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_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 | #ifdef _MSC_VER |
114 | #if defined(_M_IX86) && \ |
115 | !defined(PROTOBUF_DISABLE_LITTLE_ENDIAN_OPT_FOR_TEST) |
116 | #define PROTOBUF_LITTLE_ENDIAN 1 |
117 | #endif |
118 | #if _MSC_VER >= 1300 |
119 | // If MSVC has "/RTCc" set, it will complain about truncating casts at |
120 | // runtime. This file contains some intentional truncating casts. |
121 | #pragma runtime_checks("c", off) |
122 | #endif |
123 | #else |
124 | #include <sys/param.h> // __BYTE_ORDER |
125 | #if defined(__BYTE_ORDER) && __BYTE_ORDER == __LITTLE_ENDIAN && \ |
126 | !defined(PROTOBUF_DISABLE_LITTLE_ENDIAN_OPT_FOR_TEST) |
127 | #define PROTOBUF_LITTLE_ENDIAN 1 |
128 | #endif |
129 | #endif |
130 | #include <google/protobuf/stubs/common.h> |
131 | |
132 | |
133 | namespace google { |
134 | namespace protobuf { |
135 | |
136 | class DescriptorPool; |
137 | class MessageFactory; |
138 | |
139 | namespace io { |
140 | |
141 | // Defined in this file. |
142 | class CodedInputStream; |
143 | class CodedOutputStream; |
144 | |
145 | // Defined in other files. |
146 | class ZeroCopyInputStream; // zero_copy_stream.h |
147 | class ZeroCopyOutputStream; // zero_copy_stream.h |
148 | |
149 | // Class which reads and decodes binary data which is composed of varint- |
150 | // encoded integers and fixed-width pieces. Wraps a ZeroCopyInputStream. |
151 | // Most users will not need to deal with CodedInputStream. |
152 | // |
153 | // Most methods of CodedInputStream that return a bool return false if an |
154 | // underlying I/O error occurs or if the data is malformed. Once such a |
155 | // failure occurs, the CodedInputStream is broken and is no longer useful. |
156 | class LIBPROTOBUF_EXPORT CodedInputStream { |
157 | public: |
158 | // Create a CodedInputStream that reads from the given ZeroCopyInputStream. |
159 | explicit CodedInputStream(ZeroCopyInputStream* input); |
160 | |
161 | // Create a CodedInputStream that reads from the given flat array. This is |
162 | // faster than using an ArrayInputStream. PushLimit(size) is implied by |
163 | // this constructor. |
164 | explicit CodedInputStream(const uint8* buffer, int size); |
165 | |
166 | // Destroy the CodedInputStream and position the underlying |
167 | // ZeroCopyInputStream at the first unread byte. If an error occurred while |
168 | // reading (causing a method to return false), then the exact position of |
169 | // the input stream may be anywhere between the last value that was read |
170 | // successfully and the stream's byte limit. |
171 | ~CodedInputStream(); |
172 | |
173 | // Return true if this CodedInputStream reads from a flat array instead of |
174 | // a ZeroCopyInputStream. |
175 | inline bool IsFlat() const; |
176 | |
177 | // Skips a number of bytes. Returns false if an underlying read error |
178 | // occurs. |
179 | bool Skip(int count); |
180 | |
181 | // Sets *data to point directly at the unread part of the CodedInputStream's |
182 | // underlying buffer, and *size to the size of that buffer, but does not |
183 | // advance the stream's current position. This will always either produce |
184 | // a non-empty buffer or return false. If the caller consumes any of |
185 | // this data, it should then call Skip() to skip over the consumed bytes. |
186 | // This may be useful for implementing external fast parsing routines for |
187 | // types of data not covered by the CodedInputStream interface. |
188 | bool GetDirectBufferPointer(const void** data, int* size); |
189 | |
190 | // Like GetDirectBufferPointer, but this method is inlined, and does not |
191 | // attempt to Refresh() if the buffer is currently empty. |
192 | inline void GetDirectBufferPointerInline(const void** data, |
193 | int* size) GOOGLE_ATTRIBUTE_ALWAYS_INLINE; |
194 | |
195 | // Read raw bytes, copying them into the given buffer. |
196 | bool ReadRaw(void* buffer, int size); |
197 | |
198 | // Like ReadRaw, but reads into a string. |
199 | // |
200 | // Implementation Note: ReadString() grows the string gradually as it |
201 | // reads in the data, rather than allocating the entire requested size |
202 | // upfront. This prevents denial-of-service attacks in which a client |
203 | // could claim that a string is going to be MAX_INT bytes long in order to |
204 | // crash the server because it can't allocate this much space at once. |
205 | bool ReadString(string* buffer, int size); |
206 | // Like the above, with inlined optimizations. This should only be used |
207 | // by the protobuf implementation. |
208 | inline bool InternalReadStringInline(string* buffer, |
209 | int size) GOOGLE_ATTRIBUTE_ALWAYS_INLINE; |
210 | |
211 | |
212 | // Read a 32-bit little-endian integer. |
213 | bool ReadLittleEndian32(uint32* value); |
214 | // Read a 64-bit little-endian integer. |
215 | bool ReadLittleEndian64(uint64* value); |
216 | |
217 | // These methods read from an externally provided buffer. The caller is |
218 | // responsible for ensuring that the buffer has sufficient space. |
219 | // Read a 32-bit little-endian integer. |
220 | static const uint8* ReadLittleEndian32FromArray(const uint8* buffer, |
221 | uint32* value); |
222 | // Read a 64-bit little-endian integer. |
223 | static const uint8* ReadLittleEndian64FromArray(const uint8* buffer, |
224 | uint64* value); |
225 | |
226 | // Read an unsigned integer with Varint encoding, truncating to 32 bits. |
227 | // Reading a 32-bit value is equivalent to reading a 64-bit one and casting |
228 | // it to uint32, but may be more efficient. |
229 | bool ReadVarint32(uint32* value); |
230 | // Read an unsigned integer with Varint encoding. |
231 | bool ReadVarint64(uint64* value); |
232 | |
233 | // Read a tag. This calls ReadVarint32() and returns the result, or returns |
234 | // zero (which is not a valid tag) if ReadVarint32() fails. Also, it updates |
235 | // the last tag value, which can be checked with LastTagWas(). |
236 | // Always inline because this is only called in one place per parse loop |
237 | // but it is called for every iteration of said loop, so it should be fast. |
238 | // GCC doesn't want to inline this by default. |
239 | uint32 ReadTag() GOOGLE_ATTRIBUTE_ALWAYS_INLINE; |
240 | |
241 | // This usually a faster alternative to ReadTag() when cutoff is a manifest |
