1// © 2016 and later: Unicode, Inc. and others.
2// License & terms of use: http://www.unicode.org/copyright.html
3// Copyright (C) 2009-2012, International Business Machines
4// Corporation and others. All Rights Reserved.
5//
6// Copyright 2007 Google Inc. All Rights Reserved.
7// Author: sanjay@google.com (Sanjay Ghemawat)
8//
9// Abstract interface that consumes a sequence of bytes (ByteSink).
10//
11// Used so that we can write a single piece of code that can operate
12// on a variety of output string types.
13//
14// Various implementations of this interface are provided:
15// ByteSink:
16// CheckedArrayByteSink Write to a flat array, with bounds checking
17// StringByteSink Write to an STL string
18
19// This code is a contribution of Google code, and the style used here is
20// a compromise between the original Google code and the ICU coding guidelines.
21// For example, data types are ICU-ified (size_t,int->int32_t),
22// and API comments doxygen-ified, but function names and behavior are
23// as in the original, if possible.
24// Assertion-style error handling, not available in ICU, was changed to
25// parameter "pinning" similar to UnicodeString.
26//
27// In addition, this is only a partial port of the original Google code,
28// limited to what was needed so far. The (nearly) complete original code
29// is in the ICU svn repository at icuhtml/trunk/design/strings/contrib
30// (see ICU ticket 6765, r25517).
31
32#ifndef __BYTESTREAM_H__
33#define __BYTESTREAM_H__
34
35/**
36 * \file
37 * \brief C++ API: Interface for writing bytes, and implementation classes.
38 */
39
40#include "unicode/utypes.h"
41
42#if U_SHOW_CPLUSPLUS_API
43
44#include "unicode/uobject.h"
45#include "unicode/std_string.h"
46
47U_NAMESPACE_BEGIN
48
49/**
50 * A ByteSink can be filled with bytes.
51 * @stable ICU 4.2
52 */
53class U_COMMON_API ByteSink : public UMemory {
54public:
55 /**
56 * Default constructor.
57 * @stable ICU 4.2
58 */
59 ByteSink() { }
60 /**
61 * Virtual destructor.
62 * @stable ICU 4.2
63 */
64 virtual ~ByteSink();
65
66 /**
67 * Append "bytes[0,n-1]" to this.
68 * @param bytes the pointer to the bytes
69 * @param n the number of bytes; must be non-negative
70 * @stable ICU 4.2
71 */
72 virtual void Append(const char* bytes, int32_t n) = 0;
73
74#ifndef U_HIDE_DRAFT_API
75 /**
76 * Appends n bytes to this. Same as Append().
77 * Call AppendU8() with u8"string literals" which are const char * in C++11
78 * but const char8_t * in C++20.
79 * If the compiler does support char8_t as a distinct type,
80 * then an AppendU8() overload for that is defined and will be chosen.
81 *
82 * @param bytes the pointer to the bytes
83 * @param n the number of bytes; must be non-negative
84 * @draft ICU 67
85 */
86 inline void AppendU8(const char* bytes, int32_t n) {
87 Append(bytes, n);
88 }
89
90#if defined(__cpp_char8_t) || defined(U_IN_DOXYGEN)
91 /**
92 * Appends n bytes to this. Same as Append() but for a const char8_t * pointer.
93 * Call AppendU8() with u8"string literals" which are const char * in C++11
94 * but const char8_t * in C++20.
95 * If the compiler does support char8_t as a distinct type,
96 * then this AppendU8() overload for that is defined and will be chosen.
97 *
98 * @param bytes the pointer to the bytes
99 * @param n the number of bytes; must be non-negative
100 * @draft ICU 67
101 */
102 inline void AppendU8(const char8_t* bytes, int32_t n) {
103 Append(reinterpret_cast<const char*>(bytes), n);
104 }
105#endif
106#endif // U_HIDE_DRAFT_API
107
108 /**
109 * Returns a writable buffer for appending and writes the buffer's capacity to
110 * *result_capacity. Guarantees *result_capacity>=min_capacity.
