annotate src/de/mpiwg/dwinter/fulltextIndexer/OCRutils/.svn/text-base/ParseOcrDocument.java.svn-base @ 0:dc7622afcfea default tip

initial
author dwinter
date Wed, 03 Nov 2010 12:33:16 +0100
parents
children
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
0
dc7622afcfea initial
dwinter
parents:
diff changeset
1 /* */ package de.mpiwg.dwinter.fulltextIndexer.OCRutils;
dc7622afcfea initial
dwinter
parents:
diff changeset
2 /* */
dc7622afcfea initial
dwinter
parents:
diff changeset
3 /* */ import java.util.ArrayList;
dc7622afcfea initial
dwinter
parents:
diff changeset
4 /* */ import org.xml.sax.Attributes;
dc7622afcfea initial
dwinter
parents:
diff changeset
5 /* */ import org.xml.sax.Locator;
dc7622afcfea initial
dwinter
parents:
diff changeset
6 /* */ import org.xml.sax.SAXException;
dc7622afcfea initial
dwinter
parents:
diff changeset
7 /* */ import org.xml.sax.helpers.DefaultHandler;
dc7622afcfea initial
dwinter
parents:
diff changeset
8 /* */
dc7622afcfea initial
dwinter
parents:
diff changeset
9 /* */ public class ParseOcrDocument extends DefaultHandler
dc7622afcfea initial
dwinter
parents:
diff changeset
10 /* */ {
dc7622afcfea initial
dwinter
parents:
diff changeset
11 /* 31 */ public OCRDocument ocrDocument = new OCRDocument();
dc7622afcfea initial
dwinter
parents:
diff changeset
12 /* */ private int lineCounter;
dc7622afcfea initial
dwinter
parents:
diff changeset
13 /* */ private OCRDocument.OCRLine currentLine;
dc7622afcfea initial
dwinter
parents:
diff changeset
14 /* */ private boolean inLine;
dc7622afcfea initial
dwinter
parents:
diff changeset
15 /* */
dc7622afcfea initial
dwinter
parents:
diff changeset
16 /* */ public void startDocument()
dc7622afcfea initial
dwinter
parents:
diff changeset
17 /* */ throws SAXException
dc7622afcfea initial
dwinter
parents:
diff changeset
18 /* */ {
dc7622afcfea initial
dwinter
parents:
diff changeset
19 /* */ }
dc7622afcfea initial
dwinter
parents:
diff changeset
20 /* */
dc7622afcfea initial
dwinter
parents:
diff changeset
21 /* */ public void endDocument()
dc7622afcfea initial
dwinter
parents:
diff changeset
22 /* */ throws SAXException
dc7622afcfea initial
dwinter
parents:
diff changeset
23 /* */ {
dc7622afcfea initial
dwinter
parents:
diff changeset
24 /* */ }
dc7622afcfea initial
dwinter
parents:
diff changeset
25 /* */
dc7622afcfea initial
dwinter
parents:
diff changeset
26 /* */ public void characters(char[] c, int start, int length)
dc7622afcfea initial
dwinter
parents:
diff changeset
27 /* */ throws SAXException
dc7622afcfea initial
dwinter
parents:
diff changeset
28 /* */ {
dc7622afcfea initial
dwinter
parents:
diff changeset
29 /* 52 */ if (this.inLine)
dc7622afcfea initial
dwinter
parents:
diff changeset
30 /* 53 */ this.currentLine.content += new String(c, start, length);
dc7622afcfea initial
dwinter
parents:
diff changeset
31 /* */ }
dc7622afcfea initial
dwinter
parents:
diff changeset
32 /* */
dc7622afcfea initial
dwinter
parents:
diff changeset
33 /* */ public void ignorableWhitespace(char[] c, int start, int length)
dc7622afcfea initial
dwinter
parents:
diff changeset
34 /* */ throws SAXException
