public/html2wiki: src/org/nwoca/ssdt/tools/html2wiki/Html2Wiki.java annotate

annotate src/org/nwoca/ssdt/tools/html2wiki/Html2Wiki.java @ 16:001e43423d5d

added replace transformer to remove garbage character (ESC, unicode 001B) from "screen prints", replaced with a space.

author	ferrall@nwoca.org
date	Tue, 08 Feb 2011 08:12:33 -0500
parents	494ca5643e1a
children	a88e2f8fb117

rev	line source
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	1 package org.nwoca.ssdt.tools.html2wiki;
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	2 /*
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	3 * Html2Wiki.java
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	4 *
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	5 * Created on May 9, 2006, 3:22 PM
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	6 *
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	7 */
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	8
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	9 import java.io.*;
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	10 import java.util.Collection;
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	11 import java.util.ArrayList;
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	12 import java.util.List;
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	13 import org.apache.commons.io.FileUtils;
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	14 import java.util.regex.*;
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	15
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	16 /**
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	17 * Converter to convert HTML documents into MediaWiki test.
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	18 *
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	19 * Heavily customized to handle HTML produced by DEC DOCUMENT
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	20 * SOFTARE doctype. Breaks file into Chapters in the manner done
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	21 * by Document. Needs modification to work with other HTML files.
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	22 *
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	23 * @author SMITH
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	24 */
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	25 public class Html2Wiki {
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	26
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	27 private StringBuffer buffer;
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	28 private Collection<Transformer> transformers;
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	29 private boolean converted = false;
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	30 private static String category;
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	31
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	32 /** Creates a new instance of Html2Wiki. */
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	33 public Html2Wiki(String html) {
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	34 buffer = new StringBuffer(html);
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	35 transformers = new ArrayList<Transformer>();
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	36 transformers.add(new DeleteTransformer("<html>\|</html>\|<body>\|</body>"));
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	37 transformers.add(new DeleteTransformer("<!--.-->(\\n\|\\r)",true));
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	38 transformers.add(new DeleteTransformer("<a .*?>\|</a>"));
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	39 transformers.add(new DeleteTransformer("(?m)^\\*"));
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	40 transformers.add(new DeleteTransformer("(?m)<br>$"));
14 c8442e0eff84 Remove <caption> tags. Generlized {table} around {code} blocks. smith@nwoca.org parents: 13 diff changeset	41 transformers.add(new DeleteTransformer("<caption>.*</caption>")); // remove SDML captions (used for TOC)
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	42 transformers.add(new DeleteTransformer("<font .*?>\|</font>"));
4 22ed6d93442c Start modifying transformers to Confluence wiki syntax smith@nwoca.org parents: 2 diff changeset	43 transformers.add(new CloseTagTransformer("<li>","(\n\|\r)*(<li>\|</ul>\|</ol>\|<ul>\|<ol>)","</li>"));
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	44 transformers.add(new BadTableDataTransformer());
15 494ca5643e1a adjust h2 transformer for headings with arguments and added transformer for removing the empty contents table that is left as part of the conversion of the manuals. ferrall@nwoca.org parents: 14 diff changeset	45 transformers.add(new BadTableRowTransformer());
494ca5643e1a adjust h2 transformer for headings with arguments and added transformer for removing the empty contents table that is left as part of the conversion of the manuals. ferrall@nwoca.org parents: 14 diff changeset	46 transformers.add(new ReflowTransformer());
494ca5643e1a adjust h2 transformer for headings with arguments and added transformer for removing the empty contents table that is left as part of the conversion of the manuals. ferrall@nwoca.org parents: 14 diff changeset	47 transformers.add(new DeleteTransformer("<p>"));
8 e8ea26ab2cd7 [no commit message] ferrall@nwoca.org parents: 7 diff changeset	48 transformers.add(new ReplaceTransformer("\\{","\\{")); // Escape braces
e8ea26ab2cd7 [no commit message] ferrall@nwoca.org parents: 7 diff changeset	49 transformers.add(new ReplaceTransformer("\\}","\\}"));
7 a634b4d554d4 Minor fixups >, random smilies :), etc. Fixed blockquote. Handle escaping brackets outside pre tag. smith@nwoca.org parents: 6 diff changeset	50
a634b4d554d4 Minor fixups >, random smilies :), etc. Fixed blockquote. Handle escaping brackets outside pre tag. smith@nwoca.org parents: 6 diff changeset	51 transformers.add(new ReplaceTransformer("\\[","\\[")); // Escape brackets
a634b4d554d4 Minor fixups >, random smilies :), etc. Fixed blockquote. Handle escaping brackets outside pre tag. smith@nwoca.org parents: 6 diff changeset	52 transformers.add(new ReplaceTransformer("\\]","\\]"));
a634b4d554d4 Minor fixups >, random smilies :), etc. Fixed blockquote. Handle escaping brackets outside pre tag. smith@nwoca.org parents: 6 diff changeset	53 transformers.add(new PreTagTransformer()); // Unescape brackets inside <pre>
a634b4d554d4 Minor fixups >, random smilies :), etc. Fixed blockquote. Handle escaping brackets outside pre tag. smith@nwoca.org parents: 6 diff changeset	54 //
4 22ed6d93442c Start modifying transformers to Confluence wiki syntax smith@nwoca.org parents: 2 diff changeset	55 transformers.add(new ReplaceTransformer("<br>","\\\\"));
8 e8ea26ab2cd7 [no commit message] ferrall@nwoca.org parents: 7 diff changeset	56
e8ea26ab2cd7 [no commit message] ferrall@nwoca.org parents: 7 diff changeset	57 //replace table tag preserving border setting.
