1 | // Protocol Buffers - Google's data interchange format |
2 | // Copyright 2008 Google Inc. All rights reserved. |
3 | // https://developers.google.com/protocol-buffers/ |
4 | // |
5 | // Redistribution and use in source and binary forms, with or without |
6 | // modification, are permitted provided that the following conditions are |
7 | // met: |
8 | // |
9 | // * 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. |
18 | // |
19 | // 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. |
30 | |
31 | // Author: kenton@google.com (Kenton Varda) |
32 | // Based on original Protocol Buffers design by |
33 | // Sanjay Ghemawat, Jeff Dean, and others. |
34 | // |
35 | // This file contains the CodedInputStream and CodedOutputStream classes, |
36 | // which wrap a ZeroCopyInputStream or ZeroCopyOutputStream, respectively, |
37 | // and allow you to read or write individual pieces of data in various |
38 | // formats. In particular, these implement the varint encoding for |
39 | // integers, a simple variable-length encoding in which smaller numbers |
40 | // take fewer bytes. |
41 | // |
42 | // Typically these classes will only be used internally by the protocol |
43 | // buffer library in order to encode and decode protocol buffers. Clients |
44 | // of the library only need to know about this class if they wish to write |
45 | // custom message parsing or serialization procedures. |
46 | // |
47 | // CodedOutputStream example: |
48 | // // Write some data to "myfile". First we write a 4-byte "magic number" |
49 | // // to identify the file type, then write a length-delimited string. The |
50 | // // string is composed of a varint giving the length followed by the raw |
51 | // // bytes. |
52 | // int fd = open("myfile", O_CREAT | O_WRONLY); |
53 | // ZeroCopyOutputStream* raw_output = new FileOutputStream(fd); |
54 | // CodedOutputStream* coded_output = new CodedOutputStream(raw_output); |
55 | // |
56 | // int magic_number = 1234; |
57 | // char text[] = "Hello world!"; |
58 | // coded_output->WriteLittleEndian32(magic_number); |
59 | // coded_output->WriteVarint32(strlen(text)); |
60 | // coded_output->WriteRaw(text, strlen(text)); |
61 | // |
62 | // delete coded_output; |
63 | // delete raw_output; |
64 | // close(fd); |
65 | // |
66 | // CodedInputStream example: |
67 | // // Read a file created by the above code. |
68 | // int fd = open("myfile", O_RDONLY); |
69 | // ZeroCopyInputStream* raw_input = new FileInputStream(fd); |
70 | // CodedInputStream* coded_input = new CodedInputStream(raw_input); |
71 | // |
72 | // coded_input->ReadLittleEndian32(&magic_number); |
73 | // if (magic_number != 1234) { |
74 | // cerr << "File not in expected format." << endl; |
75 | // return; |
76 | // } |
77 | // |
78 | // uint32_t size; |
79 | // coded_input->ReadVarint32(&size); |
80 | // |
81 | // char* text = new char[size + 1]; |
82 | // coded_input->ReadRaw(buffer, size); |
83 | // text[size] = '\0'; |
84 | // |
85 | // delete coded_input; |
86 | // delete raw_input; |
87 | // close(fd); |
88 | // |
89 | // cout << "Text is: " << text << endl; |
90 | // delete [] text; |
91 | // |
92 | // For those who are interested, varint encoding is defined as follows: |
93 | // |
94 | // The encoding operates on unsigned integers of up to 64 bits in length. |
95 | // Each byte of the encoded value has the format: |
96 | // * bits 0-6: Seven bits of the number being encoded. |
97 | // * bit 7: Zero if this is the last byte in the encoding (in which |
98 | // case all remaining bits of the number are zero) or 1 if |
99 | // more bytes follow. |
100 | // The first byte contains the least-significant 7 bits of the number, the |
101 | // second byte (if present) contains the next-least-significant 7 bits, |
102 | // and so on. So, the binary number 1011000101011 would be encoded in two |
103 | // bytes as "10101011 00101100". |
104 | // |
105 | // In theory, varint could be used to encode integers of any length. |
106 | // However, for practicality we set a limit at 64 bits. The maximum encoded |
107 | // length of a number is thus 10 bytes. |
108 | |
109 | #ifndef GOOGLE_PROTOBUF_IO_CODED_STREAM_H__ |
110 | #define GOOGLE_PROTOBUF_IO_CODED_STREAM_H__ |
111 | |
112 | |
113 | #include <assert.h> |
114 | |
115 | #include <atomic> |
116 | #include <climits> |
117 | #include <cstddef> |
118 | #include <cstring> |
119 | #include <limits> |
120 | #include <string> |
121 | #include <type_traits> |
122 | #include <utility> |
123 | |
124 | #if defined(_MSC_VER) && _MSC_VER >= 1300 && !defined(__INTEL_COMPILER) |
125 | // If MSVC has "/RTCc" set, it will complain about truncating casts at |
126 | // runtime. This file contains some intentional truncating casts. |
127 | #pragma runtime_checks("c", off) |
128 | #endif |
129 | |
130 | |
131 | #include <google/protobuf/stubs/common.h> |
132 | #include <google/protobuf/stubs/logging.h> |
133 | #include <google/protobuf/stubs/strutil.h> |
134 | #include <google/protobuf/port.h> |
135 | #include <google/protobuf/stubs/port.h> |
136 | |
137 | |
138 | // Must be included last. |
139 | #include <google/protobuf/port_def.inc> |
140 | |
141 | namespace google { |
142 | namespace protobuf { |
143 | |
144 | class DescriptorPool; |
145 | class MessageFactory; |
146 | class ZeroCopyCodedInputStream; |
147 | |
148 | namespace internal { |
149 | void MapTestForceDeterministic(); |
150 | class EpsCopyByteStream; |
151 | } // namespace internal |
152 | |
153 | namespace io { |
154 | |
155 | // Defined in this file. |
156 | class CodedInputStream; |
157 | class CodedOutputStream; |
158 | |
159 | // Defined in other files. |
160 | class ZeroCopyInputStream; // zero_copy_stream.h |
161 | class ZeroCopyOutputStream; // zero_copy_stream.h |
162 | |
163 | // Class which reads and decodes binary data which is composed of varint- |
164 | // encoded integers and fixed-width pieces. Wraps a ZeroCopyInputStream. |
165 | // Most users will not need to deal with CodedInputStream. |
166 | // |
167 | // Most methods of CodedInputStream that return a bool return false if an |
168 | // underlying I/O error occurs or if the data is malformed. Once such a |
169 | // failure occurs, the CodedInputStream is broken and is no longer useful. |
170 | // After a failure, callers also should assume writes to "out" args may have |
171 | // occurred, though nothing useful can be determined from those writes. |
172 | class PROTOBUF_EXPORT CodedInputStream { |
173 | public: |
174 | // Create a CodedInputStream that reads from the given ZeroCopyInputStream. |
175 | explicit CodedInputStream(ZeroCopyInputStream* input); |
176 | |
177 | // Create a CodedInputStream that reads from the given flat array. This is |
178 | // faster than using an ArrayInputStream. PushLimit(size) is implied by |
179 | // this constructor. |
180 | explicit CodedInputStream(const uint8_t* buffer, int size); |
181 | |
182 | // Destroy the CodedInputStream and position the underlying |
183 | // ZeroCopyInputStream at the first unread byte. If an error occurred while |
184 | // reading (causing a method to return false), then the exact position of |
185 | // the input stream may be anywhere between the last value that was read |
186 | // successfully and the stream's byte limit. |
187 | ~CodedInputStream(); |
188 | |
189 | // Return true if this CodedInputStream reads from a flat array instead of |
190 | // a ZeroCopyInputStream. |
191 | inline bool IsFlat() const; |
192 | |
193 | // Skips a number of bytes. Returns false if an underlying read error |
194 | // occurs. |
195 | inline bool Skip(int count); |
196 | |
197 | // Sets *data to point directly at the unread part of the CodedInputStream's |
198 | // underlying buffer, and *size to the size of that buffer, but does not |
199 | // advance the stream's current position. This will always either produce |
200 | // a non-empty buffer or return false. If the caller consumes any of |
201 | // this data, it should then call Skip() to skip over the consumed bytes. |
202 | // This may be useful for implementing external fast parsing routines for |
203 | // types of data not covered by the CodedInputStream interface. |
204 | bool GetDirectBufferPointer(const void** data, int* size); |
205 | |
206 | // Like GetDirectBufferPointer, but this method is inlined, and does not |
207 | // attempt to Refresh() if the buffer is currently empty. |
208 | PROTOBUF_ALWAYS_INLINE |
209 | void GetDirectBufferPointerInline(const void** data, int* size); |
210 | |
211 | // Read raw bytes, copying them into the given buffer. |
212 | bool ReadRaw(void* buffer, int size); |
213 | |
214 | // Like ReadRaw, but reads into a string. |
215 | bool ReadString(std::string* buffer, int size); |
216 | |
217 | |
218 | // Read a 32-bit little-endian integer. |
219 | bool ReadLittleEndian32(uint32_t* value); |
220 | // Read a 64-bit little-endian integer. |
221 | bool ReadLittleEndian64(uint64_t* value); |
222 | |
223 | // These methods read from an externally provided buffer. The caller is |
224 | // responsible for ensuring that the buffer has sufficient space. |
225 | // Read a 32-bit little-endian integer. |
226 | static const uint8_t* ReadLittleEndian32FromArray(const uint8_t* buffer, |
227 | uint32_t* value); |
228 | // Read a 64-bit little-endian integer. |
229 | static const uint8_t* ReadLittleEndian64FromArray(const uint8_t* buffer, |
230 | uint64_t* value); |
231 | |
232 | // Read an unsigned integer with Varint encoding, truncating to 32 bits. |
233 | // Reading a 32-bit value is equivalent to reading a 64-bit one and casting |
234 | // it to uint32_t, but may be more efficient. |
235 | bool ReadVarint32(uint32_t* value); |
236 | // Read an unsigned integer with Varint encoding. |
237 | bool ReadVarint64(uint64_t* value); |
238 | |
239 | // Reads a varint off the wire into an "int". This should be used for reading |
240 | // sizes off the wire (sizes of strings, submessages, bytes fields, etc). |
241 | // |
242 | // The value from the wire is interpreted as unsigned. If its value exceeds |
243 | // the representable value of an integer on this platform, instead of |
