol = builder_->FindSymbolNotEnforcingDeps(
3903 options->GetDescriptor()->full_name());
kenton@google.com24bf56f2008-09-24 20:31:01 +00003904 if (!symbol.IsNull() && symbol.type == Symbol::MESSAGE) {
3905 options_descriptor = symbol.descriptor;
3906 } else {
3907 // The options message's descriptor was not in the builder's pool, so use
3908 // the standard version from the generated pool. We're not holding the
3909 // generated pool's mutex, so we can search it the straightforward way.
3910 options_descriptor = options->GetDescriptor();
3911 }
3912 GOOGLE_CHECK(options_descriptor);
3913
3914 // We iterate over the name parts to drill into the submessages until we find
3915 // the leaf field for the option. As we drill down we remember the current
3916 // submessage's descriptor in |descriptor| and the next field in that
3917 // submessage in |field|. We also track the fields we're drilling down
3918 // through in |intermediate_fields|. As we go, we reconstruct the full option
3919 // name in |debug_msg_name|, for use in error messages.
3920 const Descriptor* descriptor = options_descriptor;
3921 const FieldDescriptor* field = NULL;
3922 vector<const FieldDescriptor*> intermediate_fields;
3923 string debug_msg_name = "";
3924
3925 for (int i = 0; i < uninterpreted_option_->name_size(); ++i) {
3926 const string& name_part = uninterpreted_option_->name(i).name_part();
3927 if (debug_msg_name.size() > 0) {
3928 debug_msg_name += ".";
3929 }
3930 if (uninterpreted_option_->name(i).is_extension()) {
3931 debug_msg_name += "(" + name_part + ")";
3932 // Search for the extension's descriptor as an extension in the builder's
3933 // pool. Note that we use DescriptorBuilder::LookupSymbol(), not
3934 // DescriptorPool::FindExtensionByName(), for two reasons: 1) It allows
3935 // relative lookups, and 2) because we're already holding the pool's
3936 // mutex, and the latter method locks it again.
3937 Symbol symbol = builder_->LookupSymbol(name_part,
3938 options_to_interpret_->name_scope);
3939 if (!symbol.IsNull() && symbol.type == Symbol::FIELD) {
3940 field = symbol.field_descriptor;
3941 }
3942 // If we don't find the field then the field's descriptor was not in the
3943 // builder's pool, but there's no point in looking in the generated
3944 // pool. We require that you import the file that defines any extensions
3945 // you use, so they must be present in the builder's pool.
3946 } else {
3947 debug_msg_name += name_part;
kenton@google.comd37d46d2009-04-25 02:53:47 +00003948 // Search for the field's descriptor as a regular field.
3949 field = descriptor->FindFieldByName(name_part);
kenton@google.com24bf56f2008-09-24 20:31:01 +00003950 }
3951
kenton@google.com26bd9ee2008-11-21 00:06:27 +00003952 if (field == NULL) {
kenton@google.comd37d46d2009-04-25 02:53:47 +00003953 if (get_allow_unknown(builder_->pool_)) {
3954 // We can't find the option, but AllowUnknownDependencies() is enabled,
3955 // so we will just leave it as uninterpreted.
3956 AddWithoutInterpreting(*uninterpreted_option_, options);
3957 return true;
3958 } else {
3959 return AddNameError("Option \"" + debug_msg_name + "\" unknown.");
3960 }
kenton@google.com24bf56f2008-09-24 20:31:01 +00003961 } else if (field->containing_type() != descriptor) {
kenton@google.comd37d46d2009-04-25 02:53:47 +00003962 if (get_is_placeholder(field->containing_type())) {
3963 // The field is an extension of a placeholder type, so we can't
3964 // reliably verify whether it is a valid extension to use here (e.g.
3965 // we don't know if it is an extension of the correct *Options message,
3966 // or if it has a valid field number, etc.). Just leave it as
3967 // uninterpreted instead.
3968 AddWithoutInterpreting(*uninterpreted_option_, options);
3969 return true;
3970 } else {
3971 // This can only happen if, due to some insane misconfiguration of the
3972 // pools, we find the options message in one pool but the field in
3973 // another. This would probably imply a hefty bug somewhere.