dc7622afcfea initial
dwinter
parents:
diff changeset
35 /* */ {
dc7622afcfea initial
dwinter
parents:
diff changeset
36 /* */ }
dc7622afcfea initial
dwinter
parents:
diff changeset
37 /* */
dc7622afcfea initial
dwinter
parents:
diff changeset
38 /* */ public void processingInstruction(String target, String data)
dc7622afcfea initial
dwinter
parents:
diff changeset
39 /* */ throws SAXException
dc7622afcfea initial
dwinter
parents:
diff changeset
40 /* */ {
dc7622afcfea initial
dwinter
parents:
diff changeset
41 /* */ }
dc7622afcfea initial
dwinter
parents:
diff changeset
42 /* */
dc7622afcfea initial
dwinter
parents:
diff changeset
43 /* */ public void setDocumentLocator(Locator arg1)
dc7622afcfea initial
dwinter
parents:
diff changeset
44 /* */ {
dc7622afcfea initial
dwinter
parents:
diff changeset
45 /* */ }
dc7622afcfea initial
dwinter
parents:
diff changeset
46 /* */
dc7622afcfea initial
dwinter
parents:
diff changeset
47 /* */ public void endElement(String uri, String localName, String name)
dc7622afcfea initial
dwinter
parents:
diff changeset
48 /* */ throws SAXException
dc7622afcfea initial
dwinter
parents:
diff changeset
49 /* */ {
dc7622afcfea initial
dwinter
parents:
diff changeset
50 /* 76 */ if ((!name.equals("span")) ||
dc7622afcfea initial
dwinter
parents:
diff changeset
51 /* 78 */ (!this.inLine))
dc7622afcfea initial
dwinter
parents:
diff changeset
52 /* */ return;
dc7622afcfea initial
dwinter
parents:
diff changeset
53 /* 80 */ this.ocrDocument.OCRLines.add(this.currentLine);
dc7622afcfea initial
dwinter
parents:
diff changeset
54 /* 81 */ this.inLine = false;
dc7622afcfea initial
dwinter
parents:
diff changeset
55 /* */ }
dc7622afcfea initial
dwinter
parents:
diff changeset
56 /* */
dc7622afcfea initial
dwinter
parents:
diff changeset
57 /* */ public void endPrefixMapping(String prefix)
dc7622afcfea initial
dwinter
parents:
diff changeset
58 /* */ throws SAXException
dc7622afcfea initial
dwinter
parents:
diff changeset
59 /* */ {
dc7622afcfea initial
dwinter
parents:
diff changeset
60 /* */ }
dc7622afcfea initial
dwinter
parents:
diff changeset
61 /* */
dc7622afcfea initial
dwinter
parents:
diff changeset
62 /* */ public void skippedEntity(String name)
dc7622afcfea initial
dwinter
parents:
diff changeset
63 /* */ throws SAXException
dc7622afcfea initial
dwinter
parents:
diff changeset
64 /* */ {
dc7622afcfea initial
dwinter
parents:
diff changeset
65 /* */ }
dc7622afcfea initial
dwinter
parents:
diff changeset
66 /* */
dc7622afcfea initial
dwinter
parents:
diff changeset
67 /* */ public void startElement(String uri, String localName, String name, Attributes attrs)
dc7622afcfea initial
dwinter
parents:
diff changeset
68 /* */ throws SAXException
dc7622afcfea initial
dwinter
parents:
diff changeset
69 /* */ {
dc7622afcfea initial
dwinter
parents:
diff changeset
70 /* 100 */ if ((name.equals("div")) &&
dc7622afcfea initial
dwinter
parents:
diff changeset
71 /* 101 */ (attrs != null))
dc7622afcfea initial
dwinter
parents:
diff changeset
72 /* */ {
dc7622afcfea initial
dwinter
parents:
diff changeset
73 /* 103 */ int length = attrs.getLength();