244 | // truncating we return false. Truncating (as performed by ReadVarint32() |
245 | // above) is an acceptable approach for fields representing an integer, but |
246 | // when we are parsing a size from the wire, truncating the value would result |
247 | // in us misparsing the payload. |
248 | bool ReadVarintSizeAsInt(int* value); |
249 | |
250 | // Read a tag. This calls ReadVarint32() and returns the result, or returns |
251 | // zero (which is not a valid tag) if ReadVarint32() fails. Also, ReadTag |
252 | // (but not ReadTagNoLastTag) updates the last tag value, which can be checked |
253 | // with LastTagWas(). |
254 | // |
255 | // Always inline because this is only called in one place per parse loop |
256 | // but it is called for every iteration of said loop, so it should be fast. |
257 | // GCC doesn't want to inline this by default. |
258 | PROTOBUF_ALWAYS_INLINE uint32_t ReadTag() { |
259 | return last_tag_ = ReadTagNoLastTag(); |
260 | } |
261 | |
262 | PROTOBUF_ALWAYS_INLINE uint32_t ReadTagNoLastTag(); |
263 | |
264 | // This usually a faster alternative to ReadTag() when cutoff is a manifest |
265 | // constant. It does particularly well for cutoff >= 127. The first part |
266 | // of the return value is the tag that was read, though it can also be 0 in |
267 | // the cases where ReadTag() would return 0. If the second part is true |
268 | // then the tag is known to be in [0, cutoff]. If not, the tag either is |
269 | // above cutoff or is 0. (There's intentional wiggle room when tag is 0, |
270 | // because that can arise in several ways, and for best performance we want |
271 | // to avoid an extra "is tag == 0?" check here.) |
272 | PROTOBUF_ALWAYS_INLINE |
273 | std::pair<uint32_t, bool> ReadTagWithCutoff(uint32_t cutoff) { |
274 | std::pair<uint32_t, bool> result = ReadTagWithCutoffNoLastTag(cutoff); |
275 | last_tag_ = result.first; |
276 | return result; |
277 | } |
278 | |
279 | PROTOBUF_ALWAYS_INLINE |
280 | std::pair<uint32_t, bool> ReadTagWithCutoffNoLastTag(uint32_t cutoff); |
281 | |
282 | // Usually returns true if calling ReadVarint32() now would produce the given |
283 | // value. Will always return false if ReadVarint32() would not return the |
284 | // given value. If ExpectTag() returns true, it also advances past |
285 | // the varint. For best performance, use a compile-time constant as the |
286 | // parameter. |
287 | // Always inline because this collapses to a small number of instructions |
288 | // when given a constant parameter, but GCC doesn't want to inline by default. |
289 | PROTOBUF_ALWAYS_INLINE bool ExpectTag(uint32_t expected); |
290 | |
291 | // Like above, except this reads from the specified buffer. The caller is |
292 | // responsible for ensuring that the buffer is large enough to read a varint |
293 | // of the expected size. For best performance, use a compile-time constant as |
294 | // the expected tag parameter. |
295 | // |
296 | // Returns a pointer beyond the expected tag if it was found, or NULL if it |
297 | // was not. |
298 | PROTOBUF_ALWAYS_INLINE |
299 | static const uint8_t* ExpectTagFromArray(const uint8_t* buffer, |
300 | uint32_t expected); |
301 | |
302 | // Usually returns true if no more bytes can be read. Always returns false |
303 | // if more bytes can be read. If ExpectAtEnd() returns true, a subsequent |
304 | // call to LastTagWas() will act as if ReadTag() had been called and returned |
305 | // zero, and ConsumedEntireMessage() will return true. |
306 | bool ExpectAtEnd(); |
307 | |
308 | // If the last call to ReadTag() or ReadTagWithCutoff() returned the given |
309 | // value, returns true. Otherwise, returns false. |
310 | // ReadTagNoLastTag/ReadTagWithCutoffNoLastTag do not preserve the last |
311 | // returned value. |
312 | // |
313 | // This is needed because parsers for some types of embedded messages |
314 | // (with field type TYPE_GROUP) don't actually know that they've reached the |
315 | // end of a message until they see an ENDGROUP tag, which was actually part |
316 | // of the enclosing message. The enclosing message would like to check that |
317 | // tag to make sure it had the right number, so it calls LastTagWas() on |
318 | // return from the embedded parser to check. |
319 | bool LastTagWas(uint32_t expected); |
320 | void SetLastTag(uint32_t tag) { last_tag_ = tag; } |
321 | |
322 | // When parsing message (but NOT a group), this method must be called |
323 | // immediately after MergeFromCodedStream() returns (if it returns true) |
324 | // to further verify that the message ended in a legitimate way. For |
325 | // example, this verifies that parsing did not end on an end-group tag. |
326 | // It also checks for some cases where, due to optimizations, |
327 | // MergeFromCodedStream() can incorrectly return true. |
328 | bool ConsumedEntireMessage(); |
329 | void SetConsumed() { legitimate_message_end_ = true; } |
330 | |
331 | // Limits ---------------------------------------------------------- |
332 | // Limits are used when parsing length-delimited embedded messages. |
333 | // After the message's length is read, PushLimit() is used to prevent |
334 | // the CodedInputStream from reading beyond that length. Once the |
335 | // embedded message has been parsed, PopLimit() is called to undo the |
336 | // limit. |
337 | |
338 | // Opaque type used with PushLimit() and PopLimit(). Do not modify |
339 | // values of this type yourself. The only reason that this isn't a |
340 | // struct with private internals is for efficiency. |
341 | typedef int Limit; |
342 | |
343 | // Places a limit on the number of bytes that the stream may read, |
344 | // starting from the current position. Once the stream hits this limit, |
345 | // it will act like the end of the input has been reached until PopLimit() |
346 | // is called. |
347 | // |
348 | // As the names imply, the stream conceptually has a stack of limits. The |
349 | // shortest limit on the stack is always enforced, even if it is not the |
350 | // top limit. |
351 | // |
352 | // The value returned by PushLimit() is opaque to the caller, and must |
353 | // be passed unchanged to the corresponding call to PopLimit(). |
354 | Limit PushLimit(int byte_limit); |
355 | |
356 | // Pops the last limit pushed by PushLimit(). The input must be the value |
357 | // returned by that call to PushLimit(). |
358 | void PopLimit(Limit limit); |
359 | |
360 | // Returns the number of bytes left until the nearest limit on the |
361 | // stack is hit, or -1 if no limits are in place. |
362 | int BytesUntilLimit() const; |
363 | |
364 | // Returns current position relative to the beginning of the input stream. |
365 | int CurrentPosition() const; |
366 | |
367 | // Total Bytes Limit ----------------------------------------------- |
368 | // To prevent malicious users from sending excessively large messages |
369 | // and causing memory exhaustion, CodedInputStream imposes a hard limit on |
370 | // the total number of bytes it will read. |
371 | |
372 | // Sets the maximum number of bytes that this CodedInputStream will read |
373 | // before refusing to continue. To prevent servers from allocating enormous |
374 | // amounts of memory to hold parsed messages, the maximum message length |
375 | // should be limited to the shortest length that will not harm usability. |
376 | // The default limit is INT_MAX (~2GB) and apps should set shorter limits |
377 | // if possible. An error will always be printed to stderr if the limit is |
378 | // reached. |
379 | // |
380 | // Note: setting a limit less than the current read position is interpreted |
381 | // as a limit on the current position. |
382 | // |
383 | // This is unrelated to PushLimit()/PopLimit(). |
384 | void SetTotalBytesLimit(int total_bytes_limit); |
385 | |
386 | // The Total Bytes Limit minus the Current Position, or -1 if the total bytes |
387 | // limit is INT_MAX. |
388 | int BytesUntilTotalBytesLimit() const; |
389 | |
390 | // Recursion Limit ------------------------------------------------- |
391 | // To prevent corrupt or malicious messages from causing stack overflows, |
392 | // we must keep track of the depth of recursion when parsing embedded |
393 | // messages and groups. CodedInputStream keeps track of this because it |
394 | // is the only object that is passed down the stack during parsing. |
395 | |
396 | // Sets the maximum recursion depth. The default is 100. |
397 | void SetRecursionLimit(int limit); |
398 | int RecursionBudget() { return recursion_budget_; } |
399 | |
400 | static int GetDefaultRecursionLimit() { return default_recursion_limit_; } |
401 | |
402 | // Increments the current recursion depth. Returns true if the depth is |
403 | // under the limit, false if it has gone over. |
404 | bool IncrementRecursionDepth(); |
405 | |
406 | // Decrements the recursion depth if possible. |
407 | void DecrementRecursionDepth(); |
408 | |
409 | // Decrements the recursion depth blindly. This is faster than |
410 | // DecrementRecursionDepth(). It should be used only if all previous |
411 | // increments to recursion depth were successful. |
412 | void UnsafeDecrementRecursionDepth(); |
413 | |
414 | // Shorthand for make_pair(PushLimit(byte_limit), --recursion_budget_). |
415 | // Using this can reduce code size and complexity in some cases. The caller |
416 | // is expected to check that the second part of the result is non-negative (to |
417 | // bail out if the depth of recursion is too high) and, if all is well, to |
418 | // later pass the first part of the result to PopLimit() or similar. |
419 | std::pair<CodedInputStream::Limit, int> IncrementRecursionDepthAndPushLimit( |
420 | int byte_limit); |
421 | |
422 | // Shorthand for PushLimit(ReadVarint32(&length) ? length : 0). |
423 | Limit ReadLengthAndPushLimit(); |
424 | |
425 | // Helper that is equivalent to: { |
426 | // bool result = ConsumedEntireMessage(); |
427 | // PopLimit(limit); |
428 | // UnsafeDecrementRecursionDepth(); |
429 | // return result; } |
430 | // Using this can reduce code size and complexity in some cases. |
431 | // Do not use unless the current recursion depth is greater than zero. |
