1 /*
2 * Copyright (C) 2015 The Android Open Source Project
3 *
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at
7 *
8 * http://www.apache.org/licenses/LICENSE-2.0
9 *
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
15 */
16
17 #include "minikin/GraphemeBreak.h"
18
19 #include <vector>
20
21 #include <gtest/gtest.h>
22
23 #include "UnicodeUtils.h"
24
25 namespace minikin {
26
IsBreak(const char * src)27 bool IsBreak(const char* src) {
28 const size_t BUF_SIZE = 256;
29 uint16_t buf[BUF_SIZE];
30 size_t offset;
31 size_t size;
32 ParseUnicode(buf, BUF_SIZE, src, &size, &offset);
33 return GraphemeBreak::isGraphemeBreak(nullptr, buf, 0, size, offset);
34 }
35
IsBreakWithAdvances(const float * advances,const char * src)36 bool IsBreakWithAdvances(const float* advances, const char* src) {
37 const size_t BUF_SIZE = 256;
38 uint16_t buf[BUF_SIZE];
39 size_t offset;
40 size_t size;
41 ParseUnicode(buf, BUF_SIZE, src, &size, &offset);
42 return GraphemeBreak::isGraphemeBreak(advances, buf, 0, size, offset);
43 }
44
TEST(GraphemeBreak,utf16)45 TEST(GraphemeBreak, utf16) {
46 EXPECT_FALSE(IsBreak("U+D83C | U+DC31")); // emoji, U+1F431
47
48 // tests for invalid UTF-16
49 EXPECT_TRUE(IsBreak("U+D800 | U+D800")); // two leading surrogates
50 EXPECT_TRUE(IsBreak("U+DC00 | U+DC00")); // two trailing surrogates
51 EXPECT_TRUE(IsBreak("'a' | U+D800")); // lonely leading surrogate
52 EXPECT_TRUE(IsBreak("U+DC00 | 'a'")); // lonely trailing surrogate
53 EXPECT_TRUE(IsBreak("U+D800 | 'a'")); // leading surrogate followed by non-surrogate
54 EXPECT_TRUE(IsBreak("'a' | U+DC00")); // non-surrogate followed by trailing surrogate
55 }
56
TEST(GraphemeBreak,rules)57 TEST(GraphemeBreak, rules) {
58 // Rule GB1, sot ÷; Rule GB2, ÷ eot
59 EXPECT_TRUE(IsBreak("| 'a'"));
60 EXPECT_TRUE(IsBreak("'a' |"));
61
62 // Rule GB3, CR x LF
63 EXPECT_FALSE(IsBreak("U+000D | U+000A")); // CR x LF
64
65 // Rule GB4, (Control | CR | LF) ÷
66 EXPECT_TRUE(IsBreak("'a' | U+2028")); // Line separator
67 EXPECT_TRUE(IsBreak("'a' | U+000D")); // LF
68 EXPECT_TRUE(IsBreak("'a' | U+000A")); // CR
69
70 // Rule GB5, ÷ (Control | CR | LF)
71 EXPECT_TRUE(IsBreak("U+2028 | 'a'")); // Line separator
72 EXPECT_TRUE(IsBreak("U+000D | 'a'")); // LF
73 EXPECT_TRUE(IsBreak("U+000A | 'a'")); // CR
74
75 // Rule GB6, L x ( L | V | LV | LVT )
76 EXPECT_FALSE(IsBreak("U+1100 | U+1100")); // L x L
77 EXPECT_FALSE(IsBreak("U+1100 | U+1161")); // L x V
78 EXPECT_FALSE(IsBreak("U+1100 | U+AC00")); // L x LV
79 EXPECT_FALSE(IsBreak("U+1100 | U+AC01")); // L x LVT
80
81 // Rule GB7, ( LV | V ) x ( V | T )
82 EXPECT_FALSE(IsBreak("U+AC00 | U+1161")); // LV x V
83 EXPECT_FALSE(IsBreak("U+1161 | U+1161")); // V x V
84 EXPECT_FALSE(IsBreak("U+AC00 | U+11A8")); // LV x T
85 EXPECT_FALSE(IsBreak("U+1161 | U+11A8")); // V x T
86
87 // Rule GB8, ( LVT | T ) x T
88 EXPECT_FALSE(IsBreak("U+AC01 | U+11A8")); // LVT x T
89 EXPECT_FALSE(IsBreak("U+11A8 | U+11A8")); // T x T
90
91 // Other hangul pairs not counted above _are_ breaks (GB10)
92 EXPECT_TRUE(IsBreak("U+AC00 | U+1100")); // LV x L
