You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@tika.apache.org by ni...@apache.org on 2010/09/28 16:49:13 UTC
svn commit: r1002190 - in /tika/trunk/tika-parsers/src:
main/java/org/apache/tika/parser/microsoft/WordExtractor.java
test/java/org/apache/tika/parser/microsoft/POIContainerExtractionTest.java
Author: nick
Date: Tue Sep 28 14:49:13 2010
New Revision: 1002190
URL: http://svn.apache.org/viewvc?rev=1002190&view=rev
Log:
When processing .doc files, handle matching the embedded images to the character runs better. Somewhat copes with \u0001 real images vs \u0008 floating escher images, and is as good as we can probably get given the current Microsoft docs... Avoids NullPointers though! (TIKA-506)
Modified:
tika/trunk/tika-parsers/src/main/java/org/apache/tika/parser/microsoft/WordExtractor.java
tika/trunk/tika-parsers/src/test/java/org/apache/tika/parser/microsoft/POIContainerExtractionTest.java
Modified: tika/trunk/tika-parsers/src/main/java/org/apache/tika/parser/microsoft/WordExtractor.java
URL: http://svn.apache.org/viewvc/tika/trunk/tika-parsers/src/main/java/org/apache/tika/parser/microsoft/WordExtractor.java?rev=1002190&r1=1002189&r2=1002190&view=diff
==============================================================================
--- tika/trunk/tika-parsers/src/main/java/org/apache/tika/parser/microsoft/WordExtractor.java (original)
+++ tika/trunk/tika-parsers/src/main/java/org/apache/tika/parser/microsoft/WordExtractor.java Tue Sep 28 14:49:13 2010
@@ -19,8 +19,11 @@ package org.apache.tika.parser.microsoft
import java.io.FileNotFoundException;
import java.io.IOException;
import java.util.ArrayList;
-import java.util.Iterator;
+import java.util.HashMap;
+import java.util.HashSet;
import java.util.List;
+import java.util.Map;
+import java.util.Set;
import org.apache.poi.hwpf.HWPFDocument;
import org.apache.poi.hwpf.HWPFOldDocument;
@@ -69,9 +72,7 @@ public class WordExtractor extends Abstr
// the pictures should be in order, and may be directly
// placed or referenced from an anchor
PicturesTable pictureTable = document.getPicturesTable();
- CountingIterator<Picture> pictures = new CountingIterator<Picture>(
- pictureTable.getAllPictures().iterator()
- );
+ PicturesSource pictures = new PicturesSource(document);
// Do the main paragraph text
Range r = document.getRange();
@@ -96,11 +97,11 @@ public class WordExtractor extends Abstr
addTextIfAny(xhtml, "footer", wordExtractor.getFooterText());
// Handle any pictures that we haven't output yet
- while(pictures.hasNext()) {
- Picture p = pictures.next();
+ for(Picture p = pictures.nextUnclaimed(); p != null; ) {
handlePictureCharacterRun(
- null, p, pictures.getCount(), xhtml
+ null, p, pictures, xhtml
);
+ p = pictures.nextUnclaimed();
}
// Handle any embeded office documents
@@ -118,7 +119,7 @@ public class WordExtractor extends Abstr
}
private int handleParagraph(Paragraph p, int parentTableLevel, Range r, HWPFDocument document,
- CountingIterator<Picture> pictures, PicturesTable pictureTable, XHTMLContentHandler xhtml)
+ PicturesSource pictures, PicturesTable pictureTable, XHTMLContentHandler xhtml)
throws SAXException, IOException, TikaException {
// Note - a poi bug means we can't currently properly recurse
// into nested tables, so currently we don't
@@ -162,15 +163,20 @@ public class WordExtractor extends Abstr
CharacterRun cr = p.getCharacterRun(j);
if(cr.text().equals("\u0013")) {
- j += handleSpecialCharacterRuns(p, j, tas.isHeading(), xhtml);
- } else if(cr.text().equals("\u0008")) {
- // Floating Picture
- Picture picture = pictures.next();
- handlePictureCharacterRun(cr, picture, pictures.getCount(), xhtml);
+ j += handleSpecialCharacterRuns(p, j, tas.isHeading(), pictures, xhtml);
+ } else if(cr.text().startsWith("\u0008")) {
+ // Floating Picture(s)
+ for(int pn=0; pn<cr.text().length(); pn++) {
+ // Assume they're in the order from the unclaimed list...