dc7622afcfea initial
dwinter
parents:
diff changeset
74 /* */
dc7622afcfea initial
dwinter
parents:
diff changeset
75 /* 106 */ for (int i = 0; i < length; ++i)
dc7622afcfea initial
dwinter
parents:
diff changeset
76 /* */ {
dc7622afcfea initial
dwinter
parents:
diff changeset
77 /* 108 */ if ((!attrs.getLocalName(i).equals("class")) ||
dc7622afcfea initial
dwinter
parents:
diff changeset
78 /* 110 */ (!attrs.getValue(i).equals("ocr_page"))) continue;
dc7622afcfea initial
dwinter
parents:
diff changeset
79 /* 111 */ doPage(attrs);
dc7622afcfea initial
dwinter
parents:
diff changeset
80 /* */ }
dc7622afcfea initial
dwinter
parents:
diff changeset
81 /* */
dc7622afcfea initial
dwinter
parents:
diff changeset
82 /* */ }
dc7622afcfea initial
dwinter
parents:
diff changeset
83 /* */
dc7622afcfea initial
dwinter
parents:
diff changeset
84 /* 117 */ if ((!name.equals("span")) ||
dc7622afcfea initial
dwinter
parents:
diff changeset
85 /* 118 */ (attrs == null))
dc7622afcfea initial
dwinter
parents:
diff changeset
86 /* */ return;
dc7622afcfea initial
dwinter
parents:
diff changeset
87 /* 120 */ int length = attrs.getLength();
dc7622afcfea initial
dwinter
parents:
diff changeset
88 /* */
dc7622afcfea initial
dwinter
parents:
diff changeset
89 /* 123 */ for (int i = 0; i < length; ++i)
dc7622afcfea initial
dwinter
parents:
diff changeset
90 /* */ {
dc7622afcfea initial
dwinter
parents:
diff changeset
91 /* 125 */ if ((!attrs.getLocalName(i).equals("class")) ||
dc7622afcfea initial
dwinter
parents:
diff changeset
92 /* 127 */ (!attrs.getValue(i).equals("ocr_line"))) continue;
dc7622afcfea initial
dwinter
parents:
diff changeset
93 /* 128 */ doLine(attrs);
dc7622afcfea initial
dwinter
parents:
diff changeset
94 /* */ }
dc7622afcfea initial
dwinter
parents:
diff changeset
95 /* */ }
dc7622afcfea initial
dwinter
parents:
diff changeset
96 /* */
dc7622afcfea initial
dwinter
parents:
diff changeset
97 /* */ private void doPage(Attributes attrs)
dc7622afcfea initial
dwinter
parents:
diff changeset
98 /* */ {
dc7622afcfea initial
dwinter
parents:
diff changeset
99 /* 136 */ int length = attrs.getLength();
dc7622afcfea initial
dwinter
parents:
diff changeset
100 /* */
dc7622afcfea initial
dwinter
parents:
diff changeset
101 /* 139 */ for (int i = 0; i < length; ++i)
dc7622afcfea initial
dwinter
parents:
diff changeset
102 /* */ {
dc7622afcfea initial
dwinter
parents:
diff changeset
103 /* 141 */ if (!attrs.getLocalName(i).equals("title"))
dc7622afcfea initial
dwinter
parents:
diff changeset
104 /* */ continue;
dc7622afcfea initial
dwinter
parents:
diff changeset
105 /* 143 */ String title = attrs.getValue(i);
dc7622afcfea initial
dwinter
parents:
diff changeset
106 /* 144 */ String[] splitted = title.split(" ");
dc7622afcfea initial
dwinter
parents:
diff changeset
107 /* 145 */ String dimension = splitted[(splitted.length - 2)] + " " + splitted[(splitted.length - 1)];
dc7622afcfea initial
dwinter
parents:
diff changeset
108 /* 146 */ this.ocrDocument.pageDimension = dimension;
dc7622afcfea initial
dwinter
parents:
diff changeset
109 /* */ }
dc7622afcfea initial
