|
57363
|
1 |
package build.tools.fixuppandoc;
|
|
|
2 |
|
|
|
3 |
import java.io.BufferedReader;
|
|
|
4 |
import java.io.BufferedWriter;
|
|
|
5 |
import java.io.IOException;
|
|
|
6 |
import java.io.InputStreamReader;
|
|
|
7 |
import java.io.OutputStreamWriter;
|
|
|
8 |
import java.io.PrintStream;
|
|
|
9 |
import java.io.PrintWriter;
|
|
|
10 |
import java.io.Reader;
|
|
|
11 |
import java.io.Writer;
|
|
|
12 |
import java.nio.file.Files;
|
|
|
13 |
import java.nio.file.Path;
|
|
|
14 |
import java.util.ArrayList;
|
|
|
15 |
import java.util.LinkedHashMap;
|
|
|
16 |
import java.util.LinkedHashSet;
|
|
|
17 |
import java.util.List;
|
|
|
18 |
import java.util.Locale;
|
|
|
19 |
import java.util.Map;
|
|
|
20 |
import java.util.Objects;
|
|
|
21 |
import java.util.Set;
|
|
|
22 |
import java.util.regex.Pattern;
|
|
|
23 |
|
|
|
24 |
/**
|
|
|
25 |
* Fixup HTML generated by pandoc.
|
|
|
26 |
*
|
|
|
27 |
* <h2>{@code <html>}</h2>
|
|
|
28 |
*
|
|
|
29 |
* Replace the existing element with {@code <html lang="en">}, removing references to XML.
|
|
|
30 |
*
|
|
|
31 |
* <h2>{@code <main>}</h2>
|
|
|
32 |
*
|
|
|
33 |
* {@code <main>} is inserted if palpable content is found that is not with a
|
|
|
34 |
* section such as {@code header}, {@code footer}, {@code aside}.
|
|
|
35 |
*
|
|
|
36 |
* {@code </main>} is inserted if {@code <main>} was inserted and a section
|
|
|
37 |
* is started that should not be included in the main section.
|
|
|
38 |
*
|
|
|
39 |
* <h2>Tables: row headings</h2>
|
|
|
40 |
*
|
|
|
41 |
* {@code scope="row"} is added to the {@code <td>} elements in the first
|
|
|
42 |
* column whose cell contents are all different and therefore which can be
|
|
|
43 |
* used to identify the row. In case of ambiguity, a column containing
|
|
|
44 |
* a {@code <th>} whose contents begin <em>name</em> is preferred.
|
|
|
45 |
*
|
|
|
46 |
*
|
|
|
47 |
* <h2>{@code <meta name="generator">}</h2>
|
|
|
48 |
*
|
|
|
49 |
* Update the content string, to indicate it has been processed by this program.
|
|
|
50 |
*
|
|
|
51 |
*/
|
|
|
52 |
public class Main {
|
|
|
53 |
/**
|
|
|
54 |
* Runs the program.
|
|
|
55 |
*
|
|
|
56 |
* <pre>
|
|
57365
|
57 |
* java build.tools.fixuppandoc.Main [-o output-file] [input-file]
|
|
57363
|
58 |
* </pre>
|
|
|
59 |
*
|
|
|
60 |
* If no input file is specified, the program will read from standard input.
|
|
|
61 |
* If no output file is specified, the program will write to standard output.
|
|
|
62 |
* Any error messages will be written to the standard error stream.
|
|
|
63 |
*
|
|
|
64 |
* @param args the command-line arguments
|
|
|
65 |
*/
|
|
|
66 |
public static void main(String... args) {
|
|
|
67 |
try {
|
|
|
68 |
new Main().run(args);
|
|
|
69 |
} catch (IOException | IllegalArgumentException e) {
|
|
|
70 |
System.err.println(e);
|
|
|
71 |
System.exit(1);
|
|
|
72 |
} catch (Throwable t) {
|
|
|
73 |
t.printStackTrace(System.err);
|
|
|
74 |
System.exit(1);
|
|
|
75 |
}
|
|
|
76 |
}
|
|
|
77 |
|
|
|
78 |
private void run(String... args) throws IOException {
|
|
|
79 |
Path inFile = null;
|
|
|
80 |
Path outFile = null;
|
|
|
81 |
|
|
|
82 |
for (int i = 0; i < args.length; i++) {
|
|
|
83 |
String arg = args[i];
|
|
|
84 |
if (arg.equals("-o") && i + 1 < args.length) {
|
|
|
85 |
outFile = Path.of(args[++i]);
|
|
|
86 |
} else if (arg.startsWith("-")) {
|
|
|
87 |
throw new IllegalArgumentException(arg);
|
|
|
88 |
} else if (inFile == null) {
|
|
|
89 |
inFile = Path.of(arg);
|
|
|
90 |
} else {
|
|
|
91 |
throw new IllegalArgumentException(arg);
|
|
|
92 |
}
|
|
|
93 |
}
|
|
|
94 |
|
|
|
95 |
new Fixup().run(inFile, outFile);
|
|
|
96 |
}
|
|
|
97 |
|
|
|
98 |
/**
|
|
|
99 |
* A class to read HTML, copying input to output, modifying
|
|
|
100 |
* fragments as needed.