242 | // constant. It does particularly well for cutoff >= 127. The first part |
243 | // of the return value is the tag that was read, though it can also be 0 in |
244 | // the cases where ReadTag() would return 0. If the second part is true |
245 | // then the tag is known to be in [0, cutoff]. If not, the tag either is |
246 | // above cutoff or is 0. (There's intentional wiggle room when tag is 0, |
247 | // because that can arise in several ways, and for best performance we want |
248 | // to avoid an extra "is tag == 0?" check here.) |
249 | inline std::pair<uint32, bool> ReadTagWithCutoff(uint32 cutoff) |
250 | GOOGLE_ATTRIBUTE_ALWAYS_INLINE; |
251 | |
252 | // Usually returns true if calling ReadVarint32() now would produce the given |
253 | // value. Will always return false if ReadVarint32() would not return the |
254 | // given value. If ExpectTag() returns true, it also advances past |
255 | // the varint. For best performance, use a compile-time constant as the |
256 | // parameter. |
257 | // Always inline because this collapses to a small number of instructions |
258 | // when given a constant parameter, but GCC doesn't want to inline by default. |
259 | bool ExpectTag(uint32 expected) GOOGLE_ATTRIBUTE_ALWAYS_INLINE; |
260 | |
261 | // Like above, except this reads from the specified buffer. The caller is |
262 | // responsible for ensuring that the buffer is large enough to read a varint |
263 | // of the expected size. For best performance, use a compile-time constant as |
264 | // the expected tag parameter. |
265 | // |
266 | // Returns a pointer beyond the expected tag if it was found, or NULL if it |
267 | // was not. |
268 | static const uint8* ExpectTagFromArray( |
269 | const uint8* buffer, |
270 | uint32 expected) GOOGLE_ATTRIBUTE_ALWAYS_INLINE; |
271 | |
272 | // Usually returns true if no more bytes can be read. Always returns false |
273 | // if more bytes can be read. If ExpectAtEnd() returns true, a subsequent |
274 | // call to LastTagWas() will act as if ReadTag() had been called and returned |
275 | // zero, and ConsumedEntireMessage() will return true. |
276 | bool ExpectAtEnd(); |
277 | |
278 | // If the last call to ReadTag() or ReadTagWithCutoff() returned the |
279 | // given value, returns true. Otherwise, returns false; |
280 | // |
281 | // This is needed because parsers for some types of embedded messages |
282 | // (with field type TYPE_GROUP) don't actually know that they've reached the |
283 | // end of a message until they see an ENDGROUP tag, which was actually part |
284 | // of the enclosing message. The enclosing message would like to check that |
285 | // tag to make sure it had the right number, so it calls LastTagWas() on |
286 | // return from the embedded parser to check. |
287 | bool LastTagWas(uint32 expected); |
288 | |
289 | // When parsing message (but NOT a group), this method must be called |
290 | // immediately after MergeFromCodedStream() returns (if it returns true) |
291 | // to further verify that the message ended in a legitimate way. For |
292 | // example, this verifies that parsing did not end on an end-group tag. |
293 | // It also checks for some cases where, due to optimizations, |
294 | // MergeFromCodedStream() can incorrectly return true. |
295 | bool ConsumedEntireMessage(); |
296 | |
297 | // Limits ---------------------------------------------------------- |
298 | // Limits are used when parsing length-delimited embedded messages. |
299 | // After the message's length is read, PushLimit() is used to prevent |
300 | // the CodedInputStream from reading beyond that length. Once the |
301 | // embedded message has been parsed, PopLimit() is called to undo the |
302 | // limit. |
303 | |
304 | // Opaque type used with PushLimit() and PopLimit(). Do not modify |
305 | // values of this type yourself. The only reason that this isn't a |
306 | // struct with private internals is for efficiency. |
307 | typedef int Limit; |
308 | |
309 | // Places a limit on the number of bytes that the stream may read, |
310 | // starting from the current position. Once the stream hits this limit, |
311 | // it will act like the end of the input has been reached until PopLimit() |
312 | // is called. |
313 | // |
314 | // As the names imply, the stream conceptually has a stack of limits. The |
315 | // shortest limit on the stack is always enforced, even if it is not the |
316 | // top limit. |
317 | // |
318 | // The value returned by PushLimit() is opaque to the caller, and must |
319 | // be passed unchanged to the corresponding call to PopLimit(). |
320 | Limit PushLimit(int byte_limit); |
321 | |
322 | // Pops the last limit pushed by PushLimit(). The input must be the value |
323 | // returned by that call to PushLimit(). |
324 | void PopLimit(Limit limit); |
325 | |
326 | // Returns the number of bytes left until the nearest limit on the |
327 | // stack is hit, or -1 if no limits are in place. |
328 | int BytesUntilLimit() const; |
329 | |
330 | // Returns current position relative to the beginning of the input stream. |
331 | int CurrentPosition() const; |
332 | |
333 | // Total Bytes Limit ----------------------------------------------- |
334 | // To prevent malicious users from sending excessively large messages |
335 | // and causing integer overflows or memory exhaustion, CodedInputStream |
336 | // imposes a hard limit on the total number of bytes it will read. |
337 | |
338 | // Sets the maximum number of bytes that this CodedInputStream will read |
339 | // before refusing to continue. To prevent integer overflows in the |
340 | // protocol buffers implementation, as well as to prevent servers from |
341 | // allocating enormous amounts of memory to hold parsed messages, the |
342 | // maximum message length should be limited to the shortest length that |
343 | // will not harm usability. The theoretical shortest message that could |
344 | // cause integer overflows is 512MB. The default limit is 64MB. Apps |
345 | // should set shorter limits if possible. If warning_threshold is not -1, |
346 | // a warning will be printed to stderr after warning_threshold bytes are |
347 | // read. For backwards compatibility all negative values get squashed to -1, |
348 | // as other negative values might have special internal meanings. |
349 | // An error will always be printed to stderr if the limit is reached. |
