| 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: [email protected] (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 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 <string> |
| 120 |
#include <type_traits> |
| 121 |
#include <utility> |
| 122 |
|
| 123 |
#ifdef _MSC_VER |
| 124 |
// Assuming windows is always little-endian. |
| 125 |
#if !defined(PROTOBUF_DISABLE_LITTLE_ENDIAN_OPT_FOR_TEST) |
| 126 |
#define PROTOBUF_LITTLE_ENDIAN 1 |
| 127 |
#endif |
| 128 |
#if _MSC_VER >= 1300 && !defined(__INTEL_COMPILER) |
| 129 |
// If MSVC has "/RTCc" set, it will complain about truncating casts at |
| 130 |
// runtime. This file contains some intentional truncating casts. |
| 131 |
#pragma runtime_checks("c", off) |
| 132 |
#endif |
| 133 |
#else |
| 134 |
#include <sys/param.h> // __BYTE_ORDER |
| 135 |
#if ((defined(__LITTLE_ENDIAN__) && !defined(__BIG_ENDIAN__)) || \ |
| 136 |
(defined(__BYTE_ORDER) && __BYTE_ORDER == __LITTLE_ENDIAN)) && \ |
| 137 |
!defined(PROTOBUF_DISABLE_LITTLE_ENDIAN_OPT_FOR_TEST) |
| 138 |
#define PROTOBUF_LITTLE_ENDIAN 1 |
| 139 |
#endif |
| 140 |
#endif |
| 141 |
#include <google/protobuf/stubs/common.h> |
| 142 |
#include <google/protobuf/stubs/logging.h> |
| 143 |
#include <google/protobuf/stubs/strutil.h> |
| 144 |
#include <google/protobuf/port.h> |
| 145 |
#include <google/protobuf/stubs/port.h> |
| 146 |
|
| 147 |
|
| 148 |
#include <google/protobuf/port_def.inc> |
| 149 |
|
| 150 |
namespace google { |
| 151 |
namespace protobuf { |
| 152 |
|
| 153 |
class DescriptorPool; |
| 154 |
class MessageFactory; |
| 155 |
class ZeroCopyCodedInputStream; |
| 156 |
|
| 157 |
namespace internal { |
| 158 |
void MapTestForceDeterministic(); |
| 159 |
class EpsCopyByteStream; |
| 160 |
} // namespace internal |
| 161 |
|
| 162 |
namespace io { |
| 163 |
|
| 164 |
// Defined in this file. |
| 165 |
class CodedInputStream; |
| 166 |
class CodedOutputStream; |
| 167 |
|
| 168 |
// Defined in other files. |
| 169 |
class ZeroCopyInputStream; // zero_copy_stream.h |
| 170 |
class ZeroCopyOutputStream; // zero_copy_stream.h |
| 171 |
|
| 172 |
// Class which reads and decodes binary data which is composed of varint- |
| 173 |
// encoded integers and fixed-width pieces. Wraps a ZeroCopyInputStream. |
| 174 |
// Most users will not need to deal with CodedInputStream. |
| 175 |
// |
| 176 |
// Most methods of CodedInputStream that return a bool return false if an |
| 177 |
// underlying I/O error occurs or if the data is malformed. Once such a |
| 178 |
// failure occurs, the CodedInputStream is broken and is no longer useful. |
| 179 |
// After a failure, callers also should assume writes to "out" args may have |
| 180 |
// occurred, though nothing useful can be determined from those writes. |
| 181 |
class PROTOBUF_EXPORT CodedInputStream { |
| 182 |
public: |
| 183 |
// Create a CodedInputStream that reads from the given ZeroCopyInputStream. |
| 184 |
explicit CodedInputStream(ZeroCopyInputStream* input); |
| 185 |
|
| 186 |
// Create a CodedInputStream that reads from the given flat array. This is |
| 187 |
// faster than using an ArrayInputStream. PushLimit(size) is implied by |
| 188 |
// this constructor. |
| 189 |
explicit CodedInputStream(const uint8* buffer, int size); |
| 190 |
|
| 191 |
// Destroy the CodedInputStream and position the underlying |
| 192 |
// ZeroCopyInputStream at the first unread byte. If an error occurred while |
| 193 |
// reading (causing a method to return false), then the exact position of |
| 194 |
// the input stream may be anywhere between the last value that was read |
| 195 |
// successfully and the stream's byte limit. |
| 196 |
~CodedInputStream(); |
| 197 |
|
| 198 |
// Return true if this CodedInputStream reads from a flat array instead of |
| 199 |
// a ZeroCopyInputStream. |
| 200 |
inline bool IsFlat() const; |
| 201 |
|
| 202 |
// Skips a number of bytes. Returns false if an underlying read error |
| 203 |
// occurs. |
| 204 |
inline bool Skip(int count); |
| 205 |
|
| 206 |
// Sets *data to point directly at the unread part of the CodedInputStream's |
| 207 |
// underlying buffer, and *size to the size of that buffer, but does not |
| 208 |
// advance the stream's current position. This will always either produce |
| 209 |
// a non-empty buffer or return false. If the caller consumes any of |
| 210 |
// this data, it should then call Skip() to skip over the consumed bytes. |
| 211 |
// This may be useful for implementing external fast parsing routines for |
| 212 |
// types of data not covered by the CodedInputStream interface. |
| 213 |
bool GetDirectBufferPointer(const void** data, int* size); |
| 214 |
|
| 215 |
// Like GetDirectBufferPointer, but this method is inlined, and does not |
| 216 |
// attempt to Refresh() if the buffer is currently empty. |
| 217 |
PROTOBUF_ALWAYS_INLINE |
| 218 |
void GetDirectBufferPointerInline(const void** data, int* size); |
| 219 |
|
| 220 |
// Read raw bytes, copying them into the given buffer. |
| 221 |
bool ReadRaw(void* buffer, int size); |
| 222 |
|
| 223 |
// Like ReadRaw, but reads into a string. |
| 224 |
bool ReadString(std::string* buffer, int size); |
| 225 |
|
| 226 |
|
| 227 |
// Read a 32-bit little-endian integer. |
| 228 |
bool ReadLittleEndian32(uint32* value); |
| 229 |
// Read a 64-bit little-endian integer. |
| 230 |
bool ReadLittleEndian64(uint64* value); |
| 231 |
|
| 232 |
// These methods read from an externally provided buffer. The caller is |
| 233 |
// responsible for ensuring that the buffer has sufficient space. |
| 234 |
// Read a 32-bit little-endian integer. |
| 235 |
static const uint8* ReadLittleEndian32FromArray(const uint8* buffer, |
| 236 |
uint32* value); |
| 237 |
// Read a 64-bit little-endian integer. |
| 238 |
static const uint8* ReadLittleEndian64FromArray(const uint8* buffer, |
| 239 |
uint64* value); |
| 240 |
|
| 241 |
// Read an unsigned integer with Varint encoding, truncating to 32 bits. |
| 242 |
// Reading a 32-bit value is equivalent to reading a 64-bit one and casting |
| 243 |
// it to uint32, but may be more efficient. |
| 244 |
bool ReadVarint32(uint32* value); |
| 245 |
// Read an unsigned integer with Varint encoding. |
| 246 |
bool ReadVarint64(uint64* value); |
| 247 |
|
| 248 |
// Reads a varint off the wire into an "int". This should be used for reading |
| 249 |
// sizes off the wire (sizes of strings, submessages, bytes fields, etc). |
| 250 |
// |
| 251 |
// The value from the wire is interpreted as unsigned. If its value exceeds |
| 252 |
// the representable value of an integer on this platform, instead of |
| 253 |
// truncating we return false. Truncating (as performed by ReadVarint32() |
| 254 |
// above) is an acceptable approach for fields representing an integer, but |
| 255 |
// when we are parsing a size from the wire, truncating the value would result |
| 256 |
// in us misparsing the payload. |
| 257 |
bool ReadVarintSizeAsInt(int* value); |
| 258 |
|
| 259 |
// Read a tag. This calls ReadVarint32() and returns the result, or returns |
| 260 |
// zero (which is not a valid tag) if ReadVarint32() fails. Also, ReadTag |
| 261 |
// (but not ReadTagNoLastTag) updates the last tag value, which can be checked |
| 262 |
// with LastTagWas(). |
| 263 |
// |
| 264 |
// Always inline because this is only called in one place per parse loop |
| 265 |
// but it is called for every iteration of said loop, so it should be fast. |
| 266 |
// GCC doesn't want to inline this by default. |
| 267 |
PROTOBUF_ALWAYS_INLINE uint32 ReadTag() { |
| 268 |
return last_tag_ = ReadTagNoLastTag(); |
| 269 |
} |
| 270 |
|
| 271 |
PROTOBUF_ALWAYS_INLINE uint32 ReadTagNoLastTag(); |
| 272 |
|
| 273 |
// This usually a faster alternative to ReadTag() when cutoff is a manifest |
| 274 |
// constant. It does particularly well for cutoff >= 127. The first part |
| 275 |
// of the return value is the tag that was read, though it can also be 0 in |
| 276 |
// the cases where ReadTag() would return 0. If the second part is true |
| 277 |
// then the tag is known to be in [0, cutoff]. If not, the tag either is |
| 278 |
// above cutoff or is 0. (There's intentional wiggle room when tag is 0, |
| 279 |
// because that can arise in several ways, and for best performance we want |
| 280 |
// to avoid an extra "is tag == 0?" check here.) |
| 281 |
PROTOBUF_ALWAYS_INLINE |
| 282 |
std::pair<uint32, bool> ReadTagWithCutoff(uint32 cutoff) { |
| 283 |
std::pair<uint32, bool> result = ReadTagWithCutoffNoLastTag(cutoff); |
| 284 |
last_tag_ = result.first; |
| 285 |
return result; |
| 286 |
} |
| 287 |
|
| 288 |
PROTOBUF_ALWAYS_INLINE |
| 289 |
std::pair<uint32, bool> ReadTagWithCutoffNoLastTag(uint32 cutoff); |
| 290 |
|
| 291 |
// Usually returns true if calling ReadVarint32() now would produce the given |
| 292 |
// value. Will always return false if ReadVarint32() would not return the |
| 293 |
// given value. If ExpectTag() returns true, it also advances past |
| 294 |
// the varint. For best performance, use a compile-time constant as the |
| 295 |
// parameter. |
| 296 |
// Always inline because this collapses to a small number of instructions |
| 297 |
// when given a constant parameter, but GCC doesn't want to inline by default. |
| 298 |
PROTOBUF_ALWAYS_INLINE bool ExpectTag(uint32 expected); |
| 299 |
|
| 300 |
// Like above, except this reads from the specified buffer. The caller is |
| 301 |
