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 // An Encoder is used to encode a sequence of blocks of characters into
13 // a sequence of blocks of bytes. Following instantiation of an encoder,
14 // sequential blocks of characters are converted into blocks of bytes through
15 // calls to the GetBytes method. The encoder maintains state between the
16 // conversions, allowing it to correctly encode character sequences that span
19 // Instances of specific implementations of the Encoder abstract base
20 // class are typically obtained through calls to the GetEncoder method
21 // of Encoding objects.
23 public abstract class Encoder
25 internal EncoderFallback _fallback = null;
27 internal EncoderFallbackBuffer _fallbackBuffer = null;
31 // We don't call default reset because default reset probably isn't good if we aren't initialized.
34 public EncoderFallback 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 EncoderFallbackBuffer FallbackBuffer
63 if (_fallbackBuffer == null)
65 if (_fallback != null)
66 _fallbackBuffer = _fallback.CreateFallbackBuffer();
68 _fallbackBuffer = EncoderFallback.ReplacementFallback.CreateFallbackBuffer();
71 return _fallbackBuffer;
75 internal bool InternalHasFallbackBuffer
79 return _fallbackBuffer != null;
85 // Normally if we call GetBytes() and an error is thrown we don't change the state of the encoder. 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 GetBytes() throws an error, then they need to call Reset().
90 // Virtual implementation has to call GetBytes with flush and a big enough buffer to clear a 0 char string
91 // We avoid GetMaxByteCount() because a) we can't call the base encoder and b) it might be really big.
92 public virtual void Reset()
94 char[] charTemp = { };
95 byte[] byteTemp = new byte[GetByteCount(charTemp, 0, 0, true)];
96 GetBytes(charTemp, 0, 0, byteTemp, 0, true);
97 if (_fallbackBuffer != null)
98 _fallbackBuffer.Reset();
101 // Returns the number of bytes the next call to GetBytes will
102 // produce if presented with the given range of characters and the given
103 // value of the flush parameter. The returned value takes into
104 // account the state in which the encoder was left following the last call
105 // to GetBytes. The state of the encoder is not affected by a call
108 public abstract int GetByteCount(char[] chars, int index, int count, bool flush);
110 // We expect this to be the workhorse for NLS encodings
111 // unfortunately for existing overrides, it has to call the [] version,
112 // which is really slow, so avoid this method if you might be calling external encodings.
113 [CLSCompliant(false)]
114 public virtual unsafe int GetByteCount(char* chars, int count, bool flush)
116 // Validate input parameters
118 throw new ArgumentNullException(nameof(chars),
119 SR.ArgumentNull_Array);
122 throw new ArgumentOutOfRangeException(nameof(count),
123 SR.ArgumentOutOfRange_NeedNonNegNum);
124 Contract.EndContractBlock();
126 char[] arrChar = new char[count];
129 for (index = 0; index < count; index++)
130 arrChar[index] = chars[index];
132 return GetByteCount(arrChar, 0, count, flush);
135 // Encodes a range of characters in a character array into a range of bytes
136 // in a byte array. The method encodes charCount characters from
137 // chars starting at index charIndex, storing the resulting
138 // bytes in bytes starting at index byteIndex. The encoding
139 // takes into account the state in which the encoder was left following the
140 // last call to this method. The flush parameter indicates whether
141 // the encoder should flush any shift-states and partial characters at the
142 // end of the conversion. To ensure correct termination of a sequence of
143 // blocks of encoded bytes, the last call to GetBytes should specify
144 // a value of true for the flush parameter.
146 // An exception occurs if the byte array is not large enough to hold the
147 // complete encoding of the characters. The GetByteCount method can
148 // be used to determine the exact number of bytes that will be produced for
149 // a given range of characters. Alternatively, the GetMaxByteCount
150 // method of the Encoding that produced this encoder can be used to
151 // determine the maximum number of bytes that will be produced for a given
152 // number of characters, regardless of the actual character values.
154 public abstract int GetBytes(char[] chars, int charIndex, int charCount,
155 byte[] bytes, int byteIndex, bool flush);
157 // We expect this to be the workhorse for NLS Encodings, but for existing
158 // ones we need a working (if slow) default implementation)
160 // WARNING WARNING WARNING
162 // WARNING: If this breaks it could be a security threat. Obviously we
163 // call this internally, so you need to make sure that your pointers, counts
164 // and indexes are correct when you call this method.
166 // In addition, we have internal code, which will be marked as "safe" calling
167 // this code. However this code is dependent upon the implementation of an
168 // external GetBytes() method, which could be overridden by a third party and
169 // the results of which cannot be guaranteed. We use that result to copy
170 // the byte[] to our byte* output buffer. If the result count was wrong, we
171 // could easily overflow our output buffer. Therefore we do an extra test
172 // when we copy the buffer so that we don't overflow byteCount either.