3974 return AddNameError("Option field \"" + debug_msg_name +
3975 "\" is not a field or extension of message \"" +
3976 descriptor->name() + "\".");
3977 }
kenton@google.com24bf56f2008-09-24 20:31:01 +00003978 } else if (field->is_repeated()) {
3979 return AddNameError("Option field \"" + debug_msg_name +
3980 "\" is repeated. Repeated options are not "
3981 "supported.");
3982 } else if (i < uninterpreted_option_->name_size() - 1) {
3983 if (field->cpp_type() != FieldDescriptor::CPPTYPE_MESSAGE) {
3984 return AddNameError("Option \"" + debug_msg_name +
3985 "\" is an atomic type, not a message.");
3986 } else {
3987 // Drill down into the submessage.
3988 intermediate_fields.push_back(field);
3989 descriptor = field->message_type();
3990 }
kenton@google.com24bf56f2008-09-24 20:31:01 +00003991 }
3992 }
3993
3994 // We've found the leaf field. Now we use UnknownFieldSets to set its value
3995 // on the options message. We do so because the message may not yet know
3996 // about its extension fields, so we may not be able to set the fields
3997 // directly. But the UnknownFieldSets will serialize to the same wire-format
3998 // message, so reading that message back in once the extension fields are
3999 // known will populate them correctly.
4000
4001 // First see if the option is already set.
4002 if (!ExamineIfOptionIsSet(
4003 intermediate_fields.begin(),
4004 intermediate_fields.end(),
4005 field, debug_msg_name,
4006 options->GetReflection()->GetUnknownFields(*options))) {
4007 return false; // ExamineIfOptionIsSet() already added the error.
4008 }
4009
4010
4011 // First set the value on the UnknownFieldSet corresponding to the
4012 // innermost message.
4013 scoped_ptr<UnknownFieldSet> unknown_fields(new UnknownFieldSet());
kenton@google.comd37d46d2009-04-25 02:53:47 +00004014 if (!SetOptionValue(field, unknown_fields.get())) {
kenton@google.com24bf56f2008-09-24 20:31:01 +00004015 return false; // SetOptionValue() already added the error.
4016 }
4017
4018 // Now wrap the UnknownFieldSet with UnknownFieldSets corresponding to all
4019 // the intermediate messages.
4020 for (vector<const FieldDescriptor*>::reverse_iterator iter =
4021 intermediate_fields.rbegin();
4022 iter != intermediate_fields.rend(); ++iter) {
4023 scoped_ptr<UnknownFieldSet> parent_unknown_fields(new UnknownFieldSet());
4024 switch ((*iter)->type()) {
4025 case FieldDescriptor::TYPE_MESSAGE: {
4026 io::StringOutputStream outstr(
kenton@google.comd37d46d2009-04-25 02:53:47 +00004027 parent_unknown_fields->AddLengthDelimited((*iter)->number()));
kenton@google.com24bf56f2008-09-24 20:31:01 +00004028 io::CodedOutputStream out(&outstr);
kenton@google.comd37d46d2009-04-25 02:53:47 +00004029 internal::WireFormat::SerializeUnknownFields(*unknown_fields, &out);
4030 GOOGLE_CHECK(!out.HadError())
kenton@google.com24bf56f2008-09-24 20:31:01 +00004031 << "Unexpected failure while serializing option submessage "
4032 << debug_msg_name << "\".";
4033 break;
4034 }
4035
4036 case FieldDescriptor::TYPE_GROUP: {
kenton@google.comd37d46d2009-04-25 02:53:47 +00004037 parent_unknown_fields->AddGroup((*iter)->number())
4038 ->MergeFrom(*unknown_fields);
kenton@google.com24bf56f2008-09-24 20:31:01 +00004039 break;
4040 }
4041
4042 default:
4043 GOOGLE_LOG(FATAL) << "Invalid wire type for CPPTYPE_MESSAGE: "
4044 << (*iter)->type();
4045 return false;
4046 }
4047 unknown_fields.reset(parent_unknown_fields.release());
4048 }
4049
4050 // Now merge the UnknownFieldSet corresponding to the top-level message into
4051 // the options message.