|
|
|
101 |
*/
|
|
|
102 |
class Fixup extends HtmlParser {
|
|
|
103 |
/** The output stream. */
|
|
|
104 |
PrintWriter out;
|
|
|
105 |
|
|
|
106 |
/** A stream for reporting errors. */
|
|
|
107 |
PrintStream err = System.err;
|
|
|
108 |
|
|
|
109 |
/**
|
|
|
110 |
* Flag to indicate when {@code <main>} is permitted around palpable content.
|
|
|
111 |
* Set within {@code <body>}; disabled within elements in which {@code <main>}
|
|
|
112 |
* is not permitted.
|
|
|
113 |
*/
|
|
|
114 |
boolean allowMain = false;
|
|
|
115 |
|
|
|
116 |
/**
|
|
|
117 |
* Flag to indicate that {@code <main>} is required.
|
|
|
118 |
* Set on {@code <body>}; reset when {@code <main>} is either found or generated.
|
|
|
119 |
*/
|
|
|
120 |
boolean needMain = false;
|
|
|
121 |
|
|
|
122 |
/**
|
|
|
123 |
* Flag to indicate that {@code </main>} is required.
|
|
|
124 |
* Set if {@code <main>} is generated.
|
|
|
125 |
* Reset when a start or end element is found that requires that {@code </main>}
|
|
|
126 |
* needs to be generated if necessary.
|
|
|
127 |
*/
|
|
|
128 |
boolean needEndMain = false;
|
|
|
129 |
|
|
|
130 |
/**
|
|
|
131 |
* Handler for {@code <table>} elements.
|
|
|
132 |
*/
|
|
|
133 |
Table table;
|
|
|
134 |
|
|
|
135 |
/**
|
|
|
136 |
* Run the program, copying an input file to an output file.
|
|
|
137 |
* If the input file is {@code null}, input is read from the standard input.
|
|
|
138 |
* If the output file is {@code null}, output is written to the standard output.
|
|
|
139 |
*
|
|
|
140 |
* @param inFile the input file
|
|
|
141 |
* @param outFile the output file
|
|
|
142 |
* @throws IOException if an IO error occurs
|
|
|
143 |
*/
|
|
|
144 |
void run(Path inFile, Path outFile) throws IOException {
|
|
|
145 |
try (Writer out = openWriter(outFile)) {
|
|
|
146 |
this.out = new PrintWriter(out);
|
|
|
147 |
if (inFile != null) {
|
|
|
148 |
read(inFile);
|
|
|
149 |
} else {
|
|
|
150 |
read(new BufferedReader(new InputStreamReader(System.in)));
|
|
|
151 |
}
|
|
|
152 |
}
|
|
|
153 |
}
|
|
|
154 |
|
|
|
155 |
/**
|
|
|
156 |
* Returns a writer for a file, or for the standard output if the file is {@code null}.
|
|
|
157 |
*
|
|
|
158 |
* @param file the file
|
|
|
159 |
* @return the writer
|
|
|
160 |
* @throws IOException if an IO error occurs
|
|
|
161 |
*/
|
|
|
162 |
private Writer openWriter(Path file) throws IOException {
|
|
|
163 |
if (file != null) {
|
|
|
164 |
return Files.newBufferedWriter(file);
|
|
|
165 |
} else {
|
|
|
166 |
return new BufferedWriter(new OutputStreamWriter(System.out) {
|
|
|
167 |
@Override
|
|
|
168 |
public void close() throws IOException {
|
|
|
169 |
flush();
|
|
|
170 |
}
|
|
|
171 |
});
|
|
|
172 |
}
|
|
|
173 |
}
|
|
|
174 |
|
|
|
175 |
@Override
|
|
|
176 |
protected void error(Path file, int lineNumber, String message) {
|
|
|
177 |
err.print(file == null ? "<stdin>" : file);
|
|
|
178 |
if (lineNumber > 0) {
|
|
|
179 |
err.print(":");
|
|
|
180 |
err.print(lineNumber);
|
|
|
181 |
}
|
|
|
182 |
err.print(": ");
|
|
|
183 |
err.println(message);
|
|
|
184 |
}
|
|
|
185 |
|
|
|
186 |
@Override
|
|
|
187 |
protected void error(Path file, int lineNumber, Throwable t) {
|
|
|
188 |
error(file, lineNumber, t.toString());
|
|
|
189 |
t.printStackTrace(err);
|
|
|
190 |
}
|
|
|
191 |
|
|
|
192 |
/**
|
|
|
193 |
* The buffer in which input is stored until an appropriate action can be determined.
|
|
|
194 |
* Using the buffer ensures that the output exactly matches the input, except where
|
|
|
195 |
* it is intentionally modified.