93 EXPECT_TRUE(IsBreak("U+AC01 | U+1100")); // LVT x L
94 EXPECT_TRUE(IsBreak("U+11A8 | U+1100")); // T x L
95 EXPECT_TRUE(IsBreak("U+11A8 | U+AC00")); // T x LV
96 EXPECT_TRUE(IsBreak("U+11A8 | U+AC01")); // T x LVT
97
98 // Rule GB12 and Rule GB13, Regional_Indicator x Regional_Indicator
99 EXPECT_FALSE(IsBreak("U+1F1FA | U+1F1F8"));
100 EXPECT_TRUE(IsBreak("U+1F1FA U+1F1F8 | U+1F1FA U+1F1F8")); // Regional indicator pair (flag)
101 EXPECT_FALSE(IsBreak("U+1F1FA | U+1F1F8 U+1F1FA U+1F1F8")); // Regional indicator pair (flag)
102 EXPECT_FALSE(IsBreak("U+1F1FA U+1F1F8 U+1F1FA | U+1F1F8")); // Regional indicator pair (flag)
103
104 EXPECT_TRUE(IsBreak("U+1F1FA U+1F1F8 | U+1F1FA")); // Regional indicator pair (flag)
105 EXPECT_FALSE(IsBreak("U+1F1FA | U+1F1F8 U+1F1FA")); // Regional indicator pair (flag)
106 // Same case as the two above, knowing that the first two characters ligate, which is what
107 // would typically happen.
108 const float firstPairLigated[] = {1.0, 0.0, 0.0, 0.0, 1.0, 0.0}; // Two entries per codepoint
109 EXPECT_TRUE(IsBreakWithAdvances(firstPairLigated, "U+1F1FA U+1F1F8 | U+1F1FA"));
110 EXPECT_FALSE(IsBreakWithAdvances(firstPairLigated, "U+1F1FA | U+1F1F8 U+1F1FA"));
111 // Repeat the tests, But now the font doesn't have a ligature for the first two characters,
112 // while it does have a ligature for the last two. This could happen for fonts that do not
113 // support some (potentially encoded later than they were developed) flags.
114 const float secondPairLigated[] = {1.0, 0.0, 1.0, 0.0, 0.0, 0.0};
115 EXPECT_FALSE(IsBreakWithAdvances(secondPairLigated, "U+1F1FA U+1F1F8 | U+1F1FA"));
116 EXPECT_TRUE(IsBreakWithAdvances(secondPairLigated, "U+1F1FA | U+1F1F8 U+1F1FA"));
117
118 EXPECT_TRUE(IsBreak("'a' U+1F1FA U+1F1F8 | U+1F1FA")); // Regional indicator pair (flag)
119 EXPECT_FALSE(IsBreak("'a' U+1F1FA | U+1F1F8 U+1F1FA")); // Regional indicator pair (flag)
120
121 EXPECT_TRUE(
122 IsBreak("'a' U+1F1FA U+1F1F8 | U+1F1FA U+1F1F8")); // Regional indicator pair (flag)
123 EXPECT_FALSE(
124 IsBreak("'a' U+1F1FA | U+1F1F8 U+1F1FA U+1F1F8")); // Regional indicator pair (flag)
125 EXPECT_FALSE(
126 IsBreak("'a' U+1F1FA U+1F1F8 U+1F1FA | U+1F1F8")); // Regional indicator pair (flag)
127
128 // Rule GB9, x (Extend | ZWJ)
129 EXPECT_FALSE(IsBreak("'a' | U+0301")); // combining accent
130 EXPECT_FALSE(IsBreak("'a' | U+200D")); // ZWJ
131 // Rule GB9a, x SpacingMark
132 EXPECT_FALSE(IsBreak("U+0915 | U+093E")); // KA, AA (spacing mark)
133 // Rule GB9b, Prepend x
134 // see tailoring test for prepend, as current ICU doesn't have any characters in the class
135
136 // Rule GB999, Any ÷ Any
137 EXPECT_TRUE(IsBreak("'a' | 'b'"));
138 EXPECT_TRUE(IsBreak("'f' | 'i'")); // probable ligature
139 EXPECT_TRUE(IsBreak("U+0644 | U+0627")); // probable ligature, lam + alef
140 EXPECT_TRUE(IsBreak("U+4E00 | U+4E00")); // CJK ideographs
141 EXPECT_TRUE(IsBreak("'a' | U+1F1FA U+1F1F8")); // Regional indicator pair (flag)
142 EXPECT_TRUE(IsBreak("U+1F1FA U+1F1F8 | 'a'")); // Regional indicator pair (flag)
143
144 // Extended rule for emoji tag sequence.
