|
OLD | NEW |
---|---|
(Empty) | |
1 // Copyright (c) 2012 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 "content/renderer/hyphenator/hyphenator.h" | |
6 | |
7 #include <string> | |
8 | |
9 #include "base/file_util.h" | |
10 #include "base/logging.h" | |
11 #include "base/memory/scoped_ptr.h" | |
12 #include "base/string_util.h" | |
13 #include "base/utf_string_conversions.h" | |
14 #include "third_party/hyphen/hyphen.h" | |
15 #include "unicode/uscript.h" | |
16 | |
17 namespace { | |
18 | |
19 // A class that converts a sequence of UTF-8 characters to UTF-16 ones and holds | |
20 // only the length of converted UTF-16 characters. This class is used for | |
21 // creating a mapping from the position of a UTF-8 string to a position of a | |
22 // UTF-16 string without unnecessary conversions. Even though the following | |
23 // snippet produces the same mapping, it needs to convert same characters many | |
24 // times. This class incrementally counts the number of converted UTF-16 | |
25 // characters to avoid this problem. | |
26 // | |
27 // scoped_array<size_t> position(new size_t[text.length()]); | |
28 // for (size_t i = 0; i < text.length(); ++i) | |
29 // position[i] = UTF8ToUTF16(text.substr(0, i)).length(); | |
30 // | |
31 class UTF16TextLength { | |
32 public: | |
33 UTF16TextLength(); | |
34 ~UTF16TextLength(); | |
35 | |
36 // Returns the current position. | |
37 int utf16_length() const { return utf16_length_; } | |
38 | |
39 // Appends one UTF-8 character to this converter and advances the converted | |
40 // position. This converter increases the position by one when it finishes | |
41 // reading a BMP character and increases by two when it finish reading a | |
42 // non-BMP character. | |
43 void Append(char c); | |
44 | |
45 private: | |
46 // The length of the converted UTF-16 text. | |
47 int utf16_length_; | |
48 | |
49 // The buffer that stores UTF-8 characters being converted. | |
50 std::string utf8_text_; | |
51 }; | |
jochen (gone - plz use gerrit)
2012/07/02 15:33:01
DISALLOW_COPY_AND_ASSIGN
Hironori Bono
2012/07/10 09:22:09
Done. Thanks for noticing it.
| |
52 | |
53 UTF16TextLength::UTF16TextLength() | |
54 : utf16_length_(0) { | |
55 } | |
56 | |
57 UTF16TextLength::~UTF16TextLength() { | |
58 } | |
59 | |
60 void UTF16TextLength::Append(char c) { | |
61 // Append the given character and try converting the UTF-8 characters in this | |
62 // buffer to Unicode codepoints. If this buffer includes a Unicode codepoint, | |
63 // get the number of UTF-16 characters representing this codepoint and advance | |
64 // the position. | |
65 int code = 0; | |
66 int index = 0; | |
67 utf8_text_.push_back(c); | |
68 U8_NEXT(utf8_text_.data(), index, static_cast<int>(utf8_text_.length()), | |
69 code); | |
70 if (code != U_SENTINEL) { | |
71 utf8_text_.clear(); | |
72 utf16_length_ += U16_LENGTH(code); | |
73 } | |
74 } | |
75 | |
76 // A class that encapsulates a hyphenation query. This class owns resources | |
77 // temporarily needed for hyphenating one word, and deletes them when it is | |
78 // deleted as listed in the following snippet. | |
79 // | |
80 // std::vector<int> hyphens; | |
81 // QUery query(UTF8ToUTF16("hyphenate")); | |
82 // query.Hyphenate(dict, &hyphens); | |
83 // | |
84 class Query { | |
85 public: | |
86 explicit Query(const string16& word); | |
87 ~Query(); | |
88 | |
89 // Hyphenates a word with the specified dictionary. This function hyphenates | |
90 // the word provided to its constructor and returns a list of hyphenation | |
91 // points, positions where we can insert hyphens. The following snippet shows | |
92 // how to insert hyphens with hyphenation points returned by this function. | |
93 // | |
94 // string16 word(UTF8ToUTF16("hyphenate")); | |
95 // std::vector<int> hyphens; | |
96 // Query query(word); | |
97 // query.Hyphenate(dict, &hyphens); | |
98 // for (std::vector<int>::const_reverse_iterator it = hyphens.rbegin(); | |
99 // it != hyphens.rend(); ++it) { | |
100 // word.insert(*it, 1, '-'); | |
101 // } | |
tony
2012/06/28 20:06:44
I don't think the code example is that useful as a
Hironori Bono
2012/07/10 09:22:09
Done. I have removed this redundant example.
| |
102 // | |
103 bool Hyphenate(HyphenDict* dictionary, std::vector<int>* hyphen_offsets); | |
104 | |
105 private: | |
106 // A word to be hyphenated. | |
107 std::string word_utf8_; | |
108 | |
109 // Return variables from the hyphen library. | |
110 scoped_array<char> hyphen_vector_; | |
111 char** rep_; | |
112 int* pos_; | |
113 int* cut_; | |
114 }; | |
jochen (gone - plz use gerrit)
2012/07/02 15:33:01
DISALLOW_COPY_AND_ASSIGN
Hironori Bono
2012/07/10 09:22:09
Done. Thanks for noticing it.