432 | bool DecrementRecursionDepthAndPopLimit(Limit limit); |
433 | |
434 | // Helper that is equivalent to: { |
435 | // bool result = ConsumedEntireMessage(); |
436 | // PopLimit(limit); |
437 | // return result; } |
438 | // Using this can reduce code size and complexity in some cases. |
439 | bool CheckEntireMessageConsumedAndPopLimit(Limit limit); |
440 | |
441 | // Extension Registry ---------------------------------------------- |
442 | // ADVANCED USAGE: 99.9% of people can ignore this section. |
443 | // |
444 | // By default, when parsing extensions, the parser looks for extension |
445 | // definitions in the pool which owns the outer message's Descriptor. |
446 | // However, you may call SetExtensionRegistry() to provide an alternative |
447 | // pool instead. This makes it possible, for example, to parse a message |
448 | // using a generated class, but represent some extensions using |
449 | // DynamicMessage. |
450 | |
451 | // Set the pool used to look up extensions. Most users do not need to call |
452 | // this as the correct pool will be chosen automatically. |
453 | // |
454 | // WARNING: It is very easy to misuse this. Carefully read the requirements |
455 | // below. Do not use this unless you are sure you need it. Almost no one |
456 | // does. |
457 | // |
458 | // Let's say you are parsing a message into message object m, and you want |
459 | // to take advantage of SetExtensionRegistry(). You must follow these |
460 | // requirements: |
461 | // |
462 | // The given DescriptorPool must contain m->GetDescriptor(). It is not |
463 | // sufficient for it to simply contain a descriptor that has the same name |
464 | // and content -- it must be the *exact object*. In other words: |
465 | // assert(pool->FindMessageTypeByName(m->GetDescriptor()->full_name()) == |
466 | // m->GetDescriptor()); |
467 | // There are two ways to satisfy this requirement: |
468 | // 1) Use m->GetDescriptor()->pool() as the pool. This is generally useless |
469 | // because this is the pool that would be used anyway if you didn't call |
470 | // SetExtensionRegistry() at all. |
471 | // 2) Use a DescriptorPool which has m->GetDescriptor()->pool() as an |
472 | // "underlay". Read the documentation for DescriptorPool for more |
473 | // information about underlays. |
474 | // |
475 | // You must also provide a MessageFactory. This factory will be used to |
476 | // construct Message objects representing extensions. The factory's |
477 | // GetPrototype() MUST return non-NULL for any Descriptor which can be found |
478 | // through the provided pool. |
479 | // |
480 | // If the provided factory might return instances of protocol-compiler- |
481 | // generated (i.e. compiled-in) types, or if the outer message object m is |
482 | // a generated type, then the given factory MUST have this property: If |
483 | // GetPrototype() is given a Descriptor which resides in |
484 | // DescriptorPool::generated_pool(), the factory MUST return the same |
485 | // prototype which MessageFactory::generated_factory() would return. That |
486 | // is, given a descriptor for a generated type, the factory must return an |
487 | // instance of the generated class (NOT DynamicMessage). However, when |
488 | // given a descriptor for a type that is NOT in generated_pool, the factory |
489 | // is free to return any implementation. |
490 | // |
491 | // The reason for this requirement is that generated sub-objects may be |
492 | // accessed via the standard (non-reflection) extension accessor methods, |
493 | // and these methods will down-cast the object to the generated class type. |
494 | // If the object is not actually of that type, the results would be undefined. |
495 | // On the other hand, if an extension is not compiled in, then there is no |
496 | // way the code could end up accessing it via the standard accessors -- the |
497 | // only way to access the extension is via reflection. When using reflection, |
498 | // DynamicMessage and generated messages are indistinguishable, so it's fine |
499 | // if these objects are represented using DynamicMessage. |
500 | // |
501 | // Using DynamicMessageFactory on which you have called |
502 | // SetDelegateToGeneratedFactory(true) should be sufficient to satisfy the |
503 | // above requirement. |
504 | // |
505 | // If either pool or factory is NULL, both must be NULL. |
506 | // |
507 | // Note that this feature is ignored when parsing "lite" messages as they do |
508 | // not have descriptors. |
509 | void SetExtensionRegistry(const DescriptorPool* pool, |
510 | MessageFactory* factory); |
511 | |
512 | // Get the DescriptorPool set via SetExtensionRegistry(), or NULL if no pool |
513 | // has been provided. |
514 | const DescriptorPool* GetExtensionPool(); |
515 | |
516 | // Get the MessageFactory set via SetExtensionRegistry(), or NULL if no |
517 | // factory has been provided. |
518 | MessageFactory* GetExtensionFactory(); |
519 | |
520 | private: |
521 | GOOGLE_DISALLOW_EVIL_CONSTRUCTORS(CodedInputStream); |
522 | |
523 | const uint8_t* buffer_; |
524 | const uint8_t* buffer_end_; // pointer to the end of the buffer. |
525 | ZeroCopyInputStream* input_; |
526 | int total_bytes_read_; // total bytes read from input_, including |
527 | // the current buffer |
528 | |
529 | // If total_bytes_read_ surpasses INT_MAX, we record the extra bytes here |
530 | // so that we can BackUp() on destruction. |
531 | int overflow_bytes_; |
532 | |
533 | // LastTagWas() stuff. |
534 | uint32_t last_tag_; // result of last ReadTag() or ReadTagWithCutoff(). |
535 | |
536 | // This is set true by ReadTag{Fallback/Slow}() if it is called when exactly |
537 | // at EOF, or by ExpectAtEnd() when it returns true. This happens when we |
538 | // reach the end of a message and attempt to read another tag. |
539 | bool legitimate_message_end_; |
540 | |
541 | // See EnableAliasing(). |
542 | bool aliasing_enabled_; |
543 | |
544 | // Limits |
545 | Limit current_limit_; // if position = -1, no limit is applied |
546 | |
547 | // For simplicity, if the current buffer crosses a limit (either a normal |
548 | // limit created by PushLimit() or the total bytes limit), buffer_size_ |
549 | // only tracks the number of bytes before that limit. This field |
550 | // contains the number of bytes after it. Note that this implies that if |
551 | // buffer_size_ == 0 and buffer_size_after_limit_ > 0, we know we've |
552 | // hit a limit. However, if both are zero, it doesn't necessarily mean |
553 | // we aren't at a limit -- the buffer may have ended exactly at the limit. |
554 | int buffer_size_after_limit_; |
555 | |
556 | // Maximum number of bytes to read, period. This is unrelated to |
557 | // current_limit_. Set using SetTotalBytesLimit(). |
558 | int total_bytes_limit_; |
559 | |
560 | // Current recursion budget, controlled by IncrementRecursionDepth() and |
561 | // similar. Starts at recursion_limit_ and goes down: if this reaches |
562 | // -1 we are over budget. |
563 | int recursion_budget_; |
564 | // Recursion depth limit, set by SetRecursionLimit(). |
565 | int recursion_limit_; |
566 | |
567 | // See SetExtensionRegistry(). |
568 | const DescriptorPool* extension_pool_; |
569 | MessageFactory* extension_factory_; |
570 | |
571 | // Private member functions. |
572 | |
573 | // Fallback when Skip() goes past the end of the current buffer. |
574 | bool SkipFallback(int count, int original_buffer_size); |
575 | |
576 | // Advance the buffer by a given number of bytes. |
577 | void Advance(int amount); |
578 | |
579 | // Back up input_ to the current buffer position. |
580 | void BackUpInputToCurrentPosition(); |
581 | |
582 | // Recomputes the value of buffer_size_after_limit_. Must be called after |
583 | // current_limit_ or total_bytes_limit_ changes. |
584 | void RecomputeBufferLimits(); |
585 | |
586 | // Writes an error message saying that we hit total_bytes_limit_. |
587 | void PrintTotalBytesLimitError(); |
588 | |
589 | // Called when the buffer runs out to request more data. Implies an |
590 | // Advance(BufferSize()). |
591 | bool Refresh(); |
592 | |
593 | // When parsing varints, we optimize for the common case of small values, and |
594 | // then optimize for the case when the varint fits within the current buffer |
595 | // piece. The Fallback method is used when we can't use the one-byte |
596 | // optimization. The Slow method is yet another fallback when the buffer is |
597 | // not large enough. Making the slow path out-of-line speeds up the common |
598 | // case by 10-15%. The slow path is fairly uncommon: it only triggers when a |
599 | // message crosses multiple buffers. Note: ReadVarint32Fallback() and |
600 | // ReadVarint64Fallback() are called frequently and generally not inlined, so |
601 | // they have been optimized to avoid "out" parameters. The former returns -1 |
602 | // if it fails and the uint32_t it read otherwise. The latter has a bool |
603 | // indicating success or failure as part of its return type. |
604 | int64_t ReadVarint32Fallback(uint32_t first_byte_or_zero); |
605 | int ReadVarintSizeAsIntFallback(); |
606 | std::pair<uint64_t, bool> ReadVarint64Fallback(); |
607 | bool ReadVarint32Slow(uint32_t* value); |
608 | bool ReadVarint64Slow(uint64_t* value); |
609 | int ReadVarintSizeAsIntSlow(); |
610 | bool ReadLittleEndian32Fallback(uint32_t* value); |
611 | bool ReadLittleEndian64Fallback(uint64_t* value); |
612 | |
613 | // Fallback/slow methods for reading tags. These do not update last_tag_, |
614 | // but will set legitimate_message_end_ if we are at the end of the input |
615 | // stream. |
616 | uint32_t ReadTagFallback(uint32_t first_byte_or_zero); |
617 | uint32_t ReadTagSlow(); |
618 | bool ReadStringFallback(std::string* buffer, int size); |
619 | |
620 | // Return the size of the buffer. |
621 | int BufferSize() const; |
622 | |
623 | static const int kDefaultTotalBytesLimit = INT_MAX; |
624 | |
625 | static int default_recursion_limit_; // 100 by default. |
626 | |
627 | friend class google::protobuf::ZeroCopyCodedInputStream; |
628 | friend class google::protobuf::internal::EpsCopyByteStream; |