// responsible for ensuring that the buffer is large enough to read a varint |
| 302 |
// of the expected size. For best performance, use a compile-time constant as |
| 303 |
// the expected tag parameter. |
| 304 |
// |
| 305 |
// Returns a pointer beyond the expected tag if it was found, or NULL if it |
| 306 |
// was not. |
| 307 |
PROTOBUF_ALWAYS_INLINE |
| 308 |
static const uint8* ExpectTagFromArray(const uint8* buffer, uint32 expected); |
| 309 |
|
| 310 |
// Usually returns true if no more bytes can be read. Always returns false |
| 311 |
// if more bytes can be read. If ExpectAtEnd() returns true, a subsequent |
| 312 |
// call to LastTagWas() will act as if ReadTag() had been called and returned |
| 313 |
// zero, and ConsumedEntireMessage() will return true. |
| 314 |
bool ExpectAtEnd(); |
| 315 |
|
| 316 |
// If the last call to ReadTag() or ReadTagWithCutoff() returned the given |
| 317 |
// value, returns true. Otherwise, returns false. |
| 318 |
// ReadTagNoLastTag/ReadTagWithCutoffNoLastTag do not preserve the last |
| 319 |
// returned value. |
| 320 |
// |
| 321 |
// This is needed because parsers for some types of embedded messages |
| 322 |
// (with field type TYPE_GROUP) don't actually know that they've reached the |
| 323 |
// end of a message until they see an ENDGROUP tag, which was actually part |
| 324 |
// of the enclosing message. The enclosing message would like to check that |
| 325 |
// tag to make sure it had the right number, so it calls LastTagWas() on |
| 326 |
// return from the embedded parser to check. |
| 327 |
bool LastTagWas(uint32 expected); |
| 328 |
void SetLastTag(uint32 tag) { last_tag_ = tag; } |
| 329 |
|
| 330 |
// When parsing message (but NOT a group), this method must be called |
| 331 |
// immediately after MergeFromCodedStream() returns (if it returns true) |
| 332 |
// to further verify that the message ended in a legitimate way. For |
| 333 |
// example, this verifies that parsing did not end on an end-group tag. |
| 334 |
// It also checks for some cases where, due to optimizations, |
| 335 |
// MergeFromCodedStream() can incorrectly return true. |
| 336 |
bool ConsumedEntireMessage(); |
| 337 |
void SetConsumed() { legitimate_message_end_ = true; } |
| 338 |
|
| 339 |
// Limits ---------------------------------------------------------- |
| 340 |
// Limits are used when parsing length-delimited embedded messages. |
| 341 |
// After the message's length is read, PushLimit() is used to prevent |
| 342 |
// the CodedInputStream from reading beyond that length. Once the |
| 343 |
// embedded message has been parsed, PopLimit() is called to undo the |
| 344 |
// limit. |
| 345 |
|
| 346 |
// Opaque type used with PushLimit() and PopLimit(). Do not modify |
| 347 |
// values of this type yourself. The only reason that this isn't a |
| 348 |
// struct with private internals is for efficiency. |
| 349 |
typedef int Limit; |
| 350 |
|
| 351 |
// Places a limit on the number of bytes that the stream may read, |
| 352 |
// starting from the current position. Once the stream hits this limit, |
| 353 |
// it will act like the end of the input has been reached until PopLimit() |
| 354 |
// is called. |
| 355 |
// |
| 356 |
// As the names imply, the stream conceptually has a stack of limits. The |
| 357 |
// shortest limit on the stack is always enforced, even if it is not the |
| 358 |
// top limit. |
| 359 |
// |
| 360 |
// The value returned by PushLimit() is opaque to the caller, and must |
| 361 |
// be passed unchanged to the corresponding call to PopLimit(). |
| 362 |
Limit PushLimit(int byte_limit); |
| 363 |
|
| 364 |
// Pops the last limit pushed by PushLimit(). The input must be the value |
| 365 |
// returned by that call to PushLimit(). |
| 366 |
void PopLimit(Limit limit); |
| 367 |
|
| 368 |
// Returns the number of bytes left until the nearest limit on the |
| 369 |
// stack is hit, or -1 if no limits are in place. |
| 370 |
int BytesUntilLimit() const; |
| 371 |
|
| 372 |
// Returns current position relative to the beginning of the input stream. |
| 373 |
int CurrentPosition() const; |
| 374 |
|
| 375 |
// Total Bytes Limit ----------------------------------------------- |
| 376 |
// To prevent malicious users from sending excessively large messages |
| 377 |
// and causing memory exhaustion, CodedInputStream imposes a hard limit on |
| 378 |
// the total number of bytes it will read. |
| 379 |
|
| 380 |
// Sets the maximum number of bytes that this CodedInputStream will read |
| 381 |
// before refusing to continue. To prevent servers from allocating enormous |
| 382 |
// amounts of memory to hold parsed messages, the maximum message length |
| 383 |
// should be limited to the shortest length that will not harm usability. |
| 384 |
// The default limit is INT_MAX (~2GB) and apps should set shorter limits |
| 385 |
// if possible. An error will always be printed to stderr if the limit is |
| 386 |
// reached. |
| 387 |
// |
| 388 |
// Note: setting a limit less than the current read position is interpreted |
| 389 |
// as a limit on the current position. |
| 390 |
// |
| 391 |
// This is unrelated to PushLimit()/PopLimit(). |
| 392 |
void SetTotalBytesLimit(int total_bytes_limit); |
| 393 |
|
| 394 |
PROTOBUF_DEPRECATED_MSG( |
| 395 |
"Please use the single parameter version of SetTotalBytesLimit(). The " |
| 396 |
"second parameter is ignored.") |
| 397 |
void SetTotalBytesLimit(int total_bytes_limit, int) { |
| 398 |
SetTotalBytesLimit(total_bytes_limit); |
| 399 |
} |
| 400 |
|
| 401 |
// The Total Bytes Limit minus the Current Position, or -1 if the total bytes |
| 402 |
// limit is INT_MAX. |
| 403 |
int BytesUntilTotalBytesLimit() const; |
| 404 |
|
| 405 |
// Recursion Limit ------------------------------------------------- |
| 406 |
// To prevent corrupt or malicious messages from causing stack overflows, |
| 407 |
// we must keep track of the depth of recursion when parsing embedded |
| 408 |
// messages and groups. CodedInputStream keeps track of this because it |
| 409 |
// is the only object that is passed down the stack during parsing. |
| 410 |
|
| 411 |
// Sets the maximum recursion depth. The default is 100. |
| 412 |
void SetRecursionLimit(int limit); |
| 413 |
int RecursionBudget() { return recursion_budget_; } |
| 414 |
|
| 415 |
static int GetDefaultRecursionLimit() { return default_recursion_limit_; } |
| 416 |
|
| 417 |
// Increments the current recursion depth. Returns true if the depth is |
| 418 |
// under the limit, false if it has gone over. |
| 419 |
bool IncrementRecursionDepth(); |
| 420 |
|
| 421 |
// Decrements the recursion depth if possible. |
| 422 |
void DecrementRecursionDepth(); |
| 423 |
|
| 424 |
// Decrements the recursion depth blindly. This is faster than |
| 425 |
// DecrementRecursionDepth(). It should be used only if all previous |
| 426 |
// increments to recursion depth were successful. |
| 427 |
void UnsafeDecrementRecursionDepth(); |
| 428 |
|
| 429 |
// Shorthand for make_pair(PushLimit(byte_limit), --recursion_budget_). |
| 430 |
// Using this can reduce code size and complexity in some cases. The caller |
| 431 |
// is expected to check that the second part of the result is non-negative (to |
| 432 |
// bail out if the depth of recursion is too high) and, if all is well, to |
| 433 |
// later pass the first part of the result to PopLimit() or similar. |
| 434 |
std::pair<CodedInputStream::Limit, int> IncrementRecursionDepthAndPushLimit( |
| 435 |
int byte_limit); |
| 436 |
|
| 437 |
// Shorthand for PushLimit(ReadVarint32(&length) ? length : 0). |
| 438 |
Limit ReadLengthAndPushLimit(); |
| 439 |
|
| 440 |
// Helper that is equivalent to: { |
| 441 |
// bool result = ConsumedEntireMessage(); |
| 442 |
// PopLimit(limit); |
| 443 |
// UnsafeDecrementRecursionDepth(); |
| 444 |
// return result; } |
| 445 |
// Using this can reduce code size and complexity in some cases. |
| 446 |
// Do not use unless the current recursion depth is greater than zero. |
| 447 |
bool DecrementRecursionDepthAndPopLimit(Limit limit); |
| 448 |
|
| 449 |
// Helper that is equivalent to: { |
| 450 |
// bool result = ConsumedEntireMessage(); |
| 451 |
// PopLimit(limit); |
| 452 |
// return result; } |
| 453 |
// Using this can reduce code size and complexity in some cases. |
| 454 |
bool CheckEntireMessageConsumedAndPopLimit(Limit limit); |
| 455 |
|
| 456 |
// Extension Registry ---------------------------------------------- |
| 457 |
// ADVANCED USAGE: 99.9% of people can ignore this section. |
| 458 |
// |
| 459 |
// By default, when parsing extensions, the parser looks for extension |
| 460 |
// definitions in the pool which owns the outer message's Descriptor. |
| 461 |
// However, you may call SetExtensionRegistry() to provide an alternative |
| 462 |
// pool instead. This makes it possible, for example, to parse a message |
| 463 |
// using a generated class, but represent some extensions using |
| 464 |
// DynamicMessage. |
| 465 |
|
| 466 |
// Set the pool used to look up extensions. Most users do not need to call |
| 467 |
// this as the correct pool will be chosen automatically. |
| 468 |
// |
| 469 |
// WARNING: It is very easy to misuse this. Carefully read the requirements |
| 470 |
// below. Do not use this unless you are sure you need it. Almost no one |
| 471 |
// does. |
| 472 |
// |
| 473 |
// Let's say you are parsing a message into message object m, and you want |
| 474 |
