1 // Licensed to the .NET Foundation under one or more agreements.
2 // The .NET Foundation licenses this file to you under the MIT license.
3 // See the LICENSE file in the project root for more information.
7 using System.Diagnostics;
8 using System.Diagnostics.Contracts;
12 // A Decoder is used to decode a sequence of blocks of bytes into a
13 // sequence of blocks of characters. Following instantiation of a decoder,
14 // sequential blocks of bytes are converted into blocks of characters through
15 // calls to the GetChars method. The decoder maintains state between the
16 // conversions, allowing it to correctly decode byte sequences that span
19 // Instances of specific implementations of the Decoder abstract base
20 // class are typically obtained through calls to the GetDecoder method
21 // of Encoding objects.
23 public abstract class Decoder
25 internal DecoderFallback _fallback = null;
27 internal DecoderFallbackBuffer _fallbackBuffer = null;
31 // We don't call default reset because default reset probably isn't good if we aren't initialized.
34 public DecoderFallback Fallback
44 throw new ArgumentNullException(nameof(value));
45 Contract.EndContractBlock();
47 // Can't change fallback if buffer is wrong
48 if (_fallbackBuffer != null && _fallbackBuffer.Remaining > 0)
49 throw new ArgumentException(
50 SR.Argument_FallbackBufferNotEmpty, nameof(value));
53 _fallbackBuffer = null;
57 // Note: we don't test for threading here because async access to Encoders and Decoders
58 // doesn't work anyway.
59 public DecoderFallbackBuffer FallbackBuffer
63 if (_fallbackBuffer == null)
65 if (_fallback != null)
66 _fallbackBuffer = _fallback.CreateFallbackBuffer();
68 _fallbackBuffer = DecoderFallback.ReplacementFallback.CreateFallbackBuffer();
71 return _fallbackBuffer;
75 internal bool InternalHasFallbackBuffer
79 return _fallbackBuffer != null;
85 // Normally if we call GetChars() and an error is thrown we don't change the state of the Decoder. This
86 // would allow the caller to correct the error condition and try again (such as if they need a bigger buffer.)
88 // If the caller doesn't want to try again after GetChars() throws an error, then they need to call Reset().
90 // Virtual implementation has to call GetChars with flush and a big enough buffer to clear a 0 byte string
91 // We avoid GetMaxCharCount() because a) we can't call the base encoder and b) it might be really big.
92 public virtual void Reset()
94 byte[] byteTemp = Array.Empty<byte>();
95 char[] charTemp = new char[GetCharCount(byteTemp, 0, 0, true)];
96 GetChars(byteTemp, 0, 0, charTemp, 0, true);
97 _fallbackBuffer?.Reset();
100 // Returns the number of characters the next call to GetChars will
101 // produce if presented with the given range of bytes. The returned value
102 // takes into account the state in which the decoder was left following the
103 // last call to GetChars. The state of the decoder is not affected
104 // by a call to this method.
106 public abstract int GetCharCount(byte[] bytes, int index, int count);
108 public virtual int GetCharCount(byte[] bytes, int index, int count, bool flush)
110 return GetCharCount(bytes, index, count);
113 // We expect this to be the workhorse for NLS Encodings, but for existing
114 // ones we need a working (if slow) default implementation)
115 [CLSCompliant(false)]
116 public virtual unsafe int GetCharCount(byte* bytes, int count, bool flush)
118 // Validate input parameters
120 throw new ArgumentNullException(nameof(bytes),
121 SR.ArgumentNull_Array);
124 throw new ArgumentOutOfRangeException(nameof(count),
125 SR.ArgumentOutOfRange_NeedNonNegNum);
126 Contract.EndContractBlock();
128 byte[] arrbyte = new byte[count];
131 for (index = 0; index < count; index++)
132 arrbyte[index] = bytes[index];
134 return GetCharCount(arrbyte, 0, count);
137 // Decodes a range of bytes in a byte array into a range of characters
138 // in a character array. The method decodes byteCount bytes from
139 // bytes starting at index byteIndex, storing the resulting
140 // characters in chars starting at index charIndex. The
141 // decoding takes into account the state in which the decoder was left
142 // following the last call to this method.
