icuSources/i18n/csrutf8.cpp

   1 /*
   2  **********************************************************************
   3  *   Copyright (C) 2005-2008, International Business Machines
   4  *   Corporation and others.  All Rights Reserved.
   5  **********************************************************************
   6  */
   7
   8 #include "unicode/utypes.h"
   9
  10 #if !UCONFIG_NO_CONVERSION
  11
  12 #include "csrutf8.h"
  13
  14 U_NAMESPACE_BEGIN
  15
  16 CharsetRecog_UTF8::~CharsetRecog_UTF8()
  17 {
  18     // nothing to do
  19 }
  20
  21 const char *CharsetRecog_UTF8::getName() const
  22 {
  23     return "UTF-8";
  24 }
  25
  26 int32_t CharsetRecog_UTF8::match(InputText* det) {
  27     bool hasBOM = FALSE;
  28     int32_t numValid = 0;
  29     int32_t numInvalid = 0;
  30     const uint8_t *input = det->fRawInput;
  31     int32_t i;
  32     int32_t trailBytes = 0;
  33     int32_t confidence;
  34
  35     if (det->fRawLength >= 3 &&
  36         input[0] == 0xEF && input[1] == 0xBB && input[2] == 0xBF) {
  37             hasBOM = TRUE;
  38     }
  39
  40     // Scan for multi-byte sequences
  41     for (i=0; i < det->fRawLength; i += 1) {
  42         int32_t b = input[i];
  43
  44         if ((b & 0x80) == 0) {
  45             continue;   // ASCII
  46         }
  47
  48         // Hi bit on char found.  Figure out how long the sequence should be
  49         if ((b & 0x0E0) == 0x0C0) {
  50             trailBytes = 1;
  51         } else if ((b & 0x0F0) == 0x0E0) {
  52             trailBytes = 2;
  53         } else if ((b & 0x0F8) == 0xF0) {
  54             trailBytes = 3;
  55         } else {
  56             numInvalid += 1;
  57
  58             if (numInvalid > 5) {
  59                 break;
  60             }
  61
  62             trailBytes = 0;
  63         }
  64
  65         // Verify that we've got the right number of trail bytes in the sequence
  66         for (;;) {
  67             i += 1;
  68
  69             if (i >= det->fRawLength) {
  70                 break;
  71             }
  72
  73             b = input[i];
  74
  75             if ((b & 0xC0) != 0x080) {
  76                 numInvalid += 1;
  77                 break;
  78             }
  79
  80             if (--trailBytes == 0) {
  81                 numValid += 1;
  82                 break;
  83             }
  84         }
  85
  86     }
  87
  88     // Cook up some sort of confidence score, based on presense of a BOM
  89     //    and the existence of valid and/or invalid multi-byte sequences.
  90     confidence = 0;
  91     if (hasBOM && numInvalid == 0) {
  92         confidence = 100;
  93     } else if (hasBOM && numValid > numInvalid*10) {
  94         confidence = 80;
  95     } else if (numValid > 3 && numInvalid == 0) {
  96         confidence = 100;
  97     } else if (numValid > 0 && numInvalid == 0) {
  98         confidence = 80;
  99     } else if (numValid == 0 && numInvalid == 0) {
 100         // Plain ASCII.
 101         confidence = 10;
 102     } else if (numValid > numInvalid*10) {
 103         // Probably corruput utf-8 data.  Valid sequences aren't likely by chance.
 104         confidence = 25;
 105     }
 106
 107     return confidence;
 108 }
 109
 110 U_NAMESPACE_END
 111 #endif