10 2fb5084b1564 [no commit message] ferrall@nwoca.org parents: 9 diff changeset	58 transformers.add(new TagTransformer("<table\\sborder=(\\d).*?>", true, "{table:border=", "\|width=75%}"));
8 e8ea26ab2cd7 [no commit message] ferrall@nwoca.org parents: 7 diff changeset	59
4 22ed6d93442c Start modifying transformers to Confluence wiki syntax smith@nwoca.org parents: 2 diff changeset	60 transformers.add(new ReplaceTransformer("<table.*?>\|</table>","{table}"));
22ed6d93442c Start modifying transformers to Confluence wiki syntax smith@nwoca.org parents: 2 diff changeset	61 transformers.add(new ReplaceTransformer("<tr>\|</tr>","{tr}"));
5 d34f4d408ef9 [no commit message] ferrall@nwoca.org parents: 4 diff changeset	62 transformers.add(new ReplaceTransformer("<td.*?>\|</td>","{td}"));
d34f4d408ef9 [no commit message] ferrall@nwoca.org parents: 4 diff changeset	63 transformers.add(new ReplaceTransformer("<th.*?>\|</th>","{th}"));
4 22ed6d93442c Start modifying transformers to Confluence wiki syntax smith@nwoca.org parents: 2 diff changeset	64 transformers.add(new ReplaceTransformer("<ol.*?>\|</ol>","{ol}"));
22ed6d93442c Start modifying transformers to Confluence wiki syntax smith@nwoca.org parents: 2 diff changeset	65 transformers.add(new ReplaceTransformer("<ul.*?>\|</ul>","{ul}"));
22ed6d93442c Start modifying transformers to Confluence wiki syntax smith@nwoca.org parents: 2 diff changeset	66 transformers.add(new ReplaceTransformer("<li>","{li}"));
13 cf58f4b9902b clean up ending {li} on line by itself smith@nwoca.org parents: 12 diff changeset	67 transformers.add(new ReplaceTransformer("\\n\\s*</li>","{li}\n")); // remove leading space from </li>
cf58f4b9902b clean up ending {li} on line by itself smith@nwoca.org parents: 12 diff changeset	68 transformers.add(new ReplaceTransformer("</li>","{li}\n")); // Replace remaining </li>
16 001e43423d5d added replace transformer to remove garbage character (ESC, unicode 001B) from "screen prints", replaced with a space. ferrall@nwoca.org parents: 15 diff changeset	69 transformers.add(new ReplaceTransformer("\\u001B"," ")); // Replace ASCII ESC character
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	70
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	71 transformers.add(new ChapterTransformer(category));
4 22ed6d93442c Start modifying transformers to Confluence wiki syntax smith@nwoca.org parents: 2 diff changeset	72 transformers.add(new TagTransformer("<pre>(.*?)</pre>", true, "{code}","{code}"));
22ed6d93442c Start modifying transformers to Confluence wiki syntax smith@nwoca.org parents: 2 diff changeset	73 transformers.add(new TagTransformer("<center>(.*?)</center>", true, "{center}","{center}"));
22ed6d93442c Start modifying transformers to Confluence wiki syntax smith@nwoca.org parents: 2 diff changeset	74 transformers.add(new TagTransformer("<em>(.?)</em>", "","*"));
12 c1d94c623854 ferrall@nwoca.org parents: 11 diff changeset	75 transformers.add(new TagTransformer("<strong>(.?)</strong>", true, "","*"));
9 ccb40d1cb213 [no commit message] ferrall@nwoca.org parents: 8 diff changeset	76 transformers.add(new TagTransformer("<u>(.*?)</u>" , "+","+"));
