1 // Copyright (c) 2011 The Chromium Authors. All rights reserved.
2 // Use of this source code is governed by a BSD-style license that can be
3 // found in the LICENSE file.
4 
5 #include <math.h>
6 #include <stdarg.h>
7 #include <stddef.h>
8 
9 #include <limits>
10 #include <sstream>
11 
12 #include "base/format_macros.h"
13 #include "base/i18n/icu_string_conversions.h"
14 #include "base/logging.h"
15 #include "base/macros.h"
16 #include "base/strings/string_piece.h"
17 #include "base/strings/stringprintf.h"
18 #include "base/strings/utf_string_conversions.h"
19 #include "build/build_config.h"
20 #include "testing/gtest/include/gtest/gtest.h"
21 
22 namespace base {
23 
24 namespace {
25 
26 // Given a null-terminated string of wchar_t with each wchar_t representing
27 // a UTF-16 code unit, returns a string16 made up of wchar_t's in the input.
28 // Each wchar_t should be <= 0xFFFF and a non-BMP character (> U+FFFF)
29 // should be represented as a surrogate pair (two UTF-16 units)
30 // *even* where wchar_t is 32-bit (Linux and Mac).
31 //
32 // This is to help write tests for functions with string16 params until
33 // the C++ 0x UTF-16 literal is well-supported by compilers.
BuildString16(const wchar_t * s)34 string16 BuildString16(const wchar_t* s) {
35 #if defined(WCHAR_T_IS_UTF16)
36   return string16(s);
37 #elif defined(WCHAR_T_IS_UTF32)
38   string16 u16;
39   while (*s != 0) {
40     DCHECK_LE(static_cast<unsigned int>(*s), 0xFFFFu);
41     u16.push_back(*s++);
42   }
43   return u16;
44 #endif
45 }
46 
47 }  // namespace
48 
49 // kConverterCodepageCases is not comprehensive. There are a number of cases
50 // to add if we really want to have a comprehensive coverage of various
51 // codepages and their 'idiosyncrasies'. Currently, the only implementation
52 // for CodepageTo* and *ToCodepage uses ICU, which has a very extensive
53 // set of tests for the charset conversion. So, we can get away with a
54 // relatively small number of cases listed below.
55 //
56 // Note about |u16_wide| in the following struct.
57 // On Windows, the field is always identical to |wide|. On Mac and Linux,
58 // it's identical as long as there's no character outside the
59 // BMP (<= U+FFFF). When there is, it is different from |wide| and
60 // is not a real wide string (UTF-32 string) in that each wchar_t in
61 // the string is a UTF-16 code unit zero-extended to be 32-bit
62 // even when the code unit belongs to a surrogate pair.
63 // For instance, a Unicode string (U+0041 U+010000) is represented as
64 // L"\x0041\xD800\xDC00" instead of L"\x0041\x10000".
65 // To avoid the clutter, |u16_wide| will be set to NULL
66 // if it's identical to |wide| on *all* platforms.
67 
68 static const struct {
69   const char* codepage_name;
70   const char* encoded;
71   OnStringConversionError::Type on_error;
72   bool success;
73   const wchar_t* wide;
74   const wchar_t* u16_wide;
75 } kConvertCodepageCases[] = {
76     // Test a case where the input cannot be decoded, using SKIP, FAIL
77     // and SUBSTITUTE error handling rules. "A7 41" is valid, but "A6" isn't.