|
|
|
196 |
*/
|
|
|
197 |
private StringBuilder buffer = new StringBuilder();
|
|
|
198 |
|
|
|
199 |
@Override
|
|
|
200 |
public int nextChar() throws IOException {
|
|
|
201 |
if (ch > 0) {
|
|
|
202 |
buffer.append((char) ch);
|
|
|
203 |
}
|
|
|
204 |
return super.nextChar();
|
|
|
205 |
}
|
|
|
206 |
|
|
|
207 |
@Override
|
|
|
208 |
protected void doctype(String s) {
|
|
|
209 |
flushBuffer();
|
|
|
210 |
}
|
|
|
211 |
|
|
|
212 |
@Override
|
|
|
213 |
protected void startElement(String name, Map<String,String> attrs, boolean selfClosing) {
|
|
|
214 |
switch (name) {
|
|
|
215 |
case "html":
|
|
|
216 |
// replace the existing <html> fragment
|
|
|
217 |
out.write("<html lang=\"en\">");
|
|
|
218 |
buffer.setLength(0);
|
|
|
219 |
break;
|
|
|
220 |
|
|
|
221 |
case "meta":
|
|
|
222 |
// update the meta-data for the generator
|
|
|
223 |
if (Objects.equals(attrs.get("name"), "generator")) {
|
|
|
224 |
out.write(buffer.toString()
|
|
|
225 |
.replaceAll("(content=\"[^\"]*)(\")", "$1,fixuphtml$2"));
|
|
|
226 |
buffer.setLength(0);
|
|
|
227 |
}
|
|
|
228 |
break;
|
|
|
229 |
|
|
|
230 |
case "article":
|
|
|
231 |
case "aside":
|
|
|
232 |
case "footer":
|
|
|
233 |
case "header":
|
|
|
234 |
case "nav":
|
|
|
235 |
// starting one of these elements will terminate <main> if one is being
|
|
|
236 |
// inserted
|
|
|
237 |
if (needEndMain) {
|
|
|
238 |
out.write("</main>");
|
|
|
239 |
needEndMain = false;
|
|
|
240 |
}
|
|
|
241 |
// <main> is not permitted within these elements
|
|
|
242 |
allowMain = false;
|
|
|
243 |
break;
|
|
|
244 |
|
|
|
245 |
case "body":
|
|
|
246 |
// within <body>, <main> is both permitted and required
|
|
|
247 |
allowMain = true;
|
|
|
248 |
needMain = true;
|
|
|
249 |
break;
|
|
|
250 |
|
|
|
251 |
case "main":
|
|
|
252 |
// an explicit <main> found in the input; no need to add one
|
|
|
253 |
needMain = false;
|
|
|
254 |
break;
|
|
|
255 |
|
|
|
256 |
case "table":
|
|
|
257 |
// The entire content of a <table> is buffered, until it can be
|
|
|
258 |
// determined in which column of the table contains the cells
|
|
|
259 |
// that can be used to identify the row.
|
|
|
260 |
if (table == null) {
|
|
|
261 |
table = new Table();
|
|
|
262 |
} else {
|
|
|
263 |
// tables containing nested tables are not updated
|
|
|
264 |
table.simple = false;
|
|
|
265 |
}
|
|
|
266 |
table.nestDepth++;
|
|
|
267 |
break;
|
|
|
268 |
|
|
|
269 |
case "thead":
|
|
|
270 |
case "tbody":
|
|
|
271 |
if (table != null) {
|
|
|
272 |
table.endCell();
|
|
|
273 |
}
|
|
|
274 |
break;
|
|
|
275 |
|
|
|
276 |
case "tr":
|
|
|
277 |
if (table != null) {
|
|
|
278 |
table.endCell();
|
|
|
279 |
table.nextCellColumnIndex = 0;
|
|
|
280 |
}
|
|
|
281 |
break;
|
|
|
282 |
|
|
|
283 |
case "td":
|
|
|
284 |
case "th":
|
|
|
285 |
if (table != null) {
|
|
|
286 |
if (attrs.containsKey("rowspan")
|
|
|
287 |
|| attrs.containsKey("colspan")
|
|
|
288 |
|| attrs.containsKey("scope")) {
|
|
|
289 |
// tables containing spanning cells and tables that already
|
|
|
290 |
// contain scope attributes are not updated
|
|
|
291 |
table.simple = false;
|
|
|
292 |
}
|
|
|
293 |
table.startCell(name);
|
|
|
294 |
}
|
|
|
295 |
break;
|
|
|
296 |
}
|
|
|
297 |
|
|
|
298 |
// by default, the content is deemed to be palpable content, and so
|
|
|
299 |
// insert <main> if it is permitted and one is still required,
|
|
|
300 |
// while also ensuring that it does not appear before <body>
|
|
|
301 |
if (allowMain && needMain && !name.equals("body")) {
|
|
|
302 |
out.write("<main>");
|
|
|
303 |
needMain = false;
|
|
|
304 |
needEndMain = true;
|
|
|
305 |
}
|
|
|
306 |
|
|
|
307 |
flushBuffer();
|
|
|
308 |
}
|
|
|
309 |
|
|
|
310 |
@Override
|
|
|
311 |
protected void endElement(String name) {
|
|
|
312 |
switch (name) {
|
|
|
313 |
case "article":
|
|
|
314 |
case "aside":
|
|
|
315 |
case "footer":
|
|
|
316 |
case "header":
|
|
|
317 |
case "nav":
|
|
|
318 |
// The code does not handle nested elements of these kinds, but could.
|
|
|
319 |
// So, assuming they are not nested, ending these elements implies
|
|
|
320 |
// that <main> is once again permitted.