// to take advantage of SetExtensionRegistry(). You must follow these |
| 475 |
// requirements: |
| 476 |
// |
| 477 |
// The given DescriptorPool must contain m->GetDescriptor(). It is not |
| 478 |
// sufficient for it to simply contain a descriptor that has the same name |
| 479 |
// and content -- it must be the *exact object*. In other words: |
| 480 |
// assert(pool->FindMessageTypeByName(m->GetDescriptor()->full_name()) == |
| 481 |
// m->GetDescriptor()); |
| 482 |
// There are two ways to satisfy this requirement: |
| 483 |
// 1) Use m->GetDescriptor()->pool() as the pool. This is generally useless |
| 484 |
// because this is the pool that would be used anyway if you didn't call |
| 485 |
// SetExtensionRegistry() at all. |
| 486 |
// 2) Use a DescriptorPool which has m->GetDescriptor()->pool() as an |
| 487 |
// "underlay". Read the documentation for DescriptorPool for more |
| 488 |
// information about underlays. |
| 489 |
// |
| 490 |
// You must also provide a MessageFactory. This factory will be used to |
| 491 |
// construct Message objects representing extensions. The factory's |
| 492 |
// GetPrototype() MUST return non-NULL for any Descriptor which can be found |
| 493 |
// through the provided pool. |
| 494 |
// |
| 495 |
// If the provided factory might return instances of protocol-compiler- |
| 496 |
// generated (i.e. compiled-in) types, or if the outer message object m is |
| 497 |
// a generated type, then the given factory MUST have this property: If |
| 498 |
// GetPrototype() is given a Descriptor which resides in |
| 499 |
// DescriptorPool::generated_pool(), the factory MUST return the same |
| 500 |
// prototype which MessageFactory::generated_factory() would return. That |
| 501 |
// is, given a descriptor for a generated type, the factory must return an |
| 502 |
// instance of the generated class (NOT DynamicMessage). However, when |
| 503 |
// given a descriptor for a type that is NOT in generated_pool, the factory |
| 504 |
// is free to return any implementation. |
| 505 |
// |
| 506 |
// The reason for this requirement is that generated sub-objects may be |
| 507 |
// accessed via the standard (non-reflection) extension accessor methods, |
| 508 |
// and these methods will down-cast the object to the generated class type. |
| 509 |
// If the object is not actually of that type, the results would be undefined. |
| 510 |
// On the other hand, if an extension is not compiled in, then there is no |
| 511 |
// way the code could end up accessing it via the standard accessors -- the |
| 512 |
// only way to access the extension is via reflection. When using reflection, |
| 513 |
// DynamicMessage and generated messages are indistinguishable, so it's fine |
| 514 |
// if these objects are represented using DynamicMessage. |
| 515 |
// |
| 516 |
// Using DynamicMessageFactory on which you have called |
| 517 |
// SetDelegateToGeneratedFactory(true) should be sufficient to satisfy the |
| 518 |
// above requirement. |
| 519 |
// |
| 520 |
// If either pool or factory is NULL, both must be NULL. |
| 521 |
// |
| 522 |
// Note that this feature is ignored when parsing "lite" messages as they do |
| 523 |
// not have descriptors. |
| 524 |
void SetExtensionRegistry(const DescriptorPool* pool, |
| 525 |
MessageFactory* factory); |
| 526 |
|
| 527 |
// Get the DescriptorPool set via SetExtensionRegistry(), or NULL if no pool |
| 528 |
// has been provided. |
| 529 |
const DescriptorPool* GetExtensionPool(); |
| 530 |
|
| 531 |
// Get the MessageFactory set via SetExtensionRegistry(), or NULL if no |
| 532 |
// factory has been provided. |
| 533 |
MessageFactory* GetExtensionFactory(); |
| 534 |
|
| 535 |
private: |
| 536 |
GOOGLE_DISALLOW_EVIL_CONSTRUCTORS(CodedInputStream); |
| 537 |
|
| 538 |
const uint8* buffer_; |
| 539 |
const uint8* buffer_end_; // pointer to the end of the buffer. |
| 540 |
ZeroCopyInputStream* input_; |
| 541 |
int total_bytes_read_; // total bytes read from input_, including |
| 542 |
// the current buffer |
| 543 |
|
| 544 |
// If total_bytes_read_ surpasses INT_MAX, we record the extra bytes here |
| 545 |
// so that we can BackUp() on destruction. |
| 546 |
int overflow_bytes_; |
| 547 |
|
| 548 |
// LastTagWas() stuff. |
| 549 |
uint32 last_tag_; // result of last ReadTag() or ReadTagWithCutoff(). |
| 550 |
|
| 551 |
// This is set true by ReadTag{Fallback/Slow}() if it is called when exactly |
| 552 |
// at EOF, or by ExpectAtEnd() when it returns true. This happens when we |
| 553 |
// reach the end of a message and attempt to read another tag. |
| 554 |
bool legitimate_message_end_; |
| 555 |
|
| 556 |
// See EnableAliasing(). |
| 557 |
bool aliasing_enabled_; |
| 558 |
|
| 559 |
// Limits |
| 560 |
Limit current_limit_; // if position = -1, no limit is applied |
| 561 |
|
| 562 |
// For simplicity, if the current buffer crosses a limit (either a normal |
| 563 |
// limit created by PushLimit() or the total bytes limit), buffer_size_ |
| 564 |
// only tracks the number of bytes before that limit. This field |
| 565 |
// contains the number of bytes after it. Note that this implies that if |
| 566 |
// buffer_size_ == 0 and buffer_size_after_limit_ > 0, we know we've |
| 567 |
// hit a limit. However, if both are zero, it doesn't necessarily mean |
| 568 |
// we aren't at a limit -- the buffer may have ended exactly at the limit. |
| 569 |
int buffer_size_after_limit_; |
| 570 |
|
| 571 |
// Maximum number of bytes to read, period. This is unrelated to |
| 572 |
// current_limit_. Set using SetTotalBytesLimit(). |
| 573 |
int total_bytes_limit_; |
| 574 |
|
| 575 |
// Current recursion budget, controlled by IncrementRecursionDepth() and |
| 576 |
// similar. Starts at recursion_limit_ and goes down: if this reaches |
| 577 |
// -1 we are over budget. |
| 578 |
int recursion_budget_; |
| 579 |
// Recursion depth limit, set by SetRecursionLimit(). |
| 580 |
int recursion_limit_; |
| 581 |
|
| 582 |
// See SetExtensionRegistry(). |
| 583 |
const DescriptorPool* extension_pool_; |
| 584 |
MessageFactory* extension_factory_; |
| 585 |
|
| 586 |
// Private member functions. |
| 587 |
|
| 588 |
// Fallback when Skip() goes past the end of the current buffer. |
| 589 |
bool SkipFallback(int count, int original_buffer_size); |
| 590 |
|
| 591 |
// Advance the buffer by a given number of bytes. |
| 592 |
void Advance(int amount); |
| 593 |
|
| 594 |
// Back up input_ to the current buffer position. |
| 595 |
void BackUpInputToCurrentPosition(); |
| 596 |
|
| 597 |
// Recomputes the value of buffer_size_after_limit_. Must be called after |
| 598 |
// current_limit_ or total_bytes_limit_ changes. |
| 599 |
void RecomputeBufferLimits(); |
| 600 |
|
| 601 |
// Writes an error message saying that we hit total_bytes_limit_. |
| 602 |
void PrintTotalBytesLimitError(); |
| 603 |
|
| 604 |
// Called when the buffer runs out to request more data. Implies an |
| 605 |
// Advance(BufferSize()). |
| 606 |
bool Refresh(); |
| 607 |
|
| 608 |
// When parsing varints, we optimize for the common case of small values, and |
| 609 |
// then optimize for the case when the varint fits within the current buffer |
| 610 |
// piece. The Fallback method is used when we can't use the one-byte |
| 611 |
// optimization. The Slow method is yet another fallback when the buffer is |
| 612 |
// not large enough. Making the slow path out-of-line speeds up the common |
| 613 |
// case by 10-15%. The slow path is fairly uncommon: it only triggers when a |
| 614 |
// message crosses multiple buffers. Note: ReadVarint32Fallback() and |
| 615 |
// ReadVarint64Fallback() are called frequently and generally not inlined, so |
| 616 |
// they have been optimized to avoid "out" parameters. The former returns -1 |
| 617 |
// if it fails and the uint32 it read otherwise. The latter has a bool |
| 618 |
// indicating success or failure as part of its return type. |
| 619 |
int64 ReadVarint32Fallback(uint32 first_byte_or_zero); |
| 620 |
int ReadVarintSizeAsIntFallback(); |
| 621 |
std::pair<uint64, bool> ReadVarint64Fallback(); |
| 622 |
bool ReadVarint32Slow(uint32* value); |
| 623 |
bool ReadVarint64Slow(uint64* value); |
| 624 |
int ReadVarintSizeAsIntSlow(); |
| 625 |
bool ReadLittleEndian32Fallback(uint32* value); |
| 626 |
bool ReadLittleEndian64Fallback(uint64* value); |
| 627 |
|
| 628 |
// Fallback/slow methods for reading tags. These do not update last_tag_, |
| 629 |
// but will set legitimate_message_end_ if we are at the end of the input |
| 630 |
// stream. |
| 631 |
uint32 ReadTagFallback(uint32 first_byte_or_zero); |
| 632 |
uint32 ReadTagSlow(); |
| 633 |
bool ReadStringFallback(std::string* buffer, int size); |
| 634 |
|
| 635 |
// Return the size of the buffer. |
| 636 |
int BufferSize() const; |
| 637 |
|
| 638 |
static const int kDefaultTotalBytesLimit = INT_MAX; |
| 639 |
|
| 640 |
static int default_recursion_limit_; // 100 by default. |
| 641 |
|
| 642 |
friend class google::protobuf::ZeroCopyCodedInputStream; |
| 643 |
friend class google::protobuf::internal::EpsCopyByteStream; |
| 644 |
}; |
| 645 |
|
| 646 |
// EpsCopyOutputStream wraps a ZeroCopyOutputStream and exposes a new stream, |
| 647 |
// which has the property you can write kSlopBytes (16 bytes) from the current |
| 648 |
// position without bounds checks. The cursor into the stream is managed by |
| 649 |
// the user of the class and is an explicit parameter in the methods. Careful |
| 650 |