145 EXPECT_TRUE(IsBreak("'a' | U+1F3F4 'a'"));
146 EXPECT_TRUE(IsBreak("'a' U+1F3F4 | 'a'"));
147
148 // Immediate tag_term after tag_base.
149 EXPECT_TRUE(IsBreak("'a' | U+1F3F4 U+E007F 'a'"));
150 EXPECT_FALSE(IsBreak("U+1F3F4 | U+E007F"));
151 EXPECT_TRUE(IsBreak("'a' U+1F3F4 U+E007F | 'a'"));
152
153 // Flag sequence
154 // U+1F3F4 U+E0067 U+E0062 U+E0073 U+E0063 U+E0074 U+E007F is emoji tag sequence for the flag
155 // of Scotland.
156 // U+1F3F4 is WAVING BLACK FLAG. This can be a tag_base character.
157 // U+E0067 is TAG LATIN SMALL LETTER G. This can be a part of tag_spec.
158 // U+E0062 is TAG LATIN SMALL LETTER B. This can be a part of tag_spec.
159 // U+E0073 is TAG LATIN SMALL LETTER S. This can be a part of tag_spec.
160 // U+E0063 is TAG LATIN SMALL LETTER C. This can be a part of tag_spec.
161 // U+E0074 is TAG LATIN SMALL LETTER T. This can be a part of tag_spec.
162 // U+E007F is CANCEL TAG. This is a tag_term character.
163 EXPECT_TRUE(IsBreak("'a' | U+1F3F4 U+E0067 U+E0062 U+E0073 U+E0063 U+E0074 U+E007F"));
164 EXPECT_FALSE(IsBreak("U+1F3F4 | U+E0067 U+E0062 U+E0073 U+E0063 U+E0074 U+E007F"));
165 EXPECT_FALSE(IsBreak("U+1F3F4 U+E0067 | U+E0062 U+E0073 U+E0063 U+E0074 U+E007F"));
166 EXPECT_FALSE(IsBreak("U+1F3F4 U+E0067 U+E0062 | U+E0073 U+E0063 U+E0074 U+E007F"));
167 EXPECT_FALSE(IsBreak("U+1F3F4 U+E0067 U+E0062 U+E0073 | U+E0063 U+E0074 U+E007F"));
168 EXPECT_FALSE(IsBreak("U+1F3F4 U+E0067 U+E0062 U+E0073 U+E0063 | U+E0074 U+E007F"));
169 EXPECT_FALSE(IsBreak("U+1F3F4 U+E0067 U+E0062 U+E0073 U+E0063 U+E0074 | U+E007F"));
170 EXPECT_TRUE(IsBreak("U+1F3F4 U+E0067 U+E0062 U+E0073 U+E0063 U+E0074 U+E007F | 'a'"));
171 }
172
TEST(GraphemeBreak,tailoring)173 TEST(GraphemeBreak, tailoring) {
174 // control characters that we interpret as "extend"
175 EXPECT_FALSE(IsBreak("'a' | U+00AD")); // soft hyphen
176 EXPECT_FALSE(IsBreak("'a' | U+200B")); // zwsp
177 EXPECT_FALSE(IsBreak("'a' | U+200E")); // lrm
178 EXPECT_FALSE(IsBreak("'a' | U+202A")); // lre
179 EXPECT_FALSE(IsBreak("'a' | U+E0041")); // tag character
180
181 // UTC-approved characters for the Prepend class
182 EXPECT_FALSE(IsBreak("U+06DD | U+0661")); // arabic subtending mark + digit one
183
184 EXPECT_TRUE(IsBreak("U+0E01 | U+0E33")); // Thai sara am
185
186 // virama is not a grapheme break, but "pure killer" is
187 EXPECT_FALSE(IsBreak("U+0915 | U+094D U+0915")); // Devanagari ka+virama+ka
188 EXPECT_FALSE(IsBreak("U+0915 U+094D | U+0915")); // Devanagari ka+virama+ka
189 EXPECT_FALSE(IsBreak("U+0E01 | U+0E3A U+0E01")); // thai phinthu = pure killer
190 EXPECT_TRUE(IsBreak("U+0E01 U+0E3A | U+0E01")); // thai phinthu = pure killer
191
192 // Repetition of above tests, but with a given advances array that implies everything
193 // became just one cluster.