4052 options->GetReflection()->MutableUnknownFields(options)->MergeFrom(
4053 *unknown_fields);
4054
4055 return true;
4056}
4057
kenton@google.comd37d46d2009-04-25 02:53:47 +00004058void DescriptorBuilder::OptionInterpreter::AddWithoutInterpreting(
4059 const UninterpretedOption& uninterpreted_option, Message* options) {
4060 const FieldDescriptor* field =
4061 options->GetDescriptor()->FindFieldByName("uninterpreted_option");
4062 GOOGLE_CHECK(field != NULL);
4063
4064 options->GetReflection()->AddMessage(options, field)
4065 ->CopyFrom(uninterpreted_option);
4066}
4067
kenton@google.com24bf56f2008-09-24 20:31:01 +00004068bool DescriptorBuilder::OptionInterpreter::ExamineIfOptionIsSet(
4069 vector<const FieldDescriptor*>::const_iterator intermediate_fields_iter,
4070 vector<const FieldDescriptor*>::const_iterator intermediate_fields_end,
4071 const FieldDescriptor* innermost_field, const string& debug_msg_name,
4072 const UnknownFieldSet& unknown_fields) {
kenton@google.comd37d46d2009-04-25 02:53:47 +00004073 // We do linear searches of the UnknownFieldSet and its sub-groups. This
4074 // should be fine since it's unlikely that any one options structure will
4075 // contain more than a handful of options.
4076
kenton@google.com24bf56f2008-09-24 20:31:01 +00004077 if (intermediate_fields_iter == intermediate_fields_end) {
4078 // We're at the innermost submessage.
kenton@google.comd37d46d2009-04-25 02:53:47 +00004079 for (int i = 0; i < unknown_fields.field_count(); i++) {
4080 if (unknown_fields.field(i).number() == innermost_field->number()) {
4081 return AddNameError("Option \"" + debug_msg_name +
4082 "\" was already set.");
4083 }
kenton@google.com24bf56f2008-09-24 20:31:01 +00004084 }
kenton@google.comd37d46d2009-04-25 02:53:47 +00004085 return true;
kenton@google.com24bf56f2008-09-24 20:31:01 +00004086 }
4087
kenton@google.comd37d46d2009-04-25 02:53:47 +00004088 for (int i = 0; i < unknown_fields.field_count(); i++) {
4089 if (unknown_fields.field(i).number() ==
4090 (*intermediate_fields_iter)->number()) {
4091 const UnknownField* unknown_field = &unknown_fields.field(i);
4092 FieldDescriptor::Type type = (*intermediate_fields_iter)->type();
4093 // Recurse into the next submessage.
kenton@google.comd37d46d2009-04-25 02:53:47 +00004094 switch (type) {
4095 case FieldDescriptor::TYPE_MESSAGE:
4096 if (unknown_field->type() == UnknownField::TYPE_LENGTH_DELIMITED) {
4097 UnknownFieldSet intermediate_unknown_fields;
4098 if (intermediate_unknown_fields.ParseFromString(
4099 unknown_field->length_delimited()) &&
kenton@google.com80b1d622009-07-29 01:13:20 +00004100 !ExamineIfOptionIsSet(intermediate_fields_iter + 1,
kenton@google.comd37d46d2009-04-25 02:53:47 +00004101 intermediate_fields_end,
4102 innermost_field, debug_msg_name,
4103 intermediate_unknown_fields)) {
4104 return false; // Error already added.
4105 }
kenton@google.com24bf56f2008-09-24 20:31:01 +00004106 }
kenton@google.comd37d46d2009-04-25 02:53:47 +00004107 break;
kenton@google.com24bf56f2008-09-24 20:31:01 +00004108
kenton@google.comd37d46d2009-04-25 02:53:47 +00004109 case FieldDescriptor::TYPE_GROUP:
4110 if (unknown_field->type() == UnknownField::TYPE_GROUP) {
kenton@google.com80b1d622009-07-29 01:13:20 +00004111 if (!ExamineIfOptionIsSet(intermediate_fields_iter + 1,
kenton@google.comd37d46d2009-04-25 02:53:47 +00004112 intermediate_fields_end,
4113 innermost_field, debug_msg_name,
4114 unknown_field->group())) {
4115 return false; // Error already added.