// use of this class, ie. keep ptr a local variable, eliminates the need to |
| 651 |
// for the compiler to sync the ptr value between register and memory. |
| 652 |
class PROTOBUF_EXPORT EpsCopyOutputStream { |
| 653 |
public: |
| 654 |
enum { kSlopBytes = 16 }; |
| 655 |
|
| 656 |
// Initialize from a stream. |
| 657 |
EpsCopyOutputStream(ZeroCopyOutputStream* stream, bool deterministic, |
| 658 |
uint8** pp) |
| 659 |
: end_(buffer_), |
| 660 |
stream_(stream), |
| 661 |
is_serialization_deterministic_(deterministic) { |
| 662 |
*pp = buffer_; |
| 663 |
} |
| 664 |
|
| 665 |
// Only for array serialization. No overflow protection, end_ will be the |
| 666 |
// pointed to the end of the array. When using this the total size is already |
| 667 |
// known, so no need to maintain the slop region. |
| 668 |
EpsCopyOutputStream(void* data, int size, bool deterministic) |
| 669 |
: end_(static_cast<uint8*>(data) + size), |
| 670 |
buffer_end_(nullptr), |
| 671 |
stream_(nullptr), |
| 672 |
is_serialization_deterministic_(deterministic) {} |
| 673 |
|
| 674 |
// Initialize from stream but with the first buffer already given (eager). |
| 675 |
EpsCopyOutputStream(void* data, int size, ZeroCopyOutputStream* stream, |
| 676 |
bool deterministic, uint8** pp) |
| 677 |
: stream_(stream), is_serialization_deterministic_(deterministic) { |
| 678 |
*pp = SetInitialBuffer(data, size); |
| 679 |
} |
| 680 |
|
| 681 |
// Flush everything that's written into the underlying ZeroCopyOutputStream |
| 682 |
// and trims the underlying stream to the location of ptr. |
| 683 |
uint8* Trim(uint8* ptr); |
| 684 |
|
| 685 |
// After this it's guaranteed you can safely write kSlopBytes to ptr. This |
| 686 |
// will never fail! The underlying stream can produce an error. Use HadError |
| 687 |
// to check for errors. |
| 688 |
PROTOBUF_MUST_USE_RESULT uint8* EnsureSpace(uint8* ptr) { |
| 689 |
if (PROTOBUF_PREDICT_FALSE(ptr >= end_)) { |
| 690 |
return EnsureSpaceFallback(ptr); |
| 691 |
} |
| 692 |
return ptr; |
| 693 |
} |
| 694 |
|
| 695 |
uint8* WriteRaw(const void* data, int size, uint8* ptr) { |
| 696 |
if (PROTOBUF_PREDICT_FALSE(end_ - ptr < size)) { |
| 697 |
return WriteRawFallback(data, size, ptr); |
| 698 |
} |
| 699 |
std::memcpy(ptr, data, size); |
| 700 |
return ptr + size; |
| 701 |
} |
| 702 |
// Writes the buffer specified by data, size to the stream. Possibly by |
| 703 |
// aliasing the buffer (ie. not copying the data). The caller is responsible |
| 704 |
// to make sure the buffer is alive for the duration of the |
| 705 |
// ZeroCopyOutputStream. |
| 706 |
uint8* WriteRawMaybeAliased(const void* data, int size, uint8* ptr) { |
| 707 |
if (aliasing_enabled_) { |
| 708 |
return WriteAliasedRaw(data, size, ptr); |
| 709 |
} else { |
| 710 |
return WriteRaw(data, size, ptr); |
| 711 |
} |
| 712 |
} |
| 713 |
|
| 714 |
|
| 715 |
uint8* WriteStringMaybeAliased(uint32 num, const std::string& s, uint8* ptr) { |
| 716 |
std::ptrdiff_t size = s.size(); |
| 717 |
if (PROTOBUF_PREDICT_FALSE( |
| 718 |
size >= 128 || end_ - ptr + 16 - TagSize(num << 3) - 1 < size)) { |
| 719 |
return WriteStringMaybeAliasedOutline(num, s, ptr); |
| 720 |
} |
| 721 |
ptr = UnsafeVarint((num << 3) | 2, ptr); |
| 722 |
*ptr++ = static_cast<uint8>(size); |
| 723 |
std::memcpy(ptr, s.data(), size); |
| 724 |
return ptr + size; |
| 725 |
} |
| 726 |
uint8* WriteBytesMaybeAliased(uint32 num, const std::string& s, uint8* ptr) { |
| 727 |
return WriteStringMaybeAliased(num, s, ptr); |
| 728 |
} |
| 729 |
|
| 730 |
template <typename T> |
| 731 |
PROTOBUF_ALWAYS_INLINE uint8* WriteString(uint32 num, const T& s, |
| 732 |
uint8* ptr) { |
| 733 |
std::ptrdiff_t size = s.size(); |
| 734 |
if (PROTOBUF_PREDICT_FALSE( |
| 735 |
size >= 128 || end_ - ptr + 16 - TagSize(num << 3) - 1 < size)) { |
| 736 |
return WriteStringOutline(num, s, ptr); |
| 737 |
} |
| 738 |
ptr = UnsafeVarint((num << 3) | 2, ptr); |
| 739 |
*ptr++ = static_cast<uint8>(size); |
| 740 |
std::memcpy(ptr, s.data(), size); |
| 741 |
return ptr + size; |
| 742 |
} |
| 743 |
template <typename T> |
| 744 |
uint8* WriteBytes(uint32 num, const T& s, uint8* ptr) { |
| 745 |
return WriteString(num, s, ptr); |
| 746 |
} |
| 747 |
|
| 748 |
template <typename T> |
| 749 |
PROTOBUF_ALWAYS_INLINE uint8* WriteInt32Packed(int num, const T& r, int size, |
| 750 |
uint8* ptr) { |
| 751 |
return WriteVarintPacked(num, r, size, ptr, Encode64); |
| 752 |
} |
| 753 |
template <typename T> |
| 754 |
PROTOBUF_ALWAYS_INLINE uint8* WriteUInt32Packed(int num, const T& r, int size, |
| 755 |
uint8* ptr) { |
| 756 |
return WriteVarintPacked(num, r, size, ptr, Encode32); |
| 757 |
} |
| 758 |
template <typename T> |
| 759 |
PROTOBUF_ALWAYS_INLINE uint8* WriteSInt32Packed(int num, const T& r, int size, |
| 760 |
uint8* ptr) { |
| 761 |
return WriteVarintPacked(num, r, size, ptr, ZigZagEncode32); |
| 762 |
} |
| 763 |
template <typename T> |
| 764 |
PROTOBUF_ALWAYS_INLINE uint8* WriteInt64Packed(int num, const T& r, int size, |
| 765 |
uint8* ptr) { |
| 766 |
return WriteVarintPacked(num, r, size, ptr, Encode64); |
| 767 |
} |
| 768 |
template <typename T> |
| 769 |
PROTOBUF_ALWAYS_INLINE uint8* WriteUInt64Packed(int num, const T& r, int size, |
| 770 |
uint8* ptr) { |
| 771 |
return WriteVarintPacked(num, r, size, ptr, Encode64); |
| 772 |
} |
| 773 |
template <typename T> |
| 774 |
PROTOBUF_ALWAYS_INLINE uint8* WriteSInt64Packed(int num, const T& r, int size, |
| 775 |
uint8* ptr) { |
| 776 |
return WriteVarintPacked(num, r, size, ptr, ZigZagEncode64); |
| 777 |
} |
| 778 |
template <typename T> |
| 779 |
PROTOBUF_ALWAYS_INLINE uint8* WriteEnumPacked(int num, const T& r, int size, |
| 780 |
uint8* ptr) { |
| 781 |
return WriteVarintPacked(num, r, size, ptr, Encode64); |
| 782 |
} |
| 783 |
|
| 784 |
template <typename T> |
| 785 |
PROTOBUF_ALWAYS_INLINE uint8* WriteFixedPacked(int num, const T& r, |
| 786 |
uint8* ptr) { |
| 787 |
ptr = EnsureSpace(ptr); |
| 788 |
constexpr auto element_size = sizeof(typename T::value_type); |
| 789 |
auto size = r.size() * element_size; |
| 790 |
ptr = WriteLengthDelim(num, size, ptr); |
| 791 |
return WriteRawLittleEndian<element_size>(r.data(), static_cast<int>(size), |
| 792 |
ptr); |
| 793 |
} |
| 794 |
|
| 795 |
// Returns true if there was an underlying I/O error since this object was |
| 796 |
// created. |
| 797 |
bool HadError() const { return had_error_; } |
| 798 |
|
| 799 |
// Instructs the EpsCopyOutputStream to allow the underlying |
| 800 |
// ZeroCopyOutputStream to hold pointers to the original structure instead of |
| 801 |
// copying, if it supports it (i.e. output->AllowsAliasing() is true). If the |
| 802 |
// underlying stream does not support aliasing, then enabling it has no |
| 803 |
// affect. For now, this only affects the behavior of |
| 804 |
// WriteRawMaybeAliased(). |
| 805 |
// |
| 806 |
// NOTE: It is caller's responsibility to ensure that the chunk of memory |
| 807 |
// remains live until all of the data has been consumed from the stream. |
| 808 |
void EnableAliasing(bool enabled); |
| 809 |
|
| 810 |
// See documentation on CodedOutputStream::SetSerializationDeterministic. |
| 811 |
void SetSerializationDeterministic(bool value) { |
| 812 |
is_serialization_deterministic_ = value; |
| 813 |
} |
| 814 |
|
| 815 |
// See documentation on CodedOutputStream::IsSerializationDeterministic. |
| 816 |
bool IsSerializationDeterministic() const { |
| 817 |
return is_serialization_deterministic_; |
| 818 |
} |
| 819 |
|
| 820 |
// The number of bytes written to the stream at position ptr, relative to the |
| 821 |
// stream's overall position. |
| 822 |
int64 ByteCount(uint8* ptr) const; |
| 823 |
|
| 824 |
|
| 825 |
private: |
| 826 |
uint8* end_; |
| 827 |
uint8* buffer_end_ = buffer_; |
| 828 |
uint8 buffer_[2 * kSlopBytes]; |
| 829 |
ZeroCopyOutputStream* stream_; |
| 830 |
bool had_error_ = false; |
| 831 |
bool aliasing_enabled_ = false; // See EnableAliasing(). |
| 832 |
bool is_serialization_deterministic_; |
| 833 |
|
| 834 |
uint8* EnsureSpaceFallback(uint8* ptr); |
| 835 |
inline uint8* Next(); |
| 836 |
int Flush(uint8* ptr); |
| 837 |
std::ptrdiff_t GetSize(uint8* ptr) const { |
| 838 |
GOOGLE_DCHECK(ptr <= end_ + kSlopBytes); // NOLINT |
| 839 |
return end_ + kSlopBytes - ptr; |
| 840 |
} |
| 841 |
|
| 842 |
uint8* Error() { |
| 843 |
had_error_ = true; |
| 844 |
// We use the patch buffer to always guarantee space to write to. |
| 845 |
end_ = buffer_ + kSlopBytes; |
| 846 |
return buffer_; |
| 847 |
} |
| 848 |
|
| 849 |
static constexpr int TagSize(uint32 tag) { |
| 850 |
return (tag < (1 << 7)) |
| 851 |
? 1 |
| 852 |
: (tag < (1 << 14)) |
| 853 |
? 2 |
| 854 |
: (tag < (1 << 21)) ? 3 : (tag < (1 << 28)) ? 4 : 5; |
| 855 |
} |
| 856 |
|
| 857 |
PROTOBUF_ALWAYS_INLINE uint8* WriteTag(uint32 num, uint32 wt, uint8* ptr) { |
| 858 |
GOOGLE_DCHECK(ptr < end_); // NOLINT |
| 859 |
return UnsafeVarint((num << 3) | wt, ptr); |
| 860 |
} |
| 861 |
|
| 862 |
PROTOBUF_ALWAYS_INLINE uint8* WriteLengthDelim(int num, uint32 size, |
| 863 |
uint8* ptr) { |
| 864 |
ptr = WriteTag(num, 2, ptr); |
| 865 |
return UnsafeWriteSize(size, ptr); |
| 866 |
} |
| 867 |
|
| 868 |
uint8* WriteRawFallback(const void* data, int size, uint8* ptr); |
| 869 |
|
| 870 |
uint8* WriteAliasedRaw(const void* data, int size, uint8* ptr); |
| 871 |
|
| 872 |
uint8* WriteStringMaybeAliasedOutline(uint32 num, const std::string& s, |
| 873 |