dwinter
parents:
diff changeset
110 /* */
dc7622afcfea initial
dwinter
parents:
diff changeset
111 /* 149 */ this.lineCounter = 0;
dc7622afcfea initial
dwinter
parents:
diff changeset
112 /* */ }
dc7622afcfea initial
dwinter
parents:
diff changeset
113 /* */
dc7622afcfea initial
dwinter
parents:
diff changeset
114 /* */ private void doLine(Attributes attrs)
dc7622afcfea initial
dwinter
parents:
diff changeset
115 /* */ {
dc7622afcfea initial
dwinter
parents:
diff changeset
116 /* */ OCRDocument doc = this.ocrDocument;
dc7622afcfea initial
dwinter
parents:
diff changeset
117 this.currentLine = doc.new OCRLine();
dc7622afcfea initial
dwinter
parents:
diff changeset
118 /* */
dc7622afcfea initial
dwinter
parents:
diff changeset
119 /* 156 */ this.inLine = true;
dc7622afcfea initial
dwinter
parents:
diff changeset
120 /* 157 */ this.currentLine.content = new String();
dc7622afcfea initial
dwinter
parents:
diff changeset
121 /* */
dc7622afcfea initial
dwinter
parents:
diff changeset
122 /* 159 */ int length = attrs.getLength();
dc7622afcfea initial
dwinter
parents:
diff changeset
123 /* 160 */ this.currentLine.lineNumber = String.valueOf(this.lineCounter);
dc7622afcfea initial
dwinter
parents:
diff changeset
124 /* 161 */ this.lineCounter += 1;
dc7622afcfea initial
dwinter
parents:
diff changeset
125 /* 162 */ this.currentLine.bbox = "0 0";
dc7622afcfea initial
dwinter
parents:
diff changeset
126 /* 163 */ for (int i = 0; i < length; ++i)
dc7622afcfea initial
dwinter
parents:
diff changeset
127 /* */ {
dc7622afcfea initial
dwinter
parents:
diff changeset
128 /* 165 */ if (!attrs.getLocalName(i).equals("title"))
dc7622afcfea initial
dwinter
parents:
diff changeset
129 /* */ continue;
dc7622afcfea initial
dwinter
parents:
diff changeset
130 /* 167 */ String title = attrs.getValue(i);
dc7622afcfea initial
dwinter
parents:
diff changeset
131 /* */
dc7622afcfea initial
dwinter
parents:
diff changeset
132 /* 169 */ String dimension = title.replace("bbox ", "");
dc7622afcfea initial
dwinter
parents:
diff changeset
133 /* 170 */ this.currentLine.bbox = dimension;
dc7622afcfea initial
dwinter
parents:
diff changeset
134 /* */ }
dc7622afcfea initial
dwinter
parents:
diff changeset
135 /* */ }
dc7622afcfea initial
dwinter
parents:
diff changeset
136 /* */
dc7622afcfea initial
dwinter
parents:
diff changeset
137 /* */ public void startPrefixMapping(String prefix, String uri)
dc7622afcfea initial
dwinter
parents:
diff changeset
138 /* */ throws SAXException
dc7622afcfea initial
dwinter
parents:
diff changeset
139 /* */ {
dc7622afcfea initial
dwinter
parents:
diff changeset
140 /* */ }
dc7622afcfea initial
dwinter
parents:
diff changeset
141 /* */ }
dc7622afcfea initial
dwinter
parents:
diff changeset
142
dc7622afcfea initial
dwinter
parents:
diff changeset
143 /* Location: /private/tmp/fulltextIndexer.jar
dc7622afcfea initial
dwinter
parents:
diff changeset
144 * Qualified Name: de.mpiwg.dwinter.fulltextIndexer.OCRutils.ParseOcrDocument
dc7622afcfea initial
dwinter
parents:
diff changeset
145 * JD-Core Version: 0.5.4
dc7622afcfea initial
dwinter
parents:
diff changeset
146 */