629 | }; |
630 | |
631 | // EpsCopyOutputStream wraps a ZeroCopyOutputStream and exposes a new stream, |
632 | // which has the property you can write kSlopBytes (16 bytes) from the current |
633 | // position without bounds checks. The cursor into the stream is managed by |
634 | // the user of the class and is an explicit parameter in the methods. Careful |
635 | // use of this class, ie. keep ptr a local variable, eliminates the need to |
636 | // for the compiler to sync the ptr value between register and memory. |
637 | class PROTOBUF_EXPORT EpsCopyOutputStream { |
638 | public: |
639 | enum { kSlopBytes = 16 }; |
640 | |
641 | // Initialize from a stream. |
642 | EpsCopyOutputStream(ZeroCopyOutputStream* stream, bool deterministic, |
643 | uint8_t** pp) |
644 | : end_(buffer_), |
645 | stream_(stream), |
646 | is_serialization_deterministic_(deterministic) { |
647 | *pp = buffer_; |
648 | } |
649 | |
650 | // Only for array serialization. No overflow protection, end_ will be the |
651 | // pointed to the end of the array. When using this the total size is already |
652 | // known, so no need to maintain the slop region. |
653 | EpsCopyOutputStream(void* data, int size, bool deterministic) |
654 | : end_(static_cast<uint8_t*>(data) + size), |
655 | buffer_end_(nullptr), |
656 | stream_(nullptr), |
657 | is_serialization_deterministic_(deterministic) {} |
658 | |
659 | // Initialize from stream but with the first buffer already given (eager). |
660 | EpsCopyOutputStream(void* data, int size, ZeroCopyOutputStream* stream, |
661 | bool deterministic, uint8_t** pp) |
662 | : stream_(stream), is_serialization_deterministic_(deterministic) { |
663 | *pp = SetInitialBuffer(data, size); |
664 | } |
665 | |
666 | // Flush everything that's written into the underlying ZeroCopyOutputStream |
667 | // and trims the underlying stream to the location of ptr. |
668 | uint8_t* Trim(uint8_t* ptr); |
669 | |
670 | // After this it's guaranteed you can safely write kSlopBytes to ptr. This |
671 | // will never fail! The underlying stream can produce an error. Use HadError |
672 | // to check for errors. |
673 | PROTOBUF_NODISCARD uint8_t* EnsureSpace(uint8_t* ptr) { |
674 | if (PROTOBUF_PREDICT_FALSE(ptr >= end_)) { |
675 | return EnsureSpaceFallback(ptr); |
676 | } |
677 | return ptr; |
678 | } |
679 | |
680 | uint8_t* WriteRaw(const void* data, int size, uint8_t* ptr) { |
681 | if (PROTOBUF_PREDICT_FALSE(end_ - ptr < size)) { |
682 | return WriteRawFallback(data, size, ptr); |
683 | } |
684 | std::memcpy(dest: ptr, src: data, n: size); |
685 | return ptr + size; |
686 | } |
687 | // Writes the buffer specified by data, size to the stream. Possibly by |
688 | // aliasing the buffer (ie. not copying the data). The caller is responsible |
689 | // to make sure the buffer is alive for the duration of the |
690 | // ZeroCopyOutputStream. |
691 | #ifndef NDEBUG |
692 | PROTOBUF_NOINLINE |
693 | #endif |
694 | uint8_t* WriteRawMaybeAliased(const void* data, int size, uint8_t* ptr) { |
695 | if (aliasing_enabled_) { |
696 | return WriteAliasedRaw(data, size, ptr); |
697 | } else { |
698 | return WriteRaw(data, size, ptr); |
699 | } |
700 | } |
701 | |
702 | |
703 | #ifndef NDEBUG |
704 | PROTOBUF_NOINLINE |
705 | #endif |
706 | uint8_t* WriteStringMaybeAliased(uint32_t num, const std::string& s, |
707 | uint8_t* ptr) { |
708 | std::ptrdiff_t size = s.size(); |
709 | if (PROTOBUF_PREDICT_FALSE( |
710 | size >= 128 || end_ - ptr + 16 - TagSize(num << 3) - 1 < size)) { |
711 | return WriteStringMaybeAliasedOutline(num, s, ptr); |
712 | } |
713 | ptr = UnsafeVarint(value: (num << 3) | 2, ptr); |
714 | *ptr++ = static_cast<uint8_t>(size); |
715 | std::memcpy(dest: ptr, src: s.data(), n: size); |
716 | return ptr + size; |
717 | } |
718 | uint8_t* WriteBytesMaybeAliased(uint32_t num, const std::string& s, |
719 | uint8_t* ptr) { |
720 | return WriteStringMaybeAliased(num, s, ptr); |
721 | } |
722 | |
723 | template <typename T> |
724 | PROTOBUF_ALWAYS_INLINE uint8_t* WriteString(uint32_t num, const T& s, |
725 | uint8_t* ptr) { |
726 | std::ptrdiff_t size = s.size(); |
727 | if (PROTOBUF_PREDICT_FALSE( |
728 | size >= 128 || end_ - ptr + 16 - TagSize(num << 3) - 1 < size)) { |
729 | return WriteStringOutline(num, s, ptr); |
730 | } |
731 | ptr = UnsafeVarint(value: (num << 3) | 2, ptr); |
732 | *ptr++ = static_cast<uint8_t>(size); |
733 | std::memcpy(dest: ptr, src: s.data(), n: size); |
734 | return ptr + size; |
735 | } |
736 | template <typename T> |
737 | #ifndef NDEBUG |
738 | PROTOBUF_NOINLINE |
739 | #endif |
740 | uint8_t* WriteBytes(uint32_t num, const T& s, uint8_t* ptr) { |
741 | return WriteString(num, s, ptr); |
742 | } |
743 | |
744 | template <typename T> |
745 | PROTOBUF_ALWAYS_INLINE uint8_t* WriteInt32Packed(int num, const T& r, |
746 | int size, uint8_t* ptr) { |
747 | return WriteVarintPacked(num, r, size, ptr, Encode64); |
748 | } |
749 | template <typename T> |
750 | PROTOBUF_ALWAYS_INLINE uint8_t* WriteUInt32Packed(int num, const T& r, |
751 | int size, uint8_t* ptr) { |
752 | return WriteVarintPacked(num, r, size, ptr, Encode32); |
753 | } |
754 | template <typename T> |
755 | PROTOBUF_ALWAYS_INLINE uint8_t* WriteSInt32Packed(int num, const T& r, |
756 | int size, uint8_t* ptr) { |
757 | return WriteVarintPacked(num, r, size, ptr, ZigZagEncode32); |
758 | } |
759 | template <typename T> |
760 | PROTOBUF_ALWAYS_INLINE uint8_t* WriteInt64Packed(int num, const T& r, |
761 | int size, uint8_t* ptr) { |
762 | return WriteVarintPacked(num, r, size, ptr, Encode64); |
763 | } |
764 | template <typename T> |
765 | PROTOBUF_ALWAYS_INLINE uint8_t* WriteUInt64Packed(int num, const T& r, |
766 | int size, uint8_t* ptr) { |
767 | return WriteVarintPacked(num, r, size, ptr, Encode64); |
768 | } |
769 | template <typename T> |
770 | PROTOBUF_ALWAYS_INLINE uint8_t* WriteSInt64Packed(int num, const T& r, |
771 | int size, uint8_t* ptr) { |
772 | return WriteVarintPacked(num, r, size, ptr, ZigZagEncode64); |
773 | } |
774 | template <typename T> |
775 | PROTOBUF_ALWAYS_INLINE uint8_t* WriteEnumPacked(int num, const T& r, int size, |
776 | uint8_t* ptr) { |
777 | return WriteVarintPacked(num, r, size, ptr, Encode64); |
778 | } |
779 | |
780 | template <typename T> |
781 | PROTOBUF_ALWAYS_INLINE uint8_t* WriteFixedPacked(int num, const T& r, |
782 | uint8_t* ptr) { |
783 | ptr = EnsureSpace(ptr); |
784 | constexpr auto element_size = sizeof(typename T::value_type); |
785 | auto size = r.size() * element_size; |
786 | ptr = WriteLengthDelim(num, size, ptr); |
787 | return WriteRawLittleEndian<element_size>(r.data(), static_cast<int>(size), |
788 | ptr); |
789 | } |
790 | |
791 | // Returns true if there was an underlying I/O error since this object was |
792 | // created. |
793 | bool HadError() const { return had_error_; } |
794 | |
795 | // Instructs the EpsCopyOutputStream to allow the underlying |
796 | // ZeroCopyOutputStream to hold pointers to the original structure instead of |
797 | // copying, if it supports it (i.e. output->AllowsAliasing() is true). If the |
798 | // underlying stream does not support aliasing, then enabling it has no |
799 | // affect. For now, this only affects the behavior of |
800 | // WriteRawMaybeAliased(). |
801 | // |
802 | // NOTE: It is caller's responsibility to ensure that the chunk of memory |
803 | // remains live until all of the data has been consumed from the stream. |
804 | void EnableAliasing(bool enabled); |
805 | |
806 | // See documentation on CodedOutputStream::SetSerializationDeterministic. |
807 | void SetSerializationDeterministic(bool value) { |
808 | is_serialization_deterministic_ = value; |
809 | } |
810 | |
811 | // See documentation on CodedOutputStream::IsSerializationDeterministic. |
812 | bool IsSerializationDeterministic() const { |
813 | return is_serialization_deterministic_; |
814 | } |
815 | |
816 | // The number of bytes written to the stream at position ptr, relative to the |
817 | // stream's overall position. |
818 | int64_t ByteCount(uint8_t* ptr) const; |
819 | |
820 | |
821 | private: |
822 | uint8_t* end_; |
823 | uint8_t* buffer_end_ = buffer_; |
824 | uint8_t buffer_[2 * kSlopBytes]; |
825 | ZeroCopyOutputStream* stream_; |
826 | bool had_error_ = false; |
827 | bool aliasing_enabled_ = false; // See EnableAliasing(). |
828 | bool is_serialization_deterministic_; |
829 | bool skip_check_consistency = false; |
830 | |
831 | uint8_t* EnsureSpaceFallback(uint8_t* ptr); |
832 | inline uint8_t* Next(); |
833 | int Flush(uint8_t* ptr); |
834 | std::ptrdiff_t GetSize(uint8_t* ptr) const { |
835 | GOOGLE_DCHECK(ptr <= end_ + kSlopBytes); // NOLINT |
836 | return end_ + kSlopBytes - ptr; |
837 | } |
838 | |
839 | uint8_t* Error() { |
840 | had_error_ = true; |
841 | // We use the patch buffer to always guarantee space to write to. |
842 | end_ = buffer_ + kSlopBytes; |
843 | return buffer_; |
844 | } |
845 | |
846 | static constexpr int TagSize(uint32_t tag) { |
847 | return (tag < (1 << 7)) ? 1 |
848 | : (tag < (1 << 14)) ? 2 |
849 | : (tag < (1 << 21)) ? 3 |
850 | : (tag < (1 << 28)) ? 4 |
851 | : 5; |
852 | } |
853 | |
854 | PROTOBUF_ALWAYS_INLINE uint8_t* WriteTag(uint32_t num, uint32_t wt, |
855 | uint8_t* ptr) { |
856 | GOOGLE_DCHECK(ptr < end_); // NOLINT |
857 | return UnsafeVarint(value: (num << 3) | wt, ptr); |
858 | } |
859 | |
860 | PROTOBUF_ALWAYS_INLINE uint8_t* WriteLengthDelim(int num, uint32_t size, |
861 | uint8_t* ptr) { |
862 | ptr = WriteTag(num, wt: 2, ptr); |
863 | return UnsafeWriteSize(value: size, ptr); |
864 | } |
865 | |
866 | uint8_t* WriteRawFallback(const void* data, int size, uint8_t* ptr); |
867 | |
868 | uint8_t* WriteAliasedRaw(const void* data, int size, uint8_t* ptr); |
869 | |
870 | uint8_t* WriteStringMaybeAliasedOutline(uint32_t num, const std::string& s, |
871 | uint8_t* ptr); |
872 | uint8_t* WriteStringOutline(uint32_t num, const std::string& s, uint8_t* ptr); |
873 | |
874 | template <typename T, typename E> |
875 | PROTOBUF_ALWAYS_INLINE uint8_t* WriteVarintPacked(int num, const T& r, |
876 | int size, uint8_t* ptr, |
877 | const E& encode) { |