uint8* ptr); |
| 874 |
uint8* WriteStringOutline(uint32 num, const std::string& s, uint8* ptr); |
| 875 |
|
| 876 |
template <typename T, typename E> |
| 877 |
PROTOBUF_ALWAYS_INLINE uint8* WriteVarintPacked(int num, const T& r, int size, |
| 878 |
uint8* ptr, const E& encode) { |
| 879 |
ptr = EnsureSpace(ptr); |
| 880 |
ptr = WriteLengthDelim(num, size, ptr); |
| 881 |
auto it = r.data(); |
| 882 |
auto end = it + r.size(); |
| 883 |
do { |
| 884 |
ptr = EnsureSpace(ptr); |
| 885 |
ptr = UnsafeVarint(encode(*it++), ptr); |
| 886 |
} while (it < end); |
| 887 |
return ptr; |
| 888 |
} |
| 889 |
|
| 890 |
static uint32 Encode32(uint32 v) { return v; } |
| 891 |
static uint64 Encode64(uint64 v) { return v; } |
| 892 |
static uint32 ZigZagEncode32(int32 v) { |
| 893 |
return (static_cast<uint32>(v) << 1) ^ static_cast<uint32>(v >> 31); |
| 894 |
} |
| 895 |
static uint64 ZigZagEncode64(int64 v) { |
| 896 |
return (static_cast<uint64>(v) << 1) ^ static_cast<uint64>(v >> 63); |
| 897 |
} |
| 898 |
|
| 899 |
template <typename T> |
| 900 |
PROTOBUF_ALWAYS_INLINE static uint8* UnsafeVarint(T value, uint8* ptr) { |
| 901 |
static_assert(std::is_unsigned<T>::value, |
| 902 |
"Varint serialization must be unsigned"); |
| 903 |
if (value < 0x80) { |
| 904 |
ptr[0] = static_cast<uint8>(value); |
| 905 |
return ptr + 1; |
| 906 |
} |
| 907 |
ptr[0] = static_cast<uint8>(value | 0x80); |
| 908 |
value >>= 7; |
| 909 |
if (value < 0x80) { |
| 910 |
ptr[1] = static_cast<uint8>(value); |
| 911 |
return ptr + 2; |
| 912 |
} |
| 913 |
ptr++; |
| 914 |
do { |
| 915 |
*ptr = static_cast<uint8>(value | 0x80); |
| 916 |
value >>= 7; |
| 917 |
++ptr; |
| 918 |
} while (PROTOBUF_PREDICT_FALSE(value >= 0x80)); |
| 919 |
*ptr++ = static_cast<uint8>(value); |
| 920 |
return ptr; |
| 921 |
} |
| 922 |
|
| 923 |
PROTOBUF_ALWAYS_INLINE static uint8* UnsafeWriteSize(uint32 value, |
| 924 |
uint8* ptr) { |
| 925 |
while (PROTOBUF_PREDICT_FALSE(value >= 0x80)) { |
| 926 |
*ptr = static_cast<uint8>(value | 0x80); |
| 927 |
value >>= 7; |
| 928 |
++ptr; |
| 929 |
} |
| 930 |
*ptr++ = static_cast<uint8>(value); |
| 931 |
return ptr; |
| 932 |
} |
| 933 |
|
| 934 |
template <int S> |
| 935 |
uint8* WriteRawLittleEndian(const void* data, int size, uint8* ptr); |
| 936 |
#ifndef PROTOBUF_LITTLE_ENDIAN |
| 937 |
uint8* WriteRawLittleEndian32(const void* data, int size, uint8* ptr); |
| 938 |
uint8* WriteRawLittleEndian64(const void* data, int size, uint8* ptr); |
| 939 |
#endif |
| 940 |
|
| 941 |
// These methods are for CodedOutputStream. Ideally they should be private |
| 942 |
// but to match current behavior of CodedOutputStream as close as possible |
| 943 |
// we allow it some functionality. |
| 944 |
public: |
| 945 |
uint8* SetInitialBuffer(void* data, int size) { |
| 946 |
auto ptr = static_cast<uint8*>(data); |
| 947 |
if (size > kSlopBytes) { |
| 948 |
end_ = ptr + size - kSlopBytes; |
| 949 |
buffer_end_ = nullptr; |
| 950 |
return ptr; |
| 951 |
} else { |
| 952 |
end_ = buffer_ + size; |
| 953 |
buffer_end_ = ptr; |
| 954 |
return buffer_; |
| 955 |
} |
| 956 |
} |
| 957 |
|
| 958 |
private: |
| 959 |
// Needed by CodedOutputStream HadError. HadError needs to flush the patch |
| 960 |
// buffers to ensure there is no error as of yet. |
| 961 |
uint8* FlushAndResetBuffer(uint8*); |
| 962 |
|
| 963 |
// The following functions mimic the old CodedOutputStream behavior as close |
| 964 |
// as possible. They flush the current state to the stream, behave as |
| 965 |
// the old CodedOutputStream and then return to normal operation. |
| 966 |
bool Skip(int count, uint8** pp); |
| 967 |
bool GetDirectBufferPointer(void** data, int* size, uint8** pp); |
| 968 |
uint8* GetDirectBufferForNBytesAndAdvance(int size, uint8** pp); |
| 969 |
|
| 970 |
friend class CodedOutputStream; |
| 971 |
}; |
| 972 |
|
| 973 |
template <> |
| 974 |
inline uint8* EpsCopyOutputStream::WriteRawLittleEndian<1>(const void* data, |
| 975 |
int size, |
| 976 |
uint8* ptr) { |
| 977 |
return WriteRaw(data, size, ptr); |
| 978 |
} |
| 979 |
template <> |
| 980 |
inline uint8* EpsCopyOutputStream::WriteRawLittleEndian<4>(const void* data, |
| 981 |
int size, |
| 982 |
uint8* ptr) { |
| 983 |
#ifdef PROTOBUF_LITTLE_ENDIAN |
| 984 |
return WriteRaw(data, size, ptr); |
| 985 |
#else |
| 986 |
return WriteRawLittleEndian32(data, size, ptr); |
| 987 |
#endif |
| 988 |
} |
| 989 |
template <> |
| 990 |
inline uint8* EpsCopyOutputStream::WriteRawLittleEndian<8>(const void* data, |
| 991 |
int size, |
| 992 |
uint8* ptr) { |
| 993 |
#ifdef PROTOBUF_LITTLE_ENDIAN |
| 994 |
return WriteRaw(data, size, ptr); |
| 995 |
#else |
| 996 |
return WriteRawLittleEndian64(data, size, ptr); |
| 997 |
#endif |
| 998 |
} |
| 999 |
|
| 1000 |
// Class which encodes and writes binary data which is composed of varint- |
| 1001 |
// encoded integers and fixed-width pieces. Wraps a ZeroCopyOutputStream. |
| 1002 |
// Most users will not need to deal with CodedOutputStream. |
| 1003 |
// |
| 1004 |
// Most methods of CodedOutputStream which return a bool return false if an |
| 1005 |
// underlying I/O error occurs. Once such a failure occurs, the |
| 1006 |
// CodedOutputStream is broken and is no longer useful. The Write* methods do |
| 1007 |
// not return the stream status, but will invalidate the stream if an error |
| 1008 |
// occurs. The client can probe HadError() to determine the status. |
| 1009 |
// |
| 1010 |
// Note that every method of CodedOutputStream which writes some data has |
| 1011 |
// a corresponding static "ToArray" version. These versions write directly |
| 1012 |
// to the provided buffer, returning a pointer past the last written byte. |
| 1013 |
// They require that the buffer has sufficient capacity for the encoded data. |
| 1014 |
// This allows an optimization where we check if an output stream has enough |
| 1015 |
// space for an entire message before we start writing and, if there is, we |
| 1016 |
// call only the ToArray methods to avoid doing bound checks for each |
| 1017 |
// individual value. |
| 1018 |
// i.e., in the example above: |
| 1019 |
// |
| 1020 |
// CodedOutputStream* coded_output = new CodedOutputStream(raw_output); |
| 1021 |
// int magic_number = 1234; |
| 1022 |
// char text[] = "Hello world!"; |
| 1023 |
// |
| 1024 |
// int coded_size = sizeof(magic_number) + |
| 1025 |
// CodedOutputStream::VarintSize32(strlen(text)) + |
| 1026 |
// strlen(text); |
| 1027 |
// |
| 1028 |
// uint8* buffer = |
| 1029 |
// coded_output->GetDirectBufferForNBytesAndAdvance(coded_size); |
| 1030 |
// if (buffer != nullptr) { |
| 1031 |
// // The output stream has enough space in the buffer: write directly to |
| 1032 |
// // the array. |
| 1033 |
// buffer = CodedOutputStream::WriteLittleEndian32ToArray(magic_number, |
| 1034 |
// buffer); |
| 1035 |
// buffer = CodedOutputStream::WriteVarint32ToArray(strlen(text), buffer); |
| 1036 |
// buffer = CodedOutputStream::WriteRawToArray(text, strlen(text), buffer); |
| 1037 |
// } else { |
| 1038 |
// // Make bound-checked writes, which will ask the underlying stream for |
| 1039 |
// // more space as needed. |
| 1040 |
// coded_output->WriteLittleEndian32(magic_number); |
| 1041 |
// coded_output->WriteVarint32(strlen(text)); |
| 1042 |
// coded_output->WriteRaw(text, strlen(text)); |
| 1043 |
// } |
| 1044 |
// |
| 1045 |
// delete coded_output; |
| 1046 |
class PROTOBUF_EXPORT CodedOutputStream { |
| 1047 |
public: |
| 1048 |
// Create an CodedOutputStream that writes to the given ZeroCopyOutputStream. |
| 1049 |
explicit CodedOutputStream(ZeroCopyOutputStream* stream) |
| 1050 |
: CodedOutputStream(stream, true) {} |
| 1051 |
CodedOutputStream(ZeroCopyOutputStream* stream, bool do_eager_refresh); |
| 1052 |
|
| 1053 |
// Destroy the CodedOutputStream and position the underlying |
| 1054 |
// ZeroCopyOutputStream immediately after the last byte written. |
| 1055 |
~CodedOutputStream(); |
| 1056 |
|
| 1057 |
// Returns true if there was an underlying I/O error since this object was |
| 1058 |
// created. On should call Trim before this function in order to catch all |
| 1059 |
// errors. |
| 1060 |
bool HadError() { |
| 1061 |
cur_ = impl_.FlushAndResetBuffer(cur_); |
| 1062 |
GOOGLE_DCHECK(cur_); |
| 1063 |
return impl_.HadError(); |
| 1064 |
} |
| 1065 |
|
| 1066 |
// Trims any unused space in the underlying buffer so that its size matches |
| 1067 |
// the number of bytes written by this stream. The underlying buffer will |
| 1068 |
// automatically be trimmed when this stream is destroyed; this call is only |
| 1069 |
// necessary if the underlying buffer is accessed *before* the stream is |
| 1070 |
// destroyed. |
| 1071 |
void Trim() { cur_ = impl_.Trim(cur_); } |
| 1072 |
|
| 1073 |
// Skips a number of bytes, leaving the bytes unmodified in the underlying |
| 1074 |
// buffer. Returns false if an underlying write error occurs. This is |
| 1075 |
// mainly useful with GetDirectBufferPointer(). |
| 1076 |
// Note of caution, the skipped bytes may contain uninitialized data. The |
| 1077 |
// caller must make sure that the skipped bytes are properly initialized, |
| 1078 |
// otherwise you might leak bytes from your heap. |
| 1079 |
bool Skip(int count) { return impl_.Skip(count, &cur_); } |
| 1080 |
|
| 1081 |
// Sets *data to point directly at the unwritten part of the |
| 1082 |
// CodedOutputStream's underlying buffer, and *size to the size of that |
| 1083 |
