blob: a929df2329e8f32d90920b8daec4f0fa6723e334 [file] [log] [blame]
Jon Skeet0aac0e42009-09-09 18:48:02 +01001#region Copyright notice and license
Jon Skeet60c059b2008-10-23 21:17:56 +01002// Protocol Buffers - Google's data interchange format
3// Copyright 2008 Google Inc. All rights reserved.
4// http://github.com/jskeet/dotnet-protobufs/
5// Original C++/Java/Python code:
Jon Skeet68036862008-10-22 13:30:34 +01006// http://code.google.com/p/protobuf/
7//
Jon Skeet60c059b2008-10-23 21:17:56 +01008// Redistribution and use in source and binary forms, with or without
9// modification, are permitted provided that the following conditions are
10// met:
Jon Skeet68036862008-10-22 13:30:34 +010011//
Jon Skeet60c059b2008-10-23 21:17:56 +010012// * Redistributions of source code must retain the above copyright
13// notice, this list of conditions and the following disclaimer.
14// * Redistributions in binary form must reproduce the above
15// copyright notice, this list of conditions and the following disclaimer
16// in the documentation and/or other materials provided with the
17// distribution.
18// * Neither the name of Google Inc. nor the names of its
19// contributors may be used to endorse or promote products derived from
20// this software without specific prior written permission.
Jon Skeet68036862008-10-22 13:30:34 +010021//
Jon Skeet60c059b2008-10-23 21:17:56 +010022// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
23// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
24// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
25// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
26// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
27// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
28// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
29// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
30// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
31// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
32// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
Jon Skeet0aac0e42009-09-09 18:48:02 +010033#endregion
34
Jon Skeet68036862008-10-22 13:30:34 +010035using System;
36using System.Collections.Generic;
37using System.IO;
38using System.Text;
39using Google.ProtocolBuffers.Descriptors;
40
41namespace Google.ProtocolBuffers {
42
43 /// <summary>
44 /// Readings and decodes protocol message fields.
45 /// </summary>
46 /// <remarks>
47 /// This class contains two kinds of methods: methods that read specific
48 /// protocol message constructs and field types (e.g. ReadTag and
49 /// ReadInt32) and methods that read low-level values (e.g.
50 /// ReadRawVarint32 and ReadRawBytes). If you are reading encoded protocol
51 /// messages, you should use the former methods, but if you are reading some
52 /// other format of your own design, use the latter. The names of the former
53 /// methods are taken from the protocol buffer type names, not .NET types.
54 /// (Hence ReadFloat instead of ReadSingle, and ReadBool instead of ReadBoolean.)
55 ///
56 /// TODO(jonskeet): Consider whether recursion and size limits shouldn't be readonly,
57 /// set at construction time.
58 /// </remarks>
59 public sealed class CodedInputStream {
60 private readonly byte[] buffer;
61 private int bufferSize;
62 private int bufferSizeAfterLimit = 0;
63 private int bufferPos = 0;
64 private readonly Stream input;
65 private uint lastTag = 0;
66
Jon Skeet2178b932009-06-25 07:52:07 +010067 internal const int DefaultRecursionLimit = 64;
68 internal const int DefaultSizeLimit = 64 << 20; // 64MB
69 internal const int BufferSize = 4096;
Jon Skeet68036862008-10-22 13:30:34 +010070
71 /// <summary>
72 /// The total number of bytes read before the current buffer. The
73 /// total bytes read up to the current position can be computed as
74 /// totalBytesRetired + bufferPos.
75 /// </summary>
76 private int totalBytesRetired = 0;
77
78 /// <summary>
79 /// The absolute position of the end of the current message.
80 /// </summary>
81 private int currentLimit = int.MaxValue;
82
83 /// <summary>
84 /// <see cref="SetRecursionLimit"/>
85 /// </summary>
86 private int recursionDepth = 0;
87 private int recursionLimit = DefaultRecursionLimit;
88
89 /// <summary>
90 /// <see cref="SetSizeLimit"/>
91 /// </summary>
92 private int sizeLimit = DefaultSizeLimit;
93
94 #region Construction
95 /// <summary>
96 /// Creates a new CodedInputStream reading data from the given
97 /// stream.
98 /// </summary>
99 public static CodedInputStream CreateInstance(Stream input) {
100 return new CodedInputStream(input);
101 }
102
103 /// <summary>
104 /// Creates a new CodedInputStream reading data from the given
105 /// byte array.
106 /// </summary>
107 public static CodedInputStream CreateInstance(byte[] buf) {
108 return new CodedInputStream(buf);
109 }
110
111 private CodedInputStream(byte[] buffer) {
112 this.buffer = buffer;
113 this.bufferSize = buffer.Length;
114 this.input = null;
115 }
116
117 private CodedInputStream(Stream input) {
118 this.buffer = new byte[BufferSize];
119 this.bufferSize = 0;
120 this.input = input;
121 }
122 #endregion
123
124 #region Validation
125 /// <summary>
126 /// Verifies that the last call to ReadTag() returned the given tag value.
