|
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 "chrome/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 | |
16 namespace { | |
17 | |
18 // A class that converts a sequence of UT-8 characters to UTF-16 ones and counts | |
19 // the length of converted UTF-16 characters. This class is used for creating a | |
20 // mapping from the position of a UTF-8 string to a position of a UTF-16 string | |
21 // converted from the UTF-8 string. Even though the following snippet produces | |
22 // the same mapping, it needs to convert same characters many times. This class | |
23 // incrementally counts the number of converted UTF-16 characters to avoid | |
24 // unnecessary conversions. | |
25 // | |
26 // scoped_array<size_t> position(new size_t[text.length()]); | |
27 // for (size_t i = 0; i < text.length(); ++i) | |
28 // position[i] = UTF8ToUTF16(text.substr(0, i)).length(); | |
29 // | |
30 class UTF16Position { | |
31 public: | |
32 UTF16Position(); | |
33 ~UTF16Position(); | |
34 | |
35 // Returns the current position. | |
36 int position() const { return position_; } | |
37 | |
38 // Appends one UTF-8 character to this converter and updates the converted | |
39 // position. This converter increases the position by one when it finishes | |
40 // reading a BMP character and increases by twi when it finish reading a | |
41 // non-BMP character. | |
42 bool Append(char c); | |
43 | |
44 private: | |
45 // The number of converted UTF-16 characters. | |
46 int position_; | |
47 | |
48 // A buffer that stores UTF-8 sharacters being converted. If this buffer does | |
49 // not have any characters being converted, this value becomes 0xffffffff. | |
50 // (This is an invalid UTF-8 character.) | |
51 uint32 buf_; | |
52 }; | |
53 | |
54 UTF16Position::UTF16Position() | |
55 : position_(0), | |
56 buf_(0xffffffff) { | |
57 } | |
58 | |
59 UTF16Position::~UTF16Position() { | |
60 } | |
61 | |
62 bool UTF16Position::Append(char c) { | |
tony
2012/03/05 18:59:00
Can we use U8_FWD_1 in unicode/utf8.h for this ins
Hironori Bono
2012/06/28 09:32:49
Thanks for your advice. Yes, it is better to use U
| |
63 // Rotate our character buffer and append the input character. | |
64 buf_ = (buf_ << 8) | c; | |
65 | |
66 // Increase the position by one when we finish reading a BMP character | |
67 // (U+0000,...,U+FFFF), UTF-16 needs one character to represent a BMP | |
68 // character. | |
69 if ((buf_ & 0x80) == 0 || | |
70 (buf_ & 0xe0c0) == 0xc080 || | |
71 (buf_ & 0xf0c0c0) == 0xe08080) { | |
72 ++position_; | |
73 buf_ = 0xffffffff; | |
74 return true; | |
75 } | |
76 // UTF-16 needs two characters to represent a unicode chracter between U+10000 | |
77 // and U+10FFFF. | |
78 if ((buf_ & 0xf8c0c0c0) == 0xf0808080) { | |
79 position_ += 2; | |
80 buf_ = 0xffffffff; | |
81 return true; | |
82 } | |
83 // UTF-16 cannot represent a Unicode character above U+10FFFF. We should abort | |
84 // this conversion. | |
85 return false; | |
86 } | |
87 | |
88 // A class that encapsulates a hyphenation query. THis class owns resources | |
tony
2012/03/05 18:59:00
Nit: THis -> This
Hironori Bono
2012/06/28 09:32:49
Done.
| |
89 // temporarily needed for hyphenating one word, and deletes them when it is | |
90 // deleted as listed in the following snippet. | |
91 // | |
92 // void GetHyphenationPoints(HyphenDict* dict, | |
93 // const string16& word, | |
94 // std::vector<int>* hyphens) { | |
95 // scoped_ptr<Query> query(word); | |
96 // query.Hyphenate(dict, hyphens); | |
97 // } | |
98 // | |
99 class Query { | |
100 public: | |
101 explicit Query(const string16& word); | |
102 ~Query(); | |
103 | |
104 // Hyphenates a word with the specified dictionary. This function hyphenates | |
105 // the word provided to its constructor and returns a list of hyphenation | |
106 // points, positions where we can insert hyphens. The following snippet shows | |
107 // how to insert hyphens with hyphenation points returned by this function. | |
108 // | |
109 // std::vector<int> hyphens; | |
110 // GetHyphenationPoints(dict, word, &hyphens); | |
tony
2012/03/05 18:59:00
This snippet doesn't seem to match the class.
Hironori Bono
2012/06/28 09:32:49
Done. Thanks for noticing it. I have updated this
| |
111 // for (std::vector<int>::const_reverse_iterator it = hyphens.rbegin(); | |
112 // it != hyphens.rend(); ++it) { | |
113 // word.insert(*it, 1, '-'); | |
114 // } | |
115 // | |
116 bool Hyphenate(HyphenDict* dictionary, std::vector<int>* hyphens); | |
117 | |
118 private: | |
119 // A word to be hyphenated. | |
120 std::string word_; | |
tony
2012/03/05 18:59:00
Nit: Maybe name this word_utf8_ to make the encodi
Hironori Bono
2012/06/28 09:32:49
Done. Thanks for you suggestion.