|
|
|
321 |
allowMain = true;
|
|
|
322 |
break;
|
|
|
323 |
|
|
|
324 |
case "body":
|
|
|
325 |
// The document is nearly done; insert <main> if needed
|
|
|
326 |
if (needEndMain) {
|
|
|
327 |
out.write("</main>");
|
|
|
328 |
needEndMain = false;
|
|
|
329 |
}
|
|
|
330 |
break;
|
|
|
331 |
|
|
|
332 |
case "table":
|
|
|
333 |
// if the table is finished, analyze it and write it out
|
|
|
334 |
if (table != null) {
|
|
|
335 |
if (--table.nestDepth == 0) {
|
|
|
336 |
table.add(buffer.toString());
|
|
|
337 |
table.write(out);
|
|
|
338 |
table = null;
|
|
|
339 |
buffer.setLength(0);
|
|
|
340 |
}
|
|
|
341 |
}
|
|
|
342 |
break;
|
|
|
343 |
|
|
|
344 |
case "thead":
|
|
|
345 |
case "tbody":
|
|
|
346 |
case "tr":
|
|
|
347 |
case "td":
|
|
|
348 |
case "th":
|
|
|
349 |
// ending any of these elements implicity or explicitly ends the
|
|
|
350 |
// current cell
|
|
|
351 |
table.endCell();
|
|
|
352 |
break;
|
|
|
353 |
|
|
|
354 |
}
|
|
|
355 |
flushBuffer();
|
|
|
356 |
}
|
|
|
357 |
|
|
|
358 |
@Override
|
|
|
359 |
protected void content(String content) {
|
|
|
360 |
if (table != null) {
|
|
|
361 |
table.content(content);
|
|
|
362 |
} else if (allowMain && needMain && !content.isBlank()) {
|
|
|
363 |
// insert <main> if required and if we have palpable content
|
|
|
364 |
out.write("<main>");
|
|
|
365 |
needMain = false;
|
|
|
366 |
needEndMain = true;
|
|
|
367 |
}
|
|
|
368 |
flushBuffer();
|
|
|
369 |
}
|
|
|
370 |
|
|
|
371 |
@Override
|
|
|
372 |
protected void comment(String comment) {
|
|
|
373 |
flushBuffer();
|
|
|
374 |
}
|
|
|
375 |
|
|
|
376 |
/**
|
|
|
377 |
* Flushes the buffer, either by adding it into a table, if one is
|
|
|
378 |
* in progress, or by writing it out.
|
|
|
379 |
*/
|
|
|
380 |
private void flushBuffer() {
|
|
|
381 |
String s = buffer.toString();
|
|
|
382 |
if (table != null) {
|
|
|
383 |
table.add(s);
|
|
|
384 |
} else {
|
|
|
385 |
out.write(s);
|
|
|
386 |
}
|
|
|
387 |
buffer.setLength(0);
|
|
|
388 |
|
|
|
389 |
}
|
|
|
390 |
}
|
|
|
391 |
|
|
|
392 |
/**
|
|
|
393 |
* Storage for the content of a {@code <table>} element} until we can determine
|
|
|
394 |
* whether we should add {@code scope="row"} to the cells in a given column,
|
|
|
395 |
* and if so, which column.
|
|
|
396 |
*
|
|
|
397 |
* The column with the highest number of unique entries is selected;
|
|
|
398 |
* in case of ambiguity, a column whose heading begins "name" is chosen.
|
|
|
399 |
*
|
|
|
400 |
* Only "simple" tables are supported. Tables with any of the following
|
|
|
401 |
* features are not considered "simple" and will not be modified:
|
|
|
402 |
* <ul>
|
|
|
403 |
* <li>Tables containing nested tables</li>
|
|
|
404 |
* <li>Tables containing cells that use "rowspan" and "colspan" attributes</li>
|
|
|
405 |
* <li>Tables containing cells that already use "scope" attributes</li>
|
|
|
406 |
* </ul>
|
|
|
407 |
*/
|
|
|
408 |
class Table {
|
|
|
409 |
/**
|
|
|
410 |
* A fragment of HTML in this table.
|
|
|
411 |
*/
|
|
|
412 |
class Entry {
|
|
|
413 |
/** The fragment. */
|
|
|
414 |
final String html;
|
|
|
415 |
/** The column for a {@code <td>} fragment, or -1. */
|
|
|
416 |
final int column;
|
|
|
417 |
|
|
|
418 |
Entry(String html, int column) {
|
|
|
419 |
this.html = html;
|
|
|
420 |
this.column = column;
|
|
|
421 |
}
|
|
|
422 |
}
|
|
|
423 |
|
|
|
424 |
/** Whether or not this is a "simple" table. */
|
|
|
425 |
boolean simple = true;
|
|
|
426 |
|
|
|
427 |
/** The nesting depth of the current table, within enclosing tables. */
|
|
|
428 |
int nestDepth;
|
|
|
429 |
|
|
|
430 |
/** A list of the HTML fragments that make up this table. */
|
|
|
431 |
List<Entry> entries;
|
|
|
432 |
|
|
|
433 |
/** The plain text contents of each column, used to determine the primary column. */
|
|
|
434 |
List<Set<String>> columnContents;
|
|
|
435 |
|
|
|
436 |
/** The column index of the next cell to be found. */
|
|
|
437 |
int nextCellColumnIndex;
|
|
|
438 |
|
|
|
439 |
/** A flag to mark the start of a {@code <td>} cell. */
|
|
|
440 |
boolean startTDCell;
|
|
|
441 |
|
|
|
442 |
/** The column index of the current cell, or -1 if not in a cell. */
|
|
|
443 |
int currCellColumnIndex;
|
|
|
444 |
|
|
|
445 |
/** The plain text contents of the current column. */
|
|
|
446 |
Set<String> currColumnContents;
|
|
|
447 |
|
|
|
448 |
/** The plain text content of the current cell. */
|
|
|
449 |
StringBuilder currCellContent;
|
|
|
450 |
|
|
|
451 |
/** The kind ({@code th} or {@code td}) of the current cell. */
|
|
|
452 |
String currCellKind;
|
|
|
453 |
|
|
|
454 |
/**
|
|
|
455 |
* The index of the column, if any, containing a heading beginning "name".