127 /// This is used to verify that a nested group ended with the correct
128 /// end tag.
129 /// </summary>
130 /// <exception cref="InvalidProtocolBufferException">The last
131 /// tag read was not the one specified</exception>
Jon Skeetd6dd0a42009-06-05 22:00:05 +0100132 [CLSCompliant(false)]
Jon Skeet68036862008-10-22 13:30:34 +0100133 public void CheckLastTagWas(uint value) {
134 if (lastTag != value) {
135 throw InvalidProtocolBufferException.InvalidEndTag();
136 }
137 }
138 #endregion
139
140 #region Reading of tags etc
141 /// <summary>
142 /// Attempt to read a field tag, returning 0 if we have reached the end
143 /// of the input data. Protocol message parsers use this to read tags,
144 /// since a protocol message may legally end wherever a tag occurs, and
145 /// zero is not a valid tag number.
146 /// </summary>
Jon Skeetd6dd0a42009-06-05 22:00:05 +0100147 [CLSCompliant(false)]
Jon Skeet68036862008-10-22 13:30:34 +0100148 public uint ReadTag() {
Jon Skeet2e6dc122009-05-29 06:34:52 +0100149 if (IsAtEnd) {
Jon Skeet68036862008-10-22 13:30:34 +0100150 lastTag = 0;
151 return 0;
152 }
153
154 lastTag = ReadRawVarint32();
155 if (lastTag == 0) {
156 // If we actually read zero, that's not a valid tag.
157 throw InvalidProtocolBufferException.InvalidTag();
158 }
159 return lastTag;
160 }
161
162 /// <summary>
163 /// Read a double field from the stream.
164 /// </summary>
165 public double ReadDouble() {
Jon Skeetb49d3c72009-11-03 16:51:01 +0000166#if SILVERLIGHT2 || COMPACT_FRAMEWORK_35
Jon Skeet3c808862009-09-09 13:22:36 +0100167 byte[] bytes = ReadRawBytes(8);
168 return BitConverter.ToDouble(bytes, 0);
169#else
Jon Skeet68036862008-10-22 13:30:34 +0100170 return BitConverter.Int64BitsToDouble((long) ReadRawLittleEndian64());
Jon Skeet3c808862009-09-09 13:22:36 +0100171#endif
Jon Skeet68036862008-10-22 13:30:34 +0100172 }
173
174 /// <summary>
175 /// Read a float field from the stream.
176 /// </summary>
177 public float ReadFloat() {
178 // TODO(jonskeet): Test this on different endiannesses
179 uint raw = ReadRawLittleEndian32();
180 byte[] rawBytes = BitConverter.GetBytes(raw);
181 return BitConverter.ToSingle(rawBytes, 0);
182 }
183
184 /// <summary>
185 /// Read a uint64 field from the stream.
186 /// </summary>
Jon Skeetd6dd0a42009-06-05 22:00:05 +0100187 [CLSCompliant(false)]
Jon Skeet68036862008-10-22 13:30:34 +0100188 public ulong ReadUInt64() {
189 return ReadRawVarint64();
190 }
191
192 /// <summary>
193 /// Read an int64 field from the stream.
194 /// </summary>
195 public long ReadInt64() {
196 return (long) ReadRawVarint64();
197 }
198
199 /// <summary>
200 /// Read an int32 field from the stream.
201 /// </summary>
202 public int ReadInt32() {
203 return (int) ReadRawVarint32();
204 }
205
206 /// <summary>
207 /// Read a fixed64 field from the stream.
208 /// </summary>
Jon Skeetd6dd0a42009-06-05 22:00:05 +0100209 [CLSCompliant(false)]
Jon Skeet68036862008-10-22 13:30:34 +0100210 public ulong ReadFixed64() {
211 return ReadRawLittleEndian64();
212 }
213
214 /// <summary>
215 /// Read a fixed32 field from the stream.
216 /// </summary>
Jon Skeetd6dd0a42009-06-05 22:00:05 +0100217 [CLSCompliant(false)]
Jon Skeet68036862008-10-22 13:30:34 +0100218 public uint ReadFixed32() {
219 return ReadRawLittleEndian32();
220 }
221
222 /// <summary>
223 /// Read a bool field from the stream.
224 /// </summary>
225 public bool ReadBool() {
226 return ReadRawVarint32() != 0;
227 }
228
229 /// <summary>
230 /// Reads a string field from the stream.
231 /// </summary>
232 public String ReadString() {
233 int size = (int) ReadRawVarint32();
Jon Skeet6a60ac32009-01-27 14:47:35 +0000234 // No need to read any data for an empty string.
235 if (size == 0) {
236 return "";
237 }
238 if (size <= bufferSize - bufferPos) {
Jon Skeet68036862008-10-22 13:30:34 +0100239 // Fast path: We already have the bytes in a contiguous buffer, so
240 // just copy directly from it.