4116 }
kenton@google.com24bf56f2008-09-24 20:31:01 +00004117 }
kenton@google.comd37d46d2009-04-25 02:53:47 +00004118 break;
kenton@google.com24bf56f2008-09-24 20:31:01 +00004119
kenton@google.comd37d46d2009-04-25 02:53:47 +00004120 default:
4121 GOOGLE_LOG(FATAL) << "Invalid wire type for CPPTYPE_MESSAGE: " << type;
4122 return false;
4123 }
kenton@google.com24bf56f2008-09-24 20:31:01 +00004124 }
4125 }
4126 return true;
4127}
4128
4129bool DescriptorBuilder::OptionInterpreter::SetOptionValue(
4130 const FieldDescriptor* option_field,
kenton@google.comd37d46d2009-04-25 02:53:47 +00004131 UnknownFieldSet* unknown_fields) {
kenton@google.com24bf56f2008-09-24 20:31:01 +00004132 // We switch on the CppType to validate.
4133 switch (option_field->cpp_type()) {
4134
4135 case FieldDescriptor::CPPTYPE_INT32:
4136 if (uninterpreted_option_->has_positive_int_value()) {
4137 if (uninterpreted_option_->positive_int_value() >
4138 static_cast<uint64>(kint32max)) {
4139 return AddValueError("Value out of range for int32 option \"" +
4140 option_field->full_name() + "\".");
4141 } else {
kenton@google.comd37d46d2009-04-25 02:53:47 +00004142 SetInt32(option_field->number(),
4143 uninterpreted_option_->positive_int_value(),
4144 option_field->type(), unknown_fields);
kenton@google.com24bf56f2008-09-24 20:31:01 +00004145 }
4146 } else if (uninterpreted_option_->has_negative_int_value()) {
4147 if (uninterpreted_option_->negative_int_value() <
4148 static_cast<int64>(kint32min)) {
4149 return AddValueError("Value out of range for int32 option \"" +
4150 option_field->full_name() + "\".");
4151 } else {
kenton@google.comd37d46d2009-04-25 02:53:47 +00004152 SetInt32(option_field->number(),
4153 uninterpreted_option_->negative_int_value(),
4154 option_field->type(), unknown_fields);
kenton@google.com24bf56f2008-09-24 20:31:01 +00004155 }
4156 } else {
4157 return AddValueError("Value must be integer for int32 option \"" +
4158 option_field->full_name() + "\".");
4159 }
4160 break;
4161
4162 case FieldDescriptor::CPPTYPE_INT64:
4163 if (uninterpreted_option_->has_positive_int_value()) {
4164 if (uninterpreted_option_->positive_int_value() >
4165 static_cast<uint64>(kint64max)) {
4166 return AddValueError("Value out of range for int64 option \"" +
4167 option_field->full_name() + "\".");
4168 } else {
kenton@google.comd37d46d2009-04-25 02:53:47 +00004169 SetInt64(option_field->number(),
4170 uninterpreted_option_->positive_int_value(),
4171 option_field->type(), unknown_fields);
kenton@google.com24bf56f2008-09-24 20:31:01 +00004172 }
4173 } else if (uninterpreted_option_->has_negative_int_value()) {
kenton@google.comd37d46d2009-04-25 02:53:47 +00004174 SetInt64(option_field->number(),
4175 uninterpreted_option_->negative_int_value(),
4176 option_field->type(), unknown_fields);
kenton@google.com24bf56f2008-09-24 20:31:01 +00004177 } else {
4178 return AddValueError("Value must be integer for int64 option \"" +
4179 option_field->full_name() + "\".");
4180 }
4181 break;
4182
4183 case FieldDescriptor::CPPTYPE_UINT32:
4184 if (uninterpreted_option_->has_positive_int_value()) {
4185 if (uninterpreted_option_->positive_int_value() > kuint32max) {
4186 return AddValueError("Value out of range for uint32 option \"" +
4187 option_field->name() + "\".");
4188 } else {
kenton@google.comd37d46d2009-04-25 02:53:47 +00004189 SetUInt32(option_field->number(),
4190 uninterpreted_option_->positive_int_value(),
4191 option_field->type(), unknown_fields);
kenton@google.com24bf56f2008-09-24 20:31:01 +00004192 }
4193 } else {
4194 return AddValueError("Value must be non-negative integer for uint32 "