194 const float conjoined[] = {1.0, 0.0, 0.0};
195 EXPECT_FALSE(IsBreakWithAdvances(conjoined,
196 "U+0915 | U+094D U+0915")); // Devanagari ka+virama+ka
197 EXPECT_FALSE(IsBreakWithAdvances(conjoined,
198 "U+0915 U+094D | U+0915")); // Devanagari ka+virama+ka
199 EXPECT_FALSE(IsBreakWithAdvances(conjoined,
200 "U+0E01 | U+0E3A U+0E01")); // thai phinthu = pure killer
201 EXPECT_TRUE(IsBreakWithAdvances(conjoined,
202 "U+0E01 U+0E3A | U+0E01")); // thai phinthu = pure killer
203
204 // Repetition of above tests, but with a given advances array that the virama did not
205 // form a cluster with the following consonant. The difference is that there is now
206 // a grapheme break after the virama in ka+virama+ka.
207 const float separate[] = {1.0, 0.0, 1.0};
208 EXPECT_FALSE(IsBreakWithAdvances(separate,
209 "U+0915 | U+094D U+0915")); // Devanagari ka+virama+ka
210 EXPECT_TRUE(IsBreakWithAdvances(separate,
211 "U+0915 U+094D | U+0915")); // Devanagari ka+virama+ka
212 EXPECT_FALSE(IsBreakWithAdvances(separate,
213 "U+0E01 | U+0E3A U+0E01")); // thai phinthu = pure killer
214 EXPECT_TRUE(IsBreakWithAdvances(separate,
215 "U+0E01 U+0E3A | U+0E01")); // thai phinthu = pure killer
216
217 // suppress grapheme breaks in zwj emoji sequences
218 EXPECT_FALSE(IsBreak("U+1F469 U+200D | U+2764 U+FE0F U+200D U+1F48B U+200D U+1F468"));
219 EXPECT_FALSE(IsBreak("U+1F469 U+200D U+2764 U+FE0F U+200D | U+1F48B U+200D U+1F468"));
220 EXPECT_FALSE(IsBreak("U+1F469 U+200D U+2764 U+FE0F U+200D U+1F48B U+200D | U+1F468"));
221 EXPECT_FALSE(IsBreak("U+1F468 U+200D | U+1F469 U+200D U+1F466"));
222 EXPECT_FALSE(IsBreak("U+1F468 U+200D U+1F469 U+200D | U+1F466"));
223 EXPECT_FALSE(IsBreak("U+1F469 U+200D | U+1F469 U+200D U+1F467 U+200D U+1F466"));
224 EXPECT_FALSE(IsBreak("U+1F469 U+200D U+1F469 U+200D | U+1F467 U+200D U+1F466"));
225 EXPECT_FALSE(IsBreak("U+1F469 U+200D U+1F469 U+200D U+1F467 U+200D | U+1F466"));
226 EXPECT_FALSE(IsBreak("U+1F441 U+200D | U+1F5E8"));
227
228 // Do not break before and after zwj with all kind of emoji characters.
