Commit | Line | Data |
---|---|---|
0ffa4754 NR |
1 | package be.nikiroo.fanfix.supported; |
2 | ||
3 | import java.io.BufferedReader; | |
4 | import java.io.ByteArrayInputStream; | |
5 | import java.io.IOException; | |
6 | import java.io.InputStreamReader; | |
7 | import java.net.URL; | |
8 | import java.util.ArrayList; | |
9 | import java.util.List; | |
10 | ||
11 | import be.nikiroo.fanfix.Instance; | |
12 | import be.nikiroo.fanfix.bundles.Config; | |
13 | import be.nikiroo.fanfix.bundles.StringId; | |
14 | import be.nikiroo.fanfix.data.Chapter; | |
15 | import be.nikiroo.fanfix.data.Paragraph; | |
16 | import be.nikiroo.fanfix.data.Paragraph.ParagraphType; | |
17 | import be.nikiroo.utils.Image; | |
18 | import be.nikiroo.utils.Progress; | |
19 | import be.nikiroo.utils.StringUtils; | |
20 | ||
21 | /** | |
22 | * Helper class for {@link BasicSupport}, mostly dedicated to {@link Paragraph} | |
8d59ce07 | 23 | * and text formating for the {@link BasicSupport} class. |
0ffa4754 NR |
24 | * |
25 | * @author niki | |
26 | */ | |
8d59ce07 | 27 | public class BasicSupportPara { |
0ffa4754 NR |
28 | // quote chars |
29 | private static char openQuote = Instance.getTrans().getCharacter( | |
30 | StringId.OPEN_SINGLE_QUOTE); | |
31 | private static char closeQuote = Instance.getTrans().getCharacter( | |
32 | StringId.CLOSE_SINGLE_QUOTE); | |
33 | private static char openDoubleQuote = Instance.getTrans().getCharacter( | |
34 | StringId.OPEN_DOUBLE_QUOTE); | |
35 | private static char closeDoubleQuote = Instance.getTrans().getCharacter( | |
36 | StringId.CLOSE_DOUBLE_QUOTE); | |
37 | ||
8d59ce07 NR |
38 | // used by this class: |
39 | BasicSupportHelper bsHelper; | |
40 | BasicSupportImages bsImages; | |
41 | ||
42 | public BasicSupportPara(BasicSupportHelper bsHelper, BasicSupportImages bsImages) { | |
43 | this.bsHelper = bsHelper; | |
44 | this.bsImages = bsImages; | |
45 | } | |
46 | ||
0ffa4754 NR |
47 | /** |
48 | * Create a {@link Chapter} object from the given information, formatting | |
49 | * the content as it should be. | |
50 | * | |
51 | * @param support | |
52 | * the linked {@link BasicSupport} | |
53 | * @param source | |
54 | * the source of the story | |
55 | * @param number | |
56 | * the chapter number | |
57 | * @param name | |
58 | * the chapter name | |
59 | * @param content | |
60 | * the chapter content | |
61 | * @param pg | |
62 | * the optional progress reporter | |
63 | * @param html | |
64 | * TRUE if the input content is in HTML mode | |
65 | * | |
66 | * @return the {@link Chapter} | |
67 | * | |
68 | * @throws IOException | |
69 | * in case of I/O error | |
70 | */ | |
8d59ce07 | 71 | public Chapter makeChapter(BasicSupport support, URL source, |
0ffa4754 NR |
72 | int number, String name, String content, boolean html, Progress pg) |
73 | throws IOException { | |
74 | // Chapter name: process it correctly, then remove the possible | |
75 | // redundant "Chapter x: " in front of it, or "-" (as in | |
76 | // "Chapter 5: - Fun!" after the ": " was automatically added) | |
8d59ce07 | 77 | String chapterName = processPara(name, false) |
0ffa4754 | 78 | .getContent().trim(); |
13fdb89a | 79 | for (String lang : Instance.getConfig().getList(Config.CONF_CHAPTER)) { |
0ffa4754 | 80 | String chapterWord = Instance.getConfig().getStringX( |
13fdb89a | 81 | Config.CONF_CHAPTER, lang); |
0ffa4754 NR |
82 | if (chapterName.startsWith(chapterWord)) { |
83 | chapterName = chapterName.substring(chapterWord.length()) | |
84 | .trim(); | |
85 | break; | |
86 | } | |
87 | } | |
88 | ||
89 | if (chapterName.startsWith(Integer.toString(number))) { | |
90 | chapterName = chapterName.substring( | |
91 | Integer.toString(number).length()).trim(); | |
92 | } | |
93 | ||