// buffer, but does not advance the stream's current position. This will |
| 1084 |
// always either produce a non-empty buffer or return false. If the caller |
| 1085 |
// writes any data to this buffer, it should then call Skip() to skip over |
| 1086 |
// the consumed bytes. This may be useful for implementing external fast |
| 1087 |
// serialization routines for types of data not covered by the |
| 1088 |
// CodedOutputStream interface. |
| 1089 |
bool GetDirectBufferPointer(void** data, int* size) { |
| 1090 |
return impl_.GetDirectBufferPointer(data, size, &cur_); |
| 1091 |
} |
| 1092 |
|
| 1093 |
// If there are at least "size" bytes available in the current buffer, |
| 1094 |
// returns a pointer directly into the buffer and advances over these bytes. |
| 1095 |
// The caller may then write directly into this buffer (e.g. using the |
| 1096 |
// *ToArray static methods) rather than go through CodedOutputStream. If |
| 1097 |
// there are not enough bytes available, returns NULL. The return pointer is |
| 1098 |
// invalidated as soon as any other non-const method of CodedOutputStream |
| 1099 |
// is called. |
| 1100 |
inline uint8* GetDirectBufferForNBytesAndAdvance(int size) { |
| 1101 |
return impl_.GetDirectBufferForNBytesAndAdvance(size, &cur_); |
| 1102 |
} |
| 1103 |
|
| 1104 |
// Write raw bytes, copying them from the given buffer. |
| 1105 |
void WriteRaw(const void* buffer, int size) { |
| 1106 |
cur_ = impl_.WriteRaw(buffer, size, cur_); |
| 1107 |
} |
| 1108 |
// Like WriteRaw() but will try to write aliased data if aliasing is |
| 1109 |
// turned on. |
| 1110 |
void WriteRawMaybeAliased(const void* data, int size); |
| 1111 |
// Like WriteRaw() but writing directly to the target array. |
| 1112 |
// This is _not_ inlined, as the compiler often optimizes memcpy into inline |
| 1113 |
// copy loops. Since this gets called by every field with string or bytes |
| 1114 |
// type, inlining may lead to a significant amount of code bloat, with only a |
| 1115 |
// minor performance gain. |
| 1116 |
static uint8* WriteRawToArray(const void* buffer, int size, uint8* target); |
| 1117 |
|
| 1118 |
// Equivalent to WriteRaw(str.data(), str.size()). |
| 1119 |
void WriteString(const std::string& str); |
| 1120 |
// Like WriteString() but writing directly to the target array. |
| 1121 |
static uint8* WriteStringToArray(const std::string& str, uint8* target); |
| 1122 |
// Write the varint-encoded size of str followed by str. |
| 1123 |
static uint8* WriteStringWithSizeToArray(const std::string& str, |
| 1124 |
uint8* target); |
| 1125 |
|
| 1126 |
|
| 1127 |
// Write a 32-bit little-endian integer. |
| 1128 |
void WriteLittleEndian32(uint32 value) { |
| 1129 |
cur_ = impl_.EnsureSpace(cur_); |
| 1130 |
SetCur(WriteLittleEndian32ToArray(value, Cur())); |
| 1131 |
} |
| 1132 |
// Like WriteLittleEndian32() but writing directly to the target array. |
| 1133 |
static uint8* WriteLittleEndian32ToArray(uint32 value, uint8* target); |
| 1134 |
// Write a 64-bit little-endian integer. |
| 1135 |
void WriteLittleEndian64(uint64 value) { |
| 1136 |
cur_ = impl_.EnsureSpace(cur_); |
| 1137 |
SetCur(WriteLittleEndian64ToArray(value, Cur())); |
| 1138 |
} |
| 1139 |
// Like WriteLittleEndian64() but writing directly to the target array. |
| 1140 |
static uint8* WriteLittleEndian64ToArray(uint64 value, uint8* target); |
| 1141 |
|
| 1142 |
// Write an unsigned integer with Varint encoding. Writing a 32-bit value |
| 1143 |
// is equivalent to casting it to uint64 and writing it as a 64-bit value, |
| 1144 |
// but may be more efficient. |
| 1145 |
void WriteVarint32(uint32 value); |
| 1146 |
// Like WriteVarint32() but writing directly to the target array. |
| 1147 |
static uint8* WriteVarint32ToArray(uint32 value, uint8* target); |
| 1148 |
// Write an unsigned integer with Varint encoding. |
| 1149 |
void WriteVarint64(uint64 value); |
| 1150 |
// Like WriteVarint64() but writing directly to the target array. |
| 1151 |
static uint8* WriteVarint64ToArray(uint64 value, uint8* target); |
| 1152 |
|
| 1153 |
// Equivalent to WriteVarint32() except when the value is negative, |
| 1154 |
// in which case it must be sign-extended to a full 10 bytes. |
| 1155 |
void WriteVarint32SignExtended(int32 value); |
| 1156 |
// Like WriteVarint32SignExtended() but writing directly to the target array. |
| 1157 |
static uint8* WriteVarint32SignExtendedToArray(int32 value, uint8* target); |
| 1158 |
|
| 1159 |
// This is identical to WriteVarint32(), but optimized for writing tags. |
| 1160 |
// In particular, if the input is a compile-time constant, this method |
| 1161 |
// compiles down to a couple instructions. |
| 1162 |
// Always inline because otherwise the aforementioned optimization can't work, |
| 1163 |
// but GCC by default doesn't want to inline this. |
| 1164 |
void WriteTag(uint32 value); |
| 1165 |
// Like WriteTag() but writing directly to the target array. |
| 1166 |
PROTOBUF_ALWAYS_INLINE |
| 1167 |
static uint8* WriteTagToArray(uint32 value, uint8* target); |
| 1168 |
|
| 1169 |
// Returns the number of bytes needed to encode the given value as a varint. |
| 1170 |
static size_t VarintSize32(uint32 value); |
| 1171 |
// Returns the number of bytes needed to encode the given value as a varint. |
| 1172 |
static size_t VarintSize64(uint64 value); |
| 1173 |
|
| 1174 |
// If negative, 10 bytes. Otherwise, same as VarintSize32(). |
| 1175 |
static size_t VarintSize32SignExtended(int32 value); |
| 1176 |
|
| 1177 |
// Compile-time equivalent of VarintSize32(). |
| 1178 |
template <uint32 Value> |
| 1179 |
struct StaticVarintSize32 { |
| 1180 |
static const size_t value = |
| 1181 |
(Value < (1 << 7)) |
| 1182 |
? 1 |
| 1183 |
: (Value < (1 << 14)) |
| 1184 |
? 2 |
| 1185 |
: (Value < (1 << 21)) ? 3 : (Value < (1 << 28)) ? 4 : 5; |
| 1186 |
}; |
| 1187 |
|
| 1188 |
// Returns the total number of bytes written since this object was created. |
| 1189 |
int ByteCount() const { |
| 1190 |
return static_cast<int>(impl_.ByteCount(cur_) - start_count_); |
| 1191 |
} |
| 1192 |
|
| 1193 |
// Instructs the CodedOutputStream to allow the underlying |
| 1194 |
// ZeroCopyOutputStream to hold pointers to the original structure instead of |
| 1195 |
// copying, if it supports it (i.e. output->AllowsAliasing() is true). If the |
| 1196 |
// underlying stream does not support aliasing, then enabling it has no |
| 1197 |
// affect. For now, this only affects the behavior of |
| 1198 |
// WriteRawMaybeAliased(). |
| 1199 |
// |
| 1200 |
// NOTE: It is caller's responsibility to ensure that the chunk of memory |
| 1201 |
// remains live until all of the data has been consumed from the stream. |
| 1202 |
void EnableAliasing(bool enabled) { impl_.EnableAliasing(enabled); } |
| 1203 |
|
| 1204 |
// Indicate to the serializer whether the user wants derministic |
| 1205 |
// serialization. The default when this is not called comes from the global |
| 1206 |
// default, controlled by SetDefaultSerializationDeterministic. |
| 1207 |
// |
| 1208 |
// What deterministic serialization means is entirely up to the driver of the |
| 1209 |
// serialization process (i.e. the caller of methods like WriteVarint32). In |
| 1210 |
// the case of serializing a proto buffer message using one of the methods of |
| 1211 |
// MessageLite, this means that for a given binary equal messages will always |
| 1212 |
// be serialized to the same bytes. This implies: |
| 1213 |
// |
| 1214 |
// * Repeated serialization of a message will return the same bytes. |
| 1215 |
// |
| 1216 |
// * Different processes running the same binary (including on different |
| 1217 |
// machines) will serialize equal messages to the same bytes. |
| 1218 |
// |
| 1219 |
// Note that this is *not* canonical across languages. It is also unstable |
| 1220 |
// across different builds with intervening message definition changes, due to |
| 1221 |
// unknown fields. Users who need canonical serialization (e.g. persistent |
| 1222 |
// storage in a canonical form, fingerprinting) should define their own |
| 1223 |
// canonicalization specification and implement the serializer using |
| 1224 |
// reflection APIs rather than relying on this API. |
| 1225 |
void SetSerializationDeterministic(bool value) { |
| 1226 |
impl_.SetSerializationDeterministic(value); |
| 1227 |
} |
| 1228 |
|
| 1229 |
// Return whether the user wants deterministic serialization. See above. |
| 1230 |
bool IsSerializationDeterministic() const { |
| 1231 |
return impl_.IsSerializationDeterministic(); |
| 1232 |
} |
| 1233 |
|
| 1234 |
static bool IsDefaultSerializationDeterministic() { |
| 1235 |
return default_serialization_deterministic_.load( |
| 1236 |
std::memory_order_relaxed) != 0; |
| 1237 |
} |
| 1238 |
|
| 1239 |
template <typename Func> |
| 1240 |
void Serialize(const Func& func); |
| 1241 |
|
| 1242 |
uint8* Cur() const { return cur_; } |
| 1243 |
void SetCur(uint8* ptr) { cur_ = ptr; } |
| 1244 |
EpsCopyOutputStream* EpsCopy() { return &impl_; } |
| 1245 |
|
| 1246 |
private: |
| 1247 |
EpsCopyOutputStream impl_; |
| 1248 |
uint8* cur_; |
| 1249 |
int64 start_count_; |
| 1250 |
static std::atomic<bool> default_serialization_deterministic_; |
| 1251 |
|
| 1252 |
// See above. Other projects may use "friend" to allow them to call this. |
| 1253 |
// After SetDefaultSerializationDeterministic() completes, all protocol |
| 1254 |
// buffer serializations will be deterministic by default. Thread safe. |