241 String result = Encoding.UTF8.GetString(buffer, bufferPos, size);
242 bufferPos += size;
243 return result;
Jon Skeet68036862008-10-22 13:30:34 +0100244 }
Jon Skeet60fb63e2009-06-20 20:46:28 +0100245 // Slow path: Build a byte array first then copy it.
246 return Encoding.UTF8.GetString(ReadRawBytes(size), 0, size);
Jon Skeet68036862008-10-22 13:30:34 +0100247 }
248
249 /// <summary>
250 /// Reads a group field value from the stream.
251 /// </summary>
252 public void ReadGroup(int fieldNumber, IBuilder builder,
253 ExtensionRegistry extensionRegistry) {
254 if (recursionDepth >= recursionLimit) {
255 throw InvalidProtocolBufferException.RecursionLimitExceeded();
256 }
257 ++recursionDepth;
258 builder.WeakMergeFrom(this, extensionRegistry);
259 CheckLastTagWas(WireFormat.MakeTag(fieldNumber, WireFormat.WireType.EndGroup));
260 --recursionDepth;
261 }
262
263 /// <summary>
264 /// Reads a group field value from the stream and merges it into the given
265 /// UnknownFieldSet.
266 /// </summary>
267 public void ReadUnknownGroup(int fieldNumber, UnknownFieldSet.Builder builder) {
268 if (recursionDepth >= recursionLimit) {
269 throw InvalidProtocolBufferException.RecursionLimitExceeded();
270 }
271 ++recursionDepth;
272 builder.MergeFrom(this);
273 CheckLastTagWas(WireFormat.MakeTag(fieldNumber, WireFormat.WireType.EndGroup));
274 --recursionDepth;
275 }
276
277 /// <summary>
278 /// Reads an embedded message field value from the stream.
279 /// </summary>
280 public void ReadMessage(IBuilder builder, ExtensionRegistry extensionRegistry) {
281 int length = (int) ReadRawVarint32();
282 if (recursionDepth >= recursionLimit) {
283 throw InvalidProtocolBufferException.RecursionLimitExceeded();
284 }
285 int oldLimit = PushLimit(length);
286 ++recursionDepth;
287 builder.WeakMergeFrom(this, extensionRegistry);
288 CheckLastTagWas(0);
289 --recursionDepth;
290 PopLimit(oldLimit);
291 }
292
293 /// <summary>
294 /// Reads a bytes field value from the stream.
295 /// </summary>
296 public ByteString ReadBytes() {
297 int size = (int) ReadRawVarint32();
298 if (size < bufferSize - bufferPos && size > 0) {
299 // Fast path: We already have the bytes in a contiguous buffer, so
300 // just copy directly from it.
301 ByteString result = ByteString.CopyFrom(buffer, bufferPos, size);
302 bufferPos += size;
303 return result;
304 } else {
305 // Slow path: Build a byte array first then copy it.
306 return ByteString.CopyFrom(ReadRawBytes(size));
307 }
308 }
309
310 /// <summary>
311 /// Reads a uint32 field value from the stream.
312 /// </summary>
Jon Skeetd6dd0a42009-06-05 22:00:05 +0100313 [CLSCompliant(false)]
Jon Skeet68036862008-10-22 13:30:34 +0100314 public uint ReadUInt32() {
315 return ReadRawVarint32();
316 }
317
318 /// <summary>
319 /// Reads an enum field value from the stream. The caller is responsible
320 /// for converting the numeric value to an actual enum.
321 /// </summary>
322 public int ReadEnum() {
323 return (int) ReadRawVarint32();
324 }
325
326 /// <summary>
327 /// Reads an sfixed32 field value from the stream.
328 /// </summary>
329 public int ReadSFixed32() {
330 return (int) ReadRawLittleEndian32();
331 }
332
333 /// <summary>
334 /// Reads an sfixed64 field value from the stream.
335 /// </summary>
336 public long ReadSFixed64() {
337 return (long) ReadRawLittleEndian64();
338 }
339
340 /// <summary>
341 /// Reads an sint32 field value from the stream.
342 /// </summary>
343 public int ReadSInt32() {
344 return DecodeZigZag32(ReadRawVarint32());
345 }
346
347 /// <summary>
348 /// Reads an sint64 field value from the stream.
349 /// </summary>
350 public long ReadSInt64() {
351 return DecodeZigZag64(ReadRawVarint64());
352 }
353
354 /// <summary>
355 /// Reads a field of any primitive type. Enums, groups and embedded
356 /// messages are not handled by this method.