94 | while (chapterName.startsWith(":") || chapterName.startsWith("-")) { | |
95 | chapterName = chapterName.substring(1).trim(); | |
96 | } | |
97 | // | |
98 | ||
99 | Chapter chap = new Chapter(number, chapterName); | |
100 | ||
101 | if (content != null) { | |
102 | List<Paragraph> paras = makeParagraphs(support, source, content, | |
103 | html, pg); | |
104 | long words = 0; | |
105 | for (Paragraph para : paras) { | |
106 | words += para.getWords(); | |
107 | } | |
108 | chap.setParagraphs(paras); | |
109 | chap.setWords(words); | |
110 | } | |
111 | ||
112 | return chap; | |
113 | } | |
114 | ||
115 | /** | |
116 | * Check quotes for bad format (i.e., quotes with normal paragraphs inside) | |
117 | * and requotify them (i.e., separate them into QUOTE paragraphs and other | |
118 | * paragraphs (quotes or not)). | |
119 | * | |
120 | * @param para | |
121 | * the paragraph to requotify (not necessarily a quote) | |
122 | * @param html | |
123 | * TRUE if the input content is in HTML mode | |
124 | * | |
125 | * @return the correctly (or so we hope) quotified paragraphs | |
126 | */ | |
8d59ce07 | 127 | protected List<Paragraph> requotify(Paragraph para, boolean html) { |
0ffa4754 NR |
128 | List<Paragraph> newParas = new ArrayList<Paragraph>(); |
129 | ||
130 | if (para.getType() == ParagraphType.QUOTE | |
131 | && para.getContent().length() > 2) { | |
132 | String line = para.getContent(); | |
133 | boolean singleQ = line.startsWith("" + openQuote); | |
134 | boolean doubleQ = line.startsWith("" + openDoubleQuote); | |
135 | ||
136 | // Do not try when more than one quote at a time | |
137 | // (some stories are not easily readable if we do) | |
138 | if (singleQ | |
139 | && line.indexOf(closeQuote, 1) < line | |
140 | .lastIndexOf(closeQuote)) { | |
141 | newParas.add(para); | |
142 | return newParas; | |
143 | } | |
144 | if (doubleQ | |
145 | && line.indexOf(closeDoubleQuote, 1) < line | |
146 | .lastIndexOf(closeDoubleQuote)) { | |
147 | newParas.add(para); | |
148 | return newParas; | |
149 | } | |
150 | // | |
151 | ||
152 | if (!singleQ && !doubleQ) { | |
153 | line = openDoubleQuote + line + closeDoubleQuote; | |
154 | newParas.add(new Paragraph(ParagraphType.QUOTE, line, para | |
155 | .getWords())); | |
156 | } else { | |
157 | char open = singleQ ? openQuote : openDoubleQuote; | |
158 | char close = singleQ ? closeQuote : closeDoubleQuote; | |
159 | ||
160 | int posDot = -1; | |
161 | boolean inQuote = false; | |
162 | int i = 0; | |
163 | for (char car : line.toCharArray()) { | |
164 | if (car == open) { | |
165 | inQuote = true; | |
166 | } else if (car == close) { | |
167 | inQuote = false; | |
168 | } else if (car == '.' && !inQuote) { | |
169 | posDot = i; | |
170 | break; | |
171 | } | |
172 | i++; | |
173 | } | |
174 | ||
175 | if (posDot >= 0) { | |
176 | String rest = line.substring(posDot + 1).trim(); | |
177 | line = line.substring(0, posDot + 1).trim(); | |
178 | long words = 1; | |
179 | for (char car : line.toCharArray()) { | |
180 | if (car == ' ') { | |
181 | words++; | |
182 | } | |
183 | } | |
184 | newParas.add(new Paragraph(ParagraphType.QUOTE, line, words)); | |
185 | if (!rest.isEmpty()) { | |
186 | newParas.addAll(requotify(processPara(rest, html), html)); | |
187 | } | |
188 | } else { | |
189 | newParas.add(para); | |
190 | } | |
191 | } | |
192 | } else { | |
193 | newParas.add(para); | |
194 | } | |
195 | ||
196 | return newParas; | |
197 | } | |
198 | ||
199 | /** | |
200 | * Process a {@link Paragraph} from a raw line of text. | |
201 | * <p> | |
202 | * Will also fix quotes and HTML encoding if needed. | |
203 | * | |
204 | * @param line | |
205 | * the raw line | |
206 | * @param html | |
207 | * TRUE if the input content is in HTML mode | |
208 | * | |