78     {"big5", "\xA7\x41\xA6", OnStringConversionError::FAIL, false, L"",
79      nullptr},
80     {"big5", "\xA7\x41\xA6", OnStringConversionError::SKIP, true, L"\x4F60",
81      nullptr},
82     {"big5", "\xA7\x41\xA6", OnStringConversionError::SUBSTITUTE, true,
83      L"\x4F60\xFFFD", nullptr},
84     // Arabic (ISO-8859)
85     {"iso-8859-6",
86      "\xC7\xEE\xE4\xD3\xF1\xEE\xE4\xC7\xE5\xEF"
87      " "
88      "\xD9\xEE\xE4\xEE\xEA\xF2\xE3\xEF\xE5\xF2",
89      OnStringConversionError::FAIL, true,
90      L"\x0627\x064E\x0644\x0633\x0651\x064E\x0644\x0627\x0645\x064F"
91      L" "
92      L"\x0639\x064E\x0644\x064E\x064A\x0652\x0643\x064F\x0645\x0652",
93      nullptr},
94     // Chinese Simplified (GB2312)
95     {"gb2312", "\xC4\xE3\xBA\xC3", OnStringConversionError::FAIL, true,
96      L"\x4F60\x597D", nullptr},
97     // Chinese (GB18030) : 4 byte sequences mapped to BMP characters
98     {"gb18030", "\x81\x30\x84\x36\xA1\xA7", OnStringConversionError::FAIL, true,
99      L"\x00A5\x00A8", nullptr},
100     // Chinese (GB18030) : A 4 byte sequence mapped to plane 2 (U+20000)
101     {"gb18030", "\x95\x32\x82\x36\xD2\xBB", OnStringConversionError::FAIL, true,
102 #if defined(WCHAR_T_IS_UTF16)
103      L"\xD840\xDC00\x4E00",
104 #elif defined(WCHAR_T_IS_UTF32)
105      L"\x20000\x4E00",
106 #endif
107      L"\xD840\xDC00\x4E00"},
108     {"big5", "\xA7\x41\xA6\x6E", OnStringConversionError::FAIL, true,
109      L"\x4F60\x597D", nullptr},
110     // Greek (ISO-8859)
111     {"iso-8859-7",
112      "\xE3\xE5\xE9\xDC"
113      " "
114      "\xF3\xEF\xF5",
115      OnStringConversionError::FAIL, true,
116      L"\x03B3\x03B5\x03B9\x03AC"
117      L" "
118      L"\x03C3\x03BF\x03C5",
119      nullptr},
120     // Hebrew (Windows)
121     {"windows-1255", "\xF9\xD1\xC8\xEC\xE5\xC9\xED",
122      OnStringConversionError::FAIL, true,
123      L"\x05E9\x05C1\x05B8\x05DC\x05D5\x05B9\x05DD", nullptr},
124     // Korean (EUC)
125     {"euc-kr", "\xBE\xC8\xB3\xE7\xC7\xCF\xBC\xBC\xBF\xE4",
126      OnStringConversionError::FAIL, true, L"\xC548\xB155\xD558\xC138\xC694",
127      nullptr},
128     // Japanese (EUC)
129     {"euc-jp", "\xA4\xB3\xA4\xF3\xA4\xCB\xA4\xC1\xA4\xCF\xB0\xEC\x8E\xA6",
130      OnStringConversionError::FAIL, true,
131      L"\x3053\x3093\x306B\x3061\x306F\x4E00\xFF66", nullptr},
132     // Japanese (ISO-2022)
133     {"iso-2022-jp",
134      "\x1B$B"
135      "\x24\x33\x24\x73\x24\x4B\x24\x41\x24\x4F\x30\x6C"
136      "\x1B(B"
137      "ab"
138      "\x1B(J"
139      "\x5C\x7E#$"
140      "\x1B(B",
141      OnStringConversionError::FAIL, true,
142      L"\x3053\x3093\x306B\x3061\x306F\x4E00"
143      L"ab\x00A5\x203E#$",
144      nullptr},
145     // Japanese (Shift-JIS)
146     {"sjis", "\x82\xB1\x82\xF1\x82\xC9\x82\xBF\x82\xCD\x88\xEA\xA6",
147      OnStringConversionError::FAIL, true,
148      L"\x3053\x3093\x306B\x3061\x306F\x4E00\xFF66", nullptr},
149     // Russian (KOI8)
150     {"koi8-r", "\xDA\xC4\xD2\xC1\xD7\xD3\xD4\xD7\xD5\xCA\xD4\xC5",
151      OnStringConversionError::FAIL, true,
152      L"\x0437\x0434\x0440\x0430\x0432\x0441\x0442\x0432"
153      L"\x0443\x0439\x0442\x0435",
154      nullptr},
155     // Thai (windows-874)
156     {"windows-874",
157      "\xCA\xC7\xD1\xCA\xB4\xD5"
158      "\xA4\xC3\xD1\xBA",
159      OnStringConversionError::FAIL, true,
160      L"\x0E2A\x0E27\x0E31\x0E2A\x0E14\x0E35"
161      L"\x0E04\x0E23\x0e31\x0E1A",
162      nullptr},
163 };
164 
TEST(ICUStringConversionsTest,ConvertBetweenCodepageAndUTF16)165 TEST(ICUStringConversionsTest, ConvertBetweenCodepageAndUTF16) {
