1 | // © 2016 and later: Unicode, Inc. and others. |
---|---|
2 | // License & terms of use: http://www.unicode.org/copyright.html |
3 | /* |
4 | ********************************************************************** |
5 | * Copyright (C) 2005-2014, International Business Machines |
6 | * Corporation and others. All Rights Reserved. |
7 | ********************************************************************** |
8 | */ |
9 | |
10 | #include "unicode/utypes.h" |
11 | |
12 | #if !UCONFIG_NO_CONVERSION |
13 | |
14 | #include "csrutf8.h" |
15 | #include "csmatch.h" |
16 | |
17 | U_NAMESPACE_BEGIN |
18 | |
19 | CharsetRecog_UTF8::~CharsetRecog_UTF8() |
20 | { |
21 | // nothing to do |
22 | } |
23 | |
24 | const char *CharsetRecog_UTF8::getName() const |
25 | { |
26 | return "UTF-8"; |
27 | } |
28 | |
29 | UBool CharsetRecog_UTF8::match(InputText* input, CharsetMatch *results) const { |
30 | bool hasBOM = FALSE; |
31 | int32_t numValid = 0; |
32 | int32_t numInvalid = 0; |
33 | const uint8_t *inputBytes = input->fRawInput; |
34 | int32_t i; |
35 | int32_t trailBytes = 0; |
36 | int32_t confidence; |
37 | |
38 | if (input->fRawLength >= 3 && |
39 | inputBytes[0] == 0xEF && inputBytes[1] == 0xBB && inputBytes[2] == 0xBF) { |
40 | hasBOM = TRUE; |
41 | } |
42 | |
43 | // Scan for multi-byte sequences |
44 | for (i=0; i < input->fRawLength; i += 1) { |
45 | int32_t b = inputBytes[i]; |
46 | |
47 | if ((b & 0x80) == 0) { |
48 | continue; // ASCII |
49 | } |
50 | |
51 | // Hi bit on char found. Figure out how long the sequence should be |
52 | if ((b & 0x0E0) == 0x0C0) { |
53 | trailBytes = 1; |
54 | } else if ((b & 0x0F0) == 0x0E0) { |
55 | trailBytes = 2; |
56 | } else if ((b & 0x0F8) == 0xF0) { |
57 | trailBytes = 3; |
58 | } else { |
59 | numInvalid += 1; |
60 | continue; |
61 | } |
62 | |
63 | // Verify that we've got the right number of trail bytes in the sequence |
64 | for (;;) { |
65 | i += 1; |
66 | |
67 | if (i >= input->fRawLength) { |
68 | break; |
69 | } |
70 | |
71 | b = inputBytes[i]; |
72 | |
73 | if ((b & 0xC0) != 0x080) { |
74 | numInvalid += 1; |
75 | break; |
76 | } |
77 | |
78 | if (--trailBytes == 0) { |
79 | numValid += 1; |
80 | break; |
81 | } |
82 | } |
83 | |
84 | } |
85 | |
86 | // Cook up some sort of confidence score, based on presence of a BOM |
87 | // and the existence of valid and/or invalid multi-byte sequences. |
88 | confidence = 0; |
89 | if (hasBOM && numInvalid == 0) { |
90 | confidence = 100; |
91 | } else if (hasBOM && numValid > numInvalid*10) { |
92 | confidence = 80; |
93 | } else if (numValid > 3 && numInvalid == 0) { |
94 | confidence = 100; |
95 | } else if (numValid > 0 && numInvalid == 0) { |
96 | confidence = 80; |
97 | } else if (numValid == 0 && numInvalid == 0) { |
98 | // Plain ASCII. Confidence must be > 10, it's more likely than UTF-16, which |
99 | // accepts ASCII with confidence = 10. |
100 | confidence = 15; |
101 | } else if (numValid > numInvalid*10) { |
102 | // Probably corruput utf-8 data. Valid sequences aren't likely by chance. |
103 | confidence = 25; |
104 | } |
105 | |
106 | results->set(input, this, confidence); |
107 | return (confidence > 0); |
108 | } |
109 | |
110 | U_NAMESPACE_END |
111 | #endif |
112 |