|
1 package org.jahia.services.fileextraction;
2
3 import java.io.*;
4 import java.util.*;
5
6 import org.jahia.utils.*;
7 import org.jahia.services.sites.JahiaSitesSlideService;
8 import org.pdfbox.cos.COSDictionary;
9 import org.pdfbox.cos.COSName;
10 import org.pdfbox.cos.COSString;
11 import org.pdfbox.util.PDFTextStripper;
12 import org.pdfbox.util.*;
13 import org.pdfbox.pdmodel.PDDocument;
14 import org.pdfbox.pdmodel.PDDocumentInformation;
15
16 import org.apache.slide.util.conf.Configuration;
17 import org.apache.slide.util.conf.ConfigurationException;
18
19
20
27 public class PDFExtractor implements FileExtractor {
28
29 private static org.apache.log4j.Logger logger =
30 org.apache.log4j.Logger.getLogger (PDFExtractor.class);
31
32 private String  path = null;
33 private long lastModifed;
34
35 protected List instructions = new ArrayList();
36 protected Map propertyMap = new HashMap();
37
38 public PDFExtractor(){
39 }
40
41
49 public synchronized ExtractedDocument getExtractedDocument(
50 String path,
51 long lastModified,
52 InputStream fileStream,
53 String charSet)
54 throws Exception {
55
56 ExtractedDocumentImpl extDoc = new ExtractedDocumentImpl();
57
58 PDDocument pdfDocument = null;
59 try
60 {
61 pdfDocument = PDDocument.load( fileStream );
62
63 if( pdfDocument.isEncrypted() )
64 {
65 pdfDocument.decrypt( "" );
67 }
68
69 ByteArrayOutputStream out = new ByteArrayOutputStream();
71 OutputStreamWriter writer = null;
72 if ( charSet != null ) {
73 writer = new OutputStreamWriter(out,charSet);
74 } else {
75 writer = new OutputStreamWriter(out);
76 }
77 PDFTextStripper stripper = new PDFTextStripper();
78 stripper.writeText(pdfDocument, writer);
79 writer.close();
80
81 String content = out.toString(charSet);
82 if ( content == null ){
83 content = "";
84 }
85 extDoc.setContent(content);
86
87 PDDocumentInformation info = pdfDocument.getDocumentInformation();
88
104
110
111 this.configure(((JahiaSitesSlideService)JahiaSitesSlideService.getInstance()).getConfiguration().getConfiguration("pdf-property-mapping"));
112
113 COSDictionary dict = info.getDictionary();
114 Iterator iterator = dict.keyList().iterator();
115 COSName key = null;
116 Calendar cal = null;
117 String val = null;
118 while ( iterator.hasNext() ){
119 try {
120 key = (COSName)iterator.next();
121 logger.debug("Found Pdf property : key=" + key.getName());
122
123 if ( !propertyMap.containsKey(key.getName()) ) {
124 continue;
125 }
126
127 cal = this.getCalendar(info,key);
128 if ( cal == null ){
129 COSString value = (COSString)info.getDictionary().getDictionaryObject( key );
130 if( value != null )
131 {
132 val = value.getString();
133 }
134 } else {
135 val = String.valueOf(cal.getTimeInMillis());
136 }
137 logger.debug("Found Pdf property : value=" + val);
138 extDoc.setProperty((String )propertyMap.get(key.getName()),val);
139 } catch ( Throwable t) {
140 logger.debug("Error handling pdf properties", t);
141 }
142 }
143
144 } catch( Throwable t ) {
145 logger.debug(t);
146 throw new IOException(" Exception occured parsing pdf :" + t);
147 } finally {
148 if( pdfDocument != null )
149 { try {
150 pdfDocument.close();
151 } catch ( Throwable t ){
152 }
153 }
154 }
155 return extDoc;
156 }
157
158
166 public synchronized ExtractedDocument getExtractedDocument(
167 String path,
168 long lastModified,
169 InputStream fileStream)
170 throws Exception {
171 return this.getExtractedDocument(path, lastModified, fileStream, null);
172 }
173
174
182 public String getContentAsString(String path, long lastModified,
183 InputStream fileStream)
184 throws Exception {
185 return getContentAsString(path, lastModified, fileStream, null);