|
|
|
456 |
* This column is given preferential treatment when deciding the primary column.
|
|
|
457 |
*/
|
|
|
458 |
int nameColumn;
|
|
|
459 |
|
|
|
460 |
Table() {
|
|
|
461 |
entries = new ArrayList<>();
|
|
|
462 |
columnContents = new ArrayList<>();
|
|
|
463 |
}
|
|
|
464 |
|
|
|
465 |
void startCell(String name) {
|
|
|
466 |
endCell();
|
|
|
467 |
startTDCell = name.equals("td");
|
|
|
468 |
currCellColumnIndex = nextCellColumnIndex++;
|
|
|
469 |
currColumnContents = getColumn(currCellColumnIndex);
|
|
|
470 |
currCellContent = new StringBuilder();
|
|
|
471 |
currCellKind = name;
|
|
|
472 |
}
|
|
|
473 |
|
|
|
474 |
void endCell() {
|
|
|
475 |
if (currCellContent != null) {
|
|
|
476 |
String c = currCellContent.toString().trim();
|
|
|
477 |
if (Objects.equals(currCellKind, "th")
|
|
|
478 |
&& c.toLowerCase(Locale.US).startsWith("name")) {
|
|
|
479 |
nameColumn = currCellColumnIndex;
|
|
|
480 |
}
|
|
|
481 |
currColumnContents.add(c);
|
|
|
482 |
currCellContent = null;
|
|
|
483 |
currCellColumnIndex = -1;
|
|
|
484 |
currColumnContents = null;
|
|
|
485 |
}
|
|
|
486 |
}
|
|
|
487 |
|
|
|
488 |
void content(String content) {
|
|
|
489 |
if (currCellContent != null) {
|
|
|
490 |
currCellContent.append(content);
|
|
|
491 |
}
|
|
|
492 |
}
|
|
|
493 |
|
|
|
494 |
void add(String html) {
|
|
|
495 |
int index = startTDCell ? currCellColumnIndex : -1;
|
|
|
496 |
entries.add(new Entry(html, index));
|
|
|
497 |
startTDCell = false;
|
|
|
498 |
}
|
|
|
499 |
|
|
|
500 |
void write(PrintWriter out) {
|
|
|
501 |
int max = -1;
|
|
|
502 |
int maxIndex = -1;
|
|
|
503 |
int index = 0;
|
|
|
504 |
for (Set<String> c : columnContents) {
|
|
|
505 |
if (c.size() > max || c.size() == max && index == nameColumn) {
|
|
|
506 |
max = c.size();
|
|
|
507 |
maxIndex = index;
|
|
|
508 |
}
|
|
|
509 |
index++;
|
|
|
510 |
}
|
|
|
511 |
for (Entry e : entries) {
|
|
|
512 |
if (simple && e.column == maxIndex) {
|
|
|
513 |
out.write(e.html.substring(0, e.html.length() - 1));
|
|
|
514 |
out.write(" scope=\"row\">");
|
|
|
515 |
} else {
|
|
|
516 |
out.write(e.html);
|
|
|
517 |
}
|
|
|
518 |
}
|
|
|
519 |
}
|
|
|
520 |
|
|
|
521 |
private Set<String> getColumn(int index) {
|
|
|
522 |
while (columnContents.size() <= index) {
|
|
|
523 |
columnContents.add(new LinkedHashSet<>());
|
|
|
524 |
}
|
|
|
525 |
|
|
|
526 |
return columnContents.get(index);
|
|
|
527 |
}
|
|
|
528 |
}
|
|
|
529 |
|
|
|
530 |
/**
|
|
|
531 |
* A basic HTML parser.
|
|
|
532 |
* Override the protected methods as needed to get notified of significant items
|
|
|
533 |
* in any file that is read.
|
|
|
534 |
*/
|
|
|
535 |
abstract class HtmlParser {
|
|
|
536 |
|
|
|
537 |
private Path file;
|
|
|
538 |
private Reader in;
|
|
|
539 |
protected int ch;
|
|
|
540 |
private int lineNumber;
|
|
|
541 |
private boolean inScript;
|
|
|
542 |
private boolean xml;
|
|
|
543 |
|
|
|
544 |
/**
|
|
|
545 |
* Read a file.
|
|
|
546 |
* @param file the file
|
|
|
547 |
*/
|
|
|
548 |
void read(Path file) {
|
|
|
549 |
try (Reader r = Files.newBufferedReader(file)) {
|
|
|
550 |
this.file = file;
|
|
|
551 |
read(r);
|
|
|
552 |
} catch (IOException e) {
|
|
|
553 |
error(file, -1, e);
|
|
|
554 |
}
|
|
|
555 |
}
|
|
|
556 |
|
|
|
557 |
HtmlParser() { }
|
|
|
558 |
|
|
|
559 |
/**
|
|
|
560 |
* Read a stream.
