Line data Source code
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 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 : #include <string>
113 : #include <utility>
114 : #ifdef _MSC_VER
115 : // Assuming windows is always little-endian.
116 : #if !defined(PROTOBUF_DISABLE_LITTLE_ENDIAN_OPT_FOR_TEST)
117 : #define PROTOBUF_LITTLE_ENDIAN 1
118 : #endif
119 : #if _MSC_VER >= 1300
120 : // If MSVC has "/RTCc" set, it will complain about truncating casts at
121 : // runtime. This file contains some intentional truncating casts.
122 : #pragma runtime_checks("c", off)
123 : #endif
124 : #else
125 : #include <sys/param.h> // __BYTE_ORDER
126 : #if ((defined(__LITTLE_ENDIAN__) && !defined(__BIG_ENDIAN__)) || \
127 : (defined(__BYTE_ORDER) && __BYTE_ORDER == __LITTLE_ENDIAN)) && \
128 : !defined(PROTOBUF_DISABLE_LITTLE_ENDIAN_OPT_FOR_TEST)
129 : #define PROTOBUF_LITTLE_ENDIAN 1
130 : #endif
131 : #endif
132 : #include <google/protobuf/stubs/common.h>
133 :
134 : namespace google {
135 :
136 : namespace protobuf {
137 :
138 : class DescriptorPool;
139 : class MessageFactory;
140 :
141 : namespace io {
142 :
143 : // Defined in this file.
144 : class CodedInputStream;
145 : class CodedOutputStream;
146 :
147 : // Defined in other files.
148 : class ZeroCopyInputStream; // zero_copy_stream.h
149 : class ZeroCopyOutputStream; // zero_copy_stream.h
150 :
151 : // Class which reads and decodes binary data which is composed of varint-
152 : // encoded integers and fixed-width pieces. Wraps a ZeroCopyInputStream.
153 : // Most users will not need to deal with CodedInputStream.
154 : //
155 : // Most methods of CodedInputStream that return a bool return false if an
156 : // underlying I/O error occurs or if the data is malformed. Once such a
157 : // failure occurs, the CodedInputStream is broken and is no longer useful.
158 : class LIBPROTOBUF_EXPORT CodedInputStream {
159 : public:
160 : // Create a CodedInputStream that reads from the given ZeroCopyInputStream.
161 : explicit CodedInputStream(ZeroCopyInputStream* input);
162 :
163 : // Create a CodedInputStream that reads from the given flat array. This is
164 : // faster than using an ArrayInputStream. PushLimit(size) is implied by
165 : // this constructor.
166 : explicit CodedInputStream(const uint8* buffer, int size);
167 :
168 : // Destroy the CodedInputStream and position the underlying
169 : // ZeroCopyInputStream at the first unread byte. If an error occurred while
170 : // reading (causing a method to return false), then the exact position of
171 : // the input stream may be anywhere between the last value that was read
172 : // successfully and the stream's byte limit.
173 : ~CodedInputStream();
174 :
175 : // Return true if this CodedInputStream reads from a flat array instead of
176 : // a ZeroCopyInputStream.
177 : inline bool IsFlat() const;
178 :
179 : // Skips a number of bytes. Returns false if an underlying read error
180 : // occurs.
181 : bool Skip(int count);
182 :
183 : // Sets *data to point directly at the unread part of the CodedInputStream's
184 : // underlying buffer, and *size to the size of that buffer, but does not
185 : // advance the stream's current position. This will always either produce
186 : // a non-empty buffer or return false. If the caller consumes any of
187 : // this data, it should then call Skip() to skip over the consumed bytes.
188 : // This may be useful for implementing external fast parsing routines for
189 : // types of data not covered by the CodedInputStream interface.
190 : bool GetDirectBufferPointer(const void** data, int* size);
191 :
192 : // Like GetDirectBufferPointer, but this method is inlined, and does not
193 : // attempt to Refresh() if the buffer is currently empty.
194 : GOOGLE_ATTRIBUTE_ALWAYS_INLINE void GetDirectBufferPointerInline(const void** data,
195 : int* size);
196 :
197 : // Read raw bytes, copying them into the given buffer.
198 : bool ReadRaw(void* buffer, int size);
199 :
200 : // Like the above, with inlined optimizations. This should only be used
201 : // by the protobuf implementation.
202 : GOOGLE_ATTRIBUTE_ALWAYS_INLINE bool InternalReadRawInline(void* buffer, int size);
203 :
204 : // Like ReadRaw, but reads into a string.
205 : //
206 : // Implementation Note: ReadString() grows the string gradually as it
207 : // reads in the data, rather than allocating the entire requested size
208 : // upfront. This prevents denial-of-service attacks in which a client
209 : // could claim that a string is going to be MAX_INT bytes long in order to
210 : // crash the server because it can't allocate this much space at once.
211 : bool ReadString(string* buffer, int size);
212 : // Like the above, with inlined optimizations. This should only be used
213 : // by the protobuf implementation.
214 : GOOGLE_ATTRIBUTE_ALWAYS_INLINE bool InternalReadStringInline(string* buffer,
215 : int size);
216 :
217 :
218 : // Read a 32-bit little-endian integer.
219 : bool ReadLittleEndian32(uint32* value);
220 : // Read a 64-bit little-endian integer.
221 : bool ReadLittleEndian64(uint64* 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* ReadLittleEndian32FromArray(const uint8* buffer,
227 : uint32* value);
228 : // Read a 64-bit little-endian integer.
229 : static const uint8* ReadLittleEndian64FromArray(const uint8* buffer,
230 : uint64* 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, but may be more efficient.
235 : bool ReadVarint32(uint32* value);
236 : // Read an unsigned integer with Varint encoding.
237 : bool ReadVarint64(uint64* value);
238 :
239 : // Read a tag. This calls ReadVarint32() and returns the result, or returns
240 : // zero (which is not a valid tag) if ReadVarint32() fails. Also, it updates
241 : // the last tag value, which can be checked with LastTagWas().
242 : // Always inline because this is only called in one place per parse loop
243 : // but it is called for every iteration of said loop, so it should be fast.
244 : // GCC doesn't want to inline this by default.
245 : GOOGLE_ATTRIBUTE_ALWAYS_INLINE uint32 ReadTag();
246 :
247 : // This usually a faster alternative to ReadTag() when cutoff is a manifest
248 : // constant. It does particularly well for cutoff >= 127. The first part
249 : // of the return value is the tag that was read, though it can also be 0 in
250 : // the cases where ReadTag() would return 0. If the second part is true
251 : // then the tag is known to be in [0, cutoff]. If not, the tag either is
252 : // above cutoff or is 0. (There's intentional wiggle room when tag is 0,
253 : // because that can arise in several ways, and for best performance we want
254 : // to avoid an extra "is tag == 0?" check here.)
255 : GOOGLE_ATTRIBUTE_ALWAYS_INLINE std::pair<uint32, bool> ReadTagWithCutoff(
256 : uint32 cutoff);
257 :
258 : // Usually returns true if calling ReadVarint32() now would produce the given
259 : // value. Will always return false if ReadVarint32() would not return the
260 : // given value. If ExpectTag() returns true, it also advances past
261 : // the varint. For best performance, use a compile-time constant as the
262 : // parameter.