357 /// </summary>
358 public object ReadPrimitiveField(FieldType fieldType) {
359 switch (fieldType) {
360 case FieldType.Double: return ReadDouble();
361 case FieldType.Float: return ReadFloat();
362 case FieldType.Int64: return ReadInt64();
363 case FieldType.UInt64: return ReadUInt64();
364 case FieldType.Int32: return ReadInt32();
365 case FieldType.Fixed64: return ReadFixed64();
366 case FieldType.Fixed32: return ReadFixed32();
367 case FieldType.Bool: return ReadBool();
368 case FieldType.String: return ReadString();
369 case FieldType.Bytes: return ReadBytes();
370 case FieldType.UInt32: return ReadUInt32();
371 case FieldType.SFixed32: return ReadSFixed32();
372 case FieldType.SFixed64: return ReadSFixed64();
373 case FieldType.SInt32: return ReadSInt32();
374 case FieldType.SInt64: return ReadSInt64();
375 case FieldType.Group:
376 throw new ArgumentException("ReadPrimitiveField() cannot handle nested groups.");
377 case FieldType.Message:
378 throw new ArgumentException("ReadPrimitiveField() cannot handle embedded messages.");
379 // We don't handle enums because we don't know what to do if the
380 // value is not recognized.
381 case FieldType.Enum:
382 throw new ArgumentException("ReadPrimitiveField() cannot handle enums.");
383 default:
384 throw new ArgumentOutOfRangeException("Invalid field type " + fieldType);
385 }
386 }
387
388 #endregion
389
390 #region Underlying reading primitives
391
392 /// <summary>
393 /// Same code as ReadRawVarint32, but read each byte individually, checking for
394 /// buffer overflow.
395 /// </summary>
396 private uint SlowReadRawVarint32() {
397 int tmp = ReadRawByte();
398 if (tmp < 128) {
399 return (uint)tmp;
400 }
401 int result = tmp & 0x7f;
402 if ((tmp = ReadRawByte()) < 128) {
403 result |= tmp << 7;
404 } else {
405 result |= (tmp & 0x7f) << 7;
406 if ((tmp = ReadRawByte()) < 128) {
407 result |= tmp << 14;
408 } else {
409 result |= (tmp & 0x7f) << 14;
410 if ((tmp = ReadRawByte()) < 128) {
411 result |= tmp << 21;
412 } else {
413 result |= (tmp & 0x7f) << 21;
414 result |= (tmp = ReadRawByte()) << 28;
415 if (tmp >= 128) {
416 // Discard upper 32 bits.
417 for (int i = 0; i < 5; i++) {
418 if (ReadRawByte() < 128) return (uint)result;
419 }
420 throw InvalidProtocolBufferException.MalformedVarint();
421 }
422 }
423 }
424 }
425 return (uint)result;
426 }
427
428 /// <summary>
429 /// Read a raw Varint from the stream. If larger than 32 bits, discard the upper bits.
430 /// This method is optimised for the case where we've got lots of data in the buffer.
431 /// That means we can check the size just once, then just read directly from the buffer
432 /// without constant rechecking of the buffer length.
433 /// </summary>
Jon Skeetd6dd0a42009-06-05 22:00:05 +0100434 [CLSCompliant(false)]
Jon Skeet68036862008-10-22 13:30:34 +0100435 public uint ReadRawVarint32() {
436 if (bufferPos + 5 > bufferSize) {
437 return SlowReadRawVarint32();
438 }
439
440 int tmp = buffer[bufferPos++];
441 if (tmp < 128) {
442 return (uint)tmp;
443 }
444 int result = tmp & 0x7f;
445 if ((tmp = buffer[bufferPos++]) < 128) {
446 result |= tmp << 7;
447 } else {
448 result |= (tmp & 0x7f) << 7;
449 if ((tmp = buffer[bufferPos++]) < 128) {
450 result |= tmp << 14;
451 } else {
452 result |= (tmp & 0x7f) << 14;
453 if ((tmp = buffer[bufferPos++]) < 128) {
454 result |= tmp << 21;
455 } else {
456 result |= (tmp & 0x7f) << 21;
457 result |= (tmp = buffer[bufferPos++]) << 28;
458 if (tmp >= 128) {
459 // Discard upper 32 bits.
460 // Note that this has to use ReadRawByte() as we only ensure we've
461 // got at least 5 bytes at the start of the method. This lets us
462 // use the fast path in more cases, and we rarely hit this section of code.
463 for (int i = 0; i < 5; i++) {
464 if (ReadRawByte() < 128) return (uint)result;
465 }
466 throw InvalidProtocolBufferException.MalformedVarint();
467 }
468 }
469 }
470 }
471 return (uint)result;
472 }
473
474 /// <summary>
Jon Skeet2e6dc122009-05-29 06:34:52 +0100475 /// Reads a varint from the input one byte at a time, so that it does not
476 /// read any bytes after the end of the varint. If you simply wrapped the
477 /// stream in a CodedInputStream and used ReadRawVarint32(Stream)}
478 /// then you would probably end up reading past the end of the varint since
479 /// CodedInputStream buffers its input.
480 /// </summary>
481 /// <param name="input"></param>
482 /// <returns></returns>
Jon Skeetc298c892009-05-30 10:07:09 +0100483 internal static uint ReadRawVarint32(Stream input) {
Jon Skeet2e6dc122009-05-29 06:34:52 +0100484 int result = 0;
485 int offset = 0;
486 for (; offset < 32; offset += 7) {
487 int b = input.ReadByte();
488 if (b == -1) {
489 throw InvalidProtocolBufferException.TruncatedMessage();
490 }
491 result |= (b & 0x7f) << offset;
492 if ((b & 0x80) == 0) {
Jon Skeetc298c892009-05-30 10:07:09 +0100493 return (uint) result;
Jon Skeet2e6dc122009-05-29 06:34:52 +0100494 }
495 }
496 // Keep reading up to 64 bits.