350 | // |
351 | // This is unrelated to PushLimit()/PopLimit(). |
352 | // |
353 | // Hint: If you are reading this because your program is printing a |
354 | // warning about dangerously large protocol messages, you may be |
355 | // confused about what to do next. The best option is to change your |
356 | // design such that excessively large messages are not necessary. |
357 | // For example, try to design file formats to consist of many small |
358 | // messages rather than a single large one. If this is infeasible, |
359 | // you will need to increase the limit. Chances are, though, that |
360 | // your code never constructs a CodedInputStream on which the limit |
361 | // can be set. You probably parse messages by calling things like |
362 | // Message::ParseFromString(). In this case, you will need to change |
363 | // your code to instead construct some sort of ZeroCopyInputStream |
364 | // (e.g. an ArrayInputStream), construct a CodedInputStream around |
365 | // that, then call Message::ParseFromCodedStream() instead. Then |
366 | // you can adjust the limit. Yes, it's more work, but you're doing |
367 | // something unusual. |
368 | void SetTotalBytesLimit(int total_bytes_limit, int warning_threshold); |
369 | |
370 | // The Total Bytes Limit minus the Current Position, or -1 if there |
371 | // is no Total Bytes Limit. |
372 | int BytesUntilTotalBytesLimit() const; |
373 | |
374 | // Recursion Limit ------------------------------------------------- |
375 | // To prevent corrupt or malicious messages from causing stack overflows, |
376 | // we must keep track of the depth of recursion when parsing embedded |
377 | // messages and groups. CodedInputStream keeps track of this because it |
378 | // is the only object that is passed down the stack during parsing. |
379 | |
380 | // Sets the maximum recursion depth. The default is 100. |
381 | void SetRecursionLimit(int limit); |
382 | |
383 | |
384 | // Increments the current recursion depth. Returns true if the depth is |
385 | // under the limit, false if it has gone over. |
386 | bool IncrementRecursionDepth(); |
387 | |
388 | // Decrements the recursion depth. |
389 | void DecrementRecursionDepth(); |
390 | |
391 | // Extension Registry ---------------------------------------------- |
392 | // ADVANCED USAGE: 99.9% of people can ignore this section. |
393 | // |
394 | // By default, when parsing extensions, the parser looks for extension |
395 | // definitions in the pool which owns the outer message's Descriptor. |
396 | // However, you may call SetExtensionRegistry() to provide an alternative |
397 | // pool instead. This makes it possible, for example, to parse a message |
398 | // using a generated class, but represent some extensions using |
399 | // DynamicMessage. |
400 | |
401 | // Set the pool used to look up extensions. Most users do not need to call |
402 | // this as the correct pool will be chosen automatically. |
403 | // |
404 | // WARNING: It is very easy to misuse this. Carefully read the requirements |
405 | // below. Do not use this unless you are sure you need it. Almost no one |
406 | // does. |
407 | // |
408 | // Let's say you are parsing a message into message object m, and you want |
409 | // to take advantage of SetExtensionRegistry(). You must follow these |
410 | // requirements: |
411 | // |
412 | // The given DescriptorPool must contain m->GetDescriptor(). It is not |
413 | // sufficient for it to simply contain a descriptor that has the same name |
414 | // and content -- it must be the *exact object*. In other words: |
415 | // assert(pool->FindMessageTypeByName(m->GetDescriptor()->full_name()) == |
416 | // m->GetDescriptor()); |
417 | // There are two ways to satisfy this requirement: |
418 | // 1) Use m->GetDescriptor()->pool() as the pool. This is generally useless |
419 | // because this is the pool that would be used anyway if you didn't call |
420 | // SetExtensionRegistry() at all. |
421 | // 2) Use a DescriptorPool which has m->GetDescriptor()->pool() as an |
422 | // "underlay". Read the documentation for DescriptorPool for more |
423 | // information about underlays. |
424 | // |
425 | // You must also provide a MessageFactory. This factory will be used to |
426 | // construct Message objects representing extensions. The factory's |
427 | // GetPrototype() MUST return non-NULL for any Descriptor which can be found |
428 | // through the provided pool. |
429 | // |
430 | // If the provided factory might return instances of protocol-compiler- |
431 | // generated (i.e. compiled-in) types, or if the outer message object m is |
432 | // a generated type, then the given factory MUST have this property: If |
433 | // GetPrototype() is given a Descriptor which resides in |
434 | // DescriptorPool::generated_pool(), the factory MUST return the same |
435 | // prototype which MessageFactory::generated_factory() would return. That |
436 | // is, given a descriptor for a generated type, the factory must return an |
437 | // instance of the generated class (NOT DynamicMessage). However, when |
438 | // given a descriptor for a type that is NOT in generated_pool, the factory |
439 | // is free to return any implementation. |
440 | // |
441 | // The reason for this requirement is that generated sub-objects may be |
442 | // accessed via the standard (non-reflection) extension accessor methods, |
443 | // and these methods will down-cast the object to the generated class type. |
444 | // If the object is not actually of that type, the results would be undefined. |
445 | // On the other hand, if an extension is not compiled in, then there is no |
446 | // way the code could end up accessing it via the standard accessors -- the |
447 | // only way to access the extension is via reflection. When using reflection, |
448 | // DynamicMessage and generated messages are indistinguishable, so it's fine |
449 | // if these objects are represented using DynamicMessage. |
450 | // |
451 | // Using DynamicMessageFactory on which you have called |
452 | // SetDelegateToGeneratedFactory(true) should be sufficient to satisfy the |
453 | // above requirement. |
454 | // |
455 | // If either pool or factory is NULL, both must be NULL. |
456 | // |
457 | // Note that this feature is ignored when parsing "lite" messages as they do |
458 | // not have descriptors. |