|
|
|
561 |
* @param r the stream
|
|
|
562 |
*/
|
|
|
563 |
void read(Reader r) {
|
|
|
564 |
try {
|
|
|
565 |
this.in = r;
|
|
|
566 |
StringBuilder content = new StringBuilder();
|
|
|
567 |
|
|
|
568 |
startFile(file);
|
|
|
569 |
try {
|
|
|
570 |
lineNumber = 1;
|
|
|
571 |
xml = false;
|
|
|
572 |
nextChar();
|
|
|
573 |
|
|
|
574 |
while (ch != -1) {
|
|
|
575 |
if (ch == '<') {
|
|
|
576 |
content(content.toString());
|
|
|
577 |
content.setLength(0);
|
|
|
578 |
html();
|
|
|
579 |
} else {
|
|
|
580 |
content.append((char) ch);
|
|
|
581 |
if (ch == '\n') {
|
|
|
582 |
content(content.toString());
|
|
|
583 |
content.setLength(0);
|
|
|
584 |
}
|
|
|
585 |
nextChar();
|
|
|
586 |
}
|
|
|
587 |
}
|
|
|
588 |
} finally {
|
|
|
589 |
endFile();
|
|
|
590 |
}
|
|
|
591 |
} catch (IOException e) {
|
|
|
592 |
error(file, lineNumber, e);
|
|
|
593 |
} catch (Throwable t) {
|
|
|
594 |
error(file, lineNumber, t);
|
|
|
595 |
t.printStackTrace(System.err);
|
|
|
596 |
}
|
|
|
597 |
}
|
|
|
598 |
|
|
|
599 |
protected int getLineNumber() {
|
|
|
600 |
return lineNumber;
|
|
|
601 |
}
|
|
|
602 |
|
|
|
603 |
/**
|
|
|
604 |
* Called when a file has been opened, before parsing begins.
|
|
|
605 |
* This is always the first notification when reading a file.
|
|
|
606 |
* This implementation does nothing.
|
|
|
607 |
*
|
|
|
608 |
* @param file the file
|
|
|
609 |
*/
|
|
|
610 |
protected void startFile(Path file) { }
|
|
|
611 |
|
|
|
612 |
/**
|
|
|
613 |
* Called when the parser has finished reading a file.
|
|
|
614 |
* This is always the last notification when reading a file,
|
|
|
615 |
* unless any errors occur while closing the file.
|
|
|
616 |
* This implementation does nothing.
|
|
|
617 |
*/
|
|
|
618 |
protected void endFile() { }
|
|
|
619 |
|
|
|
620 |
/**
|
|
|
621 |
* Called when a doctype declaration is found, at the beginning of the file.
|
|
|
622 |
* This implementation does nothing.
|
|
|
623 |
* @param s the doctype declaration
|
|
|
624 |
*/
|
|
|
625 |
protected void doctype(String s) { }
|
|
|
626 |
|
|
|
627 |
/**
|
|
|
628 |
* Called when the opening tag of an HTML element is encountered.
|
|
|
629 |
* This implementation does nothing.
|
|
|
630 |
* @param name the name of the tag
|
|
|
631 |
* @param attrs the attribute
|
|
|
632 |
* @param selfClosing whether or not this is a self-closing tag
|
|
|
633 |
*/
|
|
|
634 |
protected void startElement(String name, Map<String,String> attrs, boolean selfClosing) { }
|
|
|
635 |
|
|
|
636 |
/**
|
|
|
637 |
* Called when the closing tag of an HTML tag is encountered.
|
|
|
638 |
* This implementation does nothing.
|
|
|
639 |
* @param name the name of the tag
|
|
|
640 |
*/
|
|
|
641 |
protected void endElement(String name) { }
|
|
|
642 |
|
|
|
643 |
/**
|
|
|
644 |
* Called for sequences of character content.
|
|
|
645 |
* @param content the character content
|
|
|
646 |
*/
|
|
|
647 |
protected void content(String content) { }
|
|
|
648 |
|
|
|
649 |
/**
|
|
|
650 |
* Called for sequences of comment.
|
|
|
651 |
* @param comment the comment
|
|
|
652 |
*/
|
|
|
653 |
protected void comment(String comment) { }
|
|
|
654 |
|
|
|
655 |
/**
|
|
|
656 |
* Called when an error has been encountered.
|
|
|
657 |
* @param file the file being read
|
|
|
658 |
* @param lineNumber the line number of line containing the error
|
|
|
659 |
* @param message a description of the error
|
|
|
660 |
*/
|
|
|
661 |
protected abstract void error(Path file, int lineNumber, String message);
|
|
|
662 |
|
|
|
663 |
/**
|
|
|
664 |
* Called when an exception has been encountered.