497 for (; offset < 64; offset += 7) {
498 int b = input.ReadByte();
499 if (b == -1) {
500 throw InvalidProtocolBufferException.TruncatedMessage();
501 }
502 if ((b & 0x80) == 0) {
Jon Skeetc298c892009-05-30 10:07:09 +0100503 return (uint) result;
Jon Skeet2e6dc122009-05-29 06:34:52 +0100504 }
505 }
506 throw InvalidProtocolBufferException.MalformedVarint();
507 }
508
509 /// <summary>
Jon Skeet68036862008-10-22 13:30:34 +0100510 /// Read a raw varint from the stream.
511 /// </summary>
Jon Skeetd6dd0a42009-06-05 22:00:05 +0100512 [CLSCompliant(false)]
Jon Skeet68036862008-10-22 13:30:34 +0100513 public ulong ReadRawVarint64() {
514 int shift = 0;
515 ulong result = 0;
516 while (shift < 64) {
517 byte b = ReadRawByte();
518 result |= (ulong)(b & 0x7F) << shift;
519 if ((b & 0x80) == 0) {
520 return result;
521 }
522 shift += 7;
523 }
524 throw InvalidProtocolBufferException.MalformedVarint();
525 }
526
527 /// <summary>
528 /// Read a 32-bit little-endian integer from the stream.
529 /// </summary>
Jon Skeetd6dd0a42009-06-05 22:00:05 +0100530 [CLSCompliant(false)]
Jon Skeet68036862008-10-22 13:30:34 +0100531 public uint ReadRawLittleEndian32() {
532 uint b1 = ReadRawByte();
533 uint b2 = ReadRawByte();
534 uint b3 = ReadRawByte();
535 uint b4 = ReadRawByte();
536 return b1 | (b2 << 8) | (b3 << 16) | (b4 << 24);
537 }
538
539 /// <summary>
540 /// Read a 64-bit little-endian integer from the stream.
541 /// </summary>
Jon Skeetd6dd0a42009-06-05 22:00:05 +0100542 [CLSCompliant(false)]
Jon Skeet68036862008-10-22 13:30:34 +0100543 public ulong ReadRawLittleEndian64() {
544 ulong b1 = ReadRawByte();
545 ulong b2 = ReadRawByte();
546 ulong b3 = ReadRawByte();
547 ulong b4 = ReadRawByte();
548 ulong b5 = ReadRawByte();
549 ulong b6 = ReadRawByte();
550 ulong b7 = ReadRawByte();
551 ulong b8 = ReadRawByte();
552 return b1 | (b2 << 8) | (b3 << 16) | (b4 << 24)
553 | (b5 << 32) | (b6 << 40) | (b7 << 48) | (b8 << 56);
554 }
555 #endregion
556
557 /// <summary>
558 /// Decode a 32-bit value with ZigZag encoding.
559 /// </summary>
560 /// <remarks>
561 /// ZigZag encodes signed integers into values that can be efficiently
562 /// encoded with varint. (Otherwise, negative values must be
563 /// sign-extended to 64 bits to be varint encoded, thus always taking
564 /// 10 bytes on the wire.)
565 /// </remarks>
Jon Skeetd6dd0a42009-06-05 22:00:05 +0100566 [CLSCompliant(false)]
Jon Skeet68036862008-10-22 13:30:34 +0100567 public static int DecodeZigZag32(uint n) {
568 return (int)(n >> 1) ^ -(int)(n & 1);
569 }
570
571 /// <summary>
572 /// Decode a 32-bit value with ZigZag encoding.
573 /// </summary>
574 /// <remarks>
575 /// ZigZag encodes signed integers into values that can be efficiently
576 /// encoded with varint. (Otherwise, negative values must be
577 /// sign-extended to 64 bits to be varint encoded, thus always taking
578 /// 10 bytes on the wire.)
579 /// </remarks>
Jon Skeetd6dd0a42009-06-05 22:00:05 +0100580 [CLSCompliant(false)]
Jon Skeet68036862008-10-22 13:30:34 +0100581 public static long DecodeZigZag64(ulong n) {
582 return (long)(n >> 1) ^ -(long)(n & 1);
583 }
584
585 /// <summary>
586 /// Set the maximum message recursion depth.
587 /// </summary>
588 /// <remarks>
589 /// In order to prevent malicious
590 /// messages from causing stack overflows, CodedInputStream limits
591 /// how deeply messages may be nested. The default limit is 64.