459 | void SetExtensionRegistry(const DescriptorPool* pool, |
460 | MessageFactory* factory); |
461 | |
462 | // Get the DescriptorPool set via SetExtensionRegistry(), or NULL if no pool |
463 | // has been provided. |
464 | const DescriptorPool* GetExtensionPool(); |
465 | |
466 | // Get the MessageFactory set via SetExtensionRegistry(), or NULL if no |
467 | // factory has been provided. |
468 | MessageFactory* GetExtensionFactory(); |
469 | |
470 | private: |
471 | GOOGLE_DISALLOW_EVIL_CONSTRUCTORS(CodedInputStream); |
472 | |
473 | ZeroCopyInputStream* input_; |
474 | const uint8* buffer_; |
475 | const uint8* buffer_end_; // pointer to the end of the buffer. |
476 | int total_bytes_read_; // total bytes read from input_, including |
477 | // the current buffer |
478 | |
479 | // If total_bytes_read_ surpasses INT_MAX, we record the extra bytes here |
480 | // so that we can BackUp() on destruction. |
481 | int overflow_bytes_; |
482 | |
483 | // LastTagWas() stuff. |
484 | uint32 last_tag_; // result of last ReadTag() or ReadTagWithCutoff(). |
485 | |
486 | // This is set true by ReadTag{Fallback/Slow}() if it is called when exactly |
487 | // at EOF, or by ExpectAtEnd() when it returns true. This happens when we |
488 | // reach the end of a message and attempt to read another tag. |
489 | bool legitimate_message_end_; |
490 | |
491 | // See EnableAliasing(). |
492 | bool aliasing_enabled_; |
493 | |
494 | // Limits |
495 | Limit current_limit_; // if position = -1, no limit is applied |
496 | |
497 | // For simplicity, if the current buffer crosses a limit (either a normal |
498 | // limit created by PushLimit() or the total bytes limit), buffer_size_ |
499 | // only tracks the number of bytes before that limit. This field |
500 | // contains the number of bytes after it. Note that this implies that if |
501 | // buffer_size_ == 0 and buffer_size_after_limit_ > 0, we know we've |
502 | // hit a limit. However, if both are zero, it doesn't necessarily mean |
503 | // we aren't at a limit -- the buffer may have ended exactly at the limit. |
504 | int buffer_size_after_limit_; |
505 | |
506 | // Maximum number of bytes to read, period. This is unrelated to |
507 | // current_limit_. Set using SetTotalBytesLimit(). |
508 | int total_bytes_limit_; |
509 | |
510 | // If positive/0: Limit for bytes read after which a warning due to size |
511 | // should be logged. |
512 | // If -1: Printing of warning disabled. Can be set by client. |
513 | // If -2: Internal: Limit has been reached, print full size when destructing. |
514 | int total_bytes_warning_threshold_; |
515 | |
516 | // Current recursion depth, controlled by IncrementRecursionDepth() and |
517 | // DecrementRecursionDepth(). |
518 | int recursion_depth_; |
519 | // Recursion depth limit, set by SetRecursionLimit(). |
520 | int recursion_limit_; |
521 | |
522 | // See SetExtensionRegistry(). |
523 | const DescriptorPool* extension_pool_; |
524 | MessageFactory* extension_factory_; |
525 | |
526 | // Private member functions. |
527 | |
528 | // Advance the buffer by a given number of bytes. |
529 | void Advance(int amount); |
530 | |
531 | // Back up input_ to the current buffer position. |
532 | void BackUpInputToCurrentPosition(); |
533 | |
534 | // Recomputes the value of buffer_size_after_limit_. Must be called after |
535 | // current_limit_ or total_bytes_limit_ changes. |
536 | void RecomputeBufferLimits(); |
537 | |
538 | // Writes an error message saying that we hit total_bytes_limit_. |
539 | void PrintTotalBytesLimitError(); |
540 | |
541 | // Called when the buffer runs out to request more data. Implies an |
542 | // Advance(BufferSize()). |
543 | bool Refresh(); |
544 | |
545 | // When parsing varints, we optimize for the common case of small values, and |
546 | // then optimize for the case when the varint fits within the current buffer |
547 | // piece. The Fallback method is used when we can't use the one-byte |
548 | // optimization. The Slow method is yet another fallback when the buffer is |
549 | // not large enough. Making the slow path out-of-line speeds up the common |
550 | // case by 10-15%. The slow path is fairly uncommon: it only triggers when a |
551 | // message crosses multiple buffers. |
552 | bool ReadVarint32Fallback(uint32* value); |
553 | bool ReadVarint64Fallback(uint64* value); |
554 | bool ReadVarint32Slow(uint32* value); |
555 | bool ReadVarint64Slow(uint64* value); |
556 | bool ReadLittleEndian32Fallback(uint32* value); |
557 | bool ReadLittleEndian64Fallback(uint64* value); |
558 | // Fallback/slow methods for reading tags. These do not update last_tag_, |
559 | // but will set legitimate_message_end_ if we are at the end of the input |
560 | // stream. |
561 | uint32 ReadTagFallback(); |
562 | uint32 ReadTagSlow(); |
563 | bool ReadStringFallback(string* buffer, int size); |
564 | |
565 | // Return the size of the buffer. |
566 | int BufferSize() const; |
567 | |
568 | static const int kDefaultTotalBytesLimit = 64 << 20; // 64MB |
569 | |
570 | static const int kDefaultTotalBytesWarningThreshold = 32 << 20; // 32MB |
571 | |
572 | static int default_recursion_limit_; // 100 by default. |
573 | }; |
574 | |
575 | // Class which encodes and writes binary data which is composed of varint- |
576 | // encoded integers and fixed-width pieces. Wraps a ZeroCopyOutputStream. |
577 | // Most users will not need to deal with CodedOutputStream. |
578 | // |
579 | // Most methods of CodedOutputStream which return a bool return false if an |
580 | // underlying I/O error occurs. Once such a failure occurs, the |
581 | // CodedOutputStream is broken and is no longer useful. The Write* methods do |
582 | // not return the stream status, but will invalidate the stream if an error |
583 | // occurs. The client can probe HadError() to determine the status. |
584 | // |
585 | // Note that every method of CodedOutputStream which writes some data has |
586 | // a corresponding static "ToArray" version. These versions write directly |
587 | // to the provided buffer, returning a pointer past the last written byte. |
588 | // They require that the buffer has sufficient capacity for the encoded data. |
589 | // This allows an optimization where we check if an output stream has enough |
590 | // space for an entire message before we start writing and, if there is, we |