263 : // Always inline because this collapses to a small number of instructions
264 : // when given a constant parameter, but GCC doesn't want to inline by default.
265 : GOOGLE_ATTRIBUTE_ALWAYS_INLINE bool ExpectTag(uint32 expected);
266 :
267 : // Like above, except this reads from the specified buffer. The caller is
268 : // responsible for ensuring that the buffer is large enough to read a varint
269 : // of the expected size. For best performance, use a compile-time constant as
270 : // the expected tag parameter.
271 : //
272 : // Returns a pointer beyond the expected tag if it was found, or NULL if it
273 : // was not.
274 : GOOGLE_ATTRIBUTE_ALWAYS_INLINE static const uint8* ExpectTagFromArray(
275 : const uint8* buffer,
276 : uint32 expected);
277 :
278 : // Usually returns true if no more bytes can be read. Always returns false
279 : // if more bytes can be read. If ExpectAtEnd() returns true, a subsequent
280 : // call to LastTagWas() will act as if ReadTag() had been called and returned
281 : // zero, and ConsumedEntireMessage() will return true.
282 : bool ExpectAtEnd();
283 :
284 : // If the last call to ReadTag() or ReadTagWithCutoff() returned the
285 : // given value, returns true. Otherwise, returns false;
286 : //
287 : // This is needed because parsers for some types of embedded messages
288 : // (with field type TYPE_GROUP) don't actually know that they've reached the
289 : // end of a message until they see an ENDGROUP tag, which was actually part
290 : // of the enclosing message. The enclosing message would like to check that
291 : // tag to make sure it had the right number, so it calls LastTagWas() on
292 : // return from the embedded parser to check.
293 : bool LastTagWas(uint32 expected);
294 :
295 : // When parsing message (but NOT a group), this method must be called
296 : // immediately after MergeFromCodedStream() returns (if it returns true)
297 : // to further verify that the message ended in a legitimate way. For
298 : // example, this verifies that parsing did not end on an end-group tag.
299 : // It also checks for some cases where, due to optimizations,
300 : // MergeFromCodedStream() can incorrectly return true.
301 : bool ConsumedEntireMessage();
302 :
303 : // Limits ----------------------------------------------------------
304 : // Limits are used when parsing length-delimited embedded messages.
305 : // After the message's length is read, PushLimit() is used to prevent
306 : // the CodedInputStream from reading beyond that length. Once the
307 : // embedded message has been parsed, PopLimit() is called to undo the
308 : // limit.
309 :
310 : // Opaque type used with PushLimit() and PopLimit(). Do not modify
311 : // values of this type yourself. The only reason that this isn't a
312 : // struct with private internals is for efficiency.
313 : typedef int Limit;
314 :
315 : // Places a limit on the number of bytes that the stream may read,
316 : // starting from the current position. Once the stream hits this limit,
317 : // it will act like the end of the input has been reached until PopLimit()
318 : // is called.
319 : //
320 : // As the names imply, the stream conceptually has a stack of limits. The
321 : // shortest limit on the stack is always enforced, even if it is not the
322 : // top limit.
323 : //
324 : // The value returned by PushLimit() is opaque to the caller, and must
325 : // be passed unchanged to the corresponding call to PopLimit().
326 : Limit PushLimit(int byte_limit);
327 :
328 : // Pops the last limit pushed by PushLimit(). The input must be the value
329 : // returned by that call to PushLimit().
330 : void PopLimit(Limit limit);
331 :
332 : // Returns the number of bytes left until the nearest limit on the
333 : // stack is hit, or -1 if no limits are in place.
334 : int BytesUntilLimit() const;
335 :
336 : // Returns current position relative to the beginning of the input stream.
337 : int CurrentPosition() const;
338 :
339 : // Total Bytes Limit -----------------------------------------------
340 : // To prevent malicious users from sending excessively large messages
341 : // and causing integer overflows or memory exhaustion, CodedInputStream
342 : // imposes a hard limit on the total number of bytes it will read.
343 :
344 : // Sets the maximum number of bytes that this CodedInputStream will read
345 : // before refusing to continue. To prevent integer overflows in the
346 : // protocol buffers implementation, as well as to prevent servers from
347 : // allocating enormous amounts of memory to hold parsed messages, the
348 : // maximum message length should be limited to the shortest length that
349 : // will not harm usability. The theoretical shortest message that could
350 : // cause integer overflows is 512MB. The default limit is 64MB. Apps
351 : // should set shorter limits if possible. If warning_threshold is not -1,
352 : // a warning will be printed to stderr after warning_threshold bytes are
353 : // read. For backwards compatibility all negative values get squashed to -1,
354 : // as other negative values might have special internal meanings.
355 : // An error will always be printed to stderr if the limit is reached.
356 : //
357 : // This is unrelated to PushLimit()/PopLimit().
358 : //
359 : // Hint: If you are reading this because your program is printing a
360 : // warning about dangerously large protocol messages, you may be
361 : // confused about what to do next. The best option is to change your
362 : // design such that excessively large messages are not necessary.
363 : // For example, try to design file formats to consist of many small
364 : // messages rather than a single large one. If this is infeasible,
365 : // you will need to increase the limit. Chances are, though, that
366 : // your code never constructs a CodedInputStream on which the limit
367 : // can be set. You probably parse messages by calling things like
368 : // Message::ParseFromString(). In this case, you will need to change
369 : // your code to instead construct some sort of ZeroCopyInputStream
370 : // (e.g. an ArrayInputStream), construct a CodedInputStream around
371 : // that, then call Message::ParseFromCodedStream() instead. Then
372 : // you can adjust the limit. Yes, it's more work, but you're doing
373 : // something unusual.
374 : void SetTotalBytesLimit(int total_bytes_limit, int warning_threshold);
375 :
376 : // The Total Bytes Limit minus the Current Position, or -1 if there
377 : // is no Total Bytes Limit.
378 : int BytesUntilTotalBytesLimit() const;
379 :
380 : // Recursion Limit -------------------------------------------------
381 : // To prevent corrupt or malicious messages from causing stack overflows,
382 : // we must keep track of the depth of recursion when parsing embedded
383 : // messages and groups. CodedInputStream keeps track of this because it
384 : // is the only object that is passed down the stack during parsing.
385 :
386 : // Sets the maximum recursion depth. The default is 100.
387 : void SetRecursionLimit(int limit);
388 :
389 :
390 : // Increments the current recursion depth. Returns true if the depth is
391 : // under the limit, false if it has gone over.
392 : bool IncrementRecursionDepth();
393 :
394 : // Decrements the recursion depth if possible.
395 : void DecrementRecursionDepth();
396 :
397 : // Decrements the recursion depth blindly. This is faster than
398 : // DecrementRecursionDepth(). It should be used only if all previous
399 : // increments to recursion depth were successful.