144 // An exception occurs if the character array is not large enough to
145 // hold the complete decoding of the bytes. The GetCharCount method
146 // can be used to determine the exact number of characters that will be
147 // produced for a given range of bytes. Alternatively, the
148 // GetMaxCharCount method of the Encoding that produced this
149 // decoder can be used to determine the maximum number of characters that
150 // will be produced for a given number of bytes, regardless of the actual
153 public abstract int GetChars(byte[] bytes, int byteIndex, int byteCount,
154 char[] chars, int charIndex);
156 public virtual int GetChars(byte[] bytes, int byteIndex, int byteCount,
157 char[] chars, int charIndex, bool flush)
159 return GetChars(bytes, byteIndex, byteCount, chars, charIndex);
162 // We expect this to be the workhorse for NLS Encodings, but for existing
163 // ones we need a working (if slow) default implementation)
165 // WARNING WARNING WARNING
167 // WARNING: If this breaks it could be a security threat. Obviously we
168 // call this internally, so you need to make sure that your pointers, counts
169 // and indexes are correct when you call this method.
171 // In addition, we have internal code, which will be marked as "safe" calling
172 // this code. However this code is dependent upon the implementation of an
173 // external GetChars() method, which could be overridden by a third party and
174 // the results of which cannot be guaranteed. We use that result to copy
175 // the char[] to our char* output buffer. If the result count was wrong, we
176 // could easily overflow our output buffer. Therefore we do an extra test
177 // when we copy the buffer so that we don't overflow charCount either.
178 [CLSCompliant(false)]
179 public virtual unsafe int GetChars(byte* bytes, int byteCount,
180 char* chars, int charCount, bool flush)
182 // Validate input parameters
183 if (chars == null || bytes == null)
184 throw new ArgumentNullException(chars == null ? nameof(chars) : nameof(bytes),
185 SR.ArgumentNull_Array);
187 if (byteCount < 0 || charCount < 0)
188 throw new ArgumentOutOfRangeException((byteCount < 0 ? nameof(byteCount) : nameof(charCount)),
189 SR.ArgumentOutOfRange_NeedNonNegNum);
190 Contract.EndContractBlock();
192 // Get the byte array to convert
193 byte[] arrByte = new byte[byteCount];
196 for (index = 0; index < byteCount; index++)
197 arrByte[index] = bytes[index];
199 // Get the char array to fill
200 char[] arrChar = new char[charCount];
203 int result = GetChars(arrByte, 0, byteCount, arrChar, 0, flush);
205 Debug.Assert(result <= charCount, "Returned more chars than we have space for");
207 // Copy the char array
208 // WARNING: We MUST make sure that we don't copy too many chars. We can't
209 // rely on result because it could be a 3rd party implementation. We need
210 // to make sure we never copy more than charCount chars no matter the value
212 if (result < charCount)
215 // We check both result and charCount so that we don't accidentally overrun
216 // our pointer buffer just because of an issue in GetChars
217 for (index = 0; index < charCount; index++)
218 chars[index] = arrChar[index];
223 // This method is used when the output buffer might not be large enough.
224 // It will decode until it runs out of bytes, and then it will return
225 // true if it the entire input was converted. In either case it
226 // will also return the number of converted bytes and output characters used.
227 // It will only throw a buffer overflow exception if the entire lenght of chars[] is
228 // too small to store the next char. (like 0 or maybe 1 or 4 for some encodings)
229 // We're done processing this buffer only if completed returns true.
231 // Might consider checking Max...Count to avoid the extra counting step.