591 | // call only the ToArray methods to avoid doing bound checks for each |
592 | // individual value. |
593 | // i.e., in the example above: |
594 | // |
595 | // CodedOutputStream coded_output = new CodedOutputStream(raw_output); |
596 | // int magic_number = 1234; |
597 | // char text[] = "Hello world!"; |
598 | // |
599 | // int coded_size = sizeof(magic_number) + |
600 | // CodedOutputStream::VarintSize32(strlen(text)) + |
601 | // strlen(text); |
602 | // |
603 | // uint8* buffer = |
604 | // coded_output->GetDirectBufferForNBytesAndAdvance(coded_size); |
605 | // if (buffer != NULL) { |
606 | // // The output stream has enough space in the buffer: write directly to |
607 | // // the array. |
608 | // buffer = CodedOutputStream::WriteLittleEndian32ToArray(magic_number, |
609 | // buffer); |
610 | // buffer = CodedOutputStream::WriteVarint32ToArray(strlen(text), buffer); |
611 | // buffer = CodedOutputStream::WriteRawToArray(text, strlen(text), buffer); |
612 | // } else { |
613 | // // Make bound-checked writes, which will ask the underlying stream for |
614 | // // more space as needed. |
615 | // coded_output->WriteLittleEndian32(magic_number); |
616 | // coded_output->WriteVarint32(strlen(text)); |
617 | // coded_output->WriteRaw(text, strlen(text)); |
618 | // } |
619 | // |
620 | // delete coded_output; |
621 | class LIBPROTOBUF_EXPORT CodedOutputStream { |
622 | public: |
623 | // Create an CodedOutputStream that writes to the given ZeroCopyOutputStream. |
624 | explicit CodedOutputStream(ZeroCopyOutputStream* output); |
625 | |
626 | // Destroy the CodedOutputStream and position the underlying |
627 | // ZeroCopyOutputStream immediately after the last byte written. |
628 | ~CodedOutputStream(); |
629 | |
630 | // Skips a number of bytes, leaving the bytes unmodified in the underlying |
631 | // buffer. Returns false if an underlying write error occurs. This is |
632 | // mainly useful with GetDirectBufferPointer(). |
633 | bool Skip(int count); |
634 | |
635 | // Sets *data to point directly at the unwritten part of the |
636 | // CodedOutputStream's underlying buffer, and *size to the size of that |
637 | // buffer, but does not advance the stream's current position. This will |
638 | // always either produce a non-empty buffer or return false. If the caller |
639 | // writes any data to this buffer, it should then call Skip() to skip over |
640 | // the consumed bytes. This may be useful for implementing external fast |
641 | // serialization routines for types of data not covered by the |
642 | // CodedOutputStream interface. |
643 | bool GetDirectBufferPointer(void** data, int* size); |
644 | |
645 | // If there are at least "size" bytes available in the current buffer, |
646 | // returns a pointer directly into the buffer and advances over these bytes. |
647 | // The caller may then write directly into this buffer (e.g. using the |
648 | // *ToArray static methods) rather than go through CodedOutputStream. If |
649 | // there are not enough bytes available, returns NULL. The return pointer is |
650 | // invalidated as soon as any other non-const method of CodedOutputStream |
651 | // is called. |
652 | inline uint8* GetDirectBufferForNBytesAndAdvance(int size); |
653 | |
654 | // Write raw bytes, copying them from the given buffer. |
655 | void WriteRaw(const void* buffer, int size); |
656 | // Like WriteRaw() but will try to write aliased data if aliasing is |
657 | // turned on. |
658 | void WriteRawMaybeAliased(const void* data, int size); |
659 | // Like WriteRaw() but writing directly to the target array. |
660 | // This is _not_ inlined, as the compiler often optimizes memcpy into inline |
661 | // copy loops. Since this gets called by every field with string or bytes |
662 | // type, inlining may lead to a significant amount of code bloat, with only a |
663 | // minor performance gain. |
664 | static uint8* WriteRawToArray(const void* buffer, int size, uint8* target); |
665 | |
666 | // Equivalent to WriteRaw(str.data(), str.size()). |
667 | void WriteString(const string& str); |
668 | // Like WriteString() but writing directly to the target array. |
669 | static uint8* WriteStringToArray(const string& str, uint8* target); |
670 | // Write the varint-encoded size of str followed by str. |
671 | static uint8* WriteStringWithSizeToArray(const string& str, uint8* target); |
672 | |
673 | |
674 | // Instructs the CodedOutputStream to allow the underlying |
675 | // ZeroCopyOutputStream to hold pointers to the original structure instead of |
676 | // copying, if it supports it (i.e. output->AllowsAliasing() is true). If the |
677 | // underlying stream does not support aliasing, then enabling it has no |
678 | // affect. For now, this only affects the behavior of |
679 | // WriteRawMaybeAliased(). |
680 | // |
681 | // NOTE: It is caller's responsibility to ensure that the chunk of memory |
682 | // remains live until all of the data has been consumed from the stream. |
683 | void EnableAliasing(bool enabled); |
684 | |
685 | // Write a 32-bit little-endian integer. |
686 | void WriteLittleEndian32(uint32 value); |
687 | // Like WriteLittleEndian32() but writing directly to the target array. |
688 | static uint8* WriteLittleEndian32ToArray(uint32 value, uint8* target); |
689 | // Write a 64-bit little-endian integer. |
690 | void WriteLittleEndian64(uint64 value); |
691 | // Like WriteLittleEndian64() but writing directly to the target array. |
692 | static uint8* WriteLittleEndian64ToArray(uint64 value, uint8* target); |
693 | |
694 | // Write an unsigned integer with Varint encoding. Writing a 32-bit value |
695 | // is equivalent to casting it to uint64 and writing it as a 64-bit value, |
696 | // but may be more efficient. |
697 | void WriteVarint32(uint32 value); |
698 | // Like WriteVarint32() but writing directly to the target array. |
699 | static uint8* WriteVarint32ToArray(uint32 value, uint8* target); |
700 | // Write an unsigned integer with Varint encoding. |
701 | void WriteVarint64(uint64 value); |
702 | // Like WriteVarint64() but writing directly to the target array. |
703 | static uint8* WriteVarint64ToArray(uint64 value, uint8* target); |
704 | |
705 | // Equivalent to WriteVarint32() except when the value is negative, |
706 | // in which case it must be sign-extended to a full 10 bytes. |
707 | void WriteVarint32SignExtended(int32 value); |