229 EXPECT_FALSE(IsBreak("U+1F431 | U+200D U+1F464"));
230 EXPECT_FALSE(IsBreak("U+1F431 U+200D | U+1F464"));
231
232 // ARABIC LETTER BEH + ZWJ + heart, not a zwj emoji sequence, so we preserve the break
233 EXPECT_TRUE(IsBreak("U+0628 U+200D | U+2764"));
234 }
235
TEST(GraphemeBreak,emojiModifiers)236 TEST(GraphemeBreak, emojiModifiers) {
237 EXPECT_FALSE(IsBreak("U+261D | U+1F3FB")); // white up pointing index + modifier
238 EXPECT_FALSE(IsBreak("U+270C | U+1F3FB")); // victory hand + modifier
239 EXPECT_FALSE(IsBreak("U+1F466 | U+1F3FB")); // boy + modifier
240 EXPECT_FALSE(IsBreak("U+1F466 | U+1F3FC")); // boy + modifier
241 EXPECT_FALSE(IsBreak("U+1F466 | U+1F3FD")); // boy + modifier
242 EXPECT_FALSE(IsBreak("U+1F466 | U+1F3FE")); // boy + modifier
243 EXPECT_FALSE(IsBreak("U+1F466 | U+1F3FF")); // boy + modifier
244 EXPECT_FALSE(IsBreak("U+1F918 | U+1F3FF")); // sign of the horns + modifier
245 EXPECT_FALSE(IsBreak("U+1F933 | U+1F3FF")); // selfie (Unicode 9) + modifier
246 // Reptition of the tests above, with the knowledge that they are ligated.
247 const float ligated1_2[] = {1.0, 0.0, 0.0};
248 const float ligated2_2[] = {1.0, 0.0, 0.0, 0.0};
249 EXPECT_FALSE(IsBreakWithAdvances(ligated1_2, "U+261D | U+1F3FB"));
250 EXPECT_FALSE(IsBreakWithAdvances(ligated1_2, "U+270C | U+1F3FB"));
251 EXPECT_FALSE(IsBreakWithAdvances(ligated2_2, "U+1F466 | U+1F3FB"));
252 EXPECT_FALSE(IsBreakWithAdvances(ligated2_2, "U+1F466 | U+1F3FC"));
253 EXPECT_FALSE(IsBreakWithAdvances(ligated2_2, "U+1F466 | U+1F3FD"));
254 EXPECT_FALSE(IsBreakWithAdvances(ligated2_2, "U+1F466 | U+1F3FE"));
255 EXPECT_FALSE(IsBreakWithAdvances(ligated2_2, "U+1F466 | U+1F3FF"));
256 EXPECT_FALSE(IsBreakWithAdvances(ligated2_2, "U+1F918 | U+1F3FF"));
257 EXPECT_FALSE(IsBreakWithAdvances(ligated2_2, "U+1F933 | U+1F3FF"));
258 // Reptition of the tests above, with the knowledge that they are not ligated.
259 const float unligated1_2[] = {1.0, 1.0, 0.0};
260 const float unligated2_2[] = {1.0, 0.0, 1.0, 0.0};
261 EXPECT_TRUE(IsBreakWithAdvances(unligated1_2, "U+261D | U+1F3FB"));
262 EXPECT_TRUE(IsBreakWithAdvances(unligated1_2, "U+270C | U+1F3FB"));
263 EXPECT_TRUE(IsBreakWithAdvances(unligated2_2, "U+1F466 | U+1F3FB"));
264 EXPECT_TRUE(IsBreakWithAdvances(unligated2_2, "U+1F466 | U+1F3FC"));
265 EXPECT_TRUE(IsBreakWithAdvances(unligated2_2, "U+1F466 | U+1F3FD"));
266 EXPECT_TRUE(IsBreakWithAdvances(unligated2_2, "U+1F466 | U+1F3FE"));
267 EXPECT_TRUE(IsBreakWithAdvances(unligated2_2, "U+1F466 | U+1F3FF"));
268 EXPECT_TRUE(IsBreakWithAdvances(unligated2_2, "U+1F918 | U+1F3FF"));
269 EXPECT_TRUE(IsBreakWithAdvances(unligated2_2, "U+1F933 | U+1F3FF"));
270
271 // adding extend characters between emoji base and modifier doesn't affect grapheme cluster
272 EXPECT_FALSE(IsBreak("U+270C U+FE0E | U+1F3FB")); // victory hand + text style + modifier
273 EXPECT_FALSE(IsBreak("U+270C U+FE0F | U+1F3FB")); // heart + emoji style + modifier
274 // Reptition of the two tests above, with the knowledge that they are ligated.
