Home | History | Annotate | Download | only in unicode
      1 // Copyright (C) 2009, International Business Machines
      2 // Corporation and others. All Rights Reserved.
      3 //
      4 // Copyright 2007 Google Inc. All Rights Reserved.
      5 // Author: sanjay (at) google.com (Sanjay Ghemawat)
      6 //
      7 // Abstract interface that consumes a sequence of bytes (ByteSink).
      8 //
      9 // Used so that we can write a single piece of code that can operate
     10 // on a variety of output string types.
     11 //
     12 // Various implementations of this interface are provided:
     13 //   ByteSink:
     14 //      CheckedArrayByteSink    Write to a flat array, with bounds checking
     15 //      StringByteSink          Write to an STL string
     16 
     17 // This code is a contribution of Google code, and the style used here is
     18 // a compromise between the original Google code and the ICU coding guidelines.
     19 // For example, data types are ICU-ified (size_t,int->int32_t),
     20 // and API comments doxygen-ified, but function names and behavior are
     21 // as in the original, if possible.
     22 // Assertion-style error handling, not available in ICU, was changed to
     23 // parameter "pinning" similar to UnicodeString.
     24 //
     25 // In addition, this is only a partial port of the original Google code,
     26 // limited to what was needed so far. The (nearly) complete original code
     27 // is in the ICU svn repository at icuhtml/trunk/design/strings/contrib
     28 // (see ICU ticket 6765, r25517).
     29 
     30 #ifndef __BYTESTREAM_H__
     31 #define __BYTESTREAM_H__
     32 
     33 /**
     34  * \file
     35  * \brief C++ API: Interface for writing bytes, and implementation classes.
     36  */
     37 
     38 #include "unicode/utypes.h"
     39 #include "unicode/uobject.h"
     40 #include "unicode/std_string.h"
     41 
     42 U_NAMESPACE_BEGIN
     43 
     44 /**
     45  * A ByteSink can be filled with bytes.
     46  * @draft ICU 4.2
     47  */
     48 class U_COMMON_API ByteSink : public UMemory {
     49 public:
     50   /**
     51    * Default constructor.
     52    * @draft ICU 4.2
     53    */
     54   ByteSink() { }
     55   /**
     56    * Virtual destructor.
     57    * @draft ICU 4.2
     58    */
     59   virtual ~ByteSink() { }
     60 
     61   /**
     62    * Append "bytes[0,n-1]" to this.
     63    * @param bytes the pointer to the bytes
     64    * @param n the number of bytes; must be non-negative
     65    * @draft ICU 4.2
     66    */
     67   virtual void Append(const char* bytes, int32_t n) = 0;
     68 
     69   /**
     70    * Returns a writable buffer for appending and writes the buffer's capacity to
     71    * *result_capacity. Guarantees *result_capacity>=min_capacity.
     72    * May return a pointer to the caller-owned scratch buffer which must have
     73    * scratch_capacity>=min_capacity.
     74    * The returned buffer is only valid until the next operation
     75    * on this ByteSink.
     76    *
     77    * After writing at most *result_capacity bytes, call Append() with the
     78    * pointer returned from this function and the number of bytes written.
     79    * Many Append() implementations will avoid copying bytes if this function
     80    * returned an internal buffer.
     81    *
     82    * Partial usage example:
     83    *  int32_t capacity;
     84    *  char* buffer = sink->GetAppendBuffer(..., &capacity);
     85    *  ... Write n bytes into buffer, with n <= capacity.
     86    *  sink->Append(buffer, n);
     87    * In many implementations, that call to Append will avoid copying bytes.
     88    *
     89    * If the ByteSink allocates or reallocates an internal buffer, it should use
     90    * the desired_capacity_hint if appropriate.
     91    * If a caller cannot provide a reasonable guess at the desired capacity,
     92    * it should pass desired_capacity_hint=0.
     93    *
     94    * If a non-scratch buffer is returned, the caller may only pass
     95    * a prefix to it to Append().
     96    * That is, it is not correct to pass an interior pointer to Append().
     97    *
     98    * The default implementation always returns the scratch buffer.
     99    *
    100    * @param min_capacity required minimum capacity of the returned buffer;
    101    *                     must be non-negative
    102    * @param desired_capacity_hint desired capacity of the returned buffer;
    103    *                              must be non-negative
    104    * @param scratch default caller-owned buffer
    105    * @param scratch_capacity capacity of the scratch buffer
    106    * @param result_capacity pointer to an integer which will be set to the
    107    *                        capacity of the returned buffer
    108    * @return a buffer with *result_capacity>=min_capacity
    109    * @draft ICU 4.2
    110    */
    111   virtual char* GetAppendBuffer(int32_t min_capacity,
    112                                 int32_t desired_capacity_hint,
    113                                 char* scratch, int32_t scratch_capacity,
    114                                 int32_t* result_capacity);
    115 
    116   /**
    117    * Flush internal buffers.