| 1255 |
// However, the meaning of "after" is subtle here: to be safe, each thread |
| 1256 |
// that wants deterministic serialization by default needs to call |
| 1257 |
// SetDefaultSerializationDeterministic() or ensure on its own that another |
| 1258 |
// thread has done so. |
| 1259 |
friend void internal::MapTestForceDeterministic(); |
| 1260 |
static void SetDefaultSerializationDeterministic() { |
| 1261 |
default_serialization_deterministic_.store(true, std::memory_order_relaxed); |
| 1262 |
} |
| 1263 |
GOOGLE_DISALLOW_EVIL_CONSTRUCTORS(CodedOutputStream); |
| 1264 |
}; |
| 1265 |
|
| 1266 |
// inline methods ==================================================== |
| 1267 |
// The vast majority of varints are only one byte. These inline |
| 1268 |
// methods optimize for that case. |
| 1269 |
|
| 1270 |
inline bool CodedInputStream::ReadVarint32(uint32* value) { |
| 1271 |
uint32 v = 0; |
| 1272 |
if (PROTOBUF_PREDICT_TRUE(buffer_ < buffer_end_)) { |
| 1273 |
v = *buffer_; |
| 1274 |
if (v < 0x80) { |
| 1275 |
*value = v; |
| 1276 |
Advance(1); |
| 1277 |
return true; |
| 1278 |
} |
| 1279 |
} |
| 1280 |
int64 result = ReadVarint32Fallback(v); |
| 1281 |
*value = static_cast<uint32>(result); |
| 1282 |
return result >= 0; |
| 1283 |
} |
| 1284 |
|
| 1285 |
inline bool CodedInputStream::ReadVarint64(uint64* value) { |
| 1286 |
if (PROTOBUF_PREDICT_TRUE(buffer_ < buffer_end_) && *buffer_ < 0x80) { |
| 1287 |
*value = *buffer_; |
| 1288 |
Advance(1); |
| 1289 |
return true; |
| 1290 |
} |
| 1291 |
std::pair<uint64, bool> p = ReadVarint64Fallback(); |
| 1292 |
*value = p.first; |
| 1293 |
return p.second; |
| 1294 |
} |
| 1295 |
|
| 1296 |
inline bool CodedInputStream::ReadVarintSizeAsInt(int* value) { |
| 1297 |
if (PROTOBUF_PREDICT_TRUE(buffer_ < buffer_end_)) { |
| 1298 |
int v = *buffer_; |
| 1299 |
if (v < 0x80) { |
| 1300 |
*value = v; |
| 1301 |
Advance(1); |
| 1302 |
return true; |
| 1303 |
} |
| 1304 |
} |
| 1305 |
*value = ReadVarintSizeAsIntFallback(); |
| 1306 |
return *value >= 0; |
| 1307 |
} |
| 1308 |
|
| 1309 |
// static |
| 1310 |
inline const uint8* CodedInputStream::ReadLittleEndian32FromArray( |
| 1311 |
const uint8* buffer, uint32* value) { |
| 1312 |
#if defined(PROTOBUF_LITTLE_ENDIAN) |
| 1313 |
memcpy(value, buffer, sizeof(*value)); |
| 1314 |
return buffer + sizeof(*value); |
| 1315 |
#else |
| 1316 |
*value = (static_cast<uint32>(buffer[0])) | |
| 1317 |
(static_cast<uint32>(buffer[1]) << 8) | |
| 1318 |
(static_cast<uint32>(buffer[2]) << 16) | |
| 1319 |
(static_cast<uint32>(buffer[3]) << 24); |
| 1320 |
return buffer + sizeof(*value); |
| 1321 |
#endif |
| 1322 |
} |
| 1323 |
// static |
| 1324 |
inline const uint8* CodedInputStream::ReadLittleEndian64FromArray( |
| 1325 |
const uint8* buffer, uint64* value) { |
| 1326 |
#if defined(PROTOBUF_LITTLE_ENDIAN) |
| 1327 |
memcpy(value, buffer, sizeof(*value)); |
| 1328 |
return buffer + sizeof(*value); |
| 1329 |
#else |
| 1330 |
uint32 part0 = (static_cast<uint32>(buffer[0])) | |
| 1331 |
(static_cast<uint32>(buffer[1]) << 8) | |
| 1332 |
(static_cast<uint32>(buffer[2]) << 16) | |
| 1333 |
(static_cast<uint32>(buffer[3]) << 24); |
| 1334 |
uint32 part1 = (static_cast<uint32>(buffer[4])) | |
| 1335 |
(static_cast<uint32>(buffer[5]) << 8) | |
| 1336 |
(static_cast<uint32>(buffer[6]) << 16) | |
| 1337 |
(static_cast<uint32>(buffer[7]) << 24); |
| 1338 |
*value = static_cast<uint64>(part0) | (static_cast<uint64>(part1) << 32); |
| 1339 |
return buffer + sizeof(*value); |
| 1340 |
#endif |
| 1341 |
} |
| 1342 |
|
| 1343 |
inline bool CodedInputStream::ReadLittleEndian32(uint32* value) { |
| 1344 |
#if defined(PROTOBUF_LITTLE_ENDIAN) |
| 1345 |
if (PROTOBUF_PREDICT_TRUE(BufferSize() >= static_cast<int>(sizeof(*value)))) { |
| 1346 |
buffer_ = ReadLittleEndian32FromArray(buffer_, value); |
| 1347 |
return true; |
| 1348 |
} else { |
| 1349 |
return ReadLittleEndian32Fallback(value); |
| 1350 |
} |
| 1351 |
#else |
| 1352 |
return ReadLittleEndian32Fallback(value); |
| 1353 |
#endif |
| 1354 |
} |
| 1355 |
|
| 1356 |
inline bool CodedInputStream::ReadLittleEndian64(uint64* value) { |
| 1357 |
#if defined(PROTOBUF_LITTLE_ENDIAN) |
| 1358 |
if (PROTOBUF_PREDICT_TRUE(BufferSize() >= static_cast<int>(sizeof(*value)))) { |
| 1359 |
buffer_ = ReadLittleEndian64FromArray(buffer_, value); |
| 1360 |
return true; |
| 1361 |
} else { |
| 1362 |
return ReadLittleEndian64Fallback(value); |
| 1363 |
} |
| 1364 |
#else |
| 1365 |
return ReadLittleEndian64Fallback(value); |
| 1366 |
#endif |
| 1367 |
} |
| 1368 |
|
| 1369 |
inline uint32 CodedInputStream::ReadTagNoLastTag() { |
| 1370 |
uint32 v = 0; |
| 1371 |
if (PROTOBUF_PREDICT_TRUE(buffer_ < buffer_end_)) { |
| 1372 |
v = *buffer_; |
| 1373 |
if (v < 0x80) { |
| 1374 |
Advance(1); |
| 1375 |
return v; |
| 1376 |
} |
| 1377 |
} |
| 1378 |
v = ReadTagFallback(v); |
| 1379 |
return v; |
| 1380 |
} |
| 1381 |
|
| 1382 |
inline std::pair<uint32, bool> CodedInputStream::ReadTagWithCutoffNoLastTag( |
| 1383 |
uint32 cutoff) { |
| 1384 |
// In performance-sensitive code we can expect cutoff to be a compile-time |
| 1385 |
// constant, and things like "cutoff >= kMax1ByteVarint" to be evaluated at |
| 1386 |
// compile time. |
| 1387 |
uint32 first_byte_or_zero = 0; |
| 1388 |
if (PROTOBUF_PREDICT_TRUE(buffer_ < buffer_end_)) { |
| 1389 |
// Hot case: buffer_ non_empty, buffer_[0] in [1, 128). |
| 1390 |
// TODO(gpike): Is it worth rearranging this? E.g., if the number of fields |
| 1391 |
// is large enough then is it better to check for the two-byte case first? |
| 1392 |
first_byte_or_zero = buffer_[0]; |
| 1393 |
if (static_cast<int8>(buffer_[0]) > 0) { |
| 1394 |
const uint32 kMax1ByteVarint = 0x7f; |
| 1395 |
uint32 tag = buffer_[0]; |
| 1396 |
Advance(1); |
| 1397 |
return std::make_pair(tag, cutoff >= kMax1ByteVarint || tag <= cutoff); |
| 1398 |
} |
| 1399 |
// Other hot case: cutoff >= 0x80, buffer_ has at least two bytes available, |
| 1400 |
// and tag is two bytes. The latter is tested by bitwise-and-not of the |
| 1401 |
// first byte and the second byte. |
| 1402 |
if (cutoff >= 0x80 && PROTOBUF_PREDICT_TRUE(buffer_ + 1 < buffer_end_) && |
| 1403 |
PROTOBUF_PREDICT_TRUE((buffer_[0] & ~buffer_[1]) >= 0x80)) { |
| 1404 |
const uint32 kMax2ByteVarint = (0x7f << 7) + 0x7f; |
| 1405 |
uint32 tag = (1u << 7) * buffer_[1] + (buffer_[0] - 0x80); |
| 1406 |
Advance(2); |
| 1407 |
// It might make sense to test for tag == 0 now, but it is so rare that |
| 1408 |
// that we don't bother. A varint-encoded 0 should be one byte unless |
| 1409 |
// the encoder lost its mind. The second part of the return value of |
| 1410 |
// this function is allowed to be either true or false if the tag is 0, |
| 1411 |
// so we don't have to check for tag == 0. We may need to check whether |
| 1412 |
// it exceeds cutoff. |
| 1413 |
bool at_or_below_cutoff = cutoff >= kMax2ByteVarint || tag <= cutoff; |
| 1414 |
return std::make_pair(tag, at_or_below_cutoff); |
| 1415 |
} |
| 1416 |
} |
| 1417 |
// Slow path |
| 1418 |
const uint32 tag = ReadTagFallback(first_byte_or_zero); |
| 1419 |
return std::make_pair(tag, static_cast<uint32>(tag - 1) < cutoff); |
| 1420 |
} |
| 1421 |
|
| 1422 |
inline bool CodedInputStream::LastTagWas(uint32 expected) { |
| 1423 |
return last_tag_ == expected; |
| 1424 |
} |
| 1425 |
|
| 1426 |
inline bool CodedInputStream::ConsumedEntireMessage() { |
| 1427 |
return legitimate_message_end_; |
| 1428 |
} |
| 1429 |
|
| 1430 |
inline bool CodedInputStream::ExpectTag(uint32 expected) { |
| 1431 |
if (expected < (1 << 7)) { |
| 1432 |
if (PROTOBUF_PREDICT_TRUE(buffer_ < buffer_end_) && |
| 1433 |
buffer_[0] == expected) { |
| 1434 |
Advance(1); |
| 1435 |
return true; |
| 1436 |
} else { |
| 1437 |
return false; |
| 1438 |
} |
| 1439 |
} else if (expected < (1 << 14)) { |
| 1440 |
if (PROTOBUF_PREDICT_TRUE(BufferSize() >= 2) && |
| 1441 |
buffer_[0] == static_cast<uint8>(expected | 0x80) && |
| 1442 |
buffer_[1] == static_cast<uint8>(expected >> 7)) { |
| 1443 |
Advance(2); |
| 1444 |
return true; |
| 1445 |
} else { |
| 1446 |
return false; |
| 1447 |
} |
| 1448 |
} else { |
| 1449 |
// Don't bother optimizing for larger values. |
| 1450 |
return false; |
| 1451 |
} |
| 1452 |
} |
| 1453 |
|
| 1454 |
inline const uint8* CodedInputStream::ExpectTagFromArray(const uint8* buffer, |
| 1455 |
uint32 expected) { |
| 1456 |
if (expected < (1 << 7)) { |
| 1457 |
if (buffer[0] == expected) { |
| 1458 |
return buffer + 1; |
| 1459 |
} |
| 1460 |
} else if (expected < (1 << 14)) { |
| 1461 |
if (buffer[0] == static_cast<uint8>(expected | 0x80) && |
| 1462 |
buffer[1] == static_cast<uint8>(expected >> 7)) { |
| 1463 |
return buffer + 2; |
| 1464 |
} |
| 1465 |
} |
| 1466 |
return nullptr; |
| 1467 |
} |
| 1468 |
|
| 1469 |
inline void CodedInputStream::GetDirectBufferPointerInline(const void** data, |
| 1470 |
int* size) { |
| 1471 |
*data = buffer_; |
| 1472 |
*size = static_cast<int>(buffer_end_ - buffer_); |
| 1473 |
} |
| 1474 |
|
| 1475 |
inline bool CodedInputStream::ExpectAtEnd() { |
| 1476 |
// If we are at a limit we know no more bytes can be read. Otherwise, it's |
| 1477 |
// hard to say without calling Refresh(), and we'd rather not do that. |
| 1478 |
|
| 1479 |
if (buffer_ == buffer_end_ && ((buffer_size_after_limit_ != 0) || |
| 1480 |
(total_bytes_read_ == current_limit_))) { |
| 1481 |
last_tag_ = 0; // Pretend we called ReadTag()... |