4 22ed6d93442c Start modifying transformers to Confluence wiki syntax smith@nwoca.org parents: 2 diff changeset	77 transformers.add(new TagTransformer("(?s)<kbd>(.*?)</kbd>", "{{", "}}"));
22ed6d93442c Start modifying transformers to Confluence wiki syntax smith@nwoca.org parents: 2 diff changeset	78 transformers.add(new TagTransformer("<h1>(.*)</h1>", "h1. ", ""));
15 494ca5643e1a adjust h2 transformer for headings with arguments and added transformer for removing the empty contents table that is left as part of the conversion of the manuals. ferrall@nwoca.org parents: 14 diff changeset	79 transformers.add(new TagTransformer("<h2.>(.)</h2>", "h2. ", ""));
4 22ed6d93442c Start modifying transformers to Confluence wiki syntax smith@nwoca.org parents: 2 diff changeset	80 transformers.add(new TagTransformer("<h3>(accessing the program\|sample run\|sample screens?\|sample reports?)</[h\|H]3>","h3.", ""));
22ed6d93442c Start modifying transformers to Confluence wiki syntax smith@nwoca.org parents: 2 diff changeset	81 transformers.add(new TagTransformer("<h3>(.*)</H3>", "h3. ", ""));
22ed6d93442c Start modifying transformers to Confluence wiki syntax smith@nwoca.org parents: 2 diff changeset	82 transformers.add(new TagTransformer("<h3>(.*)</h3>", "h3. ", ""));
22ed6d93442c Start modifying transformers to Confluence wiki syntax smith@nwoca.org parents: 2 diff changeset	83 transformers.add(new TagTransformer("<h4>(.*)</h4>", "h4. ", ""));
22ed6d93442c Start modifying transformers to Confluence wiki syntax smith@nwoca.org parents: 2 diff changeset	84 transformers.add(new TagTransformer("<h5>(.*)</h5>", "h5. ", ""));
22ed6d93442c Start modifying transformers to Confluence wiki syntax smith@nwoca.org parents: 2 diff changeset	85 transformers.add(new TagTransformer("<h6>(.*)</h6>", "h6. ", ""));
8 e8ea26ab2cd7 [no commit message] ferrall@nwoca.org parents: 7 diff changeset	86
e8ea26ab2cd7 [no commit message] ferrall@nwoca.org parents: 7 diff changeset	87 //Replace Notes with Info tags.
10 2fb5084b1564 [no commit message] ferrall@nwoca.org parents: 9 diff changeset	88 transformers.add(new ReplaceTransformer("\\{center}\\n\\{table:border=\\d.}\\n\\{tr\\}\\n\\s{2}\\{td\\}\\{center\\}\\Note\\*\\{center\\}","{info}"));
8 e8ea26ab2cd7 [no commit message] ferrall@nwoca.org parents: 7 diff changeset	89 transformers.add(new ReplaceTransformer("\\{td\\}\\n\\s{2}\\{tr\\}\\n\\{table\\}\\n\\{center\\}","{info}"));
5 d34f4d408ef9 [no commit message] ferrall@nwoca.org parents: 4 diff changeset	90
8 e8ea26ab2cd7 [no commit message] ferrall@nwoca.org parents: 7 diff changeset	91 //Remove unnecessary table surrounding code blocks.
14 c8442e0eff84 Remove <caption> tags. Generlized {table} around {code} blocks. smith@nwoca.org parents: 13 diff changeset	92 transformers.add(new ReplaceTransformer("\\{table:.\\}(\\n\|\\s\|\\{t.\\}\|\\\\S\\)*\\{code\\}","{code}"));
c8442e0eff84 Remove <caption> tags. Generlized {table} around {code} blocks. smith@nwoca.org parents: 13 diff changeset	93 transformers.add(new ReplaceTransformer("\\{code\\}(\\n\|\\{t.\\}\|\\s)*\\{table\\}","{code}"));
8 e8ea26ab2cd7 [no commit message] ferrall@nwoca.org parents: 7 diff changeset	94
e8ea26ab2cd7 [no commit message] ferrall@nwoca.org parents: 7 diff changeset	95 //Change borderStyle of code window for "screenshots" to none.