111 * May return a pointer to the caller-owned scratch buffer which must have
112 * scratch_capacity>=min_capacity.
113 * The returned buffer is only valid until the next operation
114 * on this ByteSink.
115 *
116 * After writing at most *result_capacity bytes, call Append() with the
117 * pointer returned from this function and the number of bytes written.
118 * Many Append() implementations will avoid copying bytes if this function
119 * returned an internal buffer.
120 *
121 * Partial usage example:
122 * int32_t capacity;
123 * char* buffer = sink->GetAppendBuffer(..., &capacity);
124 * ... Write n bytes into buffer, with n <= capacity.
125 * sink->Append(buffer, n);
126 * In many implementations, that call to Append will avoid copying bytes.
127 *
128 * If the ByteSink allocates or reallocates an internal buffer, it should use
129 * the desired_capacity_hint if appropriate.
130 * If a caller cannot provide a reasonable guess at the desired capacity,
131 * it should pass desired_capacity_hint=0.
132 *
133 * If a non-scratch buffer is returned, the caller may only pass
134 * a prefix to it to Append().
135 * That is, it is not correct to pass an interior pointer to Append().
136 *
137 * The default implementation always returns the scratch buffer.
138 *
139 * @param min_capacity required minimum capacity of the returned buffer;
140 * must be non-negative
141 * @param desired_capacity_hint desired capacity of the returned buffer;
142 * must be non-negative
143 * @param scratch default caller-owned buffer
144 * @param scratch_capacity capacity of the scratch buffer
145 * @param result_capacity pointer to an integer which will be set to the
146 * capacity of the returned buffer
147 * @return a buffer with *result_capacity>=min_capacity
148 * @stable ICU 4.2
149 */
150 virtual char* GetAppendBuffer(int32_t min_capacity,
151 int32_t desired_capacity_hint,
152 char* scratch, int32_t scratch_capacity,
153 int32_t* result_capacity);
154
155 /**
156 * Flush internal buffers.
157 * Some byte sinks use internal buffers or provide buffering
158 * and require calling Flush() at the end of the stream.
159 * The ByteSink should be ready for further Append() calls after Flush().
160 * The default implementation of Flush() does nothing.
161 * @stable ICU 4.2
162 */
163 virtual void Flush();
164
165private:
166 ByteSink(const ByteSink &) = delete;
167 ByteSink &operator=(const ByteSink &) = delete;
168};
169
170// -------------------------------------------------------------
171// Some standard implementations
172
173/**
174 * Implementation of ByteSink that writes to a flat byte array,
175 * with bounds-checking:
176 * This sink will not write more than capacity bytes to outbuf.
177 * If more than capacity bytes are Append()ed, then excess bytes are ignored,
178 * and Overflowed() will return true.
179 * Overflow does not cause a runtime error.
180 * @stable ICU 4.2
181 */
182class U_COMMON_API CheckedArrayByteSink : public ByteSink {
183public:
184 /**
185 * Constructs a ByteSink that will write to outbuf[0..capacity-1].
186 * @param outbuf buffer to write to
187 * @param capacity size of the buffer
188 * @stable ICU 4.2
189 */
190 CheckedArrayByteSink(char* outbuf, int32_t capacity);
191 /**
192 * Destructor.
193 * @stable ICU 4.2
194 */
195 virtual ~CheckedArrayByteSink();
196 /**
197 * Returns the sink to its original state, without modifying the buffer.
198 * Useful for reusing both the buffer and the sink for multiple streams.
199 * Resets the state to NumberOfBytesWritten()=NumberOfBytesAppended()=0
200 * and Overflowed()=FALSE.
201 * @return *this
202 * @stable ICU 4.6
203 */
204 virtual CheckedArrayByteSink& Reset();
205 /**
206 * Append "bytes[0,n-1]" to this.