878 | ptr = EnsureSpace(ptr); |
879 | ptr = WriteLengthDelim(num, size, ptr); |
880 | auto it = r.data(); |
881 | auto end = it + r.size(); |
882 | do { |
883 | ptr = EnsureSpace(ptr); |
884 | ptr = UnsafeVarint(encode(*it++), ptr); |
885 | } while (it < end); |
886 | return ptr; |
887 | } |
888 | |
889 | static uint32_t Encode32(uint32_t v) { return v; } |
890 | static uint64_t Encode64(uint64_t v) { return v; } |
891 | static uint32_t ZigZagEncode32(int32_t v) { |
892 | return (static_cast<uint32_t>(v) << 1) ^ static_cast<uint32_t>(v >> 31); |
893 | } |
894 | static uint64_t ZigZagEncode64(int64_t v) { |
895 | return (static_cast<uint64_t>(v) << 1) ^ static_cast<uint64_t>(v >> 63); |
896 | } |
897 | |
898 | template <typename T> |
899 | PROTOBUF_ALWAYS_INLINE static uint8_t* UnsafeVarint(T value, uint8_t* ptr) { |
900 | static_assert(std::is_unsigned<T>::value, |
901 | "Varint serialization must be unsigned" ); |
902 | ptr[0] = static_cast<uint8_t>(value); |
903 | if (value < 0x80) { |
904 | return ptr + 1; |
905 | } |
906 | // Turn on continuation bit in the byte we just wrote. |
907 | ptr[0] |= static_cast<uint8_t>(0x80); |
908 | value >>= 7; |
909 | ptr[1] = static_cast<uint8_t>(value); |
910 | if (value < 0x80) { |
911 | return ptr + 2; |
912 | } |
913 | ptr += 2; |
914 | do { |
915 | // Turn on continuation bit in the byte we just wrote. |
916 | ptr[-1] |= static_cast<uint8_t>(0x80); |
917 | value >>= 7; |
918 | *ptr = static_cast<uint8_t>(value); |
919 | ++ptr; |
920 | } while (value >= 0x80); |
921 | return ptr; |
922 | } |
923 | |
924 | PROTOBUF_ALWAYS_INLINE static uint8_t* UnsafeWriteSize(uint32_t value, |
925 | uint8_t* ptr) { |
926 | while (PROTOBUF_PREDICT_FALSE(value >= 0x80)) { |
927 | *ptr = static_cast<uint8_t>(value | 0x80); |
928 | value >>= 7; |
929 | ++ptr; |
930 | } |
931 | *ptr++ = static_cast<uint8_t>(value); |
932 | return ptr; |
933 | } |
934 | |
935 | template <int S> |
936 | uint8_t* WriteRawLittleEndian(const void* data, int size, uint8_t* ptr); |
937 | #if !defined(PROTOBUF_LITTLE_ENDIAN) || \ |
938 | defined(PROTOBUF_DISABLE_LITTLE_ENDIAN_OPT_FOR_TEST) |
939 | uint8_t* WriteRawLittleEndian32(const void* data, int size, uint8_t* ptr); |
940 | uint8_t* WriteRawLittleEndian64(const void* data, int size, uint8_t* ptr); |
941 | #endif |
942 | |
943 | // These methods are for CodedOutputStream. Ideally they should be private |
944 | // but to match current behavior of CodedOutputStream as close as possible |
945 | // we allow it some functionality. |
946 | public: |
947 | uint8_t* SetInitialBuffer(void* data, int size) { |
948 | auto ptr = static_cast<uint8_t*>(data); |
949 | if (size > kSlopBytes) { |
950 | end_ = ptr + size - kSlopBytes; |
951 | buffer_end_ = nullptr; |
952 | return ptr; |
953 | } else { |
954 | end_ = buffer_ + size; |
955 | buffer_end_ = ptr; |
956 | return buffer_; |
957 | } |
958 | } |
959 | |
960 | private: |
961 | // Needed by CodedOutputStream HadError. HadError needs to flush the patch |
962 | // buffers to ensure there is no error as of yet. |
963 | uint8_t* FlushAndResetBuffer(uint8_t*); |
964 | |
965 | // The following functions mimic the old CodedOutputStream behavior as close |
966 | // as possible. They flush the current state to the stream, behave as |
967 | // the old CodedOutputStream and then return to normal operation. |
968 | bool Skip(int count, uint8_t** pp); |
969 | bool GetDirectBufferPointer(void** data, int* size, uint8_t** pp); |
970 | uint8_t* GetDirectBufferForNBytesAndAdvance(int size, uint8_t** pp); |
971 | |
972 | friend class CodedOutputStream; |
973 | }; |
974 | |
975 | template <> |
976 | inline uint8_t* EpsCopyOutputStream::WriteRawLittleEndian<1>(const void* data, |
977 | int size, |
978 | uint8_t* ptr) { |
979 | return WriteRaw(data, size, ptr); |
980 | } |
981 | template <> |
982 | inline uint8_t* EpsCopyOutputStream::WriteRawLittleEndian<4>(const void* data, |
983 | int size, |
984 | uint8_t* ptr) { |
985 | #if defined(PROTOBUF_LITTLE_ENDIAN) && \ |
986 | !defined(PROTOBUF_DISABLE_LITTLE_ENDIAN_OPT_FOR_TEST) |
987 | return WriteRaw(data, size, ptr); |
988 | #else |
989 | return WriteRawLittleEndian32(data, size, ptr); |
990 | #endif |
991 | } |
992 | template <> |
993 | inline uint8_t* EpsCopyOutputStream::WriteRawLittleEndian<8>(const void* data, |
994 | int size, |
995 | uint8_t* ptr) { |
996 | #if defined(PROTOBUF_LITTLE_ENDIAN) && \ |
997 | !defined(PROTOBUF_DISABLE_LITTLE_ENDIAN_OPT_FOR_TEST) |
998 | return WriteRaw(data, size, ptr); |
999 | #else |
1000 | return WriteRawLittleEndian64(data, size, ptr); |
1001 | #endif |
1002 | } |
1003 | |
1004 | // Class which encodes and writes binary data which is composed of varint- |
1005 | // encoded integers and fixed-width pieces. Wraps a ZeroCopyOutputStream. |
1006 | // Most users will not need to deal with CodedOutputStream. |
1007 | // |
1008 | // Most methods of CodedOutputStream which return a bool return false if an |
1009 | // underlying I/O error occurs. Once such a failure occurs, the |
1010 | // CodedOutputStream is broken and is no longer useful. The Write* methods do |
1011 | // not return the stream status, but will invalidate the stream if an error |
1012 | // occurs. The client can probe HadError() to determine the status. |
1013 | // |
1014 | // Note that every method of CodedOutputStream which writes some data has |
1015 | // a corresponding static "ToArray" version. These versions write directly |
1016 | // to the provided buffer, returning a pointer past the last written byte. |
1017 | // They require that the buffer has sufficient capacity for the encoded data. |
1018 | // This allows an optimization where we check if an output stream has enough |
1019 | // space for an entire message before we start writing and, if there is, we |
1020 | // call only the ToArray methods to avoid doing bound checks for each |
1021 | // individual value. |
1022 | // i.e., in the example above: |
1023 | // |
1024 | // CodedOutputStream* coded_output = new CodedOutputStream(raw_output); |
1025 | // int magic_number = 1234; |
1026 | // char text[] = "Hello world!"; |
1027 | // |
1028 | // int coded_size = sizeof(magic_number) + |
1029 | // CodedOutputStream::VarintSize32(strlen(text)) + |
1030 | // strlen(text); |
1031 | // |
1032 | // uint8_t* buffer = |
1033 | // coded_output->GetDirectBufferForNBytesAndAdvance(coded_size); |
1034 | // if (buffer != nullptr) { |
1035 | // // The output stream has enough space in the buffer: write directly to |
1036 | // // the array. |
1037 | // buffer = CodedOutputStream::WriteLittleEndian32ToArray(magic_number, |
1038 | // buffer); |
1039 | // buffer = CodedOutputStream::WriteVarint32ToArray(strlen(text), buffer); |
1040 | // buffer = CodedOutputStream::WriteRawToArray(text, strlen(text), buffer); |
1041 | // } else { |
1042 | // // Make bound-checked writes, which will ask the underlying stream for |
1043 | // // more space as needed. |
1044 | // coded_output->WriteLittleEndian32(magic_number); |
1045 | // coded_output->WriteVarint32(strlen(text)); |
1046 | // coded_output->WriteRaw(text, strlen(text)); |
1047 | // } |
1048 | // |
1049 | // delete coded_output; |
1050 | class PROTOBUF_EXPORT CodedOutputStream { |
1051 | public: |
1052 | // Creates a CodedOutputStream that writes to the given `stream`. |
1053 | // The provided stream must publicly derive from `ZeroCopyOutputStream`. |
1054 | template <class Stream, class = typename std::enable_if<std::is_base_of< |
1055 | ZeroCopyOutputStream, Stream>::value>::type> |
1056 | explicit CodedOutputStream(Stream* stream); |
1057 | |
1058 | // Creates a CodedOutputStream that writes to the given `stream`, and does |
1059 | // an 'eager initialization' of the internal state if `eager_init` is true. |
1060 | // The provided stream must publicly derive from `ZeroCopyOutputStream`. |
1061 | template <class Stream, class = typename std::enable_if<std::is_base_of< |
1062 | ZeroCopyOutputStream, Stream>::value>::type> |
1063 | CodedOutputStream(Stream* stream, bool eager_init); |
1064 | |
1065 | // Destroy the CodedOutputStream and position the underlying |
1066 | // ZeroCopyOutputStream immediately after the last byte written. |
1067 | ~CodedOutputStream(); |
1068 | |
1069 | // Returns true if there was an underlying I/O error since this object was |
1070 | // created. On should call Trim before this function in order to catch all |
1071 | // errors. |
1072 | bool HadError() { |
1073 | cur_ = impl_.FlushAndResetBuffer(cur_); |
1074 | GOOGLE_DCHECK(cur_); |
1075 | return impl_.HadError(); |
1076 | } |
1077 | |
1078 | // Trims any unused space in the underlying buffer so that its size matches |
1079 | // the number of bytes written by this stream. The underlying buffer will |
1080 | // automatically be trimmed when this stream is destroyed; this call is only |
1081 | // necessary if the underlying buffer is accessed *before* the stream is |
1082 | // destroyed. |
1083 | void Trim() { cur_ = impl_.Trim(ptr: cur_); } |
1084 | |
1085 | // Skips a number of bytes, leaving the bytes unmodified in the underlying |
1086 | // buffer. Returns false if an underlying write error occurs. This is |
1087 | // mainly useful with GetDirectBufferPointer(). |
1088 | // Note of caution, the skipped bytes may contain uninitialized data. The |
1089 | // caller must make sure that the skipped bytes are properly initialized, |
1090 | // otherwise you might leak bytes from your heap. |
1091 | bool Skip(int count) { return impl_.Skip(count, pp: &cur_); } |
1092 | |
1093 | // Sets *data to point directly at the unwritten part of the |
1094 | // CodedOutputStream's underlying buffer, and *size to the size of that |
1095 | // buffer, but does not advance the stream's current position. This will |
1096 | // always either produce a non-empty buffer or return false. If the caller |
1097 | // writes any data to this buffer, it should then call Skip() to skip over |
1098 | // the consumed bytes. This may be useful for implementing external fast |
1099 | // serialization routines for types of data not covered by the |
1100 | // CodedOutputStream interface. |
1101 | bool GetDirectBufferPointer(void** data, int* size) { |