209 | * @return the processed {@link Paragraph} | |
210 | */ | |
8d59ce07 | 211 | protected Paragraph processPara(String line, boolean html) { |
0ffa4754 NR |
212 | if (html) { |
213 | line = StringUtils.unhtml(line).trim(); | |
214 | } | |
215 | boolean space = true; | |
216 | boolean brk = true; | |
217 | boolean quote = false; | |
218 | boolean tentativeCloseQuote = false; | |
219 | char prev = '\0'; | |
220 | int dashCount = 0; | |
221 | long words = 1; | |
222 | ||
223 | StringBuilder builder = new StringBuilder(); | |
224 | for (char car : line.toCharArray()) { | |
225 | if (car != '-') { | |
226 | if (dashCount > 0) { | |
227 | // dash, ndash and mdash: - – — | |
228 | // currently: always use mdash | |
229 | builder.append(dashCount == 1 ? '-' : '—'); | |
230 | } | |
231 | dashCount = 0; | |
232 | } | |
233 | ||
234 | if (tentativeCloseQuote) { | |
235 | tentativeCloseQuote = false; | |
236 | if (Character.isLetterOrDigit(car)) { | |
237 | builder.append("'"); | |
238 | } else { | |
239 | // handle double-single quotes as double quotes | |
240 | if (prev == car) { | |
241 | builder.append(closeDoubleQuote); | |
242 | continue; | |
243 | } | |
244 | ||
245 | builder.append(closeQuote); | |
246 | } | |
247 | } | |
248 | ||
249 | switch (car) { | |
250 | case ' ': // note: unbreakable space | |
251 | case ' ': | |
252 | case '\t': | |
253 | case '\n': // just in case | |
254 | case '\r': // just in case | |
255 | if (builder.length() > 0 | |
256 | && builder.charAt(builder.length() - 1) != ' ') { | |
257 | words++; | |
258 | } | |
259 | builder.append(' '); | |
260 | break; | |
261 | ||
262 | case '\'': | |
263 | if (space || (brk && quote)) { | |
264 | quote = true; | |
265 | // handle double-single quotes as double quotes | |
266 | if (prev == car) { | |
267 | builder.deleteCharAt(builder.length() - 1); | |
268 | builder.append(openDoubleQuote); | |
269 | } else { | |
270 | builder.append(openQuote); | |
271 | } | |
272 | } else if (prev == ' ' || prev == car) { | |
273 | // handle double-single quotes as double quotes | |
274 | if (prev == car) { | |
275 | builder.deleteCharAt(builder.length() - 1); | |
276 | builder.append(openDoubleQuote); | |
277 | } else { | |
278 | builder.append(openQuote); | |
279 | } | |
280 | } else { | |
281 | // it is a quote ("I'm off") or a 'quote' ("This | |
282 | // 'good' restaurant"...) | |
283 | tentativeCloseQuote = true; | |
284 | } | |
285 | break; | |
286 | ||
287 | case '"': | |
288 | if (space || (brk && quote)) { | |
289 | quote = true; | |
290 | builder.append(openDoubleQuote); | |
291 | } else if (prev == ' ') { | |
292 | builder.append(openDoubleQuote); | |
293 | } else { | |
294 | builder.append(closeDoubleQuote); | |
295 | } | |
296 | break; | |
297 | ||
298 | case '-': | |
299 | if (space) { | |
300 | quote = true; | |
301 | } else { | |
302 | dashCount++; | |
303 | } | |
304 | space = false; | |
305 | break; | |
306 | ||
307 | case '*': | |
308 | case '~': | |
309 | case '/': | |
310 | case '\\': | |
311 | case '<': | |
312 | case '>': | |
313 | case '=': | |
314 | case '+': | |
315 | case '_': | |
316 | case '–': | |
317 | case '—': | |
318 | space = false; | |
319 | builder.append(car); | |
320 | break; | |
321 | ||
322 | case '‘': | |
323 | case '`': | |
324 | case '‹': | |
325 | case '﹁': | |
326 | case '〈': | |
327 | case '「': | |
328 | if (space || (brk && quote)) { | |
329 | quote = true; | |
330 | builder.append(openQuote); | |
331 | } else { | |
332 | // handle double-single quotes as double quotes | |
333 | if (prev == car) { | |
334 | builder.deleteCharAt(builder.length() - 1); | |
335 | builder.append(openDoubleQuote); | |
336 | } else { | |
337 | builder.append(openQuote); | |
338 | } | |
339 | } | |
340 | space = false; | |