592 /// </remarks>
593 public int SetRecursionLimit(int limit) {
594 if (limit < 0) {
595 throw new ArgumentOutOfRangeException("Recursion limit cannot be negative: " + limit);
596 }
597 int oldLimit = recursionLimit;
598 recursionLimit = limit;
599 return oldLimit;
600 }
601
602 /// <summary>
603 /// Set the maximum message size.
604 /// </summary>
605 /// <remarks>
606 /// In order to prevent malicious messages from exhausting memory or
607 /// causing integer overflows, CodedInputStream limits how large a message may be.
608 /// The default limit is 64MB. You should set this limit as small
609 /// as you can without harming your app's functionality. Note that
610 /// size limits only apply when reading from an InputStream, not
611 /// when constructed around a raw byte array (nor with ByteString.NewCodedInput).
Jon Skeet2e6dc122009-05-29 06:34:52 +0100612 /// If you want to read several messages from a single CodedInputStream, you
613 /// can call ResetSizeCounter() after each message to avoid hitting the
614 /// size limit.
Jon Skeet68036862008-10-22 13:30:34 +0100615 /// </remarks>
616 public int SetSizeLimit(int limit) {
617 if (limit < 0) {
618 throw new ArgumentOutOfRangeException("Size limit cannot be negative: " + limit);
619 }
620 int oldLimit = sizeLimit;
621 sizeLimit = limit;
622 return oldLimit;
623 }
624
625 #region Internal reading and buffer management
626 /// <summary>
Jon Skeet2e6dc122009-05-29 06:34:52 +0100627 /// Resets the current size counter to zero (see SetSizeLimit).
628 /// </summary>
629 public void ResetSizeCounter() {
630 totalBytesRetired = 0;
631 }
632
633 /// <summary>
Jon Skeet68036862008-10-22 13:30:34 +0100634 /// Sets currentLimit to (current position) + byteLimit. This is called
635 /// when descending into a length-delimited embedded message. The previous
636 /// limit is returned.
637 /// </summary>
638 /// <returns>The old limit.</returns>
639 public int PushLimit(int byteLimit) {
640 if (byteLimit < 0) {
641 throw InvalidProtocolBufferException.NegativeSize();
642 }
643 byteLimit += totalBytesRetired + bufferPos;
644 int oldLimit = currentLimit;
645 if (byteLimit > oldLimit) {
646 throw InvalidProtocolBufferException.TruncatedMessage();
647 }
648 currentLimit = byteLimit;
649
650 RecomputeBufferSizeAfterLimit();
651
652 return oldLimit;
653 }
654
655 private void RecomputeBufferSizeAfterLimit() {
656 bufferSize += bufferSizeAfterLimit;
657 int bufferEnd = totalBytesRetired + bufferSize;
658 if (bufferEnd > currentLimit) {
659 // Limit is in current buffer.
660 bufferSizeAfterLimit = bufferEnd - currentLimit;
661 bufferSize -= bufferSizeAfterLimit;
662 } else {
663 bufferSizeAfterLimit = 0;
664 }
665 }
666
667 /// <summary>
668 /// Discards the current limit, returning the previous limit.
669 /// </summary>
670 public void PopLimit(int oldLimit) {
671 currentLimit = oldLimit;
672 RecomputeBufferSizeAfterLimit();
673 }
674
675 /// <summary>
Jon Skeet25a28582009-02-18 16:06:22 +0000676 /// Returns whether or not all the data before the limit has been read.
677 /// </summary>
678 /// <returns></returns>
679 public bool ReachedLimit {
680 get {
681 if (currentLimit == int.MaxValue) {
682 return false;
683 }
684 int currentAbsolutePosition = totalBytesRetired + bufferPos;
685 return currentAbsolutePosition >= currentLimit;
686 }
687 }
Jon Skeet2e6dc122009-05-29 06:34:52 +0100688
689 /// <summary>
690 /// Returns true if the stream has reached the end of the input. This is the
691 /// case if either the end of the underlying input source has been reached or
692 /// the stream has reached a limit created using PushLimit.
693 /// </summary>
694 public bool IsAtEnd {
695 get {
696 return bufferPos == bufferSize && !RefillBuffer(false);
697 }
698 }
Jon Skeet25a28582009-02-18 16:06:22 +0000699
700 /// <summary>
Jon Skeet68036862008-10-22 13:30:34 +0100701 /// Called when buffer is empty to read more bytes from the
702 /// input. If <paramref name="mustSucceed"/> is true, RefillBuffer() gurantees that
703 /// either there will be at least one byte in the buffer when it returns
704 /// or it will throw an exception. If <paramref name="mustSucceed"/> is false,
705 /// RefillBuffer() returns false if no more bytes were available.
706 /// </summary>
707 /// <param name="mustSucceed"></param>
708 /// <returns></returns>
709 private bool RefillBuffer(bool mustSucceed) {
710 if (bufferPos < bufferSize) {
711 throw new InvalidOperationException("RefillBuffer() called when buffer wasn't empty.");
712 }
713
714 if (totalBytesRetired + bufferSize == currentLimit) {
715 // Oops, we hit a limit.