166   for (size_t i = 0; i < arraysize(kConvertCodepageCases); ++i) {
167     SCOPED_TRACE(base::StringPrintf(
168                      "Test[%" PRIuS "]: <encoded: %s> <codepage: %s>", i,
169                      kConvertCodepageCases[i].encoded,
170                      kConvertCodepageCases[i].codepage_name));
171 
172     string16 utf16;
173     bool success = CodepageToUTF16(kConvertCodepageCases[i].encoded,
174                                    kConvertCodepageCases[i].codepage_name,
175                                    kConvertCodepageCases[i].on_error,
176                                    &utf16);
177     string16 utf16_expected;
178     if (kConvertCodepageCases[i].u16_wide == nullptr)
179       utf16_expected = BuildString16(kConvertCodepageCases[i].wide);
180     else
181       utf16_expected = BuildString16(kConvertCodepageCases[i].u16_wide);
182     EXPECT_EQ(kConvertCodepageCases[i].success, success);
183     EXPECT_EQ(utf16_expected, utf16);
184 
185     // When decoding was successful and nothing was skipped, we also check the
186     // reverse conversion. See also the corresponding comment in
187     // ConvertBetweenCodepageAndWide.
188     if (success &&
189         kConvertCodepageCases[i].on_error == OnStringConversionError::FAIL) {
190       std::string encoded;
191       success = UTF16ToCodepage(utf16, kConvertCodepageCases[i].codepage_name,
192                                 kConvertCodepageCases[i].on_error, &encoded);
193       EXPECT_EQ(kConvertCodepageCases[i].success, success);
194       EXPECT_EQ(kConvertCodepageCases[i].encoded, encoded);
195     }
196   }
197 }
198 
199 static const struct {
200   const char* encoded;
201   const char* codepage_name;
202   bool expected_success;
203   const char* expected_value;
204 } kConvertAndNormalizeCases[] = {
205   {"foo-\xe4.html", "iso-8859-1", true, "foo-\xc3\xa4.html"},
206   {"foo-\xe4.html", "iso-8859-7", true, "foo-\xce\xb4.html"},
207   {"foo-\xe4.html", "foo-bar", false, ""},
208   // HTML Encoding spec treats US-ASCII as synonymous with windows-1252
209   {"foo-\xff.html", "ascii", true, "foo-\xc3\xbf.html"},
210   {"foo.html", "ascii", true, "foo.html"},
211   {"foo-a\xcc\x88.html", "utf-8", true, "foo-\xc3\xa4.html"},
212   {"\x95\x32\x82\x36\xD2\xBB", "gb18030", true, "\xF0\xA0\x80\x80\xE4\xB8\x80"},
213   {"\xA7\x41\xA6\x6E", "big5", true, "\xE4\xBD\xA0\xE5\xA5\xBD"},
214   // Windows-1258 does have a combining character at xD2 (which is U+0309).
215   // The sequence of (U+00E2, U+0309) is also encoded as U+1EA9.
216   {"foo\xE2\xD2", "windows-1258", true, "foo\xE1\xBA\xA9"},
217   {"", "iso-8859-1", true, ""},
218 };
TEST(ICUStringConversionsTest,ConvertToUtf8AndNormalize)219 TEST(ICUStringConversionsTest, ConvertToUtf8AndNormalize) {
220   std::string result;
221   for (size_t i = 0; i < arraysize(kConvertAndNormalizeCases); ++i) {
222     SCOPED_TRACE(base::StringPrintf(
223                      "Test[%" PRIuS "]: <encoded: %s> <codepage: %s>", i,
224                      kConvertAndNormalizeCases[i].encoded,
225                      kConvertAndNormalizeCases[i].codepage_name));
226 
227     bool success = ConvertToUtf8AndNormalize(
228         kConvertAndNormalizeCases[i].encoded,
229         kConvertAndNormalizeCases[i].codepage_name, &result);
230     EXPECT_EQ(kConvertAndNormalizeCases[i].expected_success, success);
231     EXPECT_EQ(kConvertAndNormalizeCases[i].expected_value, result);
232   }
233 }
234 
235 }  // namespace base
236