207 * @param bytes the pointer to the bytes
208 * @param n the number of bytes; must be non-negative
209 * @stable ICU 4.2
210 */
211 virtual void Append(const char* bytes, int32_t n);
212 /**
213 * Returns a writable buffer for appending and writes the buffer's capacity to
214 * *result_capacity. For details see the base class documentation.
215 * @param min_capacity required minimum capacity of the returned buffer;
216 * must be non-negative
217 * @param desired_capacity_hint desired capacity of the returned buffer;
218 * must be non-negative
219 * @param scratch default caller-owned buffer
220 * @param scratch_capacity capacity of the scratch buffer
221 * @param result_capacity pointer to an integer which will be set to the
222 * capacity of the returned buffer
223 * @return a buffer with *result_capacity>=min_capacity
224 * @stable ICU 4.2
225 */
226 virtual char* GetAppendBuffer(int32_t min_capacity,
227 int32_t desired_capacity_hint,
228 char* scratch, int32_t scratch_capacity,
229 int32_t* result_capacity);
230 /**
231 * Returns the number of bytes actually written to the sink.
232 * @return number of bytes written to the buffer
233 * @stable ICU 4.2
234 */
235 int32_t NumberOfBytesWritten() const { return size_; }
236 /**
237 * Returns true if any bytes were discarded, i.e., if there was an
238 * attempt to write more than 'capacity' bytes.
239 * @return TRUE if more than 'capacity' bytes were Append()ed
240 * @stable ICU 4.2
241 */
242 UBool Overflowed() const { return overflowed_; }
243 /**
244 * Returns the number of bytes appended to the sink.
245 * If Overflowed() then NumberOfBytesAppended()>NumberOfBytesWritten()
246 * else they return the same number.
247 * @return number of bytes written to the buffer
248 * @stable ICU 4.6
249 */
250 int32_t NumberOfBytesAppended() const { return appended_; }
251private:
252 char* outbuf_;
253 const int32_t capacity_;
254 int32_t size_;
255 int32_t appended_;
256 UBool overflowed_;
257
258 CheckedArrayByteSink() = delete;
259 CheckedArrayByteSink(const CheckedArrayByteSink &) = delete;
260 CheckedArrayByteSink &operator=(const CheckedArrayByteSink &) = delete;
261};
262
263/**
264 * Implementation of ByteSink that writes to a "string".
265 * The StringClass is usually instantiated with a std::string.
266 * @stable ICU 4.2
267 */
268template<typename StringClass>
269class StringByteSink : public ByteSink {
270 public:
271 /**
272 * Constructs a ByteSink that will append bytes to the dest string.
273 * @param dest pointer to string object to append to
274 * @stable ICU 4.2
275 */
276 StringByteSink(StringClass* dest) : dest_(dest) { }
277 /**
278 * Constructs a ByteSink that reserves append capacity and will append bytes to the dest string.
279 *
280 * @param dest pointer to string object to append to
281 * @param initialAppendCapacity capacity beyond dest->length() to be reserve()d
282 * @stable ICU 60
283 */
284 StringByteSink(StringClass* dest, int32_t initialAppendCapacity) : dest_(dest) {
285 if (initialAppendCapacity > 0 &&
286 (uint32_t)initialAppendCapacity > (dest->capacity() - dest->length())) {
287 dest->reserve(dest->length() + initialAppendCapacity);
288 }
289 }
290 /**
291 * Append "bytes[0,n-1]" to this.
292 * @param data the pointer to the bytes
293 * @param n the number of bytes; must be non-negative
294 * @stable ICU 4.2
295 */
296 virtual void Append(const char* data, int32_t n) { dest_->append(data, n); }
297 private:
298 StringClass* dest_;
299
300 StringByteSink() = delete;
301 StringByteSink(const StringByteSink &) = delete;
302 StringByteSink &operator=(const StringByteSink &) = delete;
303};
304
305U_NAMESPACE_END
306
307#endif /* U_SHOW_CPLUSPLUS_API */
308
309#endif // __BYTESTREAM_H__
310