400 : void UnsafeDecrementRecursionDepth();
401 :
402 : // Shorthand for make_pair(PushLimit(byte_limit), --recursion_budget_).
403 : // Using this can reduce code size and complexity in some cases. The caller
404 : // is expected to check that the second part of the result is non-negative (to
405 : // bail out if the depth of recursion is too high) and, if all is well, to
406 : // later pass the first part of the result to PopLimit() or similar.
407 : std::pair<CodedInputStream::Limit, int> IncrementRecursionDepthAndPushLimit(
408 : int byte_limit);
409 :
410 : // Shorthand for PushLimit(ReadVarint32(&length) ? length : 0).
411 : Limit ReadLengthAndPushLimit();
412 :
413 : // Helper that is equivalent to: {
414 : // bool result = ConsumedEntireMessage();
415 : // PopLimit(limit);
416 : // UnsafeDecrementRecursionDepth();
417 : // return result; }
418 : // Using this can reduce code size and complexity in some cases.
419 : // Do not use unless the current recursion depth is greater than zero.
420 : bool DecrementRecursionDepthAndPopLimit(Limit limit);
421 :
422 : // Helper that is equivalent to: {
423 : // bool result = ConsumedEntireMessage();
424 : // PopLimit(limit);
425 : // return result; }
426 : // Using this can reduce code size and complexity in some cases.
427 : bool CheckEntireMessageConsumedAndPopLimit(Limit limit);
428 :
429 : // Extension Registry ----------------------------------------------
430 : // ADVANCED USAGE: 99.9% of people can ignore this section.
431 : //
432 : // By default, when parsing extensions, the parser looks for extension
433 : // definitions in the pool which owns the outer message's Descriptor.
434 : // However, you may call SetExtensionRegistry() to provide an alternative
435 : // pool instead. This makes it possible, for example, to parse a message
436 : // using a generated class, but represent some extensions using
437 : // DynamicMessage.
438 :
439 : // Set the pool used to look up extensions. Most users do not need to call
440 : // this as the correct pool will be chosen automatically.
441 : //
442 : // WARNING: It is very easy to misuse this. Carefully read the requirements
443 : // below. Do not use this unless you are sure you need it. Almost no one
444 : // does.
445 : //
446 : // Let's say you are parsing a message into message object m, and you want
447 : // to take advantage of SetExtensionRegistry(). You must follow these
448 : // requirements:
449 : //
450 : // The given DescriptorPool must contain m->GetDescriptor(). It is not
451 : // sufficient for it to simply contain a descriptor that has the same name
452 : // and content -- it must be the *exact object*. In other words:
453 : // assert(pool->FindMessageTypeByName(m->GetDescriptor()->full_name()) ==
454 : // m->GetDescriptor());
455 : // There are two ways to satisfy this requirement:
456 : // 1) Use m->GetDescriptor()->pool() as the pool. This is generally useless
457 : // because this is the pool that would be used anyway if you didn't call
458 : // SetExtensionRegistry() at all.
459 : // 2) Use a DescriptorPool which has m->GetDescriptor()->pool() as an
460 : // "underlay". Read the documentation for DescriptorPool for more
461 : // information about underlays.
462 : //
463 : // You must also provide a MessageFactory. This factory will be used to
464 : // construct Message objects representing extensions. The factory's
465 : // GetPrototype() MUST return non-NULL for any Descriptor which can be found
466 : // through the provided pool.
467 : //
468 : // If the provided factory might return instances of protocol-compiler-
469 : // generated (i.e. compiled-in) types, or if the outer message object m is
470 : // a generated type, then the given factory MUST have this property: If
471 : // GetPrototype() is given a Descriptor which resides in
472 : // DescriptorPool::generated_pool(), the factory MUST return the same
473 : // prototype which MessageFactory::generated_factory() would return. That
474 : // is, given a descriptor for a generated type, the factory must return an
475 : // instance of the generated class (NOT DynamicMessage). However, when
476 : // given a descriptor for a type that is NOT in generated_pool, the factory
477 : // is free to return any implementation.
478 : //
479 : // The reason for this requirement is that generated sub-objects may be
480 : // accessed via the standard (non-reflection) extension accessor methods,
481 : // and these methods will down-cast the object to the generated class type.
482 : // If the object is not actually of that type, the results would be undefined.
483 : // On the other hand, if an extension is not compiled in, then there is no
484 : // way the code could end up accessing it via the standard accessors -- the
485 : // only way to access the extension is via reflection. When using reflection,
486 : // DynamicMessage and generated messages are indistinguishable, so it's fine
487 : // if these objects are represented using DynamicMessage.
488 : //
489 : // Using DynamicMessageFactory on which you have called
490 : // SetDelegateToGeneratedFactory(true) should be sufficient to satisfy the
491 : // above requirement.
492 : //
493 : // If either pool or factory is NULL, both must be NULL.
494 : //
495 : // Note that this feature is ignored when parsing "lite" messages as they do
496 : // not have descriptors.
497 : void SetExtensionRegistry(const DescriptorPool* pool,
498 : MessageFactory* factory);
499 :
500 : // Get the DescriptorPool set via SetExtensionRegistry(), or NULL if no pool
501 : // has been provided.
502 : const DescriptorPool* GetExtensionPool();
503 :
504 : // Get the MessageFactory set via SetExtensionRegistry(), or NULL if no
505 : // factory has been provided.
506 : MessageFactory* GetExtensionFactory();
507 :
508 : private:
509 : GOOGLE_DISALLOW_EVIL_CONSTRUCTORS(CodedInputStream);
510 :
511 : const uint8* buffer_;
512 : const uint8* buffer_end_; // pointer to the end of the buffer.
513 : ZeroCopyInputStream* input_;
514 : int total_bytes_read_; // total bytes read from input_, including
515 : // the current buffer
516 :
517 : // If total_bytes_read_ surpasses INT_MAX, we record the extra bytes here
518 : // so that we can BackUp() on destruction.
519 : int overflow_bytes_;
520 :
521 : // LastTagWas() stuff.
522 : uint32 last_tag_; // result of last ReadTag() or ReadTagWithCutoff().
523 :
524 : // This is set true by ReadTag{Fallback/Slow}() if it is called when exactly
525 : // at EOF, or by ExpectAtEnd() when it returns true. This happens when we
526 : // reach the end of a message and attempt to read another tag.
527 : bool legitimate_message_end_;
528 :
529 : // See EnableAliasing().
530 : bool aliasing_enabled_;
531 :
532 : // Limits
533 : Limit current_limit_; // if position = -1, no limit is applied
534 :
535 : // For simplicity, if the current buffer crosses a limit (either a normal
536 : // limit created by PushLimit() or the total bytes limit), buffer_size_
537 : // only tracks the number of bytes before that limit. This field
538 : // contains the number of bytes after it. Note that this implies that if
539 : // buffer_size_ == 0 and buffer_size_after_limit_ > 0, we know we've
540 : // hit a limit. However, if both are zero, it doesn't necessarily mean
541 : // we aren't at a limit -- the buffer may have ended exactly at the limit.