173 [CLSCompliant(false)]
174 public virtual unsafe int GetBytes(char* chars, int charCount,
175 byte* bytes, int byteCount, bool flush)
177 // Validate input parameters
178 if (bytes == null || chars == null)
179 throw new ArgumentNullException(bytes == null ? nameof(bytes) : nameof(chars),
180 SR.ArgumentNull_Array);
182 if (charCount < 0 || byteCount < 0)
183 throw new ArgumentOutOfRangeException((charCount < 0 ? nameof(charCount) : nameof(byteCount)),
184 SR.ArgumentOutOfRange_NeedNonNegNum);
185 Contract.EndContractBlock();
187 // Get the char array to convert
188 char[] arrChar = new char[charCount];
191 for (index = 0; index < charCount; index++)
192 arrChar[index] = chars[index];
194 // Get the byte array to fill
195 byte[] arrByte = new byte[byteCount];
198 int result = GetBytes(arrChar, 0, charCount, arrByte, 0, flush);
200 Debug.Assert(result <= byteCount, "Returned more bytes than we have space for");
202 // Copy the byte array
203 // WARNING: We MUST make sure that we don't copy too many bytes. We can't
204 // rely on result because it could be a 3rd party implementation. We need
205 // to make sure we never copy more than byteCount bytes no matter the value
207 if (result < byteCount)
210 // Don't copy too many bytes!
211 for (index = 0; index < byteCount; index++)
212 bytes[index] = arrByte[index];
217 // This method is used to avoid running out of output buffer space.
218 // It will encode until it runs out of chars, and then it will return
219 // true if it the entire input was converted. In either case it
220 // will also return the number of converted chars and output bytes used.
221 // It will only throw a buffer overflow exception if the entire lenght of bytes[] is
222 // too small to store the next byte. (like 0 or maybe 1 or 4 for some encodings)
223 // We're done processing this buffer only if completed returns true.
225 // Might consider checking Max...Count to avoid the extra counting step.
227 // Note that if all of the input chars are not consumed, then we'll do a /2, which means
228 // that its likely that we didn't consume as many chars as we could have. For some
229 // applications this could be slow. (Like trying to exactly fill an output buffer from a bigger stream)
230 public virtual void Convert(char[] chars, int charIndex, int charCount,
231 byte[] bytes, int byteIndex, int byteCount, bool flush,
232 out int charsUsed, out int bytesUsed, out bool completed)
234 // Validate parameters
235 if (chars == null || bytes == null)
236 throw new ArgumentNullException((chars == null ? nameof(chars) : nameof(bytes)),
237 SR.ArgumentNull_Array);
239 if (charIndex < 0 || charCount < 0)
240 throw new ArgumentOutOfRangeException((charIndex < 0 ? nameof(charIndex) : nameof(charCount)),
241 SR.ArgumentOutOfRange_NeedNonNegNum);
243 if (byteIndex < 0 || byteCount < 0)
244 throw new ArgumentOutOfRangeException((byteIndex < 0 ? nameof(byteIndex) : nameof(byteCount)),
245 SR.ArgumentOutOfRange_NeedNonNegNum);
247 if (chars.Length - charIndex < charCount)
248 throw new ArgumentOutOfRangeException(nameof(chars),
249 SR.ArgumentOutOfRange_IndexCountBuffer);
251 if (bytes.Length - byteIndex < byteCount)
252 throw new ArgumentOutOfRangeException(nameof(bytes),
253 SR.ArgumentOutOfRange_IndexCountBuffer);
254 Contract.EndContractBlock();
256 charsUsed = charCount;
258 // Its easy to do if it won't overrun our buffer.
259 // Note: We don't want to call unsafe version because that might be an untrusted version
260 // which could be really unsafe and we don't want to mix it up.
261 while (charsUsed > 0)
263 if (GetByteCount(chars, charIndex, charsUsed, flush) <= byteCount)
265 bytesUsed = GetBytes(chars, charIndex, charsUsed, bytes, byteIndex, flush);
266 completed = (charsUsed == charCount &&
267 (_fallbackBuffer == null || _fallbackBuffer.Remaining == 0));
271 // Try again with 1/2 the count, won't flush then 'cause won't read it all
276 // Oops, we didn't have anything, we'll have to throw an overflow
277 throw new ArgumentException(SR.Argument_ConversionOverflow);
280 // Same thing, but using pointers
282 // Might consider checking Max...Count to avoid the extra counting step.
284 // Note that if all of the input chars are not consumed, then we'll do a /2, which means
285 // that its likely that we didn't consume as many chars as we could have. For some
286 // applications this could be slow. (Like trying to exactly fill an output buffer from a bigger stream)
287 [CLSCompliant(false)]
288 public virtual unsafe void Convert(char* chars, int charCount,
289 byte* bytes, int byteCount, bool flush,
290 out int charsUsed, out int bytesUsed, out bool completed)
292 // Validate input parameters
293 if (bytes == null || chars == null)
294 throw new ArgumentNullException(bytes == null ? nameof(bytes) : nameof(chars),
295 SR.ArgumentNull_Array);
296 if (charCount < 0 || byteCount < 0)
297 throw new ArgumentOutOfRangeException((charCount < 0 ? nameof(charCount) : nameof(byteCount)),
298 SR.ArgumentOutOfRange_NeedNonNegNum);
299 Contract.EndContractBlock();
301 // Get ready to do it
302 charsUsed = charCount;
304 // Its easy to do if it won't overrun our buffer.
305 while (charsUsed > 0)
307 if (GetByteCount(chars, charsUsed, flush) <= byteCount)
309 bytesUsed = GetBytes(chars, charsUsed, bytes, byteCount, flush);
310 completed = (charsUsed == charCount &&
311 (_fallbackBuffer == null || _fallbackBuffer.Remaining == 0));
315 // Try again with 1/2 the count, won't flush then 'cause won't read it all
320 // Oops, we didn't have anything, we'll have to throw an overflow
321 throw new ArgumentException(SR.Argument_ConversionOverflow);