716 if (mustSucceed) {
717 throw InvalidProtocolBufferException.TruncatedMessage();
718 } else {
719 return false;
720 }
721 }
722
723 totalBytesRetired += bufferSize;
724
725 bufferPos = 0;
726 bufferSize = (input == null) ? 0 : input.Read(buffer, 0, buffer.Length);
Jon Skeet2e6dc122009-05-29 06:34:52 +0100727 if (bufferSize < 0) {
728 throw new InvalidOperationException("Stream.Read returned a negative count");
729 }
Jon Skeet68036862008-10-22 13:30:34 +0100730 if (bufferSize == 0) {
731 if (mustSucceed) {
732 throw InvalidProtocolBufferException.TruncatedMessage();
733 } else {
734 return false;
735 }
736 } else {
737 RecomputeBufferSizeAfterLimit();
738 int totalBytesRead =
739 totalBytesRetired + bufferSize + bufferSizeAfterLimit;
740 if (totalBytesRead > sizeLimit || totalBytesRead < 0) {
741 throw InvalidProtocolBufferException.SizeLimitExceeded();
742 }
743 return true;
744 }
745 }
746
747 /// <summary>
748 /// Read one byte from the input.
749 /// </summary>
750 /// <exception cref="InvalidProtocolBufferException">
Jon Skeet2178b932009-06-25 07:52:07 +0100751 /// the end of the stream or the current limit was reached
Jon Skeet68036862008-10-22 13:30:34 +0100752 /// </exception>
753 public byte ReadRawByte() {
754 if (bufferPos == bufferSize) {
755 RefillBuffer(true);
756 }
757 return buffer[bufferPos++];
758 }
759
760 /// <summary>
761 /// Read a fixed size of bytes from the input.
762 /// </summary>
763 /// <exception cref="InvalidProtocolBufferException">
764 /// the end of the stream or the current limit was reached
765 /// </exception>
766 public byte[] ReadRawBytes(int size) {
767 if (size < 0) {
768 throw InvalidProtocolBufferException.NegativeSize();
769 }
770
771 if (totalBytesRetired + bufferPos + size > currentLimit) {
772 // Read to the end of the stream anyway.
773 SkipRawBytes(currentLimit - totalBytesRetired - bufferPos);
774 // Then fail.
775 throw InvalidProtocolBufferException.TruncatedMessage();
776 }
777
778 if (size <= bufferSize - bufferPos) {
779 // We have all the bytes we need already.
780 byte[] bytes = new byte[size];
781 Array.Copy(buffer, bufferPos, bytes, 0, size);
782 bufferPos += size;
783 return bytes;
784 } else if (size < BufferSize) {
785 // Reading more bytes than are in the buffer, but not an excessive number
786 // of bytes. We can safely allocate the resulting array ahead of time.
787
788 // First copy what we have.
789 byte[] bytes = new byte[size];
790 int pos = bufferSize - bufferPos;
791 Array.Copy(buffer, bufferPos, bytes, 0, pos);
792 bufferPos = bufferSize;
793
794 // We want to use RefillBuffer() and then copy from the buffer into our
795 // byte array rather than reading directly into our byte array because
796 // the input may be unbuffered.
797 RefillBuffer(true);
798
799 while (size - pos > bufferSize) {
800 Array.Copy(buffer, 0, bytes, pos, bufferSize);
801 pos += bufferSize;
802 bufferPos = bufferSize;
803 RefillBuffer(true);
804 }
805
806 Array.Copy(buffer, 0, bytes, pos, size - pos);
807 bufferPos = size - pos;
808
809 return bytes;
810 } else {
811 // The size is very large. For security reasons, we can't allocate the
812 // entire byte array yet. The size comes directly from the input, so a
813 // maliciously-crafted message could provide a bogus very large size in
814 // order to trick the app into allocating a lot of memory. We avoid this
815 // by allocating and reading only a small chunk at a time, so that the
816 // malicious message must actually *be* extremely large to cause
817 // problems. Meanwhile, we limit the allowed size of a message elsewhere.
818
819 // Remember the buffer markers since we'll have to copy the bytes out of
820 // it later.
821 int originalBufferPos = bufferPos;
822 int originalBufferSize = bufferSize;
823
824 // Mark the current buffer consumed.
825 totalBytesRetired += bufferSize;
826 bufferPos = 0;
827 bufferSize = 0;
828
829 // Read all the rest of the bytes we need.
830 int sizeLeft = size - (originalBufferSize - originalBufferPos);
831 List<byte[]> chunks = new List<byte[]>();
832
833 while (sizeLeft > 0) {
834 byte[] chunk = new byte[Math.Min(sizeLeft, BufferSize)];
835 int pos = 0;
836 while (pos < chunk.Length) {
837 int n = (input == null) ? -1 : input.Read(chunk, pos, chunk.Length - pos);
838 if (n <= 0) {
839 throw InvalidProtocolBufferException.TruncatedMessage();
840 }
841 totalBytesRetired += n;
842 pos += n;
843 }
844 sizeLeft -= chunk.Length;
845 chunks.Add(chunk);
846 }
847
848 // OK, got everything. Now concatenate it all into one buffer.