e8ea26ab2cd7 [no commit message] ferrall@nwoca.org parents: 7 diff changeset	96 transformers.add(new TagTransformer("\\{code\\}([\\s\\n]*?_______________)", true, "{code:borderStyle=none}", ""));
e8ea26ab2cd7 [no commit message] ferrall@nwoca.org parents: 7 diff changeset	97
e8ea26ab2cd7 [no commit message] ferrall@nwoca.org parents: 7 diff changeset	98
e8ea26ab2cd7 [no commit message] ferrall@nwoca.org parents: 7 diff changeset	99
7 a634b4d554d4 Minor fixups >, random smilies :), etc. Fixed blockquote. Handle escaping brackets outside pre tag. smith@nwoca.org parents: 6 diff changeset	100 transformers.add(new TagTransformer("<blockquote>(.*?)</blockquote>", true, "{quote}", "{quote}"));
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	101 transformers.add(new DeleteTransformer("(?s)<hr.*?>"));
8 e8ea26ab2cd7 [no commit message] ferrall@nwoca.org parents: 7 diff changeset	102 transformers.add(new ReflowTransformer("(\\{info\\})([^\\{]*)(\\{info\\})"));
12 c1d94c623854 ferrall@nwoca.org parents: 11 diff changeset	103 transformers.add(new ReflowTransformer("(\\{note\\})([^\\{]*)(\\{note\\})"));
c1d94c623854 ferrall@nwoca.org parents: 11 diff changeset	104 transformers.add(new ReflowTransformer("(\\{td\\})([^\\{]*)(\\{td\\})"));
c1d94c623854 ferrall@nwoca.org parents: 11 diff changeset	105 transformers.add(new ReflowTransformer("(\\{li\\})([^\\{]*)(\\{li\\})"));
7 a634b4d554d4 Minor fixups >, random smilies :), etc. Fixed blockquote. Handle escaping brackets outside pre tag. smith@nwoca.org parents: 6 diff changeset	106 transformers.add(new TagTransformer("<sup>(.*?)</sup>", true, "^\\[","\\]^ "));
a634b4d554d4 Minor fixups >, random smilies :), etc. Fixed blockquote. Handle escaping brackets outside pre tag. smith@nwoca.org parents: 6 diff changeset	107 transformers.add(new ReplaceTransformer("<","<"));
a634b4d554d4 Minor fixups >, random smilies :), etc. Fixed blockquote. Handle escaping brackets outside pre tag. smith@nwoca.org parents: 6 diff changeset	108 transformers.add(new ReplaceTransformer(">",">"));
a634b4d554d4 Minor fixups >, random smilies :), etc. Fixed blockquote. Handle escaping brackets outside pre tag. smith@nwoca.org parents: 6 diff changeset	109 transformers.add(new ReplaceTransformer(""","\""));
12 c1d94c623854 ferrall@nwoca.org parents: 11 diff changeset	110 transformers.add(new ReplaceTransformer("&","&"));
7 a634b4d554d4 Minor fixups >, random smilies :), etc. Fixed blockquote. Handle escaping brackets outside pre tag. smith@nwoca.org parents: 6 diff changeset	111 transformers.add(new ReplaceTransformer(":\\)",": )")); // No smilies...
13 cf58f4b9902b clean up ending {li} on line by itself smith@nwoca.org parents: 12 diff changeset	112 transformers.add(new ReplaceTransformer("(\\w)(--)(\\w)"," -- ",2)); // avoid strikeout
15 494ca5643e1a adjust h2 transformer for headings with arguments and added transformer for removing the empty contents table that is left as part of the conversion of the manuals. ferrall@nwoca.org parents: 14 diff changeset	113 transformers.add(new ReplaceTransformer("\\{table(.*?)\\}\\n\\s{2}\\{tr\\}\\n\\s{4}\\{td\\}Contents\\{td\\}\\n\\s{2}\\{tr\\}\\n\\{table\\}","")); // remove "contents" table
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	114
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	115 }
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	116
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	117 /**
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	118 * @param args the command line arguments
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	119 */
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	120 public static void main(String[] args) throws IOException {
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	121
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	122 if (args.length == 0) {
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	123 System.out.println("Usage:");
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	124 System.out.println(" Html2Wiki {inputDirectory} [Category]");
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	125 System.out.println(" default is current directory");
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	126 System.out.println(" Processes all *.html files. ");
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	127 System.out.println(" Each 'chapter' written to *.wiki");
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	128 return;
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	129 }
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	130
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	131 File inputs = new File(args[0]);
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	132
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	133 if (args.length > 1) {
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	134 category = args[1];
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	135 }
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	136
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	137 File[] inputFiles = inputs.listFiles(new HtmlFileFilter());
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	138 for (int i = 0; i < inputFiles.length; i++) {
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	139
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	140 process(inputFiles[i]);
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	141
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	142 }
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	143
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	144 }
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	145
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	146 protected static void process(File input) throws IOException {
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	147
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	148 System.out.println(input.getAbsoluteFile());
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	149
99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	150 Html2Wiki converter = new Html2Wiki(FileUtils.readFileToString(input, null));
99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	151
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	152 WikiChapter[] chapters = converter.getWikiChapters();
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	153
99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	154 System.out.format("Writing %d wiki files...\n", chapters.length);
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	155
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	156
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	157 for (int i = 0; i < chapters.length; i++) {
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	158
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	159 FileUtils.writeStringToFile(new File(input.getParent(),
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	160 generateFilename(chapters[i].getChapterName()) + ".wiki"),
99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	161 chapters[i].getContents().toString(),
99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	162 null);
99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	163
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	164 }