    118    * Some byte sinks use internal buffers or provide buffering
    119    * and require calling Flush() at the end of the stream.
    120    * The default implementation of Flush() does nothing.
    121    * @draft ICU 4.2
    122    */
    123   virtual void Flush();
    124 
    125 private:
    126   ByteSink(const ByteSink &); // copy constructor not implemented
    127   ByteSink &operator=(const ByteSink &); // assignment operator not implemented
    128 };
    129 
    130 // -------------------------------------------------------------
    131 // Some standard implementations
    132 
    133 /**
    134  * Implementation of ByteSink that writes to a flat byte array,
    135  * with bounds-checking:
    136  * This sink will not write more than capacity bytes to outbuf.
    137  * If more than capacity bytes are Append()ed, then excess bytes are ignored,
    138  * and Overflowed() will return true.
    139  * Overflow does not cause a runtime error.
    140  * @draft ICU 4.2
    141  */
    142 class U_COMMON_API CheckedArrayByteSink : public ByteSink {
    143 public:
    144   /**
    145    * Constructs a ByteSink that will write to outbuf[0..capacity-1].
    146    * @param outbuf buffer to write to
    147    * @param capacity size of the buffer
    148    * @draft ICU 4.2
    149    */
    150   CheckedArrayByteSink(char* outbuf, int32_t capacity);
    151   /**
    152    * Append "bytes[0,n-1]" to this.
    153    * @param bytes the pointer to the bytes
    154    * @param n the number of bytes; must be non-negative
    155    * @draft ICU 4.2
    156    */
    157   virtual void Append(const char* bytes, int32_t n);
    158   /**
    159    * Returns a writable buffer for appending and writes the buffer's capacity to
    160    * *result_capacity. For details see the base class documentation.
    161    * @param min_capacity required minimum capacity of the returned buffer;
    162    *                     must be non-negative
    163    * @param desired_capacity_hint desired capacity of the returned buffer;
    164    *                              must be non-negative
    165    * @param scratch default caller-owned buffer
    166    * @param scratch_capacity capacity of the scratch buffer
    167    * @param result_capacity pointer to an integer which will be set to the
    168    *                        capacity of the returned buffer
    169    * @return a buffer with *result_capacity>=min_capacity
    170    * @draft ICU 4.2
    171    */
    172   virtual char* GetAppendBuffer(int32_t min_capacity,
    173                                 int32_t desired_capacity_hint,
    174                                 char* scratch, int32_t scratch_capacity,
    175                                 int32_t* result_capacity);
    176   /**
    177    * Returns the number of bytes actually written to the sink.
    178    * @return number of bytes written to the buffer
    179    * @draft ICU 4.2
    180    */
    181   int32_t NumberOfBytesWritten() const { return size_; }
    182   /**
    183    * Returns true if any bytes were discarded, i.e., if there was an
    184    * attempt to write more than 'capacity' bytes.
    185    * @return TRUE if more than 'capacity' bytes were Append()ed
    186    * @draft ICU 4.2
    187    */
    188   UBool Overflowed() const { return overflowed_; }
    189 private:
    190   char* outbuf_;
    191   const int32_t capacity_;
    192   int32_t size_;
    193   bool overflowed_;
    194   CheckedArrayByteSink(); ///< default constructor not implemented
    195   CheckedArrayByteSink(const CheckedArrayByteSink &); ///< copy constructor not implemented
    196   CheckedArrayByteSink &operator=(const CheckedArrayByteSink &); ///< assignment operator not implemented
    197 };
    198 
    199 #if U_HAVE_STD_STRING
    200 
    201 /**
    202  * Implementation of ByteSink that writes to a "string".
    203  * The StringClass is usually instantiated with a std::string.
    204  * @draft ICU 4.2
    205  */
    206 template<typename StringClass>
    207 class StringByteSink : public ByteSink {
    208  public:
    209   /**
    210    * Constructs a ByteSink that will append bytes to the dest string.
    211    * @param dest pointer to string object to append to
    212    * @draft ICU 4.2
    213    */
    214   StringByteSink(StringClass* dest) : dest_(dest) { }
    215   /**
    216    * Append "bytes[0,n-1]" to this.
    217    * @param bytes the pointer to the bytes
    218    * @param n the number of bytes; must be non-negative
    219    * @draft ICU 4.2
    220    */
    221   virtual void Append(const char* data, int32_t n) { dest_->append(data, n); }
    222  private:
    223   StringClass* dest_;
    224   StringByteSink(); ///< default constructor not implemented
    225   StringByteSink(const StringByteSink &); ///< copy constructor not implemented
    226   StringByteSink &operator=(const StringByteSink &); ///< assignment operator not implemented
    227 };
    228 
    229 #endif
    230 
    231 U_NAMESPACE_END
    232 
    233 #endif  // __BYTESTREAM_H__
    234