275 const float ligated1_1_2[] = {1.0, 0.0, 0.0, 0.0};
276 EXPECT_FALSE(IsBreakWithAdvances(ligated1_1_2, "U+270C U+FE0E | U+1F3FB"));
277 EXPECT_FALSE(IsBreakWithAdvances(ligated1_1_2, "U+270C U+FE0F | U+1F3FB"));
278 // Reptition of the first two tests, with the knowledge that they are not ligated.
279 const float unligated1_1_2[] = {1.0, 0.0, 1.0, 0.0};
280 EXPECT_TRUE(IsBreakWithAdvances(unligated1_1_2, "U+270C U+FE0E | U+1F3FB"));
281 EXPECT_TRUE(IsBreakWithAdvances(unligated1_1_2, "U+270C U+FE0F | U+1F3FB"));
282
283 // rat is not an emoji modifer
284 EXPECT_TRUE(IsBreak("U+1F466 | U+1F400")); // boy + rat
285 }
286
TEST(GraphemeBreak,genderBalancedEmoji)287 TEST(GraphemeBreak, genderBalancedEmoji) {
288 // U+1F469 is WOMAN, U+200D is ZWJ, U+1F4BC is BRIEFCASE.
289 EXPECT_FALSE(IsBreak("U+1F469 | U+200D U+1F4BC"));
290 EXPECT_FALSE(IsBreak("U+1F469 U+200D | U+1F4BC"));
291 // The above two cases, when the ligature is not supported in the font. We now expect a break
292 // between them.
293 const float unligated2_1_2[] = {1.0, 0.0, 0.0, 1.0, 0.0};
294 EXPECT_FALSE(IsBreakWithAdvances(unligated2_1_2, "U+1F469 | U+200D U+1F4BC"));
295 EXPECT_TRUE(IsBreakWithAdvances(unligated2_1_2, "U+1F469 U+200D | U+1F4BC"));
296
297 // U+2695 has now emoji property, so should be part of ZWJ sequence.
298 EXPECT_FALSE(IsBreak("U+1F469 | U+200D U+2695"));
299 EXPECT_FALSE(IsBreak("U+1F469 U+200D | U+2695"));
300 // The above two cases, when the ligature is not supported in the font. We now expect a break
301 // between them.
302 const float unligated2_1_1[] = {1.0, 0.0, 0.0, 1.0};
303 EXPECT_FALSE(IsBreakWithAdvances(unligated2_1_1, "U+1F469 | U+200D U+2695"));
304 EXPECT_TRUE(IsBreakWithAdvances(unligated2_1_1, "U+1F469 U+200D | U+2695"));
305 }
306
TEST(GraphemeBreak,offsets)307 TEST(GraphemeBreak, offsets) {
308 uint16_t string[] = {0x0041, 0x06DD, 0x0045, 0x0301, 0x0049, 0x0301};
309 EXPECT_TRUE(GraphemeBreak::isGraphemeBreak(nullptr, string, 2, 3, 2));
310 EXPECT_FALSE(GraphemeBreak::isGraphemeBreak(nullptr, string, 2, 3, 3));
311 EXPECT_TRUE(GraphemeBreak::isGraphemeBreak(nullptr, string, 2, 3, 4));
312 EXPECT_TRUE(GraphemeBreak::isGraphemeBreak(nullptr, string, 2, 3, 5));
313 }
314
TEST(GraphemeBreak,startWithZWJ)315 TEST(GraphemeBreak, startWithZWJ) {
316 // It used to be looking before the ZWJ char even if it is the start of the text.
317 IsBreak("U+200D | U+1F5E8"); // UB sanitizer will catch if minikin looks the char before ZWJ
318 }
319
320 } // namespace minikin
321