708 | // Like WriteVarint32SignExtended() but writing directly to the target array. |
709 | static uint8* WriteVarint32SignExtendedToArray(int32 value, uint8* target); |
710 | |
711 | // This is identical to WriteVarint32(), but optimized for writing tags. |
712 | // In particular, if the input is a compile-time constant, this method |
713 | // compiles down to a couple instructions. |
714 | // Always inline because otherwise the aformentioned optimization can't work, |
715 | // but GCC by default doesn't want to inline this. |
716 | void WriteTag(uint32 value); |
717 | // Like WriteTag() but writing directly to the target array. |
718 | static uint8* WriteTagToArray( |
719 | uint32 value, uint8* target) GOOGLE_ATTRIBUTE_ALWAYS_INLINE; |
720 | |
721 | // Returns the number of bytes needed to encode the given value as a varint. |
722 | static int VarintSize32(uint32 value); |
723 | // Returns the number of bytes needed to encode the given value as a varint. |
724 | static int VarintSize64(uint64 value); |
725 | |
726 | // If negative, 10 bytes. Otheriwse, same as VarintSize32(). |
727 | static int VarintSize32SignExtended(int32 value); |
728 | |
729 | // Compile-time equivalent of VarintSize32(). |
730 | template <uint32 Value> |
731 | struct StaticVarintSize32 { |
732 | static const int value = |
733 | (Value < (1 << 7)) |
734 | ? 1 |
735 | : (Value < (1 << 14)) |
736 | ? 2 |
737 | : (Value < (1 << 21)) |
738 | ? 3 |
739 | : (Value < (1 << 28)) |
740 | ? 4 |
741 | : 5; |
742 | }; |
743 | |
744 | // Returns the total number of bytes written since this object was created. |
745 | inline int ByteCount() const; |
746 | |
747 | // Returns true if there was an underlying I/O error since this object was |
748 | // created. |
749 | bool HadError() const { return had_error_; } |
750 | |
751 | private: |
752 | GOOGLE_DISALLOW_EVIL_CONSTRUCTORS(CodedOutputStream); |
753 | |
754 | ZeroCopyOutputStream* output_; |
755 | uint8* buffer_; |
756 | int buffer_size_; |
757 | int total_bytes_; // Sum of sizes of all buffers seen so far. |
758 | bool had_error_; // Whether an error occurred during output. |
759 | bool aliasing_enabled_; // See EnableAliasing(). |
760 | |
761 | // Advance the buffer by a given number of bytes. |
762 | void Advance(int amount); |
763 | |
764 | // Called when the buffer runs out to request more data. Implies an |
765 | // Advance(buffer_size_). |
766 | bool Refresh(); |
767 | |
768 | // Like WriteRaw() but may avoid copying if the underlying |
769 | // ZeroCopyOutputStream supports it. |
770 | void WriteAliasedRaw(const void* buffer, int size); |
771 | |
772 | static uint8* WriteVarint32FallbackToArray(uint32 value, uint8* target); |
773 | |
774 | // Always-inlined versions of WriteVarint* functions so that code can be |
775 | // reused, while still controlling size. For instance, WriteVarint32ToArray() |
776 | // should not directly call this: since it is inlined itself, doing so |
777 | // would greatly increase the size of generated code. Instead, it should call |
778 | // WriteVarint32FallbackToArray. Meanwhile, WriteVarint32() is already |
779 | // out-of-line, so it should just invoke this directly to avoid any extra |
780 | // function call overhead. |
781 | static uint8* WriteVarint32FallbackToArrayInline( |
782 | uint32 value, uint8* target) GOOGLE_ATTRIBUTE_ALWAYS_INLINE; |
783 | static uint8* WriteVarint64ToArrayInline( |
784 | uint64 value, uint8* target) GOOGLE_ATTRIBUTE_ALWAYS_INLINE; |
785 | |
786 | static int VarintSize32Fallback(uint32 value); |
787 | }; |
788 | |
789 | // inline methods ==================================================== |
790 | // The vast majority of varints are only one byte. These inline |
791 | // methods optimize for that case. |
792 | |
793 | inline bool CodedInputStream::ReadVarint32(uint32* value) { |
794 | if (GOOGLE_PREDICT_TRUE(buffer_ < buffer_end_) && *buffer_ < 0x80) { |
795 | *value = *buffer_; |
796 | Advance(1); |
797 | return true; |
798 | } else { |
799 | return ReadVarint32Fallback(value); |
800 | } |
801 | } |
802 | |
803 | inline bool CodedInputStream::ReadVarint64(uint64* value) { |
804 | if (GOOGLE_PREDICT_TRUE(buffer_ < buffer_end_) && *buffer_ < 0x80) { |
805 | *value = *buffer_; |
806 | Advance(1); |
807 | return true; |
808 | } else { |
809 | return ReadVarint64Fallback(value); |
810 | } |
811 | } |
812 | |
813 | // static |
814 | inline const uint8* CodedInputStream::ReadLittleEndian32FromArray( |
815 | const uint8* buffer, |
816 | uint32* value) { |
817 | #if defined(PROTOBUF_LITTLE_ENDIAN) |
818 | memcpy(value, buffer, sizeof(*value)); |
819 | return buffer + sizeof(*value); |
820 | #else |
821 | *value = (static_cast<uint32>(buffer[0]) ) | |
822 | (static_cast<uint32>(buffer[1]) << 8) | |
823 | (static_cast<uint32>(buffer[2]) << 16) | |
824 | (static_cast<uint32>(buffer[3]) << 24); |
825 | return buffer + sizeof(*value); |
826 | #endif |
827 | } |
828 | // static |
829 | inline const uint8* CodedInputStream::ReadLittleEndian64FromArray( |
830 | const uint8* buffer, |
831 | uint64* value) { |
832 | #if defined(PROTOBUF_LITTLE_ENDIAN) |
833 | memcpy(value, buffer, sizeof(*value)); |
834 | return buffer + sizeof(*value); |
835 | #else |
836 | uint32 part0 = (static_cast<uint32>(buffer[0]) ) | |
837 | (static_cast<uint32>(buffer[1]) << 8) | |
838 | (static_cast<uint32>(buffer[2]) << 16) | |
839 | (static_cast<uint32>(buffer[3]) << 24); |
840 | uint32 part1 = (static_cast<uint32>(buffer[4]) ) | |
841 | (static_cast<uint32>(buffer[5]) << 8) | |
842 | (static_cast<uint32>(buffer[6]) << 16) | |
843 | (static_cast<uint32>(buffer[7]) << 24); |
844 | *value = static_cast<uint64>(part0) | |
845 | (static_cast<uint64>(part1) << 32); |
846 | return buffer + sizeof(*value); |
847 | #endif |
848 | } |
849 | |
850 | inline bool CodedInputStream::ReadLittleEndian32(uint32* value) { |
851 | #if defined(PROTOBUF_LITTLE_ENDIAN) |
852 | if (GOOGLE_PREDICT_TRUE(BufferSize() >= static_cast<int>(sizeof(*value)))) { |
853 | memcpy(value, buffer_, sizeof(*value)); |
854 | Advance(sizeof(*value)); |
855 | return true; |
856 | } else { |
857 | return ReadLittleEndian32Fallback(value); |
858 | } |
859 | #else |
860 | return ReadLittleEndian32Fallback(value); |
861 | #endif |
862 | } |
863 | |
864 | inline bool CodedInputStream::ReadLittleEndian64(uint64* value) { |
865 | #if defined(PROTOBUF_LITTLE_ENDIAN) |