|
|
|
665 |
* @param file the file being read
|
|
|
666 |
* @param lineNumber the line number of the line being read when the exception was found
|
|
|
667 |
* @param t the exception
|
|
|
668 |
*/
|
|
|
669 |
protected abstract void error(Path file, int lineNumber, Throwable t);
|
|
|
670 |
|
|
|
671 |
protected int nextChar() throws IOException {
|
|
|
672 |
ch = in.read();
|
|
|
673 |
if (ch == '\n')
|
|
|
674 |
lineNumber++;
|
|
|
675 |
return ch;
|
|
|
676 |
}
|
|
|
677 |
|
|
|
678 |
/**
|
|
|
679 |
* Read the start or end of an HTML tag, or an HTML comment
|
|
|
680 |
* {@literal <identifier attrs> } or {@literal </identifier> }
|
|
|
681 |
* @throws java.io.IOException if there is a problem reading the file
|
|
|
682 |
*/
|
|
|
683 |
protected void html() throws IOException {
|
|
|
684 |
nextChar();
|
|
|
685 |
if (isIdentifierStart((char) ch)) {
|
|
|
686 |
String name = readIdentifier().toLowerCase(Locale.US);
|
|
|
687 |
Map<String,String> attrs = htmlAttrs();
|
|
|
688 |
if (attrs != null) {
|
|
|
689 |
boolean selfClosing = false;
|
|
|
690 |
if (ch == '/') {
|
|
|
691 |
nextChar();
|
|
|
692 |
selfClosing = true;
|
|
|
693 |
}
|
|
|
694 |
if (ch == '>') {
|
|
|
695 |
nextChar();
|
|
|
696 |
startElement(name, attrs, selfClosing);
|
|
|
697 |
if (name.equals("script")) {
|
|
|
698 |
inScript = true;
|
|
|
699 |
}
|
|
|
700 |
return;
|
|
|
701 |
}
|
|
|
702 |
}
|
|
|
703 |
} else if (ch == '/') {
|
|
|
704 |
nextChar();
|
|
|
705 |
if (isIdentifierStart((char) ch)) {
|
|
|
706 |
String name = readIdentifier().toLowerCase(Locale.US);
|
|
|
707 |
skipWhitespace();
|
|
|
708 |
if (ch == '>') {
|
|
|
709 |
nextChar();
|
|
|
710 |
endElement(name);
|
|
|
711 |
if (name.equals("script")) {
|
|
|
712 |
inScript = false;
|
|
|
713 |
}
|
|
|
714 |
return;
|
|
|
715 |
}
|
|
|
716 |
}
|
|
|
717 |
} else if (ch == '!') {
|
|
|
718 |
nextChar();
|
|
|
719 |
if (ch == '-') {
|
|
|
720 |
nextChar();
|
|
|
721 |
if (ch == '-') {
|
|
|
722 |
nextChar();
|
|
|
723 |
StringBuilder comment = new StringBuilder();
|
|
|
724 |
while (ch != -1) {
|
|
|
725 |
int dash = 0;
|
|
|
726 |
while (ch == '-') {
|
|
|
727 |
dash++;
|
|
|
728 |
comment.append(ch);
|
|
|
729 |
nextChar();
|
|
|
730 |
}
|
|
|
731 |
// Strictly speaking, a comment should not contain "--"
|
|
|
732 |
// so dash > 2 is an error, dash == 2 implies ch == '>'
|
|
|
733 |
// See http://www.w3.org/TR/html-markup/syntax.html#syntax-comments
|
|
|
734 |
// for more details.
|
|
|
735 |
if (dash >= 2 && ch == '>') {
|
|
|
736 |
comment.setLength(comment.length() - 2);
|
|
|
737 |
comment(comment.toString());
|
|
|
738 |
nextChar();
|
|
|
739 |
return;
|
|
|
740 |
}
|
|
|
741 |
|
|
|
742 |
comment.append(ch);
|
|
|
743 |
nextChar();
|
|
|
744 |
}
|
|
|
745 |
}
|
|
|
746 |
} else if (ch == '[') {
|
|
|
747 |
nextChar();
|
|
|
748 |
if (ch == 'C') {
|
|
|
749 |
nextChar();
|
|
|
750 |
if (ch == 'D') {
|
|
|
751 |
nextChar();
|
|
|
752 |
if (ch == 'A') {
|
|
|
753 |
nextChar();
|
|
|
754 |
if (ch == 'T') {
|
|
|
755 |
nextChar();
|
|
|
756 |
if (ch == 'A') {
|
|
|
757 |
nextChar();
|
|
|
758 |
if (ch == '[') {
|
|
|
759 |
while (true) {
|
|
|
760 |
nextChar();
|
|
|
761 |
if (ch == ']') {
|
|
|
762 |
nextChar();
|
|
|
763 |
if (ch == ']') {
|
|
|
764 |
nextChar();
|
|
|
765 |
if (ch == '>') {
|
|
|
766 |
nextChar();
|
|
|
767 |
return;
|
|
|
768 |
}
|
|
|
769 |
}
|
|
|
770 |
}
|
|
|
771 |
}
|
|
|
772 |
|
|
|
773 |
}
|
|
|
774 |
}
|
|
|
775 |
}
|
|
|
776 |
}
|
|
|
777 |
}
|
|
|
778 |
}
|
|
|
779 |
} else {
|
|
|
780 |
StringBuilder sb = new StringBuilder();
|
|
|
781 |
while (ch != -1 && ch != '>') {
|
|
|
782 |
sb.append((char) ch);
|
|
|
783 |
nextChar();
|
|
|
784 |
}
|
|
|
785 |
Pattern p = Pattern.compile("(?is)doctype\\s+html\\s?.*");
|
|
|
786 |
String s = sb.toString();
|
|
|
787 |
if (p.matcher(s).matches()) {
|
|
|
788 |
doctype(s);
|
|
|
789 |
return;
|
|
|
790 |
}
|
|
|
791 |
}
|
|
|
792 |
} else if (ch == '?') {
|
|
|
793 |
nextChar();
|
|
|
794 |
if (ch == 'x') {
|
|
|
795 |
nextChar();
|
|
|
796 |
if (ch == 'm') {
|
|
|
797 |
nextChar();
|
|
|
798 |
if (ch == 'l') {
|
|
|
799 |
Map<String,String> attrs = htmlAttrs();
|
|
|
800 |
if (ch == '?') {
|
|
|
801 |
nextChar();
|
|
|
802 |
if (ch == '>') {
|
|
|
803 |
nextChar();
|
|
|
804 |
xml = true;
|
|
|
805 |
return;
|
|
|
806 |
}
|
|
|
807 |
}
|
|
|
808 |
}
|
|
|
809 |
}
|
|
|
810 |
|
|
|
811 |
}
|
|
|
812 |
}
|
|
|
813 |
|
|
|
814 |
if (!inScript) {
|
|
|
815 |
error(file, lineNumber, "bad html");
|
|
|
816 |
}
|
|
|
817 |
}
|
|
|
818 |
|
|
|
819 |
/**
|
|
|
820 |
* Read a series of HTML attributes, terminated by {@literal > }.