186 }
187
188
197 public String getContentAsString(String path, long lastModified,
198 InputStream fileStream,
199 String charSet) throws Exception {
200 this.path = path;
201 this.lastModifed = lastModified;
202 String strVal = null;
203
204 if (fileStream != null) {
205 Reader pdfReader = null;
206 try {
207 long startTime = System.currentTimeMillis();
208 pdfReader = this.getPDFReader(fileStream, charSet);
209 long elapsedTime = System.currentTimeMillis() - startTime;
210 logger.info("Finished pdf extraction with PDFBox in " +
211 elapsedTime + "ms.");
212
213 startTime = System.currentTimeMillis();
214 strVal = FileUtils.readerToString(
215 pdfReader);
216 elapsedTime = System.currentTimeMillis() - startTime;
217 logger.info("Finished reading pdf Reader to String in " +
218 elapsedTime + "ms.");
219 }
220 catch (Throwable t) {
221 logger.debug("Error extracting dpdf file " + this.path ,t);
222 }
223 finally {
224 try {
225 if (pdfReader != null) {
226 pdfReader.close();
227 }
228 }
229 catch (Throwable t) {
230 }
231 }
232 }
233 return strVal;
234 }
235
236 public Reader getPDFReader(InputStream fileStream) throws IOException {
237 return getPDFReader(fileStream, null);
238 }
239
240 public Reader getPDFReader(InputStream fileStream,
241 String charSet) throws IOException
242 {
243 Reader reader = null;
244 PDDocument pdfDocument = null;
245 try {
246 pdfDocument = PDDocument.load(fileStream);
247 if(pdfDocument.isEncrypted()) {
248 pdfDocument.decrypt("");
250 }
251 ByteArrayOutputStream out = new ByteArrayOutputStream();
253 OutputStreamWriter writer = new OutputStreamWriter(out);
254 PDFTextStripper stripper = new PDFTextStripper();
255 stripper.writeText(pdfDocument, writer);
256 writer.close();
257 byte[] contents = out.toByteArray();
258 if ( charSet != null ){
259 reader = new InputStreamReader(new ByteArrayInputStream(contents),
260 charSet);
261 } else {
262 reader = new InputStreamReader(new ByteArrayInputStream(contents));
263 }
264 }
265 catch( Throwable t )
266 {
267 logger.debug(t);
268 throw new IOException(" Exception occured parsing pdf :" + t);
269 }
270 finally
271 {
272 if( pdfDocument != null )
273 { try {
274 pdfDocument.close();
275 } catch ( Throwable t ){
276 }
277 }
278 }
279 return reader;
280 }
281
282 public void configure(Configuration configuration) throws ConfigurationException {
283 Enumeration instructions = configuration.getConfigurations("instruction");
284 while (instructions.hasMoreElements()) {
285 Configuration extract = (Configuration)instructions.nextElement();
286 String property = extract.getAttribute("property");
287 String id = extract.getAttribute("id");
288 propertyMap.put(id, property);
289 }
290 }
291
292
299 private Calendar getCalendar(PDDocumentInformation info, COSName key){
300 if ( info == null || key == null ){
301 return null;
302 }
303 Calendar retval = null;
304 COSString value = (COSString)info.getDictionary().getDictionaryObject( key );
305
306 if( value != null )
307 {
308 String strValue = value.getString();
310 int index = 0;
311 if( strValue.startsWith( "D:" ) )
312 {
313 index = 2;
314 }
315 StringBuffer buff = new StringBuffer ("D:");
316
317 for( int i=index; i<strValue.length(); i++ )
319 {
320 if ( Character.isDigit( strValue.charAt( i ) ) ){
321 buff.append( strValue.charAt( i ) );
322 } else {
323 break;
324 }
325 }
327 DateConverter converter = new DateConverter();
330 try
331 {
332 retval = converter.toCalendar( buff.toString() );
333 }
334 catch( IOException e )
335 {
336 retval = null;
337 }
338 }
340 return retval;
341 }
342 }
343 |
Popular Tags |
Java API By Example, From Geeks To Geeks.