1102 | return impl_.GetDirectBufferPointer(data, size, pp: &cur_); |
1103 | } |
1104 | |
1105 | // If there are at least "size" bytes available in the current buffer, |
1106 | // returns a pointer directly into the buffer and advances over these bytes. |
1107 | // The caller may then write directly into this buffer (e.g. using the |
1108 | // *ToArray static methods) rather than go through CodedOutputStream. If |
1109 | // there are not enough bytes available, returns NULL. The return pointer is |
1110 | // invalidated as soon as any other non-const method of CodedOutputStream |
1111 | // is called. |
1112 | inline uint8_t* GetDirectBufferForNBytesAndAdvance(int size) { |
1113 | return impl_.GetDirectBufferForNBytesAndAdvance(size, pp: &cur_); |
1114 | } |
1115 | |
1116 | // Write raw bytes, copying them from the given buffer. |
1117 | void WriteRaw(const void* buffer, int size) { |
1118 | cur_ = impl_.WriteRaw(data: buffer, size, ptr: cur_); |
1119 | } |
1120 | // Like WriteRaw() but will try to write aliased data if aliasing is |
1121 | // turned on. |
1122 | void WriteRawMaybeAliased(const void* data, int size); |
1123 | // Like WriteRaw() but writing directly to the target array. |
1124 | // This is _not_ inlined, as the compiler often optimizes memcpy into inline |
1125 | // copy loops. Since this gets called by every field with string or bytes |
1126 | // type, inlining may lead to a significant amount of code bloat, with only a |
1127 | // minor performance gain. |
1128 | static uint8_t* WriteRawToArray(const void* buffer, int size, |
1129 | uint8_t* target); |
1130 | |
1131 | // Equivalent to WriteRaw(str.data(), str.size()). |
1132 | void WriteString(const std::string& str); |
1133 | // Like WriteString() but writing directly to the target array. |
1134 | static uint8_t* WriteStringToArray(const std::string& str, uint8_t* target); |
1135 | // Write the varint-encoded size of str followed by str. |
1136 | static uint8_t* WriteStringWithSizeToArray(const std::string& str, |
1137 | uint8_t* target); |
1138 | |
1139 | |
1140 | // Write a 32-bit little-endian integer. |
1141 | void WriteLittleEndian32(uint32_t value) { |
1142 | cur_ = impl_.EnsureSpace(ptr: cur_); |
1143 | SetCur(WriteLittleEndian32ToArray(value, target: Cur())); |
1144 | } |
1145 | // Like WriteLittleEndian32() but writing directly to the target array. |
1146 | static uint8_t* WriteLittleEndian32ToArray(uint32_t value, uint8_t* target); |
1147 | // Write a 64-bit little-endian integer. |
1148 | void WriteLittleEndian64(uint64_t value) { |
1149 | cur_ = impl_.EnsureSpace(ptr: cur_); |
1150 | SetCur(WriteLittleEndian64ToArray(value, target: Cur())); |
1151 | } |
1152 | // Like WriteLittleEndian64() but writing directly to the target array. |
1153 | static uint8_t* WriteLittleEndian64ToArray(uint64_t value, uint8_t* target); |
1154 | |
1155 | // Write an unsigned integer with Varint encoding. Writing a 32-bit value |
1156 | // is equivalent to casting it to uint64_t and writing it as a 64-bit value, |
1157 | // but may be more efficient. |
1158 | void WriteVarint32(uint32_t value); |
1159 | // Like WriteVarint32() but writing directly to the target array. |
1160 | static uint8_t* WriteVarint32ToArray(uint32_t value, uint8_t* target); |
1161 | // Like WriteVarint32() but writing directly to the target array, and with |
1162 | // the less common-case paths being out of line rather than inlined. |
1163 | static uint8_t* WriteVarint32ToArrayOutOfLine(uint32_t value, |
1164 | uint8_t* target); |
1165 | // Write an unsigned integer with Varint encoding. |
1166 | void WriteVarint64(uint64_t value); |
1167 | // Like WriteVarint64() but writing directly to the target array. |
1168 | static uint8_t* WriteVarint64ToArray(uint64_t value, uint8_t* target); |
1169 | |
1170 | // Equivalent to WriteVarint32() except when the value is negative, |
1171 | // in which case it must be sign-extended to a full 10 bytes. |
1172 | void WriteVarint32SignExtended(int32_t value); |
1173 | // Like WriteVarint32SignExtended() but writing directly to the target array. |
1174 | static uint8_t* WriteVarint32SignExtendedToArray(int32_t value, |
1175 | uint8_t* target); |
1176 | |
1177 | // This is identical to WriteVarint32(), but optimized for writing tags. |
1178 | // In particular, if the input is a compile-time constant, this method |
1179 | // compiles down to a couple instructions. |
1180 | // Always inline because otherwise the aforementioned optimization can't work, |
1181 | // but GCC by default doesn't want to inline this. |
1182 | void WriteTag(uint32_t value); |
1183 | // Like WriteTag() but writing directly to the target array. |
1184 | PROTOBUF_ALWAYS_INLINE |
1185 | static uint8_t* WriteTagToArray(uint32_t value, uint8_t* target); |
1186 | |
1187 | // Returns the number of bytes needed to encode the given value as a varint. |
1188 | static size_t VarintSize32(uint32_t value); |
1189 | // Returns the number of bytes needed to encode the given value as a varint. |
1190 | static size_t VarintSize64(uint64_t value); |
1191 | |
1192 | // If negative, 10 bytes. Otherwise, same as VarintSize32(). |
1193 | static size_t VarintSize32SignExtended(int32_t value); |
1194 | |
1195 | // Same as above, plus one. The additional one comes at no compute cost. |
1196 | static size_t VarintSize32PlusOne(uint32_t value); |
1197 | static size_t VarintSize64PlusOne(uint64_t value); |
1198 | static size_t VarintSize32SignExtendedPlusOne(int32_t value); |
1199 | |
1200 | // Compile-time equivalent of VarintSize32(). |
1201 | template <uint32_t Value> |
1202 | struct StaticVarintSize32 { |
1203 | static const size_t value = (Value < (1 << 7)) ? 1 |
1204 | : (Value < (1 << 14)) ? 2 |
1205 | : (Value < (1 << 21)) ? 3 |
1206 | : (Value < (1 << 28)) ? 4 |
1207 | : 5; |
1208 | }; |
1209 | |
1210 | // Returns the total number of bytes written since this object was created. |
1211 | int ByteCount() const { |
1212 | return static_cast<int>(impl_.ByteCount(ptr: cur_) - start_count_); |
1213 | } |
1214 | |
1215 | // Instructs the CodedOutputStream to allow the underlying |
1216 | // ZeroCopyOutputStream to hold pointers to the original structure instead of |
1217 | // copying, if it supports it (i.e. output->AllowsAliasing() is true). If the |
1218 | // underlying stream does not support aliasing, then enabling it has no |
1219 | // affect. For now, this only affects the behavior of |
1220 | // WriteRawMaybeAliased(). |
1221 | // |
1222 | // NOTE: It is caller's responsibility to ensure that the chunk of memory |
1223 | // remains live until all of the data has been consumed from the stream. |
1224 | void EnableAliasing(bool enabled) { impl_.EnableAliasing(enabled); } |
1225 | |
1226 | // Indicate to the serializer whether the user wants deterministic |
1227 | // serialization. The default when this is not called comes from the global |
1228 | // default, controlled by SetDefaultSerializationDeterministic. |
1229 | // |
1230 | // What deterministic serialization means is entirely up to the driver of the |
1231 | // serialization process (i.e. the caller of methods like WriteVarint32). In |
1232 | // the case of serializing a proto buffer message using one of the methods of |
1233 | // MessageLite, this means that for a given binary equal messages will always |
1234 | // be serialized to the same bytes. This implies: |
1235 | // |
1236 | // * Repeated serialization of a message will return the same bytes. |
1237 | // |
1238 | // * Different processes running the same binary (including on different |
1239 | // machines) will serialize equal messages to the same bytes. |
1240 | // |
1241 | // Note that this is *not* canonical across languages. It is also unstable |
1242 | // across different builds with intervening message definition changes, due to |
1243 | // unknown fields. Users who need canonical serialization (e.g. persistent |
1244 | // storage in a canonical form, fingerprinting) should define their own |
1245 | // canonicalization specification and implement the serializer using |
1246 | // reflection APIs rather than relying on this API. |
1247 | void SetSerializationDeterministic(bool value) { |
1248 | impl_.SetSerializationDeterministic(value); |
1249 | } |
1250 | |
1251 | // Return whether the user wants deterministic serialization. See above. |
1252 | bool IsSerializationDeterministic() const { |
1253 | return impl_.IsSerializationDeterministic(); |
1254 | } |
1255 | |
1256 | static bool IsDefaultSerializationDeterministic() { |
1257 | return default_serialization_deterministic_.load( |
1258 | m: std::memory_order_relaxed) != 0; |
1259 | } |
1260 | |
1261 | template <typename Func> |
1262 | void Serialize(const Func& func); |
1263 | |
1264 | uint8_t* Cur() const { return cur_; } |
1265 | void SetCur(uint8_t* ptr) { cur_ = ptr; } |
1266 | EpsCopyOutputStream* EpsCopy() { return &impl_; } |
1267 | |
1268 | private: |
1269 | template <class Stream> |
1270 | void InitEagerly(Stream* stream); |
1271 | |
1272 | EpsCopyOutputStream impl_; |
1273 | uint8_t* cur_; |
1274 | int64_t start_count_; |
1275 | static std::atomic<bool> default_serialization_deterministic_; |
1276 | |
1277 | // See above. Other projects may use "friend" to allow them to call this. |
1278 | // After SetDefaultSerializationDeterministic() completes, all protocol |
1279 | // buffer serializations will be deterministic by default. Thread safe. |
1280 | // However, the meaning of "after" is subtle here: to be safe, each thread |
1281 | // that wants deterministic serialization by default needs to call |
1282 | // SetDefaultSerializationDeterministic() or ensure on its own that another |
1283 | // thread has done so. |
1284 | friend void internal::MapTestForceDeterministic(); |
1285 | static void SetDefaultSerializationDeterministic() { |
1286 | default_serialization_deterministic_.store(i: true, m: std::memory_order_relaxed); |
1287 | } |
1288 | // REQUIRES: value >= 0x80, and that (value & 7f) has been written to *target. |
1289 | static uint8_t* WriteVarint32ToArrayOutOfLineHelper(uint32_t value, |
1290 | uint8_t* target); |