542 : int buffer_size_after_limit_;
543 :
544 : // Maximum number of bytes to read, period. This is unrelated to
545 : // current_limit_. Set using SetTotalBytesLimit().
546 : int total_bytes_limit_;
547 :
548 : // If positive/0: Limit for bytes read after which a warning due to size
549 : // should be logged.
550 : // If -1: Printing of warning disabled. Can be set by client.
551 : // If -2: Internal: Limit has been reached, print full size when destructing.
552 : int total_bytes_warning_threshold_;
553 :
554 : // Current recursion budget, controlled by IncrementRecursionDepth() and
555 : // similar. Starts at recursion_limit_ and goes down: if this reaches
556 : // -1 we are over budget.
557 : int recursion_budget_;
558 : // Recursion depth limit, set by SetRecursionLimit().
559 : int recursion_limit_;
560 :
561 : // See SetExtensionRegistry().
562 : const DescriptorPool* extension_pool_;
563 : MessageFactory* extension_factory_;
564 :
565 : // Private member functions.
566 :
567 : // Advance the buffer by a given number of bytes.
568 : void Advance(int amount);
569 :
570 : // Back up input_ to the current buffer position.
571 : void BackUpInputToCurrentPosition();
572 :
573 : // Recomputes the value of buffer_size_after_limit_. Must be called after
574 : // current_limit_ or total_bytes_limit_ changes.
575 : void RecomputeBufferLimits();
576 :
577 : // Writes an error message saying that we hit total_bytes_limit_.
578 : void PrintTotalBytesLimitError();
579 :
580 : // Called when the buffer runs out to request more data. Implies an
581 : // Advance(BufferSize()).
582 : bool Refresh();
583 :
584 : // When parsing varints, we optimize for the common case of small values, and
585 : // then optimize for the case when the varint fits within the current buffer
586 : // piece. The Fallback method is used when we can't use the one-byte
587 : // optimization. The Slow method is yet another fallback when the buffer is
588 : // not large enough. Making the slow path out-of-line speeds up the common
589 : // case by 10-15%. The slow path is fairly uncommon: it only triggers when a
590 : // message crosses multiple buffers. Note: ReadVarint32Fallback() and
591 : // ReadVarint64Fallback() are called frequently and generally not inlined, so
592 : // they have been optimized to avoid "out" parameters. The former returns -1
593 : // if it fails and the uint32 it read otherwise. The latter has a bool
594 : // indicating success or failure as part of its return type.
595 : int64 ReadVarint32Fallback(uint32 first_byte_or_zero);
596 : std::pair<uint64, bool> ReadVarint64Fallback();
597 : bool ReadVarint32Slow(uint32* value);
598 : bool ReadVarint64Slow(uint64* value);
599 : bool ReadLittleEndian32Fallback(uint32* value);
600 : bool ReadLittleEndian64Fallback(uint64* value);
601 : // Fallback/slow methods for reading tags. These do not update last_tag_,
602 : // but will set legitimate_message_end_ if we are at the end of the input
603 : // stream.
604 : uint32 ReadTagFallback(uint32 first_byte_or_zero);
605 : uint32 ReadTagSlow();
606 : bool ReadStringFallback(string* buffer, int size);
607 :
608 : // Return the size of the buffer.
609 : int BufferSize() const;
610 :
611 : static const int kDefaultTotalBytesLimit = 64 << 20; // 64MB
612 :
613 : static const int kDefaultTotalBytesWarningThreshold = 32 << 20; // 32MB
614 :
615 : static int default_recursion_limit_; // 100 by default.
616 : };
617 :
618 : // Class which encodes and writes binary data which is composed of varint-
619 : // encoded integers and fixed-width pieces. Wraps a ZeroCopyOutputStream.
620 : // Most users will not need to deal with CodedOutputStream.
621 : //
622 : // Most methods of CodedOutputStream which return a bool return false if an
623 : // underlying I/O error occurs. Once such a failure occurs, the
624 : // CodedOutputStream is broken and is no longer useful. The Write* methods do
625 : // not return the stream status, but will invalidate the stream if an error
626 : // occurs. The client can probe HadError() to determine the status.
627 : //
628 : // Note that every method of CodedOutputStream which writes some data has
629 : // a corresponding static "ToArray" version. These versions write directly
630 : // to the provided buffer, returning a pointer past the last written byte.
631 : // They require that the buffer has sufficient capacity for the encoded data.
632 : // This allows an optimization where we check if an output stream has enough
633 : // space for an entire message before we start writing and, if there is, we
634 : // call only the ToArray methods to avoid doing bound checks for each
635 : // individual value.
636 : // i.e., in the example above:
637 : //
638 : // CodedOutputStream coded_output = new CodedOutputStream(raw_output);
639 : // int magic_number = 1234;
640 : // char text[] = "Hello world!";
641 : //
642 : // int coded_size = sizeof(magic_number) +
643 : // CodedOutputStream::VarintSize32(strlen(text)) +
644 : // strlen(text);
645 : //
646 : // uint8* buffer =
647 : // coded_output->GetDirectBufferForNBytesAndAdvance(coded_size);
648 : // if (buffer != NULL) {
649 : // // The output stream has enough space in the buffer: write directly to
650 : // // the array.
651 : // buffer = CodedOutputStream::WriteLittleEndian32ToArray(magic_number,
652 : // buffer);
653 : // buffer = CodedOutputStream::WriteVarint32ToArray(strlen(text), buffer);
654 : // buffer = CodedOutputStream::WriteRawToArray(text, strlen(text), buffer);
655 : // } else {
656 : // // Make bound-checked writes, which will ask the underlying stream for
657 : // // more space as needed.
658 : // coded_output->WriteLittleEndian32(magic_number);
659 : // coded_output->WriteVarint32(strlen(text));
660 : // coded_output->WriteRaw(text, strlen(text));
661 : // }
662 : //
663 : // delete coded_output;
664 : class LIBPROTOBUF_EXPORT CodedOutputStream {
665 : public:
666 : // Create an CodedOutputStream that writes to the given ZeroCopyOutputStream.
667 : explicit CodedOutputStream(ZeroCopyOutputStream* output);
668 :
669 : // Destroy the CodedOutputStream and position the underlying
670 : // ZeroCopyOutputStream immediately after the last byte written.
671 : ~CodedOutputStream();
672 :
673 : // Trims any unused space in the underlying buffer so that its size matches
674 : // the number of bytes written by this stream. The underlying buffer will
675 : // automatically be trimmed when this stream is destroyed; this call is only
676 : // necessary if the underlying buffer is accessed *before* the stream is
677 : // destroyed.
678 : void Trim();
679 :
680 : // Skips a number of bytes, leaving the bytes unmodified in the underlying
681 : // buffer. Returns false if an underlying write error occurs. This is
682 : // mainly useful with GetDirectBufferPointer().