341 | brk = false; | |
342 | break; | |
343 | ||
344 | case '’': | |
345 | case '›': | |
346 | case '﹂': | |
347 | case '〉': | |
348 | case '」': | |
349 | space = false; | |
350 | brk = false; | |
351 | // handle double-single quotes as double quotes | |
352 | if (prev == car) { | |
353 | builder.deleteCharAt(builder.length() - 1); | |
354 | builder.append(closeDoubleQuote); | |
355 | } else { | |
356 | builder.append(closeQuote); | |
357 | } | |
358 | break; | |
359 | ||
360 | case '«': | |
361 | case '“': | |
362 | case '﹃': | |
363 | case '《': | |
364 | case '『': | |
365 | if (space || (brk && quote)) { | |
366 | quote = true; | |
367 | builder.append(openDoubleQuote); | |
368 | } else { | |
369 | builder.append(openDoubleQuote); | |
370 | } | |
371 | space = false; | |
372 | brk = false; | |
373 | break; | |
374 | ||
375 | case '»': | |
376 | case '”': | |
377 | case '﹄': | |
378 | case '》': | |
379 | case '』': | |
380 | space = false; | |
381 | brk = false; | |
382 | builder.append(closeDoubleQuote); | |
383 | break; | |
384 | ||
385 | default: | |
386 | space = false; | |
387 | brk = false; | |
388 | builder.append(car); | |
389 | break; | |
390 | } | |
391 | ||
392 | prev = car; | |
393 | } | |
394 | ||
395 | if (tentativeCloseQuote) { | |
396 | tentativeCloseQuote = false; | |
397 | builder.append(closeQuote); | |
398 | } | |
399 | ||
400 | line = builder.toString().trim(); | |
401 | ||
402 | ParagraphType type = ParagraphType.NORMAL; | |
403 | if (space) { | |
404 | type = ParagraphType.BLANK; | |
405 | } else if (brk) { | |
406 | type = ParagraphType.BREAK; | |
407 | } else if (quote) { | |
408 | type = ParagraphType.QUOTE; | |
409 | } | |
410 | ||
411 | return new Paragraph(type, line, words); | |
412 | } | |
413 | ||
414 | /** | |
415 | * Convert the given content into {@link Paragraph}s. | |
416 | * | |
417 | * @param support | |
8d59ce07 NR |
418 | * the linked {@link BasicSupport} (can be NULL), |
419 | * used to download optional image content in [] | |
0ffa4754 NR |
420 | * @param source |
421 | * the source URL of the story | |
422 | * @param content | |
423 | * the textual content | |
424 | * @param html | |
425 | * TRUE if the input content is in HTML mode | |
426 | * @param pg | |
427 | * the optional progress reporter | |
428 | * | |
429 | * @return the {@link Paragraph}s | |
430 | * | |
431 | * @throws IOException | |
432 | * in case of I/O error | |
433 | */ | |
8d59ce07 | 434 | protected List<Paragraph> makeParagraphs(BasicSupport support, |
0ffa4754 NR |
435 | URL source, String content, boolean html, Progress pg) |
436 | throws IOException { | |
437 | if (pg == null) { | |
438 | pg = new Progress(); | |
439 | } | |
440 | ||
441 | if (html) { | |
442 | // Special <HR> processing: | |
443 | content = content.replaceAll("(<hr [^>]*>)|(<hr/>)|(<hr>)", | |
444 | "<br/>* * *<br/>"); | |
445 | } | |
446 | ||
447 | List<Paragraph> paras = new ArrayList<Paragraph>(); | |
448 | ||
449 | if (content != null && !content.trim().isEmpty()) { | |
450 | if (html) { | |
451 | String[] tab = content.split("(<p>|</p>|<br>|<br/>)"); | |
452 | pg.setMinMax(0, tab.length); | |
453 | int i = 1; | |
454 | for (String line : tab) { | |
455 | if (line.startsWith("[") && line.endsWith("]")) { | |
456 | pg.setName("Extracting image " + i); | |
457 | } | |
458 | paras.add(makeParagraph(support, source, line.trim(), html)); | |
459 | pg.setProgress(i++); | |
460 | } | |
0ffa4754 NR |
461 | } else { |
462 | List<String> lines = new ArrayList<String>(); | |
463 | BufferedReader buff = null; | |
464 | try { | |
465 | buff = new BufferedReader( | |
466 | new InputStreamReader(new ByteArrayInputStream( | |
467 | content.getBytes("UTF-8")), "UTF-8")); | |
468 | for (String line = buff.readLine(); line != null; line = buff | |