1291 | GOOGLE_DISALLOW_EVIL_CONSTRUCTORS(CodedOutputStream); |
1292 | }; |
1293 | |
1294 | // inline methods ==================================================== |
1295 | // The vast majority of varints are only one byte. These inline |
1296 | // methods optimize for that case. |
1297 | |
1298 | inline bool CodedInputStream::ReadVarint32(uint32_t* value) { |
1299 | uint32_t v = 0; |
1300 | if (PROTOBUF_PREDICT_TRUE(buffer_ < buffer_end_)) { |
1301 | v = *buffer_; |
1302 | if (v < 0x80) { |
1303 | *value = v; |
1304 | Advance(amount: 1); |
1305 | return true; |
1306 | } |
1307 | } |
1308 | int64_t result = ReadVarint32Fallback(first_byte_or_zero: v); |
1309 | *value = static_cast<uint32_t>(result); |
1310 | return result >= 0; |
1311 | } |
1312 | |
1313 | inline bool CodedInputStream::ReadVarint64(uint64_t* value) { |
1314 | if (PROTOBUF_PREDICT_TRUE(buffer_ < buffer_end_) && *buffer_ < 0x80) { |
1315 | *value = *buffer_; |
1316 | Advance(amount: 1); |
1317 | return true; |
1318 | } |
1319 | std::pair<uint64_t, bool> p = ReadVarint64Fallback(); |
1320 | *value = p.first; |
1321 | return p.second; |
1322 | } |
1323 | |
1324 | inline bool CodedInputStream::ReadVarintSizeAsInt(int* value) { |
1325 | if (PROTOBUF_PREDICT_TRUE(buffer_ < buffer_end_)) { |
1326 | int v = *buffer_; |
1327 | if (v < 0x80) { |
1328 | *value = v; |
1329 | Advance(amount: 1); |
1330 | return true; |
1331 | } |
1332 | } |
1333 | *value = ReadVarintSizeAsIntFallback(); |
1334 | return *value >= 0; |
1335 | } |
1336 | |
1337 | // static |
1338 | inline const uint8_t* CodedInputStream::ReadLittleEndian32FromArray( |
1339 | const uint8_t* buffer, uint32_t* value) { |
1340 | #if defined(PROTOBUF_LITTLE_ENDIAN) && \ |
1341 | !defined(PROTOBUF_DISABLE_LITTLE_ENDIAN_OPT_FOR_TEST) |
1342 | memcpy(dest: value, src: buffer, n: sizeof(*value)); |
1343 | return buffer + sizeof(*value); |
1344 | #else |
1345 | *value = (static_cast<uint32_t>(buffer[0])) | |
1346 | (static_cast<uint32_t>(buffer[1]) << 8) | |
1347 | (static_cast<uint32_t>(buffer[2]) << 16) | |
1348 | (static_cast<uint32_t>(buffer[3]) << 24); |
1349 | return buffer + sizeof(*value); |
1350 | #endif |
1351 | } |
1352 | // static |
1353 | inline const uint8_t* CodedInputStream::ReadLittleEndian64FromArray( |
1354 | const uint8_t* buffer, uint64_t* value) { |
1355 | #if defined(PROTOBUF_LITTLE_ENDIAN) && \ |
1356 | !defined(PROTOBUF_DISABLE_LITTLE_ENDIAN_OPT_FOR_TEST) |
1357 | memcpy(dest: value, src: buffer, n: sizeof(*value)); |
1358 | return buffer + sizeof(*value); |
1359 | #else |
1360 | uint32_t part0 = (static_cast<uint32_t>(buffer[0])) | |
1361 | (static_cast<uint32_t>(buffer[1]) << 8) | |
1362 | (static_cast<uint32_t>(buffer[2]) << 16) | |
1363 | (static_cast<uint32_t>(buffer[3]) << 24); |
1364 | uint32_t part1 = (static_cast<uint32_t>(buffer[4])) | |
1365 | (static_cast<uint32_t>(buffer[5]) << 8) | |
1366 | (static_cast<uint32_t>(buffer[6]) << 16) | |
1367 | (static_cast<uint32_t>(buffer[7]) << 24); |
1368 | *value = static_cast<uint64_t>(part0) | (static_cast<uint64_t>(part1) << 32); |
1369 | return buffer + sizeof(*value); |
1370 | #endif |
1371 | } |
1372 | |
1373 | inline bool CodedInputStream::ReadLittleEndian32(uint32_t* value) { |
1374 | #if defined(PROTOBUF_LITTLE_ENDIAN) && \ |
1375 | !defined(PROTOBUF_DISABLE_LITTLE_ENDIAN_OPT_FOR_TEST) |
1376 | if (PROTOBUF_PREDICT_TRUE(BufferSize() >= static_cast<int>(sizeof(*value)))) { |
1377 | buffer_ = ReadLittleEndian32FromArray(buffer: buffer_, value); |
1378 | return true; |
1379 | } else { |
1380 | return ReadLittleEndian32Fallback(value); |
1381 | } |
1382 | #else |
1383 | return ReadLittleEndian32Fallback(value); |
1384 | #endif |
1385 | } |
1386 | |
1387 | inline bool CodedInputStream::ReadLittleEndian64(uint64_t* value) { |
1388 | #if defined(PROTOBUF_LITTLE_ENDIAN) && \ |
1389 | !defined(PROTOBUF_DISABLE_LITTLE_ENDIAN_OPT_FOR_TEST) |
1390 | if (PROTOBUF_PREDICT_TRUE(BufferSize() >= static_cast<int>(sizeof(*value)))) { |
1391 | buffer_ = ReadLittleEndian64FromArray(buffer: buffer_, value); |
1392 | return true; |
1393 | } else { |
1394 | return ReadLittleEndian64Fallback(value); |
1395 | } |
1396 | #else |
1397 | return ReadLittleEndian64Fallback(value); |
1398 | #endif |
1399 | } |
1400 | |
1401 | inline uint32_t CodedInputStream::ReadTagNoLastTag() { |
1402 | uint32_t v = 0; |
1403 | if (PROTOBUF_PREDICT_TRUE(buffer_ < buffer_end_)) { |
1404 | v = *buffer_; |
1405 | if (v < 0x80) { |
1406 | Advance(amount: 1); |
1407 | return v; |
1408 | } |
1409 | } |
1410 | v = ReadTagFallback(first_byte_or_zero: v); |
1411 | return v; |
1412 | } |
1413 | |
1414 | inline std::pair<uint32_t, bool> CodedInputStream::ReadTagWithCutoffNoLastTag( |
1415 | uint32_t cutoff) { |
1416 | // In performance-sensitive code we can expect cutoff to be a compile-time |
1417 | // constant, and things like "cutoff >= kMax1ByteVarint" to be evaluated at |
1418 | // compile time. |
1419 | uint32_t first_byte_or_zero = 0; |
1420 | if (PROTOBUF_PREDICT_TRUE(buffer_ < buffer_end_)) { |
1421 | // Hot case: buffer_ non_empty, buffer_[0] in [1, 128). |
1422 | // TODO(gpike): Is it worth rearranging this? E.g., if the number of fields |
1423 | // is large enough then is it better to check for the two-byte case first? |
1424 | first_byte_or_zero = buffer_[0]; |
1425 | if (static_cast<int8_t>(buffer_[0]) > 0) { |
1426 | const uint32_t kMax1ByteVarint = 0x7f; |
1427 | uint32_t tag = buffer_[0]; |
1428 | Advance(amount: 1); |
1429 | return std::make_pair(x&: tag, y: cutoff >= kMax1ByteVarint || tag <= cutoff); |
1430 | } |
1431 | // Other hot case: cutoff >= 0x80, buffer_ has at least two bytes available, |
1432 | // and tag is two bytes. The latter is tested by bitwise-and-not of the |
1433 | // first byte and the second byte. |
1434 | if (cutoff >= 0x80 && PROTOBUF_PREDICT_TRUE(buffer_ + 1 < buffer_end_) && |
1435 | PROTOBUF_PREDICT_TRUE((buffer_[0] & ~buffer_[1]) >= 0x80)) { |
1436 | const uint32_t kMax2ByteVarint = (0x7f << 7) + 0x7f; |
1437 | uint32_t tag = (1u << 7) * buffer_[1] + (buffer_[0] - 0x80); |
1438 | Advance(amount: 2); |
1439 | // It might make sense to test for tag == 0 now, but it is so rare that |
1440 | // that we don't bother. A varint-encoded 0 should be one byte unless |
1441 | // the encoder lost its mind. The second part of the return value of |
1442 | // this function is allowed to be either true or false if the tag is 0, |
1443 | // so we don't have to check for tag == 0. We may need to check whether |
1444 | // it exceeds cutoff. |
1445 | bool at_or_below_cutoff = cutoff >= kMax2ByteVarint || tag <= cutoff; |
1446 | return std::make_pair(x&: tag, y&: at_or_below_cutoff); |
1447 | } |
1448 | } |
1449 | // Slow path |
1450 | const uint32_t tag = ReadTagFallback(first_byte_or_zero); |
1451 | return std::make_pair(x: tag, y: static_cast<uint32_t>(tag - 1) < cutoff); |
1452 | } |
1453 | |
1454 | inline bool CodedInputStream::LastTagWas(uint32_t expected) { |
1455 | return last_tag_ == expected; |
1456 | } |
1457 | |
1458 | inline bool CodedInputStream::ConsumedEntireMessage() { |
1459 | return legitimate_message_end_; |
1460 | } |
1461 | |
1462 | inline bool CodedInputStream::ExpectTag(uint32_t expected) { |
1463 | if (expected < (1 << 7)) { |
1464 | if (PROTOBUF_PREDICT_TRUE(buffer_ < buffer_end_) && |
1465 | buffer_[0] == expected) { |
1466 | Advance(amount: 1); |
1467 | return true; |
1468 | } else { |
1469 | return false; |
1470 | } |
1471 | } else if (expected < (1 << 14)) { |
1472 | if (PROTOBUF_PREDICT_TRUE(BufferSize() >= 2) && |
1473 | buffer_[0] == static_cast<uint8_t>(expected | 0x80) && |
1474 | buffer_[1] == static_cast<uint8_t>(expected >> 7)) { |
1475 | Advance(amount: 2); |
1476 | return true; |
1477 | } else { |
1478 | return false; |
1479 | } |
1480 | } else { |
1481 | // Don't bother optimizing for larger values. |
1482 | return false; |
1483 | } |
1484 | } |
1485 | |
1486 | inline const uint8_t* CodedInputStream::ExpectTagFromArray( |
1487 | const uint8_t* buffer, uint32_t expected) { |
1488 | if (expected < (1 << 7)) { |
1489 | if (buffer[0] == expected) { |
1490 | return buffer + 1; |
1491 | } |
1492 | } else if (expected < (1 << 14)) { |
1493 | if (buffer[0] == static_cast<uint8_t>(expected | 0x80) && |
1494 | buffer[1] == static_cast<uint8_t>(expected >> 7)) { |
1495 | return buffer + 2; |
1496 | } |
1497 | } |
1498 | return nullptr; |
1499 | } |
1500 | |
1501 | inline void CodedInputStream::GetDirectBufferPointerInline(const void** data, |
1502 | int* size) { |
1503 | *data = buffer_; |
1504 | *size = static_cast<int>(buffer_end_ - buffer_); |
1505 | } |
1506 | |
1507 | inline bool CodedInputStream::ExpectAtEnd() { |
1508 | // If we are at a limit we know no more bytes can be read. Otherwise, it's |
1509 | // hard to say without calling Refresh(), and we'd rather not do that. |
1510 | |
1511 | if (buffer_ == buffer_end_ && ((buffer_size_after_limit_ != 0) || |
1512 | (total_bytes_read_ == current_limit_))) { |
1513 | last_tag_ = 0; // Pretend we called ReadTag()... |
1514 | legitimate_message_end_ = true; // ... and it hit EOF. |
1515 | return true; |
1516 | } else { |
1517 | return false; |
1518 | } |
1519 | } |
1520 | |
1521 | inline int CodedInputStream::CurrentPosition() const { |
1522 | return total_bytes_read_ - (BufferSize() + buffer_size_after_limit_); |
1523 | } |
1524 | |
1525 | inline void CodedInputStream::Advance(int amount) { buffer_ += amount; } |
1526 | |
1527 | inline void CodedInputStream::SetRecursionLimit(int limit) { |
1528 | recursion_budget_ += limit - recursion_limit_; |
1529 | recursion_limit_ = limit; |
1530 | } |
1531 | |
1532 | inline bool CodedInputStream::IncrementRecursionDepth() { |
1533 | --recursion_budget_; |
1534 | return recursion_budget_ >= 0; |
1535 | } |
1536 | |
1537 | inline void CodedInputStream::DecrementRecursionDepth() { |
1538 | if (recursion_budget_ < recursion_limit_) ++recursion_budget_; |
1539 | } |
1540 | |
1541 | inline void CodedInputStream::UnsafeDecrementRecursionDepth() { |
1542 | assert(recursion_budget_ < recursion_limit_); |
1543 | ++recursion_budget_; |
1544 | } |
1545 | |
1546 | inline void CodedInputStream::SetExtensionRegistry(const DescriptorPool* pool, |