+ Picture picture = pictures.nextUnclaimed();
+
+ // Output
+ handlePictureCharacterRun(cr, picture, pictures, xhtml);
+ }
} else if(pictureTable.hasPicture(cr)) {
// Inline Picture
- Picture picture = pictures.next();
- handlePictureCharacterRun(cr, picture, pictures.getCount(), xhtml);
+ Picture picture = pictures.getFor(cr);
+ handlePictureCharacterRun(cr, picture, pictures, xhtml);
} else {
handleCharacterRun(cr, tas.isHeading(), xhtml);
}
@@ -215,19 +221,19 @@ public class WordExtractor extends Abstr
* Can be \13..text..\15 or \13..control..\14..text..\15 .
* Nesting is allowed
*/
- private int handleSpecialCharacterRuns(Paragraph p, int index, boolean skipStyling, XHTMLContentHandler xhtml)
- throws SAXException {
+ private int handleSpecialCharacterRuns(Paragraph p, int index, boolean skipStyling,
+ PicturesSource pictures, XHTMLContentHandler xhtml) throws SAXException, TikaException, IOException {
List<CharacterRun> controls = new ArrayList<CharacterRun>();
List<CharacterRun> texts = new ArrayList<CharacterRun>();
boolean has14 = false;
// Split it into before and after the 14
int i;
- for(i=index; i<p.numCharacterRuns(); i++) {
+ for(i=index+1; i<p.numCharacterRuns(); i++) {
CharacterRun cr = p.getCharacterRun(i);
if(cr.text().equals("\u0013")) {
// Nested, oh joy...
- int increment = handleSpecialCharacterRuns(p, i+1, skipStyling, xhtml);
+ int increment = handleSpecialCharacterRuns(p, i+1, skipStyling, pictures, xhtml);
i += increment;
} else if(cr.text().equals("\u0014")) {
has14 = true;
@@ -266,7 +272,12 @@ public class WordExtractor extends Abstr
} else {
// Just output the text ones
for(CharacterRun cr : texts) {
- handleCharacterRun(cr, skipStyling, xhtml);
+ if(pictures.hasPicture(cr)) {
+ Picture picture = pictures.getFor(cr);
+ handlePictureCharacterRun(cr, picture, pictures, xhtml);
+ } else {
+ handleCharacterRun(cr, skipStyling, xhtml);
+ }
}
}
} else {
@@ -281,9 +292,18 @@ public class WordExtractor extends Abstr
return i-index;
}
- private void handlePictureCharacterRun(CharacterRun cr, Picture picture, int pictureNumber, XHTMLContentHandler xhtml)
+ private void handlePictureCharacterRun(CharacterRun cr, Picture picture, PicturesSource pictures, XHTMLContentHandler xhtml)
throws SAXException, IOException, TikaException {
+ if(picture == null) {
+ // Oh dear, we've run out...
+ // Probably caused by multiple \u0008 images referencing
+ // the same real image
+ return;
+ }
+
+ // Which one is it?
String extension = picture.suggestFileExtension();
+ int pictureNumber = pictures.pictureNumber(picture);
// Make up a name for the picture
// There isn't one in the file, but we need to be able to reference
@@ -296,9 +316,14 @@ public class WordExtractor extends Abstr
// Output the img tag
xhtml.startElement("img", "src", "embedded:" + filename);
xhtml.endElement("img");
-
- TikaInputStream stream = TikaInputStream.get(picture.getContent());
- handleEmbeddedResource(stream, filename, mimeType, xhtml, false);
+
+ // Have we already output this one?