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	165
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	166 }
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	167
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	168 public static String generateFilename(String input) {
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	169 return input.replaceAll("\\\\\|/\|:\|\$\|\$", "-").replace("<br>", "");
99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	170
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	171 }
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	172
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	173 public String getWikiText() {
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	174 convert();
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	175 return buffer.toString();
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	176 }
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	177
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	178 public WikiChapter[] getWikiChapters() {
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	179
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	180 convert();
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	181
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	182 List<WikiChapter> chapters = new ArrayList<WikiChapter>();
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	183
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	184 Pattern chapterPat = Pattern.compile("<chapter>");
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	185 Matcher begin = chapterPat.matcher(buffer);
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	186 Matcher end = chapterPat.matcher(buffer);
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	187
99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	188 while (begin.find()) {
99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	189
99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	190
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	191 end.find(begin.end());
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	192
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	193 Pattern chapterNamePat = Pattern.compile("<chapter>(.*?)</chapter>");
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	194
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	195 Matcher chapterNameMatcher = chapterNamePat.matcher(buffer);
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	196
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	197 String chapterName = chapterNameMatcher.find(begin.start()) ? chapterNameMatcher.group(1) : null;
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	198
99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	199 CharSequence contents = buffer.subSequence(chapterName == null ? begin.start() : chapterNameMatcher.end(), end.hitEnd() ? buffer.length() : end.start());
99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	200
99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	201 chapters.add(new WikiChapter(chapterName, contents));
99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	202
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	203 }
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	204 return (WikiChapter[]) chapters.toArray(new WikiChapter[]{});
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	205 }
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	206
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	207 private void convert() {
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	208
99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	209 if (!converted) {
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	210 for (Transformer t : transformers) {
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	211
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	212 System.out.println(".Applying: " + t);
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	213 t.apply(buffer);
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	214
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	215 }
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	216 }
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	217 converted = true;
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	218 }
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	219
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	220 private static class HtmlFileFilter implements FileFilter {
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	221
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	222 public boolean accept(File pathname) {
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	223 return pathname.getName().toLowerCase().matches("^.*\\.html$");
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	224 }
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	225 }
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	226
2 5da2e67620f9 Upgrade to Ivy configuration and begin clean up of tests. Added FreeBSD license. smith@nwoca.org parents: 0 diff changeset	227 protected static class WikiChapter {
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	228
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	229 private String chapterName;
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	230 private CharSequence contents;
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	231
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	232 public WikiChapter(String chapterName, CharSequence contents) {
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	233 this.chapterName = chapterName.replaceAll("\\\\\|/\|:\|\$\|\$", "-").replaceAll("\\s+", " ").replaceAll("&", "and");
99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	234
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	235 this.contents = contents;
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	236 }
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	237
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	238 public String getChapterName() {
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	239 return chapterName;
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	240 }
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	241
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	242 public CharSequence getContents() {
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	243 return contents;
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	244 }
6 99f293bd507f Add "reflow" transformer to reflow paragraphs, list items, etc. smith@nwoca.org parents: 5 diff changeset	245
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	246 public String toString() {
2 5da2e67620f9 Upgrade to Ivy configuration and begin clean up of tests. Added FreeBSD license. smith@nwoca.org parents: 0 diff changeset	247 return "Chapter: " + chapterName + " Content length: " + contents.length();
0 f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	248 }
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	249 }
f8b1ea49d065 Initial version of crude HTML to WikiText converter. Customized for converting HTML files from DEC Document into Wiki markup. smith@nwoca.org parents: diff changeset	250 }

Mercurial > public > html2wiki

annotate src/org/nwoca/ssdt/tools/html2wiki/Html2Wiki.java @ 16:001e43423d5d