866 | if (GOOGLE_PREDICT_TRUE(BufferSize() >= static_cast<int>(sizeof(*value)))) { |
867 | memcpy(value, buffer_, sizeof(*value)); |
868 | Advance(sizeof(*value)); |
869 | return true; |
870 | } else { |
871 | return ReadLittleEndian64Fallback(value); |
872 | } |
873 | #else |
874 | return ReadLittleEndian64Fallback(value); |
875 | #endif |
876 | } |
877 | |
878 | inline uint32 CodedInputStream::ReadTag() { |
879 | if (GOOGLE_PREDICT_TRUE(buffer_ < buffer_end_) && buffer_[0] < 0x80) { |
880 | last_tag_ = buffer_[0]; |
881 | Advance(1); |
882 | return last_tag_; |
883 | } else { |
884 | last_tag_ = ReadTagFallback(); |
885 | return last_tag_; |
886 | } |
887 | } |
888 | |
889 | inline std::pair<uint32, bool> CodedInputStream::ReadTagWithCutoff( |
890 | uint32 cutoff) { |
891 | // In performance-sensitive code we can expect cutoff to be a compile-time |
892 | // constant, and things like "cutoff >= kMax1ByteVarint" to be evaluated at |
893 | // compile time. |
894 | if (GOOGLE_PREDICT_TRUE(buffer_ < buffer_end_)) { |
895 | // Hot case: buffer_ non_empty, buffer_[0] in [1, 128). |
896 | // TODO(gpike): Is it worth rearranging this? E.g., if the number of fields |
897 | // is large enough then is it better to check for the two-byte case first? |
898 | if (static_cast<int8>(buffer_[0]) > 0) { |
899 | const uint32 kMax1ByteVarint = 0x7f; |
900 | uint32 tag = last_tag_ = buffer_[0]; |
901 | Advance(1); |
902 | return make_pair(tag, cutoff >= kMax1ByteVarint || tag <= cutoff); |
903 | } |
904 | // Other hot case: cutoff >= 0x80, buffer_ has at least two bytes available, |
905 | // and tag is two bytes. The latter is tested by bitwise-and-not of the |
906 | // first byte and the second byte. |
907 | if (cutoff >= 0x80 && |
908 | GOOGLE_PREDICT_TRUE(buffer_ + 1 < buffer_end_) && |
909 | GOOGLE_PREDICT_TRUE((buffer_[0] & ~buffer_[1]) >= 0x80)) { |
910 | const uint32 kMax2ByteVarint = (0x7f << 7) + 0x7f; |
911 | uint32 tag = last_tag_ = (1u << 7) * buffer_[1] + (buffer_[0] - 0x80); |
912 | Advance(2); |
913 | // It might make sense to test for tag == 0 now, but it is so rare that |
914 | // that we don't bother. A varint-encoded 0 should be one byte unless |
915 | // the encoder lost its mind. The second part of the return value of |
916 | // this function is allowed to be either true or false if the tag is 0, |
917 | // so we don't have to check for tag == 0. We may need to check whether |
918 | // it exceeds cutoff. |
919 | bool at_or_below_cutoff = cutoff >= kMax2ByteVarint || tag <= cutoff; |
920 | return make_pair(tag, at_or_below_cutoff); |
921 | } |
922 | } |
923 | // Slow path |
924 | last_tag_ = ReadTagFallback(); |
925 | return make_pair(last_tag_, static_cast<uint32>(last_tag_ - 1) < cutoff); |
926 | } |
927 | |
928 | inline bool CodedInputStream::LastTagWas(uint32 expected) { |
929 | return last_tag_ == expected; |
930 | } |
931 | |
932 | inline bool CodedInputStream::ConsumedEntireMessage() { |
933 | return legitimate_message_end_; |
934 | } |
935 | |
936 | inline bool CodedInputStream::ExpectTag(uint32 expected) { |
937 | if (expected < (1 << 7)) { |
938 | if (GOOGLE_PREDICT_TRUE(buffer_ < buffer_end_) && buffer_[0] == expected) { |
939 | Advance(1); |
940 | return true; |
941 | } else { |
942 | return false; |
943 | } |
944 | } else if (expected < (1 << 14)) { |
945 | if (GOOGLE_PREDICT_TRUE(BufferSize() >= 2) && |
946 | buffer_[0] == static_cast<uint8>(expected | 0x80) && |
947 | buffer_[1] == static_cast<uint8>(expected >> 7)) { |
948 | Advance(2); |
949 | return true; |
950 | } else { |
951 | return false; |
952 | } |
953 | } else { |
954 | // Don't bother optimizing for larger values. |
955 | return false; |
956 | } |
957 | } |
958 | |
959 | inline const uint8* CodedInputStream::ExpectTagFromArray( |
960 | const uint8* buffer, uint32 expected) { |
961 | if (expected < (1 << 7)) { |
962 | if (buffer[0] == expected) { |
963 | return buffer + 1; |
964 | } |
965 | } else if (expected < (1 << 14)) { |
966 | if (buffer[0] == static_cast<uint8>(expected | 0x80) && |
967 | buffer[1] == static_cast<uint8>(expected >> 7)) { |
968 | return buffer + 2; |
969 | } |
970 | } |
971 | return NULL; |
972 | } |
973 | |
974 | inline void CodedInputStream::GetDirectBufferPointerInline(const void** data, |
975 | int* size) { |
976 | *data = buffer_; |
977 | *size = buffer_end_ - buffer_; |
978 | } |
979 | |
980 | inline bool CodedInputStream::ExpectAtEnd() { |
981 | // If we are at a limit we know no more bytes can be read. Otherwise, it's |
982 | // hard to say without calling Refresh(), and we'd rather not do that. |
983 | |
984 | if (buffer_ == buffer_end_ && |
985 | ((buffer_size_after_limit_ != 0) || |
986 | (total_bytes_read_ == current_limit_))) { |
987 | last_tag_ = 0; // Pretend we called ReadTag()... |
988 | legitimate_message_end_ = true; // ... and it hit EOF. |
989 | return true; |
990 | } else { |
991 | return false; |
992 | } |
993 | } |
994 | |
995 | inline int CodedInputStream::CurrentPosition() const { |
996 | return total_bytes_read_ - (BufferSize() + buffer_size_after_limit_); |
997 | } |
998 | |
999 | inline uint8* CodedOutputStream::GetDirectBufferForNBytesAndAdvance(int size) { |
1000 | if (buffer_size_ < size) { |
1001 | return NULL; |
1002 | } else { |
1003 | uint8* result = buffer_; |
1004 | Advance(size); |
1005 | return result; |
1006 | } |
1007 | } |
1008 | |
1009 | inline uint8* CodedOutputStream::WriteVarint32ToArray(uint32 value, |
1010 | uint8* target) { |
1011 | if (value < 0x80) { |
1012 | *target = value; |
1013 | return target + 1; |
1014 | } else { |
1015 | return WriteVarint32FallbackToArray(value, target); |
1016 | } |
1017 | } |
1018 | |
1019 | inline void CodedOutputStream::WriteVarint32SignExtended(int32 value) { |
1020 | if (value < 0) { |
1021 | WriteVarint64(static_cast<uint64>(value)); |
1022 | } else { |
1023 | WriteVarint32(static_cast<uint32>(value)); |
1024 | } |
1025 | } |
1026 | |
1027 | inline uint8* CodedOutputStream::WriteVarint32SignExtendedToArray( |
1028 | int32 value, uint8* target) { |
1029 | if (value < 0) { |
1030 | return WriteVarint64ToArray(static_cast<uint64>(value), target); |
1031 | } else { |
1032 | return WriteVarint32ToArray(static_cast<uint32>(value), target); |
1033 | } |
1034 | } |
1035 | |
1036 | inline uint8* CodedOutputStream::WriteLittleEndian32ToArray(uint32 value, |
1037 | uint8* target) { |