4195 "option \"" + option_field->full_name() + "\".");
4196 }
4197 break;
4198
4199 case FieldDescriptor::CPPTYPE_UINT64:
4200 if (uninterpreted_option_->has_positive_int_value()) {
kenton@google.comd37d46d2009-04-25 02:53:47 +00004201 SetUInt64(option_field->number(),
4202 uninterpreted_option_->positive_int_value(),
4203 option_field->type(), unknown_fields);
kenton@google.com24bf56f2008-09-24 20:31:01 +00004204 } else {
4205 return AddValueError("Value must be non-negative integer for uint64 "
4206 "option \"" + option_field->full_name() + "\".");
4207 }
4208 break;
4209
4210 case FieldDescriptor::CPPTYPE_FLOAT: {
4211 float value;
4212 if (uninterpreted_option_->has_double_value()) {
4213 value = uninterpreted_option_->double_value();
4214 } else if (uninterpreted_option_->has_positive_int_value()) {
4215 value = uninterpreted_option_->positive_int_value();
4216 } else if (uninterpreted_option_->has_negative_int_value()) {
4217 value = uninterpreted_option_->negative_int_value();
4218 } else {
4219 return AddValueError("Value must be number for float option \"" +
4220 option_field->full_name() + "\".");
4221 }
kenton@google.comd37d46d2009-04-25 02:53:47 +00004222 unknown_fields->AddFixed32(option_field->number(),
kenton@google.com80b1d622009-07-29 01:13:20 +00004223 google::protobuf::internal::WireFormatLite::EncodeFloat(value));
kenton@google.com24bf56f2008-09-24 20:31:01 +00004224 break;
4225 }
4226
4227 case FieldDescriptor::CPPTYPE_DOUBLE: {
4228 double value;
4229 if (uninterpreted_option_->has_double_value()) {
4230 value = uninterpreted_option_->double_value();
4231 } else if (uninterpreted_option_->has_positive_int_value()) {
4232 value = uninterpreted_option_->positive_int_value();
4233 } else if (uninterpreted_option_->has_negative_int_value()) {
4234 value = uninterpreted_option_->negative_int_value();
4235 } else {
4236 return AddValueError("Value must be number for double option \"" +
4237 option_field->full_name() + "\".");
4238 }
kenton@google.comd37d46d2009-04-25 02:53:47 +00004239 unknown_fields->AddFixed64(option_field->number(),
kenton@google.com80b1d622009-07-29 01:13:20 +00004240 google::protobuf::internal::WireFormatLite::EncodeDouble(value));
kenton@google.com24bf56f2008-09-24 20:31:01 +00004241 break;
4242 }
4243
4244 case FieldDescriptor::CPPTYPE_BOOL:
4245 uint64 value;
4246 if (!uninterpreted_option_->has_identifier_value()) {
4247 return AddValueError("Value must be identifier for boolean option "
4248 "\"" + option_field->full_name() + "\".");
4249 }
4250 if (uninterpreted_option_->identifier_value() == "true") {
4251 value = 1;
4252 } else if (uninterpreted_option_->identifier_value() == "false") {
4253 value = 0;
4254 } else {
4255 return AddValueError("Value must be \"true\" or \"false\" for boolean "
4256 "option \"" + option_field->full_name() + "\".");
4257 }
kenton@google.comd37d46d2009-04-25 02:53:47 +00004258 unknown_fields->AddVarint(option_field->number(), value);
kenton@google.com24bf56f2008-09-24 20:31:01 +00004259 break;
4260
4261 case FieldDescriptor::CPPTYPE_ENUM: {
4262 if (!uninterpreted_option_->has_identifier_value()) {
4263 return AddValueError("Value must be identifier for enum-valued option "
4264 "\"" + option_field->full_name() + "\".");
4265 }
4266 const EnumDescriptor* enum_type = option_field->enum_type();
4267 const string& value_name = uninterpreted_option_->identifier_value();
4268 const EnumValueDescriptor* enum_value = NULL;
4269
4270 if (enum_type->file()->pool() != DescriptorPool::generated_pool()) {
4271 // Note that the enum value's fully-qualified name is a sibling of the
4272 // enum's name, not a child of it.