| 1482 |
legitimate_message_end_ = true; // ... and it hit EOF. |
| 1483 |
return true; |
| 1484 |
} else { |
| 1485 |
return false; |
| 1486 |
} |
| 1487 |
} |
| 1488 |
|
| 1489 |
inline int CodedInputStream::CurrentPosition() const { |
| 1490 |
return total_bytes_read_ - (BufferSize() + buffer_size_after_limit_); |
| 1491 |
} |
| 1492 |
|
| 1493 |
inline void CodedInputStream::Advance(int amount) { buffer_ += amount; } |
| 1494 |
|
| 1495 |
inline void CodedInputStream::SetRecursionLimit(int limit) { |
| 1496 |
recursion_budget_ += limit - recursion_limit_; |
| 1497 |
recursion_limit_ = limit; |
| 1498 |
} |
| 1499 |
|
| 1500 |
inline bool CodedInputStream::IncrementRecursionDepth() { |
| 1501 |
--recursion_budget_; |
| 1502 |
return recursion_budget_ >= 0; |
| 1503 |
} |
| 1504 |
|
| 1505 |
inline void CodedInputStream::DecrementRecursionDepth() { |
| 1506 |
if (recursion_budget_ < recursion_limit_) ++recursion_budget_; |
| 1507 |
} |
| 1508 |
|
| 1509 |
inline void CodedInputStream::UnsafeDecrementRecursionDepth() { |
| 1510 |
assert(recursion_budget_ < recursion_limit_); |
| 1511 |
++recursion_budget_; |
| 1512 |
} |
| 1513 |
|
| 1514 |
inline void CodedInputStream::SetExtensionRegistry(const DescriptorPool* pool, |
| 1515 |
MessageFactory* factory) { |
| 1516 |
extension_pool_ = pool; |
| 1517 |
extension_factory_ = factory; |
| 1518 |
} |
| 1519 |
|
| 1520 |
inline const DescriptorPool* CodedInputStream::GetExtensionPool() { |
| 1521 |
return extension_pool_; |
| 1522 |
} |
| 1523 |
|
| 1524 |
inline MessageFactory* CodedInputStream::GetExtensionFactory() { |
| 1525 |
return extension_factory_; |
| 1526 |
} |
| 1527 |
|
| 1528 |
inline int CodedInputStream::BufferSize() const { |
| 1529 |
return static_cast<int>(buffer_end_ - buffer_); |
| 1530 |
} |
| 1531 |
|
| 1532 |
inline CodedInputStream::CodedInputStream(ZeroCopyInputStream* input) |
| 1533 |
: buffer_(nullptr), |
| 1534 |
buffer_end_(nullptr), |
| 1535 |
input_(input), |
| 1536 |
total_bytes_read_(0), |
| 1537 |
overflow_bytes_(0), |
| 1538 |
last_tag_(0), |
| 1539 |
legitimate_message_end_(false), |
| 1540 |
aliasing_enabled_(false), |
| 1541 |
current_limit_(kint32max), |
| 1542 |
buffer_size_after_limit_(0), |
| 1543 |
total_bytes_limit_(kDefaultTotalBytesLimit), |
| 1544 |
recursion_budget_(default_recursion_limit_), |
| 1545 |
recursion_limit_(default_recursion_limit_), |
| 1546 |
extension_pool_(nullptr), |
| 1547 |
extension_factory_(nullptr) { |
| 1548 |
// Eagerly Refresh() so buffer space is immediately available. |
| 1549 |
Refresh(); |
| 1550 |
} |
| 1551 |
|
| 1552 |
inline CodedInputStream::CodedInputStream(const uint8* buffer, int size) |
| 1553 |
: buffer_(buffer), |
| 1554 |
buffer_end_(buffer + size), |
| 1555 |
input_(nullptr), |
| 1556 |
total_bytes_read_(size), |
| 1557 |
overflow_bytes_(0), |
| 1558 |
last_tag_(0), |
| 1559 |
legitimate_message_end_(false), |
| 1560 |
aliasing_enabled_(false), |
| 1561 |
current_limit_(size), |
| 1562 |
buffer_size_after_limit_(0), |
| 1563 |
total_bytes_limit_(kDefaultTotalBytesLimit), |
| 1564 |
recursion_budget_(default_recursion_limit_), |
| 1565 |
recursion_limit_(default_recursion_limit_), |
| 1566 |
extension_pool_(nullptr), |
| 1567 |
extension_factory_(nullptr) { |
| 1568 |
// Note that setting current_limit_ == size is important to prevent some |
| 1569 |
// code paths from trying to access input_ and segfaulting. |
| 1570 |
} |
| 1571 |
|
| 1572 |
inline bool CodedInputStream::IsFlat() const { return input_ == nullptr; } |
| 1573 |
|
| 1574 |
inline bool CodedInputStream::Skip(int count) { |
| 1575 |
if (count < 0) return false; // security: count is often user-supplied |
| 1576 |
|
| 1577 |
const int original_buffer_size = BufferSize(); |
| 1578 |
|
| 1579 |
if (count <= original_buffer_size) { |
| 1580 |
// Just skipping within the current buffer. Easy. |
| 1581 |
Advance(count); |
| 1582 |
return true; |
| 1583 |
} |
| 1584 |
|
| 1585 |
return SkipFallback(count, original_buffer_size); |
| 1586 |
} |
| 1587 |
|
| 1588 |
inline uint8* CodedOutputStream::WriteVarint32ToArray(uint32 value, |
| 1589 |
uint8* target) { |
| 1590 |
return EpsCopyOutputStream::UnsafeVarint(value, target); |
| 1591 |
} |
| 1592 |
|
| 1593 |
inline uint8* CodedOutputStream::WriteVarint64ToArray(uint64 value, |
| 1594 |
uint8* target) { |
| 1595 |
return EpsCopyOutputStream::UnsafeVarint(value, target); |
| 1596 |
} |
| 1597 |
|
| 1598 |
inline void CodedOutputStream::WriteVarint32SignExtended(int32 value) { |
| 1599 |
WriteVarint64(static_cast<uint64>(value)); |
| 1600 |
} |
| 1601 |
|
| 1602 |
inline uint8* CodedOutputStream::WriteVarint32SignExtendedToArray( |
| 1603 |
int32 value, uint8* target) { |
| 1604 |
return WriteVarint64ToArray(static_cast<uint64>(value), target); |
| 1605 |
} |
| 1606 |
|
| 1607 |
inline uint8* CodedOutputStream::WriteLittleEndian32ToArray(uint32 value, |
| 1608 |
uint8* target) { |
| 1609 |
#if defined(PROTOBUF_LITTLE_ENDIAN) |
| 1610 |
memcpy(target, &value, sizeof(value)); |
| 1611 |
#else |
| 1612 |
target[0] = static_cast<uint8>(value); |
| 1613 |
target[1] = static_cast<uint8>(value >> 8); |
| 1614 |
target[2] = static_cast<uint8>(value >> 16); |
| 1615 |
target[3] = static_cast<uint8>(value >> 24); |
| 1616 |
#endif |
| 1617 |
return target + sizeof(value); |
| 1618 |
} |
| 1619 |
|
| 1620 |
inline uint8* CodedOutputStream::WriteLittleEndian64ToArray(uint64 value, |
| 1621 |
uint8* target) { |
| 1622 |
#if defined(PROTOBUF_LITTLE_ENDIAN) |
| 1623 |
memcpy(target, &value, sizeof(value)); |
| 1624 |
#else |
| 1625 |
uint32 part0 = static_cast<uint32>(value); |
| 1626 |
uint32 part1 = static_cast<uint32>(value >> 32); |
| 1627 |
|
| 1628 |
target[0] = static_cast<uint8>(part0); |
| 1629 |
target[1] = static_cast<uint8>(part0 >> 8); |
| 1630 |
target[2] = static_cast<uint8>(part0 >> 16); |
| 1631 |
target[3] = static_cast<uint8>(part0 >> 24); |
| 1632 |
target[4] = static_cast<uint8>(part1); |
| 1633 |
target[5] = static_cast<uint8>(part1 >> 8); |
| 1634 |
target[6] = static_cast<uint8>(part1 >> 16); |
| 1635 |
target[7] = static_cast<uint8>(part1 >> 24); |
| 1636 |
#endif |
| 1637 |
return target + sizeof(value); |
| 1638 |
} |
| 1639 |
|
| 1640 |
inline void CodedOutputStream::WriteVarint32(uint32 value) { |
| 1641 |
cur_ = impl_.EnsureSpace(cur_); |
| 1642 |
SetCur(WriteVarint32ToArray(value, Cur())); |
| 1643 |
} |
| 1644 |
|
| 1645 |
inline void CodedOutputStream::WriteVarint64(uint64 value) { |
| 1646 |
cur_ = impl_.EnsureSpace(cur_); |
| 1647 |
SetCur(WriteVarint64ToArray(value, Cur())); |
| 1648 |
} |
| 1649 |
|
| 1650 |
inline void CodedOutputStream::WriteTag(uint32 value) { WriteVarint32(value); } |
| 1651 |
|
| 1652 |
inline uint8* CodedOutputStream::WriteTagToArray(uint32 value, uint8* target) { |
| 1653 |
return WriteVarint32ToArray(value, target); |
| 1654 |
} |
| 1655 |
|
| 1656 |
inline size_t CodedOutputStream::VarintSize32(uint32 value) { |
| 1657 |
// This computes value == 0 ? 1 : floor(log2(value)) / 7 + 1 |
| 1658 |
// Use an explicit multiplication to implement the divide of |
| 1659 |
// a number in the 1..31 range. |
| 1660 |
// Explicit OR 0x1 to avoid calling Bits::Log2FloorNonZero(0), which is |
| 1661 |
// undefined. |
| 1662 |
uint32 log2value = Bits::Log2FloorNonZero(value | 0x1); |
| 1663 |
return static_cast<size_t>((log2value * 9 + 73) / 64); |
| 1664 |
} |
| 1665 |
|
| 1666 |
inline size_t CodedOutputStream::VarintSize64(uint64 value) { |
| 1667 |
// This computes value == 0 ? 1 : floor(log2(value)) / 7 + 1 |
| 1668 |
// Use an explicit multiplication to implement the divide of |
| 1669 |
// a number in the 1..63 range. |
| 1670 |
// Explicit OR 0x1 to avoid calling Bits::Log2FloorNonZero(0), which is |
| 1671 |
// undefined. |
| 1672 |
uint32 log2value = Bits::Log2FloorNonZero64(value | 0x1); |
| 1673 |
return static_cast<size_t>((log2value * 9 + 73) / 64); |
| 1674 |
} |
| 1675 |
|
| 1676 |
inline size_t CodedOutputStream::VarintSize32SignExtended(int32 value) { |
| 1677 |
if (value < 0) { |
| 1678 |
return 10; // TODO(kenton): Make this a symbolic constant. |
| 1679 |
} else { |
| 1680 |
return VarintSize32(static_cast<uint32>(value)); |
| 1681 |
} |
| 1682 |
} |
| 1683 |
|
| 1684 |
inline void CodedOutputStream::WriteString(const std::string& str) { |
| 1685 |
WriteRaw(str.data(), static_cast<int>(str.size())); |
| 1686 |
} |
| 1687 |
|
| 1688 |
inline void CodedOutputStream::WriteRawMaybeAliased(const void* data, |
| 1689 |
int size) { |
| 1690 |
cur_ = impl_.WriteRawMaybeAliased(data, size, cur_); |
| 1691 |
} |
| 1692 |
|
| 1693 |
inline uint8* CodedOutputStream::WriteRawToArray(const void* data, int size, |
| 1694 |
uint8* target) { |
| 1695 |
memcpy(target, data, size); |
| 1696 |
return target + size; |
| 1697 |
} |
| 1698 |
|
| 1699 |
inline uint8* CodedOutputStream::WriteStringToArray(const std::string& str, |
| 1700 |
uint8* target) { |
| 1701 |
return WriteRawToArray(str.data(), static_cast<int>(str.size()), target); |
| 1702 |
} |
| 1703 |
|
| 1704 |
} // namespace io |
| 1705 |
} // namespace protobuf |
| 1706 |
} // namespace google |
| 1707 |
|
| 1708 |
#if defined(_MSC_VER) && _MSC_VER >= 1300 && !defined(__INTEL_COMPILER) |
| 1709 |
#pragma runtime_checks("c", restore) |
| 1710 |
#endif // _MSC_VER && !defined(__INTEL_COMPILER) |
| 1711 |
|
| 1712 |
#include <google/protobuf/port_undef.inc> |
| 1713 |
|
| 1714 |
#endif // GOOGLE_PROTOBUF_IO_CODED_STREAM_H__ |