469 | .readLine()) { | |
470 | lines.add(line.trim()); | |
471 | } | |
472 | } finally { | |
473 | if (buff != null) { | |
474 | buff.close(); | |
475 | } | |
476 | } | |
477 | ||
478 | pg.setMinMax(0, lines.size()); | |
479 | int i = 0; | |
480 | for (String line : lines) { | |
481 | if (line.startsWith("[") && line.endsWith("]")) { | |
482 | pg.setName("Extracting image " + i); | |
483 | } | |
484 | paras.add(makeParagraph(support, source, line, html)); | |
485 | pg.setProgress(i++); | |
486 | } | |
0ffa4754 NR |
487 | } |
488 | ||
68328e17 NR |
489 | pg.done(); |
490 | pg.setName(null); | |
491 | ||
0ffa4754 NR |
492 | // Check quotes for "bad" format |
493 | List<Paragraph> newParas = new ArrayList<Paragraph>(); | |
494 | for (Paragraph para : paras) { | |
8d59ce07 | 495 | newParas.addAll(requotify(para, html)); |
0ffa4754 NR |
496 | } |
497 | paras = newParas; | |
498 | ||
499 | // Remove double blanks/brks | |
500 | fixBlanksBreaks(paras); | |
501 | } | |
502 | ||
503 | return paras; | |
504 | } | |
505 | ||
506 | /** | |
507 | * Convert the given line into a single {@link Paragraph}. | |
508 | * | |
509 | * @param support | |
8d59ce07 NR |
510 | * the linked {@link BasicSupport} (can be NULL), |
511 | * used to download optional image content in [] | |
0ffa4754 NR |
512 | * @param source |
513 | * the source URL of the story | |
514 | * @param line | |
515 | * the textual content of the paragraph | |
516 | * @param html | |
517 | * TRUE if the input content is in HTML mode | |
518 | * | |
519 | * @return the {@link Paragraph} | |
520 | */ | |
8d59ce07 | 521 | protected Paragraph makeParagraph(BasicSupport support, URL source, |
0ffa4754 NR |
522 | String line, boolean html) { |
523 | Image image = null; | |
524 | if (line.startsWith("[") && line.endsWith("]")) { | |
8d59ce07 | 525 | image = bsHelper.getImage(support, source, line |
0ffa4754 NR |
526 | .substring(1, line.length() - 1).trim()); |
527 | } | |
528 | ||
529 | if (image != null) { | |
530 | return new Paragraph(image); | |
531 | } | |
532 | ||
8d59ce07 | 533 | return processPara(line, html); |
0ffa4754 NR |
534 | } |
535 | ||
536 | /** | |
537 | * Fix the {@link ParagraphType#BLANK}s and {@link ParagraphType#BREAK}s of | |
538 | * those {@link Paragraph}s. | |
539 | * <p> | |
540 | * The resulting list will not contain a starting or trailing blank/break | |
541 | * nor 2 blanks or breaks following each other. | |
542 | * | |
543 | * @param paras | |
544 | * the list of {@link Paragraph}s to fix | |
545 | */ | |
8d59ce07 | 546 | protected void fixBlanksBreaks(List<Paragraph> paras) { |
0ffa4754 NR |
547 | boolean space = false; |
548 | boolean brk = true; | |
549 | for (int i = 0; i < paras.size(); i++) { | |
550 | Paragraph para = paras.get(i); | |
551 | boolean thisSpace = para.getType() == ParagraphType.BLANK; | |
552 | boolean thisBrk = para.getType() == ParagraphType.BREAK; | |
553 | ||
554 | if (i > 0 && space && thisBrk) { | |
555 | paras.remove(i - 1); | |
556 | i--; | |
557 | } else if ((space || brk) && (thisSpace || thisBrk)) { | |
558 | paras.remove(i); | |
559 | i--; | |
560 | } | |
561 | ||
562 | space = thisSpace; | |
563 | brk = thisBrk; | |
564 | } | |
565 | ||
566 | // Remove blank/brk at start | |
567 | if (paras.size() > 0 | |
568 | && (paras.get(0).getType() == ParagraphType.BLANK || paras.get( | |
569 | 0).getType() == ParagraphType.BREAK)) { | |
570 | paras.remove(0); | |
571 | } | |
572 | ||
573 | // Remove blank/brk at end | |
574 | int last = paras.size() - 1; | |
575 | if (paras.size() > 0 | |
576 | && (paras.get(last).getType() == ParagraphType.BLANK || paras | |
577 | .get(last).getType() == ParagraphType.BREAK)) { | |
578 | paras.remove(last); | |
579 | } | |
580 | } | |
581 | } |