683 : bool Skip(int count);
684 :
685 : // Sets *data to point directly at the unwritten part of the
686 : // CodedOutputStream's underlying buffer, and *size to the size of that
687 : // buffer, but does not advance the stream's current position. This will
688 : // always either produce a non-empty buffer or return false. If the caller
689 : // writes any data to this buffer, it should then call Skip() to skip over
690 : // the consumed bytes. This may be useful for implementing external fast
691 : // serialization routines for types of data not covered by the
692 : // CodedOutputStream interface.
693 : bool GetDirectBufferPointer(void** data, int* size);
694 :
695 : // If there are at least "size" bytes available in the current buffer,
696 : // returns a pointer directly into the buffer and advances over these bytes.
697 : // The caller may then write directly into this buffer (e.g. using the
698 : // *ToArray static methods) rather than go through CodedOutputStream. If
699 : // there are not enough bytes available, returns NULL. The return pointer is
700 : // invalidated as soon as any other non-const method of CodedOutputStream
701 : // is called.
702 : inline uint8* GetDirectBufferForNBytesAndAdvance(int size);
703 :
704 : // Write raw bytes, copying them from the given buffer.
705 : void WriteRaw(const void* buffer, int size);
706 : // Like WriteRaw() but will try to write aliased data if aliasing is
707 : // turned on.
708 : void WriteRawMaybeAliased(const void* data, int size);
709 : // Like WriteRaw() but writing directly to the target array.
710 : // This is _not_ inlined, as the compiler often optimizes memcpy into inline
711 : // copy loops. Since this gets called by every field with string or bytes
712 : // type, inlining may lead to a significant amount of code bloat, with only a
713 : // minor performance gain.
714 : static uint8* WriteRawToArray(const void* buffer, int size, uint8* target);
715 :
716 : // Equivalent to WriteRaw(str.data(), str.size()).
717 : void WriteString(const string& str);
718 : // Like WriteString() but writing directly to the target array.
719 : static uint8* WriteStringToArray(const string& str, uint8* target);
720 : // Write the varint-encoded size of str followed by str.
721 : static uint8* WriteStringWithSizeToArray(const string& str, uint8* target);
722 :
723 :
724 : // Instructs the CodedOutputStream to allow the underlying
725 : // ZeroCopyOutputStream to hold pointers to the original structure instead of
726 : // copying, if it supports it (i.e. output->AllowsAliasing() is true). If the
727 : // underlying stream does not support aliasing, then enabling it has no
728 : // affect. For now, this only affects the behavior of
729 : // WriteRawMaybeAliased().
730 : //
731 : // NOTE: It is caller's responsibility to ensure that the chunk of memory
732 : // remains live until all of the data has been consumed from the stream.
733 : void EnableAliasing(bool enabled);
734 :
735 : // Write a 32-bit little-endian integer.
736 : void WriteLittleEndian32(uint32 value);
737 : // Like WriteLittleEndian32() but writing directly to the target array.
738 : static uint8* WriteLittleEndian32ToArray(uint32 value, uint8* target);
739 : // Write a 64-bit little-endian integer.
740 : void WriteLittleEndian64(uint64 value);
741 : // Like WriteLittleEndian64() but writing directly to the target array.
742 : static uint8* WriteLittleEndian64ToArray(uint64 value, uint8* target);
743 :
744 : // Write an unsigned integer with Varint encoding. Writing a 32-bit value
745 : // is equivalent to casting it to uint64 and writing it as a 64-bit value,
746 : // but may be more efficient.
747 : void WriteVarint32(uint32 value);
748 : // Like WriteVarint32() but writing directly to the target array.
749 : static uint8* WriteVarint32ToArray(uint32 value, uint8* target);
750 : // Write an unsigned integer with Varint encoding.
751 : void WriteVarint64(uint64 value);
752 : // Like WriteVarint64() but writing directly to the target array.
753 : static uint8* WriteVarint64ToArray(uint64 value, uint8* target);
754 :
755 : // Equivalent to WriteVarint32() except when the value is negative,
756 : // in which case it must be sign-extended to a full 10 bytes.
757 : void WriteVarint32SignExtended(int32 value);
758 : // Like WriteVarint32SignExtended() but writing directly to the target array.
759 : static uint8* WriteVarint32SignExtendedToArray(int32 value, uint8* target);
760 :
761 : // This is identical to WriteVarint32(), but optimized for writing tags.
762 : // In particular, if the input is a compile-time constant, this method
763 : // compiles down to a couple instructions.
764 : // Always inline because otherwise the aformentioned optimization can't work,
765 : // but GCC by default doesn't want to inline this.
766 : void WriteTag(uint32 value);
767 : // Like WriteTag() but writing directly to the target array.
768 : GOOGLE_ATTRIBUTE_ALWAYS_INLINE static uint8* WriteTagToArray(uint32 value,
769 : uint8* target);
770 :
771 : // Returns the number of bytes needed to encode the given value as a varint.
772 : static int VarintSize32(uint32 value);
773 : // Returns the number of bytes needed to encode the given value as a varint.
774 : static int VarintSize64(uint64 value);
775 :
776 : // If negative, 10 bytes. Otheriwse, same as VarintSize32().
777 : static int VarintSize32SignExtended(int32 value);
778 :
779 : // Compile-time equivalent of VarintSize32().
780 : template <uint32 Value>
781 : struct StaticVarintSize32 {
782 : static const int value =
783 : (Value < (1 << 7))
784 : ? 1
785 : : (Value < (1 << 14))
786 : ? 2
787 : : (Value < (1 << 21))
788 : ? 3
789 : : (Value < (1 << 28))
790 : ? 4
791 : : 5;
792 : };
793 :
794 : // Returns the total number of bytes written since this object was created.
795 : inline int ByteCount() const;
796 :
797 : // Returns true if there was an underlying I/O error since this object was
798 : // created.
799 : bool HadError() const { return had_error_; }
800 :
801 : private:
802 : GOOGLE_DISALLOW_EVIL_CONSTRUCTORS(CodedOutputStream);
803 :
804 : ZeroCopyOutputStream* output_;
805 : uint8* buffer_;
806 : int buffer_size_;
807 : int total_bytes_; // Sum of sizes of all buffers seen so far.
808 : bool had_error_; // Whether an error occurred during output.
809 : bool aliasing_enabled_; // See EnableAliasing().
810 :
811 : // Advance the buffer by a given number of bytes.
812 : void Advance(int amount);
813 :
814 : // Called when the buffer runs out to request more data. Implies an
815 : // Advance(buffer_size_).
816 : bool Refresh();
817 :
818 : // Like WriteRaw() but may avoid copying if the underlying
819 : // ZeroCopyOutputStream supports it.
820 : void WriteAliasedRaw(const void* buffer, int size);
821 :
822 : // If this write might cross the end of the buffer, we compose the bytes first
823 : // then use WriteRaw().
824 : void WriteVarint32SlowPath(uint32 value);
825 :
826 : // Always-inlined versions of WriteVarint* functions so that code can be
827 : // reused, while still controlling size. For instance, WriteVarint32ToArray()
828 : // should not directly call this: since it is inlined itself, doing so
829 : // would greatly increase the size of generated code. Instead, it should call
830 : // WriteVarint32FallbackToArray. Meanwhile, WriteVarint32() is already
831 : // out-of-line, so it should just invoke this directly to avoid any extra
832 : // function call overhead.