| |
115 | |
116 Query::Query(const string16& word) | |
117 : rep_(NULL), | |
118 pos_(NULL), | |
119 cut_(NULL) { | |
120 // Remove trailing punctuation characters. WebKit does not remove these | |
121 // characters when it hyphenates a word. These characters prevent the hyphen | |
122 // library from applying some rules, i.e. they prevent the library from adding | |
123 // hyphens. | |
124 DCHECK(!word.empty()); | |
125 const char16* data = word.data(); | |
126 int length = static_cast<int>(word.length()); | |
127 while (length > 0) { | |
128 int previous = length; | |
129 int code = 0; | |
130 U16_PREV(data, 0, previous, code); | |
131 UErrorCode error = U_ZERO_ERROR; | |
132 if (uscript_getScript(code, &error) != USCRIPT_COMMON) | |
133 break; | |
134 length = previous; | |
135 } | |
136 UTF16ToUTF8(word.c_str(), length, &word_utf8_); | |
137 // Create a hyphen vector used by hnj_hyphen_hyphenate2(). We allocate a | |
138 // buffer of |word_.length()| + 5 as written in Line 112 of | |
139 // <http://cs.chromium.org/src/third_party/hyphen/hyphen.h>. | |
140 hyphen_vector_.reset(new char[word_utf8_.length() + 5]); | |
141 } | |
142 | |
143 Query::~Query() { | |
144 if (rep_) { | |
145 for (size_t i = 0; i < word_utf8_.length(); ++i) { | |
146 if (rep_[i]) | |
147 free(rep_[i]); | |
tony
2012/06/28 20:06:44
Nit: Should we be using hnj_free here and below?
Hironori Bono
2012/07/10 09:22:09
Thanks for your comment. The hyphen library uses h
| |
148 } | |
149 free(rep_); | |
150 } | |
151 if (pos_) | |
152 free(pos_); | |
153 if (cut_) | |
154 free(cut_); | |
155 } | |
156 | |
157 bool Query::Hyphenate(HyphenDict* dictionary, | |
158 std::vector<int>* hyphen_offsets) { | |
159 DCHECK(dictionary); | |
160 DCHECK(hyphen_offsets); | |
161 | |
162 int error_code = hnj_hyphen_hyphenate2(dictionary, | |
163 word_utf8_.data(), | |
164 static_cast<int>(word_utf8_.length()), | |
165 hyphen_vector_.get(), | |
166 NULL, | |
167 &rep_, | |
168 &pos_, | |
169 &cut_); | |
170 if (error_code) | |
171 return false; | |
172 | |
173 // WebKit needs hyphenation points counted in UTF-16 characters. On the other | |
174 // hand, the hyphen library returns hyphenation points counted in UTF-8 | |
175 // characters. We increamentally convert hyphenation points in UTF-8 | |
176 // characters to hyphenation points in UTF-16 characters and write the | |
177 // converted hyphenation points to the output vector. | |
178 UTF16TextLength text_length; | |
179 hyphen_offsets->clear(); | |
180 for (size_t i = 0; i < word_utf8_.length(); ++i) { | |
181 text_length.Append(word_utf8_[i]); | |
182 if (hyphen_vector_[i] & 1) | |
183 hyphen_offsets->push_back(text_length.utf16_length()); | |
184 } | |
185 return !hyphen_offsets->empty(); | |
186 } | |
187 | |
188 } // namespace | |
189 | |
190 Hyphenator::Hyphenator(base::PlatformFile file) | |
191 : dictionary_(NULL), | |
192 rule_file_(file), | |
193 result_(0) { | |
194 } | |
195 | |
196 Hyphenator::~Hyphenator() { | |
197 if (dictionary_) | |
198 hnj_hyphen_free(dictionary_); | |
199 } | |
200 | |
201 bool Hyphenator::Initialize() { | |
202 if (dictionary_) | |
203 return true; | |
204 | |
205 rule_map_.reset(new file_util::MemoryMappedFile); | |
206 if (!rule_map_->Initialize(rule_file_)) | |
tony
2012/06/28 20:06:44
If this is in the sandboxed renderer process, how
Hironori Bono
2012/07/10 09:22:09
Our spellchecker opens dictionary in a browser pro
| |
207 return false; | |
208 | |
209 dictionary_ = hnj_hyphen_load(rule_map_->data(), rule_map_->length()); | |
210 return !!dictionary_; | |
211 } | |
212 | |
213 size_t Hyphenator::ComputeLastHyphenLocation(const string16& word, | |
214 size_t before_index) { | |
215 if (!dictionary_ || word.empty()) | |
216 return 0; | |
217 | |
218 // Call the hyphen library to get all hyphenation points, i.e. positions where | |
219 // we can insert hyphens. When WebKit finds a line-break, it calls this | |
220 // function twice or more with the same word to find the best hyphenation | |
221 // point. To avoid calling the hyphen library twice or more with the same | |
222 // word, we cache the last query. | |
223 if (word_ != word) { | |
224 word_ = word; | |
225 Query query(word); | |
226 result_ = query.Hyphenate(dictionary_, &hyphen_offsets_); | |
227 } | |
228 if (!result_) | |
229 return 0; | |
230 for (std::vector<int>::reverse_iterator it = hyphen_offsets_.rbegin(); | |
231 it != hyphen_offsets_.rend(); ++it) { | |
232 if (static_cast<size_t>(*it) < before_index) | |
233 return *it; | |
234 } | |
235 return 0; | |
236 } | |
OLD | NEW |