233 // Note that if all of the input bytes are not consumed, then we'll do a /2, which means
234 // that its likely that we didn't consume as many bytes as we could have. For some
235 // applications this could be slow. (Like trying to exactly fill an output buffer from a bigger stream)
236 public virtual void Convert(byte[] bytes, int byteIndex, int byteCount,
237 char[] chars, int charIndex, int charCount, bool flush,
238 out int bytesUsed, out int charsUsed, out bool completed)
240 // Validate parameters
241 if (bytes == null || chars == null)
242 throw new ArgumentNullException((bytes == null ? nameof(bytes) : nameof(chars)),
243 SR.ArgumentNull_Array);
245 if (byteIndex < 0 || byteCount < 0)
246 throw new ArgumentOutOfRangeException((byteIndex < 0 ? nameof(byteIndex) : nameof(byteCount)),
247 SR.ArgumentOutOfRange_NeedNonNegNum);
249 if (charIndex < 0 || charCount < 0)
250 throw new ArgumentOutOfRangeException((charIndex < 0 ? nameof(charIndex) : nameof(charCount)),
251 SR.ArgumentOutOfRange_NeedNonNegNum);
253 if (bytes.Length - byteIndex < byteCount)
254 throw new ArgumentOutOfRangeException(nameof(bytes),
255 SR.ArgumentOutOfRange_IndexCountBuffer);
257 if (chars.Length - charIndex < charCount)
258 throw new ArgumentOutOfRangeException(nameof(chars),
259 SR.ArgumentOutOfRange_IndexCountBuffer);
260 Contract.EndContractBlock();
262 bytesUsed = byteCount;
264 // Its easy to do if it won't overrun our buffer.
265 while (bytesUsed > 0)
267 if (GetCharCount(bytes, byteIndex, bytesUsed, flush) <= charCount)
269 charsUsed = GetChars(bytes, byteIndex, bytesUsed, chars, charIndex, flush);
270 completed = (bytesUsed == byteCount &&
271 (_fallbackBuffer == null || _fallbackBuffer.Remaining == 0));
275 // Try again with 1/2 the count, won't flush then 'cause won't read it all
280 // Oops, we didn't have anything, we'll have to throw an overflow
281 throw new ArgumentException(SR.Argument_ConversionOverflow);
284 // This is the version that uses *.
285 // We're done processing this buffer only if completed returns true.
287 // Might consider checking Max...Count to avoid the extra counting step.
289 // Note that if all of the input bytes are not consumed, then we'll do a /2, which means
290 // that its likely that we didn't consume as many bytes as we could have. For some
291 // applications this could be slow. (Like trying to exactly fill an output buffer from a bigger stream)
292 [CLSCompliant(false)]
293 public virtual unsafe void Convert(byte* bytes, int byteCount,
294 char* chars, int charCount, bool flush,
295 out int bytesUsed, out int charsUsed, out bool completed)
297 // Validate input parameters
298 if (chars == null || bytes == null)
299 throw new ArgumentNullException(chars == null ? nameof(chars) : nameof(bytes),
300 SR.ArgumentNull_Array);
302 if (byteCount < 0 || charCount < 0)
303 throw new ArgumentOutOfRangeException((byteCount < 0 ? nameof(byteCount) : nameof(charCount)),
304 SR.ArgumentOutOfRange_NeedNonNegNum);
305 Contract.EndContractBlock();
307 // Get ready to do it
308 bytesUsed = byteCount;
310 // Its easy to do if it won't overrun our buffer.
311 while (bytesUsed > 0)
313 if (GetCharCount(bytes, bytesUsed, flush) <= charCount)
315 charsUsed = GetChars(bytes, bytesUsed, chars, charCount, flush);
316 completed = (bytesUsed == byteCount &&
317 (_fallbackBuffer == null || _fallbackBuffer.Remaining == 0));
321 // Try again with 1/2 the count, won't flush then 'cause won't read it all
326 // Oops, we didn't have anything, we'll have to throw an overflow
327 throw new ArgumentException(SR.Argument_ConversionOverflow);