+ // (Only expose each individual image once)
+ if(! pictures.hasOutput(picture)) {
+ TikaInputStream stream = TikaInputStream.get(picture.getContent());
+ handleEmbeddedResource(stream, filename, mimeType, xhtml, false);
+ pictures.recordOutput(picture);
+ }
}
/**
@@ -387,28 +412,82 @@ public class WordExtractor extends Abstr
}
}
- private static class CountingIterator<T> implements Iterator<T> {
- private Iterator<T> parent;
- private int count = 0;
- private CountingIterator(Iterator<T> parent) {
- this.parent = parent;
- }
-
- public boolean hasNext() {
- return parent.hasNext();
- }
-
- public T next() {
- count++;
- return parent.next();
- }
-
- public int getCount() {
- return count;
- }
-
- public void remove() {
- throw new UnsupportedOperationException();
- }
+ /**
+ * Provides access to the pictures both by offset, iteration
+ * over the un-claimed, and peeking forward
+ */
+ private static class PicturesSource {
+ private PicturesTable picturesTable;
+ private Set<Picture> output = new HashSet<Picture>();
+ private Map<Integer,Picture> lookup;
+ private List<Picture> nonU1based;
+ private List<Picture> all;
+ private int pn = 0;
+
+ private PicturesSource(HWPFDocument doc) {
+ picturesTable = doc.getPicturesTable();
+ all = picturesTable.getAllPictures();
+
+ // Compute the Offset-Picture lookup
+ lookup = new HashMap<Integer, Picture>();
+ for(Picture p : all) {
+ // TODO Make this nicer when POI 3.7 is out
+ String name = p.suggestFullFileName();
+ if(name.indexOf('.') > -1)
+ name = name.substring(0, name.indexOf('.'));
+ int offset = Integer.parseInt(name, 16);
+ lookup.put(offset, p);
+ }
+
+ // Work out which Pictures aren't referenced by
+ // a \u0001 in the main text
+ // These are \u0008 escher floating ones, ones
+ // found outside the normal text, and who
+ // knows what else...
+ nonU1based = new ArrayList<Picture>();
+ nonU1based.addAll(all);
+ Range r = doc.getRange();
+ for(int i=0; i<r.numCharacterRuns(); i++) {
+ CharacterRun cr = r.getCharacterRun(i);
+ if(picturesTable.hasPicture(cr)) {
+ Picture p = getFor(cr);
+ int at = nonU1based.indexOf(p);
+ nonU1based.set(at, null);
+ }
+ }
+ }
+
+ private boolean hasPicture(CharacterRun cr) {
+ return picturesTable.hasPicture(cr);
+ }
+
+ private void recordOutput(Picture picture) {
+ output.add(picture);
+ }
+ private boolean hasOutput(Picture picture) {
+ return output.contains(picture);
+ }
+
+ private int pictureNumber(Picture picture) {
+ return all.indexOf(picture) + 1;
+ }
+
+ private Picture getFor(CharacterRun cr) {
+ return lookup.get(cr.getPicOffset());
+ }
+
+ /**
+ * Return the next unclaimed one, used towards
+ * the end
+ */
+ private Picture nextUnclaimed() {
+ Picture p = null;
+ while(pn < nonU1based.size()) {
+ p = nonU1based.get(pn);
+ pn++;
+ if(p != null) return p;
+ }
+ return null;
+ }
}
}
Modified: tika/trunk/tika-parsers/src/test/java/org/apache/tika/parser/microsoft/POIContainerExtractionTest.java
URL: http://svn.apache.org/viewvc/tika/trunk/tika-parsers/src/test/java/org/apache/tika/parser/microsoft/POIContainerExtractionTest.java?rev=1002190&r1=1002189&r2=1002190&view=diff
==============================================================================
--- tika/trunk/tika-parsers/src/test/java/org/apache/tika/parser/microsoft/POIContainerExtractionTest.java (original)
+++ tika/trunk/tika-parsers/src/test/java/org/apache/tika/parser/microsoft/POIContainerExtractionTest.java Tue Sep 28 14:49:13 2010
@@ -158,15 +158,24 @@ public class POIContainerExtractionTest
assertEquals(9, handler.mediaTypes.size());
// Filenames are a bit iffy...