4273 string fully_qualified_name = enum_type->full_name();
4274 fully_qualified_name.resize(fully_qualified_name.size() -
4275 enum_type->name().size());
4276 fully_qualified_name += value_name;
4277
4278 // Search for the enum value's descriptor in the builder's pool. Note
kenton@google.com26bd9ee2008-11-21 00:06:27 +00004279 // that we use DescriptorBuilder::FindSymbolNotEnforcingDeps(), not
kenton@google.com24bf56f2008-09-24 20:31:01 +00004280 // DescriptorPool::FindEnumValueByName() because we're already holding
4281 // the pool's mutex, and the latter method locks it again.
kenton@google.com26bd9ee2008-11-21 00:06:27 +00004282 Symbol symbol =
4283 builder_->FindSymbolNotEnforcingDeps(fully_qualified_name);
kenton@google.com24bf56f2008-09-24 20:31:01 +00004284 if (!symbol.IsNull() && symbol.type == Symbol::ENUM_VALUE) {
4285 if (symbol.enum_value_descriptor->type() != enum_type) {
4286 return AddValueError("Enum type \"" + enum_type->full_name() +
4287 "\" has no value named \"" + value_name + "\" for option \"" +
4288 option_field->full_name() +
4289 "\". This appears to be a value from a sibling type.");
4290 } else {
4291 enum_value = symbol.enum_value_descriptor;
4292 }
4293 }
4294 } else {
4295 // The enum type is in the generated pool, so we can search for the
4296 // value there.
4297 enum_value = enum_type->FindValueByName(value_name);
4298 }
4299
4300 if (enum_value == NULL) {
4301 return AddValueError("Enum type \"" +
4302 option_field->enum_type()->full_name() +
4303 "\" has no value named \"" + value_name + "\" for "
4304 "option \"" + option_field->full_name() + "\".");
4305 } else {
4306 // Sign-extension is not a problem, since we cast directly from int32 to
4307 // uint64, without first going through uint32.
kenton@google.comd37d46d2009-04-25 02:53:47 +00004308 unknown_fields->AddVarint(option_field->number(),
4309 static_cast<uint64>(static_cast<int64>(enum_value->number())));
kenton@google.com24bf56f2008-09-24 20:31:01 +00004310 }
4311 break;
4312 }
4313
4314 case FieldDescriptor::CPPTYPE_STRING:
4315 if (!uninterpreted_option_->has_string_value()) {
4316 return AddValueError("Value must be quoted string for string option "
4317 "\"" + option_field->full_name() + "\".");
4318 }
4319 // The string has already been unquoted and unescaped by the parser.