1547 | MessageFactory* factory) { |
1548 | extension_pool_ = pool; |
1549 | extension_factory_ = factory; |
1550 | } |
1551 | |
1552 | inline const DescriptorPool* CodedInputStream::GetExtensionPool() { |
1553 | return extension_pool_; |
1554 | } |
1555 | |
1556 | inline MessageFactory* CodedInputStream::GetExtensionFactory() { |
1557 | return extension_factory_; |
1558 | } |
1559 | |
1560 | inline int CodedInputStream::BufferSize() const { |
1561 | return static_cast<int>(buffer_end_ - buffer_); |
1562 | } |
1563 | |
1564 | inline CodedInputStream::CodedInputStream(ZeroCopyInputStream* input) |
1565 | : buffer_(nullptr), |
1566 | buffer_end_(nullptr), |
1567 | input_(input), |
1568 | total_bytes_read_(0), |
1569 | overflow_bytes_(0), |
1570 | last_tag_(0), |
1571 | legitimate_message_end_(false), |
1572 | aliasing_enabled_(false), |
1573 | current_limit_(std::numeric_limits<int32_t>::max()), |
1574 | buffer_size_after_limit_(0), |
1575 | total_bytes_limit_(kDefaultTotalBytesLimit), |
1576 | recursion_budget_(default_recursion_limit_), |
1577 | recursion_limit_(default_recursion_limit_), |
1578 | extension_pool_(nullptr), |
1579 | extension_factory_(nullptr) { |
1580 | // Eagerly Refresh() so buffer space is immediately available. |
1581 | Refresh(); |
1582 | } |
1583 | |
1584 | inline CodedInputStream::CodedInputStream(const uint8_t* buffer, int size) |
1585 | : buffer_(buffer), |
1586 | buffer_end_(buffer + size), |
1587 | input_(nullptr), |
1588 | total_bytes_read_(size), |
1589 | overflow_bytes_(0), |
1590 | last_tag_(0), |
1591 | legitimate_message_end_(false), |
1592 | aliasing_enabled_(false), |
1593 | current_limit_(size), |
1594 | buffer_size_after_limit_(0), |
1595 | total_bytes_limit_(kDefaultTotalBytesLimit), |
1596 | recursion_budget_(default_recursion_limit_), |
1597 | recursion_limit_(default_recursion_limit_), |
1598 | extension_pool_(nullptr), |
1599 | extension_factory_(nullptr) { |
1600 | // Note that setting current_limit_ == size is important to prevent some |
1601 | // code paths from trying to access input_ and segfaulting. |
1602 | } |
1603 | |
1604 | inline bool CodedInputStream::IsFlat() const { return input_ == nullptr; } |
1605 | |
1606 | inline bool CodedInputStream::Skip(int count) { |
1607 | if (count < 0) return false; // security: count is often user-supplied |
1608 | |
1609 | const int original_buffer_size = BufferSize(); |
1610 | |
1611 | if (count <= original_buffer_size) { |
1612 | // Just skipping within the current buffer. Easy. |
1613 | Advance(amount: count); |
1614 | return true; |
1615 | } |
1616 | |
1617 | return SkipFallback(count, original_buffer_size); |
1618 | } |
1619 | |
1620 | template <class Stream, class> |
1621 | inline CodedOutputStream::CodedOutputStream(Stream* stream) |
1622 | : impl_(stream, IsDefaultSerializationDeterministic(), &cur_), |
1623 | start_count_(stream->ByteCount()) { |
1624 | InitEagerly(stream); |
1625 | } |
1626 | |
1627 | template <class Stream, class> |
1628 | inline CodedOutputStream::CodedOutputStream(Stream* stream, bool eager_init) |
1629 | : impl_(stream, IsDefaultSerializationDeterministic(), &cur_), |
1630 | start_count_(stream->ByteCount()) { |
1631 | if (eager_init) { |
1632 | InitEagerly(stream); |
1633 | } |
1634 | } |
1635 | |
1636 | template <class Stream> |
1637 | inline void CodedOutputStream::InitEagerly(Stream* stream) { |
1638 | void* data; |
1639 | int size; |
1640 | if (PROTOBUF_PREDICT_TRUE(stream->Next(&data, &size) && size > 0)) { |
1641 | cur_ = impl_.SetInitialBuffer(data, size); |
1642 | } |
1643 | } |
1644 | |
1645 | inline uint8_t* CodedOutputStream::WriteVarint32ToArray(uint32_t value, |
1646 | uint8_t* target) { |
1647 | return EpsCopyOutputStream::UnsafeVarint(value, ptr: target); |
1648 | } |
1649 | |
1650 | inline uint8_t* CodedOutputStream::WriteVarint32ToArrayOutOfLine( |
1651 | uint32_t value, uint8_t* target) { |
1652 | target[0] = static_cast<uint8_t>(value); |
1653 | if (value < 0x80) { |
1654 | return target + 1; |
1655 | } else { |
1656 | return WriteVarint32ToArrayOutOfLineHelper(value, target); |
1657 | } |
1658 | } |
1659 | |
1660 | inline uint8_t* CodedOutputStream::WriteVarint64ToArray(uint64_t value, |
1661 | uint8_t* target) { |
1662 | return EpsCopyOutputStream::UnsafeVarint(value, ptr: target); |
1663 | } |
1664 | |
1665 | inline void CodedOutputStream::WriteVarint32SignExtended(int32_t value) { |
1666 | WriteVarint64(value: static_cast<uint64_t>(value)); |
1667 | } |
1668 | |
1669 | inline uint8_t* CodedOutputStream::WriteVarint32SignExtendedToArray( |
1670 | int32_t value, uint8_t* target) { |
1671 | return WriteVarint64ToArray(value: static_cast<uint64_t>(value), target); |
1672 | } |
1673 | |
1674 | inline uint8_t* CodedOutputStream::WriteLittleEndian32ToArray(uint32_t value, |
1675 | uint8_t* target) { |
1676 | #if defined(PROTOBUF_LITTLE_ENDIAN) && \ |
1677 | !defined(PROTOBUF_DISABLE_LITTLE_ENDIAN_OPT_FOR_TEST) |
1678 | memcpy(dest: target, src: &value, n: sizeof(value)); |
1679 | #else |
1680 | target[0] = static_cast<uint8_t>(value); |
1681 | target[1] = static_cast<uint8_t>(value >> 8); |
1682 | target[2] = static_cast<uint8_t>(value >> 16); |
1683 | target[3] = static_cast<uint8_t>(value >> 24); |
1684 | #endif |
1685 | return target + sizeof(value); |
1686 | } |
1687 | |
1688 | inline uint8_t* CodedOutputStream::WriteLittleEndian64ToArray(uint64_t value, |
1689 | uint8_t* target) { |
1690 | #if defined(PROTOBUF_LITTLE_ENDIAN) && \ |
1691 | !defined(PROTOBUF_DISABLE_LITTLE_ENDIAN_OPT_FOR_TEST) |
1692 | memcpy(dest: target, src: &value, n: sizeof(value)); |
1693 | #else |
1694 | uint32_t part0 = static_cast<uint32_t>(value); |
1695 | uint32_t part1 = static_cast<uint32_t>(value >> 32); |
1696 | |
1697 | target[0] = static_cast<uint8_t>(part0); |
1698 | target[1] = static_cast<uint8_t>(part0 >> 8); |
1699 | target[2] = static_cast<uint8_t>(part0 >> 16); |
1700 | target[3] = static_cast<uint8_t>(part0 >> 24); |
1701 | target[4] = static_cast<uint8_t>(part1); |
1702 | target[5] = static_cast<uint8_t>(part1 >> 8); |
1703 | target[6] = static_cast<uint8_t>(part1 >> 16); |
1704 | target[7] = static_cast<uint8_t>(part1 >> 24); |
1705 | #endif |
1706 | return target + sizeof(value); |
1707 | } |
1708 | |
1709 | inline void CodedOutputStream::WriteVarint32(uint32_t value) { |
1710 | cur_ = impl_.EnsureSpace(ptr: cur_); |
1711 | SetCur(WriteVarint32ToArray(value, target: Cur())); |
1712 | } |
1713 | |
1714 | inline void CodedOutputStream::WriteVarint64(uint64_t value) { |
1715 | cur_ = impl_.EnsureSpace(ptr: cur_); |
1716 | SetCur(WriteVarint64ToArray(value, target: Cur())); |
1717 | } |
1718 | |
1719 | inline void CodedOutputStream::WriteTag(uint32_t value) { |
1720 | WriteVarint32(value); |
1721 | } |
1722 | |
1723 | inline uint8_t* CodedOutputStream::WriteTagToArray(uint32_t value, |
1724 | uint8_t* target) { |
1725 | return WriteVarint32ToArray(value, target); |
1726 | } |
1727 | |
1728 | inline size_t CodedOutputStream::VarintSize32(uint32_t value) { |
1729 | // This computes value == 0 ? 1 : floor(log2(value)) / 7 + 1 |
1730 | // Use an explicit multiplication to implement the divide of |
1731 | // a number in the 1..31 range. |
1732 | // Explicit OR 0x1 to avoid calling Bits::Log2FloorNonZero(0), which is |
1733 | // undefined. |
1734 | uint32_t log2value = Bits::Log2FloorNonZero(n: value | 0x1); |
1735 | return static_cast<size_t>((log2value * 9 + 73) / 64); |
1736 | } |
1737 | |
1738 | inline size_t CodedOutputStream::VarintSize32PlusOne(uint32_t value) { |
1739 | // Same as above, but one more. |
1740 | uint32_t log2value = Bits::Log2FloorNonZero(n: value | 0x1); |
1741 | return static_cast<size_t>((log2value * 9 + 73 + 64) / 64); |
1742 | } |
1743 | |
1744 | inline size_t CodedOutputStream::VarintSize64(uint64_t value) { |
1745 | // This computes value == 0 ? 1 : floor(log2(value)) / 7 + 1 |
1746 | // Use an explicit multiplication to implement the divide of |
1747 | // a number in the 1..63 range. |
1748 | // Explicit OR 0x1 to avoid calling Bits::Log2FloorNonZero(0), which is |
1749 | // undefined. |
1750 | uint32_t log2value = Bits::Log2FloorNonZero64(n: value | 0x1); |
1751 | return static_cast<size_t>((log2value * 9 + 73) / 64); |
1752 | } |
1753 | |
1754 | inline size_t CodedOutputStream::VarintSize64PlusOne(uint64_t value) { |
1755 | // Same as above, but one more. |
1756 | uint32_t log2value = Bits::Log2FloorNonZero64(n: value | 0x1); |
1757 | return static_cast<size_t>((log2value * 9 + 73 + 64) / 64); |
1758 | } |
1759 | |
1760 | inline size_t CodedOutputStream::VarintSize32SignExtended(int32_t value) { |
1761 | return VarintSize64(value: static_cast<uint64_t>(int64_t{value})); |
1762 | } |
1763 | |
1764 | inline size_t CodedOutputStream::VarintSize32SignExtendedPlusOne( |
1765 | int32_t value) { |
1766 | return VarintSize64PlusOne(value: static_cast<uint64_t>(int64_t{value})); |
1767 | } |
1768 | |
1769 | inline void CodedOutputStream::WriteString(const std::string& str) { |
1770 | WriteRaw(buffer: str.data(), size: static_cast<int>(str.size())); |
1771 | } |
1772 | |
1773 | inline void CodedOutputStream::WriteRawMaybeAliased(const void* data, |
1774 | int size) { |
1775 | cur_ = impl_.WriteRawMaybeAliased(data, size, ptr: cur_); |
1776 | } |
1777 | |
1778 | inline uint8_t* CodedOutputStream::WriteRawToArray(const void* data, int size, |
1779 | uint8_t* target) { |
1780 | memcpy(dest: target, src: data, n: size); |
1781 | return target + size; |
1782 | } |
1783 | |
1784 | inline uint8_t* CodedOutputStream::WriteStringToArray(const std::string& str, |
1785 | uint8_t* target) { |
1786 | return WriteRawToArray(data: str.data(), size: static_cast<int>(str.size()), target); |
1787 | } |
1788 | |
1789 | } // namespace io |
1790 | } // namespace protobuf |
1791 | } // namespace google |
1792 | |
1793 | #if defined(_MSC_VER) && _MSC_VER >= 1300 && !defined(__INTEL_COMPILER) |
1794 | #pragma runtime_checks("c", restore) |
1795 | #endif // _MSC_VER && !defined(__INTEL_COMPILER) |
1796 | |
1797 | #include <google/protobuf/port_undef.inc> |
1798 | |
1799 | #endif // GOOGLE_PROTOBUF_IO_CODED_STREAM_H__ |
1800 | |