833 : GOOGLE_ATTRIBUTE_ALWAYS_INLINE static uint8* WriteVarint64ToArrayInline(
834 : uint64 value, uint8* target);
835 :
836 : static int VarintSize32Fallback(uint32 value);
837 : };
838 :
839 : // inline methods ====================================================
840 : // The vast majority of varints are only one byte. These inline
841 : // methods optimize for that case.
842 :
843 4166152 : inline bool CodedInputStream::ReadVarint32(uint32* value) {
844 4166152 : uint32 v = 0;
845 4166152 : if (GOOGLE_PREDICT_TRUE(buffer_ < buffer_end_)) {
846 4167786 : v = *buffer_;
847 4167786 : if (v < 0x80) {
848 4125649 : *value = v;
849 4125649 : Advance(1);
850 4125674 : return true;
851 : }
852 : }
853 40839 : int64 result = ReadVarint32Fallback(v);
854 42225 : *value = static_cast<uint32>(result);
855 42225 : return result >= 0;
856 : }
857 :
858 1069 : inline bool CodedInputStream::ReadVarint64(uint64* value) {
859 1069 : if (GOOGLE_PREDICT_TRUE(buffer_ < buffer_end_) && *buffer_ < 0x80) {
860 974 : *value = *buffer_;
861 974 : Advance(1);
862 974 : return true;
863 : }
864 95 : std::pair<uint64, bool> p = ReadVarint64Fallback();
865 95 : *value = p.first;
866 95 : return p.second;
867 : }
868 :
869 : // static
870 : inline const uint8* CodedInputStream::ReadLittleEndian32FromArray(
871 : const uint8* buffer,
872 : uint32* value) {
873 : #if defined(PROTOBUF_LITTLE_ENDIAN)
874 : memcpy(value, buffer, sizeof(*value));
875 : return buffer + sizeof(*value);
876 : #else
877 : *value = (static_cast<uint32>(buffer[0]) ) |
878 : (static_cast<uint32>(buffer[1]) << 8) |
879 : (static_cast<uint32>(buffer[2]) << 16) |
880 : (static_cast<uint32>(buffer[3]) << 24);
881 : return buffer + sizeof(*value);
882 : #endif
883 : }
884 : // static
885 : inline const uint8* CodedInputStream::ReadLittleEndian64FromArray(
886 : const uint8* buffer,
887 : uint64* value) {
888 : #if defined(PROTOBUF_LITTLE_ENDIAN)
889 : memcpy(value, buffer, sizeof(*value));
890 : return buffer + sizeof(*value);
891 : #else
892 : uint32 part0 = (static_cast<uint32>(buffer[0]) ) |
893 : (static_cast<uint32>(buffer[1]) << 8) |
894 : (static_cast<uint32>(buffer[2]) << 16) |
895 : (static_cast<uint32>(buffer[3]) << 24);
896 : uint32 part1 = (static_cast<uint32>(buffer[4]) ) |
897 : (static_cast<uint32>(buffer[5]) << 8) |
898 : (static_cast<uint32>(buffer[6]) << 16) |
899 : (static_cast<uint32>(buffer[7]) << 24);
900 : *value = static_cast<uint64>(part0) |
901 : (static_cast<uint64>(part1) << 32);
902 : return buffer + sizeof(*value);
903 : #endif
904 : }
905 :
906 18 : inline bool CodedInputStream::ReadLittleEndian32(uint32* value) {
907 : #if defined(PROTOBUF_LITTLE_ENDIAN)
908 27 : if (GOOGLE_PREDICT_TRUE(BufferSize() >= static_cast<int>(sizeof(*value)))) {
909 9 : memcpy(value, buffer_, sizeof(*value));
910 9 : Advance(sizeof(*value));
911 9 : return true;
912 : } else {
913 0 : return ReadLittleEndian32Fallback(value);
914 : }
915 : #else
916 : return ReadLittleEndian32Fallback(value);
917 : #endif
918 : }
919 :
920 155 : inline bool CodedInputStream::ReadLittleEndian64(uint64* value) {
921 : #if defined(PROTOBUF_LITTLE_ENDIAN)
922 166 : if (GOOGLE_PREDICT_TRUE(BufferSize() >= static_cast<int>(sizeof(*value)))) {
923 144 : memcpy(value, buffer_, sizeof(*value));
924 144 : Advance(sizeof(*value));
925 144 : return true;
926 : } else {
927 0 : return ReadLittleEndian64Fallback(value);
928 : }
929 : #else
930 : return ReadLittleEndian64Fallback(value);
931 : #endif
932 : }
933 :
934 : inline uint32 CodedInputStream::ReadTag() {
935 94 : uint32 v = 0;
936 94 : if (GOOGLE_PREDICT_TRUE(buffer_ < buffer_end_)) {
937 48 : v = *buffer_;
938 48 : if (v < 0x80) {
939 30 : last_tag_ = v;
940 30 : Advance(1);
941 : return v;
942 : }
943 : }
944 64 : last_tag_ = ReadTagFallback(v);
945 : return last_tag_;
946 : }
947 :
948 : inline std::pair<uint32, bool> CodedInputStream::ReadTagWithCutoff(
949 : uint32 cutoff) {
950 : // In performance-sensitive code we can expect cutoff to be a compile-time
951 : // constant, and things like "cutoff >= kMax1ByteVarint" to be evaluated at
952 : // compile time.
953 6902612 : uint32 first_byte_or_zero = 0;
954 6902612 : if (GOOGLE_PREDICT_TRUE(buffer_ < buffer_end_)) {
955 : // Hot case: buffer_ non_empty, buffer_[0] in [1, 128).
956 : // TODO(gpike): Is it worth rearranging this? E.g., if the number of fields
957 : // is large enough then is it better to check for the two-byte case first?
958 4096423 : first_byte_or_zero = buffer_[0];
959 4096423 : if (static_cast<int8>(buffer_[0]) > 0) {
960 4095974 : const uint32 kMax1ByteVarint = 0x7f;
961 4095974 : uint32 tag = last_tag_ = buffer_[0];
962 4095974 : Advance(1);
963 4095918 : return std::make_pair(tag, cutoff >= kMax1ByteVarint || tag <= cutoff);
964 : }
965 : // Other hot case: cutoff >= 0x80, buffer_ has at least two bytes available,
966 : // and tag is two bytes. The latter is tested by bitwise-and-not of the
967 : // first byte and the second byte.
968 449 : if (cutoff >= 0x80 &&
969 898 : GOOGLE_PREDICT_TRUE(buffer_ + 1 < buffer_end_) &&
970 449 : GOOGLE_PREDICT_TRUE((buffer_[0] & ~buffer_[1]) >= 0x80)) {
971 378 : const uint32 kMax2ByteVarint = (0x7f << 7) + 0x7f;
972 378 : uint32 tag = last_tag_ = (1u << 7) * buffer_[1] + (buffer_[0] - 0x80);
973 378 : Advance(2);
974 : // It might make sense to test for tag == 0 now, but it is so rare that
975 : // that we don't bother. A varint-encoded 0 should be one byte unless
976 : // the encoder lost its mind. The second part of the return value of
977 : // this function is allowed to be either true or false if the tag is 0,
978 : // so we don't have to check for tag == 0. We may need to check whether
979 : // it exceeds cutoff.