kenton@google.comd37d46d2009-04-25 02:53:47 +00004320 unknown_fields->AddLengthDelimited(option_field->number(),
kenton@google.com24bf56f2008-09-24 20:31:01 +00004321 uninterpreted_option_->string_value());
4322 break;
4323
4324 case FieldDescriptor::CPPTYPE_MESSAGE:
liujisi@google.com33165fe2010-11-02 13:14:58 +00004325 if (!SetAggregateOption(option_field, unknown_fields)) {
4326 return false;
4327 }
kenton@google.com24bf56f2008-09-24 20:31:01 +00004328 break;
4329 }
4330
4331 return true;
4332}
4333
liujisi@google.com33165fe2010-11-02 13:14:58 +00004334class DescriptorBuilder::OptionInterpreter::AggregateOptionFinder
4335 : public TextFormat::Finder {
4336 public:
4337 DescriptorBuilder* builder_;
4338
4339 virtual const FieldDescriptor* FindExtension(
4340 Message* message, const string& name) const {
4341 if (builder_->pool_->mutex_ != NULL) {
4342 builder_->pool_->mutex_->AssertHeld();
4343 }
4344 Symbol result = builder_->LookupSymbolNoPlaceholder(
4345 name, message->GetDescriptor()->full_name());
4346 if (result.type == Symbol::FIELD &&
4347 result.field_descriptor->is_extension()) {
4348 return result.field_descriptor;
4349 } else {
4350 return NULL;
4351 }
4352 }
4353};
4354
4355// A custom error collector to record any text-format parsing errors
4356namespace {
4357class AggregateErrorCollector : public io::ErrorCollector {
4358 public:
4359 string error_;
4360
4361 virtual void AddError(int line, int column, const string& message) {
4362 if (!error_.empty()) {
4363 error_ += "; ";
4364 }
4365 error_ += message;
4366 }
4367
4368 virtual void AddWarning(int line, int column, const string& message) {
4369 // Ignore warnings
4370 }
4371};
4372}
4373
4374// We construct a dynamic message of the type corresponding to
4375// option_field, parse the supplied text-format string into this
4376// message, and serialize the resulting message to produce the value.
4377bool DescriptorBuilder::OptionInterpreter::SetAggregateOption(
4378 const FieldDescriptor* option_field,
4379 UnknownFieldSet* unknown_fields) {
4380 if (!uninterpreted_option_->has_aggregate_value()) {
4381 return AddValueError("Option \"" + option_field->full_name() +
4382 "\" is a message. To set the entire message, use "
4383 "syntax like \"" + option_field->name() +
4384 " = { <proto text format> }\". "
4385 "To set fields within it, use "
4386 "syntax like \"" + option_field->name() +
4387 ".foo = value\".");
4388 }
4389
4390 const Descriptor* type = option_field->message_type();
4391 scoped_ptr<Message> dynamic(dynamic_factory_.GetPrototype(type)->New());
4392 GOOGLE_CHECK(dynamic.get() != NULL)
4393 << "Could not create an instance of " << option_field->DebugString();
4394
4395 AggregateErrorCollector collector;
4396 AggregateOptionFinder finder;
4397 finder.builder_ = builder_;
4398 TextFormat::Parser parser;
4399 parser.RecordErrorsTo(&collector);
4400 parser.SetFinder(&finder);
4401 if (!parser.ParseFromString(uninterpreted_option_->aggregate_value(),
4402 dynamic.get())) {
4403 AddValueError("Error while parsing option value for \"" +
4404 option_field->name() + "\": " + collector.error_);
4405 return false;
4406 } else {
4407 string serial;
4408 dynamic->SerializeToString(&serial); // Never fails
4409 unknown_fields->AddLengthDelimited(option_field->number(), serial);
4410 return true;
4411 }
4412}
4413
kenton@google.comd37d46d2009-04-25 02:53:47 +00004414void DescriptorBuilder::OptionInterpreter::SetInt32(int number, int32 value,
4415 FieldDescriptor::Type type, UnknownFieldSet* unknown_fields) {
kenton@google.com24bf56f2008-09-24 20:31:01 +00004416 switch (type) {
4417 case FieldDescriptor::TYPE_INT32:
kenton@google.comd37d46d2009-04-25 02:53:47 +00004418 unknown_fields->AddVarint(number,