| |
121 | |
122 // Return variables from the hyphen library. | |
123 scoped_array<char> hyphens_; | |
124 char** rep_; | |
125 int* pos_; | |
126 int* cut_; | |
127 }; | |
128 | |
129 Query::Query(const string16& word) | |
130 : rep_(NULL), | |
131 pos_(NULL), | |
132 cut_(NULL) { | |
133 // Remove trailing punctuation characters. WebKit does not remove these | |
134 // characters when it hyphenates a word. These characters prevent the hyphen | |
135 // library from applying some rules, i.e. they prevent the library from adding | |
136 // hyphens. | |
137 DCHECK(!word.empty()); | |
138 static const char16 kPunctuationChars[] = { | |
139 '!', '"', '\'', ',', '.', '(', ')', ':', ';', '?', '\0', | |
140 }; | |
141 size_t pos = word.find_last_of(kPunctuationChars); | |
142 size_t length = (pos != string16::npos) ? pos : word.length(); | |
143 UTF16ToUTF8(word.c_str(), length, &word_); | |
144 hyphens_.reset(new char[word_.length() + 5]); | |
tony
2012/03/05 18:59:00
Why 5? Is that the max number of hyphens that will
Hironori Bono
2012/06/28 09:32:49
This is a magic number written in "hyphen.h" <http
| |
145 } | |
146 | |
147 Query::~Query() { | |
148 if (rep_) { | |
149 for (size_t i = 0; i < word_.length(); ++i) { | |
150 if (rep_[i]) | |
151 free(rep_[i]); | |
152 } | |
153 free(rep_); | |
154 } | |
155 if (pos_) | |
156 free(pos_); | |
157 if (cut_) | |
158 free(cut_); | |
159 } | |
160 | |
161 bool Query::Hyphenate(HyphenDict* dictionary, std::vector<int>* hyphens) { | |
162 DCHECK(dictionary); | |
163 DCHECK(hyphens); | |
164 | |
165 int result = hnj_hyphen_hyphenate2(dictionary, | |
tony
2012/03/05 18:59:00
Nit: result -> error_code?
Hironori Bono
2012/06/28 09:32:49
Done.
| |
166 word_.data(), | |
167 static_cast<int>(word_.length()), | |
168 hyphens_.get(), | |
169 NULL, | |
170 &rep_, | |
171 &pos_, | |
172 &cut_); | |
tony
2012/03/05 18:59:00
It looks like we can just pass in NULL since we do
Hironori Bono
2012/06/28 09:32:49
This function assumes these pointers are non-NULL
| |
173 if (result) | |
174 return false; | |
175 | |
176 // WebKit needs hyphenation points counted in UTF-16 characters. On the other | |
177 // hand, the hyphen library returns hyphenation points counted in UTF-8 | |
178 // characters. We increamentally convert hyphenation points in UTF-8 | |
179 // characters to hyphenation points in UTF-16 characters and write the | |
180 // converted hyphenation points to the output vector. | |
tony
2012/03/05 18:59:00
It would be a nice TODO to switch the hyphen libra
| |
181 UTF16Position converter; | |
182 hyphens->clear(); | |
183 for (size_t i = 0; i < word_.length(); ++i) { | |
184 converter.Append(word_[i]); | |
185 if (hyphens_[i] & 1) | |
tony
2012/03/05 18:59:00
hyphens and hyphens_ make this code hard to follow
Hironori Bono
2012/06/28 09:32:49
Done, I have renamed them. (I could not figure out
| |
186 hyphens->push_back(converter.position()); | |
187 } | |
188 return !hyphens->empty(); | |
189 } | |
190 | |
191 } // namespace | |
192 | |
193 Hyphenator::Hyphenator(base::PlatformFile file) | |
194 : dictionary_(NULL), | |
195 rule_file_(file), | |
196 result_(0) { | |
197 } | |
198 | |
199 Hyphenator::~Hyphenator() { | |
200 if (dictionary_) | |
201 hnj_hyphen_free(dictionary_); | |
202 } | |
203 | |
204 bool Hyphenator::Initialize() { | |
205 if (dictionary_) | |
206 return true; | |
207 | |
208 rule_map_.reset(new file_util::MemoryMappedFile); | |
209 if (!rule_map_->Initialize(rule_file_)) | |
210 return false; | |
211 | |
212 dictionary_ = hnj_hyphen_load(rule_map_->data(), rule_map_->length()); | |
213 return !!dictionary_; | |
214 } | |
215 | |
216 size_t Hyphenator::ComputeLastHyphenLocation(const string16& word, | |
217 size_t before_index) { | |
218 if (!dictionary_ || word.empty()) | |
219 return 0; | |
220 | |
221 // Call the hyphen library to get all hyphenation points, i.e. positions where | |
222 // we can insert hyphens. When WebKit finds a line-break, it calls this | |
223 // function twice or more with the same word to find the best hyphenation | |
224 // point. To avoid calling the hyphen library twice or more with the same | |
225 // word, we cache the last query. | |
226 if (word_ != word) { | |
227 word_ = word; | |
228 Query query(word); | |
229 result_ = query.Hyphenate(dictionary_, &hyphens_); | |
230 } | |
231 if (!result_) | |
232 return 0; | |
233 for (std::vector<int>::reverse_iterator it = hyphens_.rbegin(); | |
234 it != hyphens_.rend(); ++it) { | |
235 if (static_cast<size_t>(*it) < before_index) | |
236 return *it; | |
237 } | |
238 return 0; | |
239 } | |
OLD | NEW |