849 byte[] bytes = new byte[size];
850
851 // Start by copying the leftover bytes from this.buffer.
852 int newPos = originalBufferSize - originalBufferPos;
853 Array.Copy(buffer, originalBufferPos, bytes, 0, newPos);
854
855 // And now all the chunks.
856 foreach (byte[] chunk in chunks) {
857 Array.Copy(chunk, 0, bytes, newPos, chunk.Length);
858 newPos += chunk.Length;
859 }
860
861 // Done.
862 return bytes;
863 }
864 }
865
866 /// <summary>
867 /// Reads and discards a single field, given its tag value.
868 /// </summary>
869 /// <returns>false if the tag is an end-group tag, in which case
870 /// nothing is skipped. Otherwise, returns true.</returns>
Jon Skeetd6dd0a42009-06-05 22:00:05 +0100871 [CLSCompliant(false)]
Jon Skeet68036862008-10-22 13:30:34 +0100872 public bool SkipField(uint tag) {
873 switch (WireFormat.GetTagWireType(tag)) {
874 case WireFormat.WireType.Varint:
875 ReadInt32();
876 return true;
877 case WireFormat.WireType.Fixed64:
878 ReadRawLittleEndian64();
879 return true;
880 case WireFormat.WireType.LengthDelimited:
881 SkipRawBytes((int) ReadRawVarint32());
882 return true;
883 case WireFormat.WireType.StartGroup:
884 SkipMessage();
885 CheckLastTagWas(
886 WireFormat.MakeTag(WireFormat.GetTagFieldNumber(tag),
887 WireFormat.WireType.EndGroup));
888 return true;
889 case WireFormat.WireType.EndGroup:
890 return false;
891 case WireFormat.WireType.Fixed32:
892 ReadRawLittleEndian32();
893 return true;
894 default:
895 throw InvalidProtocolBufferException.InvalidWireType();
896 }
897 }
898
899 /// <summary>
900 /// Reads and discards an entire message. This will read either until EOF
901 /// or until an endgroup tag, whichever comes first.
902 /// </summary>
903 public void SkipMessage() {
904 while (true) {
905 uint tag = ReadTag();
906 if (tag == 0 || !SkipField(tag)) {
907 return;
908 }
909 }
910 }
911
912 /// <summary>
913 /// Reads and discards <paramref name="size"/> bytes.
914 /// </summary>
915 /// <exception cref="InvalidProtocolBufferException">the end of the stream
916 /// or the current limit was reached</exception>
917 public void SkipRawBytes(int size) {
918 if (size < 0) {
919 throw InvalidProtocolBufferException.NegativeSize();
920 }
921
922 if (totalBytesRetired + bufferPos + size > currentLimit) {
923 // Read to the end of the stream anyway.
924 SkipRawBytes(currentLimit - totalBytesRetired - bufferPos);
925 // Then fail.
926 throw InvalidProtocolBufferException.TruncatedMessage();
927 }
928
Jon Skeet2e6dc122009-05-29 06:34:52 +0100929 if (size <= bufferSize - bufferPos) {
Jon Skeet68036862008-10-22 13:30:34 +0100930 // We have all the bytes we need already.
931 bufferPos += size;
932 } else {
933 // Skipping more bytes than are in the buffer. First skip what we have.
934 int pos = bufferSize - bufferPos;
935 totalBytesRetired += pos;
936 bufferPos = 0;
937 bufferSize = 0;
938
939 // Then skip directly from the InputStream for the rest.
940 if (pos < size) {
Jon Skeet68036862008-10-22 13:30:34 +0100941 if (input == null) {
942 throw InvalidProtocolBufferException.TruncatedMessage();
943 }
Jon Skeetc298c892009-05-30 10:07:09 +0100944 SkipImpl(size - pos);
945 totalBytesRetired += size - pos;
946 }
947 }
948 }
949
950 /// <summary>
951 /// Abstraction of skipping to cope with streams which can't really skip.
952 /// </summary>
953 private void SkipImpl(int amountToSkip) {
954 if (input.CanSeek) {
955 long previousPosition = input.Position;
956 input.Position += amountToSkip;
957 if (input.Position != previousPosition + amountToSkip) {
958 throw InvalidProtocolBufferException.TruncatedMessage();
959 }
960 } else {
961 byte[] skipBuffer = new byte[1024];
962 while (amountToSkip > 0) {
963 int bytesRead = input.Read(skipBuffer, 0, skipBuffer.Length);
964 if (bytesRead <= 0) {
Jon Skeet68036862008-10-22 13:30:34 +0100965 throw InvalidProtocolBufferException.TruncatedMessage();
966 }
Jon Skeetc298c892009-05-30 10:07:09 +0100967 amountToSkip -= bytesRead;
Jon Skeet68036862008-10-22 13:30:34 +0100968 }
969 }
970 }
971 #endregion
972 }
973}