980 378 : bool at_or_below_cutoff = cutoff >= kMax2ByteVarint || tag <= cutoff;
981 378 : return std::make_pair(tag, at_or_below_cutoff);
982 : }
983 : }
984 : // Slow path
985 2806260 : last_tag_ = ReadTagFallback(first_byte_or_zero);
986 2808369 : return std::make_pair(last_tag_, static_cast<uint32>(last_tag_ - 1) < cutoff);
987 : }
988 :
989 : inline bool CodedInputStream::LastTagWas(uint32 expected) {
990 : return last_tag_ == expected;
991 : }
992 :
993 2795976 : inline bool CodedInputStream::ConsumedEntireMessage() {
994 2795976 : return legitimate_message_end_;
995 : }
996 :
997 : inline bool CodedInputStream::ExpectTag(uint32 expected) {
998 1517882 : if (expected < (1 << 7)) {
999 1573464 : if (GOOGLE_PREDICT_TRUE(buffer_ < buffer_end_) && buffer_[0] == expected) {
1000 31754 : Advance(1);
1001 155 : return true;
1002 : } else {
1003 1517660 : return false;
1004 : }
1005 0 : } else if (expected < (1 << 14)) {
1006 1676 : if (GOOGLE_PREDICT_TRUE(BufferSize() >= 2) &&
1007 864 : buffer_[0] == static_cast<uint8>(expected | 0x80) &&
1008 140 : buffer_[1] == static_cast<uint8>(expected >> 7)) {
1009 140 : Advance(2);
1010 0 : return true;
1011 : } else {
1012 0 : return false;
1013 : }
1014 : } else {
1015 : // Don't bother optimizing for larger values.
1016 0 : return false;
1017 : }
1018 : }
1019 :
1020 : inline const uint8* CodedInputStream::ExpectTagFromArray(
1021 : const uint8* buffer, uint32 expected) {
1022 : if (expected < (1 << 7)) {
1023 : if (buffer[0] == expected) {
1024 : return buffer + 1;
1025 : }
1026 : } else if (expected < (1 << 14)) {
1027 : if (buffer[0] == static_cast<uint8>(expected | 0x80) &&
1028 : buffer[1] == static_cast<uint8>(expected >> 7)) {
1029 : return buffer + 2;
1030 : }
1031 : }
1032 : return NULL;
1033 : }
1034 :
1035 : inline void CodedInputStream::GetDirectBufferPointerInline(const void** data,
1036 : int* size) {
1037 : *data = buffer_;
1038 : *size = buffer_end_ - buffer_;
1039 : }
1040 :
1041 2556223 : inline bool CodedInputStream::ExpectAtEnd() {
1042 : // If we are at a limit we know no more bytes can be read. Otherwise, it's
1043 : // hard to say without calling Refresh(), and we'd rather not do that.
1044 :
1045 5114063 : if (buffer_ == buffer_end_ &&
1046 5112178 : ((buffer_size_after_limit_ != 0) ||
1047 2555634 : (total_bytes_read_ == current_limit_))) {
1048 1279372 : last_tag_ = 0; // Pretend we called ReadTag()...
1049 1279372 : legitimate_message_end_ = true; // ... and it hit EOF.
1050 1278084 : return true;
1051 : } else {
1052 1278139 : return false;
1053 : }
1054 : }
1055 :
1056 2908372 : inline int CodedInputStream::CurrentPosition() const {
1057 5816744 : return total_bytes_read_ - (BufferSize() + buffer_size_after_limit_);
1058 : }
1059 :
1060 : inline uint8* CodedOutputStream::GetDirectBufferForNBytesAndAdvance(int size) {
1061 38 : if (buffer_size_ < size) {
1062 : return NULL;
1063 : } else {
1064 5 : uint8* result = buffer_;
1065 5 : Advance(size);
1066 : return result;
1067 : }
1068 : }
1069 :
1070 6685675 : inline uint8* CodedOutputStream::WriteVarint32ToArray(uint32 value,
1071 : uint8* target) {
1072 14962807 : while (value >= 0x80) {
1073 43627 : *target = static_cast<uint8>(value | 0x80);
1074 43627 : value >>= 7;
1075 43627 : ++target;
1076 : }
1077 8233505 : *target = static_cast<uint8>(value);
1078 6685675 : return target + 1;
1079 : }
1080 :
1081 8 : inline void CodedOutputStream::WriteVarint32SignExtended(int32 value) {
1082 8 : if (value < 0) {
1083 0 : WriteVarint64(static_cast<uint64>(value));
1084 : } else {
1085 8 : WriteVarint32(static_cast<uint32>(value));
1086 : }
1087 8 : }
1088 :
1089 1298462 : inline uint8* CodedOutputStream::WriteVarint32SignExtendedToArray(
1090 : int32 value, uint8* target) {
1091 1298462 : if (value < 0) {
1092 6 : return WriteVarint64ToArray(static_cast<uint64>(value), target);
1093 : } else {
1094 1318759 : return WriteVarint32ToArray(static_cast<uint32>(value), target);
1095 : }
1096 : }
1097 :
1098 18 : inline uint8* CodedOutputStream::WriteLittleEndian32ToArray(uint32 value,
1099 : uint8* target) {
1100 : #if defined(PROTOBUF_LITTLE_ENDIAN)
1101 : memcpy(target, &value, sizeof(value));
1102 : #else
1103 : target[0] = static_cast<uint8>(value);
1104 : target[1] = static_cast<uint8>(value >> 8);
1105 : target[2] = static_cast<uint8>(value >> 16);
1106 : target[3] = static_cast<uint8>(value >> 24);
1107 : #endif
1108 18 : return target + sizeof(value);
1109 : }
1110 :
1111 153 : inline uint8* CodedOutputStream::WriteLittleEndian64ToArray(uint64 value,
1112 : uint8* target) {
1113 : #if defined(PROTOBUF_LITTLE_ENDIAN)
1114 133 : memcpy(target, &value, sizeof(value));
1115 : #else
1116 : uint32 part0 = static_cast<uint32>(value);
1117 : uint32 part1 = static_cast<uint32>(value >> 32);
1118 :
1119 : target[0] = static_cast<uint8>(part0);
1120 : target[1] = static_cast<uint8>(part0 >> 8);
1121 : target[2] = static_cast<uint8>(part0 >> 16);
1122 : target[3] = static_cast<uint8>(part0 >> 24);
1123 : target[4] = static_cast<uint8>(part1);
1124 : target[5] = static_cast<uint8>(part1 >> 8);
1125 : target[6] = static_cast<uint8>(part1 >> 16);
1126 : target[7] = static_cast<uint8>(part1 >> 24);
1127 : #endif
1128 153 : return target + sizeof(value);
1129 : }
1130 :
1131 177 : inline void CodedOutputStream::WriteVarint32(uint32 value) {
1132 177 : if (buffer_size_ >= 5) {
1133 : // Fast path: We have enough bytes left in the buffer to guarantee that
1134 : // this write won't cross the end, so we can skip the checks.