|
|
|
821 |
* Each attribute is of the form {@literal identifier[=value] }.
|
|
|
822 |
* "value" may be unquoted, single-quoted, or double-quoted.
|
|
|
823 |
*/
|
|
|
824 |
private Map<String,String> htmlAttrs() throws IOException {
|
|
|
825 |
Map<String, String> map = new LinkedHashMap<>();
|
|
|
826 |
skipWhitespace();
|
|
|
827 |
|
|
|
828 |
while (isIdentifierStart((char) ch)) {
|
|
|
829 |
String name = readAttributeName().toLowerCase(Locale.US);
|
|
|
830 |
skipWhitespace();
|
|
|
831 |
String value = null;
|
|
|
832 |
if (ch == '=') {
|
|
|
833 |
nextChar();
|
|
|
834 |
skipWhitespace();
|
|
|
835 |
if (ch == '\'' || ch == '"') {
|
|
|
836 |
char quote = (char) ch;
|
|
|
837 |
nextChar();
|
|
|
838 |
StringBuilder sb = new StringBuilder();
|
|
|
839 |
while (ch != -1 && ch != quote) {
|
|
|
840 |
sb.append((char) ch);
|
|
|
841 |
nextChar();
|
|
|
842 |
}
|
|
|
843 |
value = sb.toString() // hack to replace common entities
|
|
|
844 |
.replace("<", "<")
|
|
|
845 |
.replace(">", ">")
|
|
|
846 |
.replace("&", "&");
|
|
|
847 |
nextChar();
|
|
|
848 |
} else {
|
|
|
849 |
StringBuilder sb = new StringBuilder();
|
|
|
850 |
while (ch != -1 && !isUnquotedAttrValueTerminator((char) ch)) {
|
|
|
851 |
sb.append((char) ch);
|
|
|
852 |
nextChar();
|
|
|
853 |
}
|
|
|
854 |
value = sb.toString();
|
|
|
855 |
}
|
|
|
856 |
skipWhitespace();
|
|
|
857 |
}
|
|
|
858 |
map.put(name, value);
|
|
|
859 |
}
|
|
|
860 |
|
|
|
861 |
return map;
|
|
|
862 |
}
|
|
|
863 |
|
|
|
864 |
private boolean isIdentifierStart(char ch) {
|
|
|
865 |
return Character.isUnicodeIdentifierStart(ch);
|
|
|
866 |
}
|
|
|
867 |
|
|
|
868 |
private String readIdentifier() throws IOException {
|
|
|
869 |
StringBuilder sb = new StringBuilder();
|
|
|
870 |
sb.append((char) ch);
|
|
|
871 |
nextChar();
|
|
|
872 |
while (ch != -1 && Character.isUnicodeIdentifierPart(ch)) {
|
|
|
873 |
sb.append((char) ch);
|
|
|
874 |
nextChar();
|
|
|
875 |
}
|
|
|
876 |
return sb.toString();
|
|
|
877 |
}
|
|
|
878 |
|
|
|
879 |
private String readAttributeName() throws IOException {
|
|
|
880 |
StringBuilder sb = new StringBuilder();
|
|
|
881 |
sb.append((char) ch);
|
|
|
882 |
nextChar();
|
|
|
883 |
while (ch != -1 && Character.isUnicodeIdentifierPart(ch)
|
|
|
884 |
|| ch == '-'
|
|
|
885 |
|| (xml || sb.toString().startsWith("xml")) && ch == ':') {
|
|
|
886 |
sb.append((char) ch);
|
|
|
887 |
nextChar();
|
|
|
888 |
}
|
|
|
889 |
return sb.toString();
|
|
|
890 |
}
|
|
|
891 |
|
|
|
892 |
private boolean isWhitespace(char ch) {
|
|
|
893 |
return Character.isWhitespace(ch);
|
|
|
894 |
}
|
|
|
895 |
|
|
|
896 |
private void skipWhitespace() throws IOException {
|
|
|
897 |
while (isWhitespace((char) ch)) {
|
|
|
898 |
nextChar();
|
|
|
899 |
}
|
|
|
900 |
}
|
|
|
901 |
|
|
|
902 |
private boolean isUnquotedAttrValueTerminator(char ch) {
|
|
|
903 |
switch (ch) {
|
|
|
904 |
case '\f': case '\n': case '\r': case '\t':
|
|
|
905 |
case ' ':
|
|
|
906 |
case '"': case '\'': case '`':
|
|
|
907 |
case '=': case '<': case '>':
|
|
|
908 |
return true;
|
|
|
909 |
default:
|
|
|
910 |
return false;
|
|
|
911 |
}
|
|
|
912 |
}
|
|
|
913 |
}
|
|
|
914 |
|
|
|
915 |
}
|