1038 | #if defined(PROTOBUF_LITTLE_ENDIAN) |
1039 | memcpy(target, &value, sizeof(value)); |
1040 | #else |
1041 | target[0] = static_cast<uint8>(value); |
1042 | target[1] = static_cast<uint8>(value >> 8); |
1043 | target[2] = static_cast<uint8>(value >> 16); |
1044 | target[3] = static_cast<uint8>(value >> 24); |
1045 | #endif |
1046 | return target + sizeof(value); |
1047 | } |
1048 | |
1049 | inline uint8* CodedOutputStream::WriteLittleEndian64ToArray(uint64 value, |
1050 | uint8* target) { |
1051 | #if defined(PROTOBUF_LITTLE_ENDIAN) |
1052 | memcpy(target, &value, sizeof(value)); |
1053 | #else |
1054 | uint32 part0 = static_cast<uint32>(value); |
1055 | uint32 part1 = static_cast<uint32>(value >> 32); |
1056 | |
1057 | target[0] = static_cast<uint8>(part0); |
1058 | target[1] = static_cast<uint8>(part0 >> 8); |
1059 | target[2] = static_cast<uint8>(part0 >> 16); |
1060 | target[3] = static_cast<uint8>(part0 >> 24); |
1061 | target[4] = static_cast<uint8>(part1); |
1062 | target[5] = static_cast<uint8>(part1 >> 8); |
1063 | target[6] = static_cast<uint8>(part1 >> 16); |
1064 | target[7] = static_cast<uint8>(part1 >> 24); |
1065 | #endif |
1066 | return target + sizeof(value); |
1067 | } |
1068 | |
1069 | inline void CodedOutputStream::WriteTag(uint32 value) { |
1070 | WriteVarint32(value); |
1071 | } |
1072 | |
1073 | inline uint8* CodedOutputStream::WriteTagToArray( |
1074 | uint32 value, uint8* target) { |
1075 | if (value < (1 << 7)) { |
1076 | target[0] = value; |
1077 | return target + 1; |
1078 | } else if (value < (1 << 14)) { |
1079 | target[0] = static_cast<uint8>(value | 0x80); |
1080 | target[1] = static_cast<uint8>(value >> 7); |
1081 | return target + 2; |
1082 | } else { |
1083 | return WriteVarint32FallbackToArray(value, target); |
1084 | } |
1085 | } |
1086 | |
1087 | inline int CodedOutputStream::VarintSize32(uint32 value) { |
1088 | if (value < (1 << 7)) { |
1089 | return 1; |
1090 | } else { |
1091 | return VarintSize32Fallback(value); |
1092 | } |
1093 | } |
1094 | |
1095 | inline int CodedOutputStream::VarintSize32SignExtended(int32 value) { |
1096 | if (value < 0) { |
1097 | return 10; // TODO(kenton): Make this a symbolic constant. |
1098 | } else { |
1099 | return VarintSize32(static_cast<uint32>(value)); |
1100 | } |
1101 | } |
1102 | |
1103 | inline void CodedOutputStream::WriteString(const string& str) { |
1104 | WriteRaw(str.data(), static_cast<int>(str.size())); |
1105 | } |
1106 | |
1107 | inline void CodedOutputStream::WriteRawMaybeAliased( |
1108 | const void* data, int size) { |
1109 | if (aliasing_enabled_) { |
1110 | WriteAliasedRaw(data, size); |
1111 | } else { |
1112 | WriteRaw(data, size); |
1113 | } |
1114 | } |
1115 | |
1116 | inline uint8* CodedOutputStream::WriteStringToArray( |
1117 | const string& str, uint8* target) { |
1118 | return WriteRawToArray(str.data(), static_cast<int>(str.size()), target); |
1119 | } |
1120 | |
1121 | inline int CodedOutputStream::ByteCount() const { |
1122 | return total_bytes_ - buffer_size_; |
1123 | } |
1124 | |
1125 | inline void CodedInputStream::Advance(int amount) { |
1126 | buffer_ += amount; |
1127 | } |
1128 | |
1129 | inline void CodedOutputStream::Advance(int amount) { |
1130 | buffer_ += amount; |
1131 | buffer_size_ -= amount; |
1132 | } |
1133 | |
1134 | inline void CodedInputStream::SetRecursionLimit(int limit) { |
1135 | recursion_limit_ = limit; |
1136 | } |
1137 | |
1138 | inline bool CodedInputStream::IncrementRecursionDepth() { |
1139 | ++recursion_depth_; |
1140 | return recursion_depth_ <= recursion_limit_; |
1141 | } |
1142 | |
1143 | inline void CodedInputStream::DecrementRecursionDepth() { |
1144 | if (recursion_depth_ > 0) --recursion_depth_; |
1145 | } |
1146 | |
1147 | inline void CodedInputStream::SetExtensionRegistry(const DescriptorPool* pool, |
1148 | MessageFactory* factory) { |
1149 | extension_pool_ = pool; |
1150 | extension_factory_ = factory; |
1151 | } |
1152 | |
1153 | inline const DescriptorPool* CodedInputStream::GetExtensionPool() { |
1154 | return extension_pool_; |
1155 | } |
1156 | |
1157 | inline MessageFactory* CodedInputStream::GetExtensionFactory() { |
1158 | return extension_factory_; |
1159 | } |
1160 | |
1161 | inline int CodedInputStream::BufferSize() const { |
1162 | return buffer_end_ - buffer_; |
1163 | } |
1164 | |
1165 | inline CodedInputStream::CodedInputStream(ZeroCopyInputStream* input) |
1166 | : input_(input), |
1167 | buffer_(NULL), |
1168 | buffer_end_(NULL), |
1169 | total_bytes_read_(0), |
1170 | overflow_bytes_(0), |
1171 | last_tag_(0), |
1172 | legitimate_message_end_(false), |
1173 | aliasing_enabled_(false), |
1174 | current_limit_(kint32max), |
1175 | buffer_size_after_limit_(0), |
1176 | total_bytes_limit_(kDefaultTotalBytesLimit), |
1177 | total_bytes_warning_threshold_(kDefaultTotalBytesWarningThreshold), |
1178 | recursion_depth_(0), |
1179 | recursion_limit_(default_recursion_limit_), |
1180 | extension_pool_(NULL), |
1181 | extension_factory_(NULL) { |
1182 | // Eagerly Refresh() so buffer space is immediately available. |
1183 | Refresh(); |
1184 | } |
1185 | |
1186 | inline CodedInputStream::CodedInputStream(const uint8* buffer, int size) |
1187 | : input_(NULL), |
1188 | buffer_(buffer), |
1189 | buffer_end_(buffer + size), |
1190 | total_bytes_read_(size), |
1191 | overflow_bytes_(0), |
1192 | last_tag_(0), |
1193 | legitimate_message_end_(false), |
1194 | aliasing_enabled_(false), |
1195 | current_limit_(size), |
1196 | buffer_size_after_limit_(0), |
1197 | total_bytes_limit_(kDefaultTotalBytesLimit), |
1198 | total_bytes_warning_threshold_(kDefaultTotalBytesWarningThreshold), |
1199 | recursion_depth_(0), |
1200 | recursion_limit_(default_recursion_limit_), |
1201 | extension_pool_(NULL), |
1202 | extension_factory_(NULL) { |
1203 | // Note that setting current_limit_ == size is important to prevent some |
1204 | // code paths from trying to access input_ and segfaulting. |
1205 | } |
1206 | |
1207 | inline bool CodedInputStream::IsFlat() const { |
1208 | return input_ == NULL; |
1209 | } |
1210 | |
1211 | } // namespace io |
1212 | } // namespace protobuf |
1213 | |
1214 | |
1215 | #if defined(_MSC_VER) && _MSC_VER >= 1300 |
1216 | #pragma runtime_checks("c", restore) |
1217 | #endif // _MSC_VER |
1218 | |
1219 | } // namespace google |
1220 | #endif // GOOGLE_PROTOBUF_IO_CODED_STREAM_H__ |
1221 | |