csrutf8.cpp revision c73f511526464f8e56c242df80552e9b0d94ae3d 
1/*
2 **********************************************************************
3 *   Copyright (C) 2005-2014, 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#include "csmatch.h"
14
15U_NAMESPACE_BEGIN
16
17CharsetRecog_UTF8::~CharsetRecog_UTF8()
18{
19    // nothing to do
20}
21
22const char *CharsetRecog_UTF8::getName() const
23{
24    return "UTF-8";
25}
26
27UBool CharsetRecog_UTF8::match(InputText* input, CharsetMatch *results) const {
28    bool hasBOM = FALSE;
29    int32_t numValid = 0;
30    int32_t numInvalid = 0;
31    const uint8_t *inputBytes = input->fRawInput;
32    int32_t i;
33    int32_t trailBytes = 0;
34    int32_t confidence;
35
36    if (input->fRawLength >= 3 &&
37        inputBytes[0] == 0xEF && inputBytes[1] == 0xBB && inputBytes[2] == 0xBF) {
38            hasBOM = TRUE;
39    }
40
41    // Scan for multi-byte sequences
42    for (i=0; i < input->fRawLength; i += 1) {
43        int32_t b = inputBytes[i];
44
45        if ((b & 0x80) == 0) {
46            continue;   // ASCII
47        }
48
49        // Hi bit on char found.  Figure out how long the sequence should be
50        if ((b & 0x0E0) == 0x0C0) {
51            trailBytes = 1;
52        } else if ((b & 0x0F0) == 0x0E0) {
53            trailBytes = 2;
54        } else if ((b & 0x0F8) == 0xF0) {
55            trailBytes = 3;
56        } else {
57            numInvalid += 1;
58            continue;
59        }
60
61        // Verify that we've got the right number of trail bytes in the sequence
62        for (;;) {
63            i += 1;
64
65            if (i >= input->fRawLength) {
66                break;
67            }
68
69            b = inputBytes[i];
70
71            if ((b & 0xC0) != 0x080) {
72                numInvalid += 1;
73                break;
74            }
75
76            if (--trailBytes == 0) {
77                numValid += 1;
78                break;
79            }
80        }
81
82    }
83
84    // Cook up some sort of confidence score, based on presence of a BOM
85    //    and the existence of valid and/or invalid multi-byte sequences.
86    confidence = 0;
87    if (hasBOM && numInvalid == 0) {
88        confidence = 100;
89    } else if (hasBOM && numValid > numInvalid*10) {
90        confidence = 80;
91    } else if (numValid > 3 && numInvalid == 0) {
92        confidence = 100;
93    } else if (numValid > 0 && numInvalid == 0) {
94        confidence = 80;
95    } else if (numValid == 0 && numInvalid == 0) {
96        // Plain ASCII. Confidence must be > 10, it's more likely than UTF-16, which
97        //              accepts ASCII with confidence = 10.
98        confidence = 15;
99    } else if (numValid > numInvalid*10) {
100        // Probably corruput utf-8 data.  Valid sequences aren't likely by chance.
101        confidence = 25;
102    }
103
104    results->set(input, this, confidence);
105    return (confidence > 0);
106}
107
108U_NAMESPACE_END
109#endif
110