4419 static_cast<uint64>(static_cast<int64>(value)));
kenton@google.com24bf56f2008-09-24 20:31:01 +00004420 break;
4421
4422 case FieldDescriptor::TYPE_SFIXED32:
kenton@google.comd37d46d2009-04-25 02:53:47 +00004423 unknown_fields->AddFixed32(number, static_cast<uint32>(value));
kenton@google.com24bf56f2008-09-24 20:31:01 +00004424 break;
4425
4426 case FieldDescriptor::TYPE_SINT32:
kenton@google.comd37d46d2009-04-25 02:53:47 +00004427 unknown_fields->AddVarint(number,
kenton@google.com80b1d622009-07-29 01:13:20 +00004428 google::protobuf::internal::WireFormatLite::ZigZagEncode32(value));
kenton@google.com24bf56f2008-09-24 20:31:01 +00004429 break;
4430
4431 default:
4432 GOOGLE_LOG(FATAL) << "Invalid wire type for CPPTYPE_INT32: " << type;
4433 break;
4434 }
4435}
4436
kenton@google.comd37d46d2009-04-25 02:53:47 +00004437void DescriptorBuilder::OptionInterpreter::SetInt64(int number, int64 value,
4438 FieldDescriptor::Type type, UnknownFieldSet* unknown_fields) {
kenton@google.com24bf56f2008-09-24 20:31:01 +00004439 switch (type) {
4440 case FieldDescriptor::TYPE_INT64:
kenton@google.comd37d46d2009-04-25 02:53:47 +00004441 unknown_fields->AddVarint(number, static_cast<uint64>(value));
kenton@google.com24bf56f2008-09-24 20:31:01 +00004442 break;
4443
4444 case FieldDescriptor::TYPE_SFIXED64:
kenton@google.comd37d46d2009-04-25 02:53:47 +00004445 unknown_fields->AddFixed64(number, static_cast<uint64>(value));
kenton@google.com24bf56f2008-09-24 20:31:01 +00004446 break;
4447
4448 case FieldDescriptor::TYPE_SINT64:
kenton@google.comd37d46d2009-04-25 02:53:47 +00004449 unknown_fields->AddVarint(number,
kenton@google.com80b1d622009-07-29 01:13:20 +00004450 google::protobuf::internal::WireFormatLite::ZigZagEncode64(value));
kenton@google.com24bf56f2008-09-24 20:31:01 +00004451 break;
4452
4453 default:
4454 GOOGLE_LOG(FATAL) << "Invalid wire type for CPPTYPE_INT64: " << type;
4455 break;
4456 }
4457}
4458
kenton@google.comd37d46d2009-04-25 02:53:47 +00004459void DescriptorBuilder::OptionInterpreter::SetUInt32(int number, uint32 value,
4460 FieldDescriptor::Type type, UnknownFieldSet* unknown_fields) {
kenton@google.com24bf56f2008-09-24 20:31:01 +00004461 switch (type) {
4462 case FieldDescriptor::TYPE_UINT32:
kenton@google.comd37d46d2009-04-25 02:53:47 +00004463 unknown_fields->AddVarint(number, static_cast<uint64>(value));
kenton@google.com24bf56f2008-09-24 20:31:01 +00004464 break;
4465
4466 case FieldDescriptor::TYPE_FIXED32:
kenton@google.comd37d46d2009-04-25 02:53:47 +00004467 unknown_fields->AddFixed32(number, static_cast<uint32>(value));
kenton@google.com24bf56f2008-09-24 20:31:01 +00004468 break;
4469
4470 default:
4471 GOOGLE_LOG(FATAL) << "Invalid wire type for CPPTYPE_UINT32: " << type;
4472 break;
4473 }
4474}
4475
kenton@google.comd37d46d2009-04-25 02:53:47 +00004476void DescriptorBuilder::OptionInterpreter::SetUInt64(int number, uint64 value,
4477 FieldDescriptor::Type type, UnknownFieldSet* unknown_fields) {
kenton@google.com24bf56f2008-09-24 20:31:01 +00004478 switch (type) {
4479 case FieldDescriptor::TYPE_UINT64:
kenton@google.comd37d46d2009-04-25 02:53:47 +00004480 unknown_fields->AddVarint(number, value);
kenton@google.com24bf56f2008-09-24 20:31:01 +00004481 break;
4482
4483 case FieldDescriptor::TYPE_FIXED64:
kenton@google.comd37d46d2009-04-25 02:53:47 +00004484 unknown_fields->AddFixed64(number, value);
kenton@google.com24bf56f2008-09-24 20:31:01 +00004485 break;
4486
4487 default:
4488 GOOGLE_LOG(FATAL) << "Invalid wire type for CPPTYPE_UINT64: " << type;
4489 break;
4490 }
4491}
4492
temporal40ee5512008-07-10 02:12:20 +00004493} // namespace protobuf
4494} // namespace google