1135 172 : uint8* target = buffer_;
1136 172 : uint8* end = WriteVarint32ToArray(value, target);
1137 172 : int size = end - target;
1138 172 : Advance(size);
1139 : } else {
1140 5 : WriteVarint32SlowPath(value);
1141 : }
1142 177 : }
1143 :
1144 64 : inline void CodedOutputStream::WriteTag(uint32 value) {
1145 64 : WriteVarint32(value);
1146 64 : }
1147 :
1148 : inline uint8* CodedOutputStream::WriteTagToArray(
1149 : uint32 value, uint8* target) {
1150 4099526 : return WriteVarint32ToArray(value, target);
1151 : }
1152 :
1153 4098322 : inline int CodedOutputStream::VarintSize32(uint32 value) {
1154 4144594 : if (value < (1 << 7)) {
1155 4058689 : return 1;
1156 : } else {
1157 48742 : return VarintSize32Fallback(value);
1158 : }
1159 : }
1160 :
1161 1277392 : inline int CodedOutputStream::VarintSize32SignExtended(int32 value) {
1162 1300846 : if (value < 0) {
1163 0 : return 10; // TODO(kenton): Make this a symbolic constant.
1164 : } else {
1165 1300840 : return VarintSize32(static_cast<uint32>(value));
1166 : }
1167 : }
1168 :
1169 : inline void CodedOutputStream::WriteString(const string& str) {
1170 26 : WriteRaw(str.data(), static_cast<int>(str.size()));
1171 : }
1172 :
1173 40 : inline void CodedOutputStream::WriteRawMaybeAliased(
1174 : const void* data, int size) {
1175 40 : if (aliasing_enabled_) {
1176 0 : WriteAliasedRaw(data, size);
1177 : } else {
1178 40 : WriteRaw(data, size);
1179 : }
1180 40 : }
1181 :
1182 : inline uint8* CodedOutputStream::WriteStringToArray(
1183 : const string& str, uint8* target) {
1184 1524209 : return WriteRawToArray(str.data(), static_cast<int>(str.size()), target);
1185 : }
1186 :
1187 : inline int CodedOutputStream::ByteCount() const {
1188 66 : return total_bytes_ - buffer_size_;
1189 : }
1190 :
1191 8208072 : inline void CodedInputStream::Advance(int amount) {
1192 9787652 : buffer_ += amount;
1193 8208072 : }
1194 :
1195 0 : inline void CodedOutputStream::Advance(int amount) {
1196 281 : buffer_ += amount;
1197 281 : buffer_size_ -= amount;
1198 0 : }
1199 :
1200 : inline void CodedInputStream::SetRecursionLimit(int limit) {
1201 : recursion_budget_ += limit - recursion_limit_;
1202 : recursion_limit_ = limit;
1203 : }
1204 :
1205 0 : inline bool CodedInputStream::IncrementRecursionDepth() {
1206 3084 : --recursion_budget_;
1207 0 : return recursion_budget_ >= 0;
1208 : }
1209 :
1210 : inline void CodedInputStream::DecrementRecursionDepth() {
1211 1 : if (recursion_budget_ < recursion_limit_) ++recursion_budget_;
1212 : }
1213 :
1214 0 : inline void CodedInputStream::UnsafeDecrementRecursionDepth() {
1215 0 : assert(recursion_budget_ < recursion_limit_);
1216 3083 : ++recursion_budget_;
1217 0 : }
1218 :
1219 : inline void CodedInputStream::SetExtensionRegistry(const DescriptorPool* pool,
1220 : MessageFactory* factory) {
1221 : extension_pool_ = pool;
1222 : extension_factory_ = factory;
1223 : }
1224 :
1225 : inline const DescriptorPool* CodedInputStream::GetExtensionPool() {
1226 : return extension_pool_;
1227 : }
1228 :
1229 : inline MessageFactory* CodedInputStream::GetExtensionFactory() {
1230 : return extension_factory_;
1231 : }
1232 :
1233 1533814 : inline int CodedInputStream::BufferSize() const {
1234 8669455 : return buffer_end_ - buffer_;
1235 : }
1236 :
1237 2794044 : inline CodedInputStream::CodedInputStream(ZeroCopyInputStream* input)
1238 : : buffer_(NULL),
1239 : buffer_end_(NULL),
1240 : input_(input),
1241 : total_bytes_read_(0),
1242 : overflow_bytes_(0),
1243 : last_tag_(0),
1244 : legitimate_message_end_(false),
1245 : aliasing_enabled_(false),
1246 : current_limit_(kint32max),
1247 : buffer_size_after_limit_(0),
1248 : total_bytes_limit_(kDefaultTotalBytesLimit),
1249 : total_bytes_warning_threshold_(kDefaultTotalBytesWarningThreshold),
1250 : recursion_budget_(default_recursion_limit_),
1251 : recursion_limit_(default_recursion_limit_),
1252 : extension_pool_(NULL),
1253 2794044 : extension_factory_(NULL) {
1254 : // Eagerly Refresh() so buffer space is immediately available.
1255 2794044 : Refresh();
1256 2794702 : }
1257 :
1258 830 : inline CodedInputStream::CodedInputStream(const uint8* buffer, int size)
1259 : : buffer_(buffer),
1260 830 : buffer_end_(buffer + size),
1261 : input_(NULL),
1262 : total_bytes_read_(size),
1263 : overflow_bytes_(0),
1264 : last_tag_(0),
1265 : legitimate_message_end_(false),
1266 : aliasing_enabled_(false),
1267 : current_limit_(size),
1268 : buffer_size_after_limit_(0),
1269 : total_bytes_limit_(kDefaultTotalBytesLimit),
1270 : total_bytes_warning_threshold_(kDefaultTotalBytesWarningThreshold),
1271 : recursion_budget_(default_recursion_limit_),
1272 : recursion_limit_(default_recursion_limit_),
1273 : extension_pool_(NULL),
1274 1660 : extension_factory_(NULL) {
1275 : // Note that setting current_limit_ == size is important to prevent some
1276 : // code paths from trying to access input_ and segfaulting.
1277 830 : }
1278 :
1279 : inline bool CodedInputStream::IsFlat() const {
1280 : return input_ == NULL;
1281 : }
1282 :
1283 : } // namespace io
1284 : } // namespace protobuf
1285 :
1286 :
1287 : #if defined(_MSC_VER) && _MSC_VER >= 1300
1288 : #pragma runtime_checks("c", restore)
1289 : #endif // _MSC_VER
1290 :
1291 : } // namespace google
1292 : #endif // GOOGLE_PROTOBUF_IO_CODED_STREAM_H__
|