-// for(String filename : handler.filenames)
-// assertEquals(null, filename);
+ // Should really be 3*embedded pictures then 3*icons then embedded docs
+ assertEquals("image1", handler.filenames.get(0));
+ assertEquals("image4.png", handler.filenames.get(1));
+ assertEquals("image5.jpg", handler.filenames.get(2));
+ assertEquals("image6.png", handler.filenames.get(3));
+ assertEquals("image2", handler.filenames.get(4));
+ assertEquals("image3", handler.filenames.get(5));
+ assertEquals(null, handler.filenames.get(6));
+ assertEquals(null, handler.filenames.get(7));
+ assertEquals(null, handler.filenames.get(8));
+
// But we do know their types
assertEquals(MediaType.parse("image/unknown"), handler.mediaTypes.get(0)); // Icon of embedded office doc?
- assertEquals(MediaType.parse("image/unknown"), handler.mediaTypes.get(1)); // Icon of embedded office doc?
- assertEquals(MediaType.parse("image/unknown"), handler.mediaTypes.get(2)); // Icon of embedded office doc?
- assertEquals(TYPE_PNG, handler.mediaTypes.get(3)); // Embedded image
- assertEquals(TYPE_JPG, handler.mediaTypes.get(4)); // Embedded image
- assertEquals(TYPE_PNG, handler.mediaTypes.get(5)); // Embedded image
+ assertEquals(TYPE_PNG, handler.mediaTypes.get(1)); // Embedded image - logo
+ assertEquals(TYPE_JPG, handler.mediaTypes.get(2)); // Embedded image - safe
+ assertEquals(TYPE_PNG, handler.mediaTypes.get(3)); // Embedded image - try
+ assertEquals(MediaType.parse("image/unknown"), handler.mediaTypes.get(4)); // Icon of embedded office doc?
+ assertEquals(MediaType.parse("image/unknown"), handler.mediaTypes.get(5)); // Icon of embedded office doc?
assertEquals(TYPE_DOCX, handler.mediaTypes.get(6)); // Embedded office doc
assertEquals(TYPE_PPT, handler.mediaTypes.get(7)); // Embedded office doc
assertEquals(TYPE_XLS, handler.mediaTypes.get(8)); // Embedded office doc
@@ -179,11 +188,11 @@ public class POIContainerExtractionTest
// We don't know their filenames, except for doc images + docx
assertEquals("image1", handler.filenames.get(0));
- assertEquals("image2", handler.filenames.get(1));
- assertEquals("image3", handler.filenames.get(2));
- assertEquals("image4.png", handler.filenames.get(3));
- assertEquals("image5.jpg", handler.filenames.get(4));
- assertEquals("image6.png", handler.filenames.get(5));
+ assertEquals("image4.png", handler.filenames.get(1));
+ assertEquals("image5.jpg", handler.filenames.get(2));
+ assertEquals("image6.png", handler.filenames.get(3));
+ assertEquals("image2", handler.filenames.get(4));
+ assertEquals("image3", handler.filenames.get(5));
assertEquals(null, handler.filenames.get(6));
assertEquals("image2.png", handler.filenames.get(7));
assertEquals("image3.jpeg", handler.filenames.get(8));
@@ -193,11 +202,11 @@ public class POIContainerExtractionTest
}
// But we do know their types
assertEquals(MediaType.parse("image/unknown"), handler.mediaTypes.get(0)); // Icon of embedded office doc?
- assertEquals(MediaType.parse("image/unknown"), handler.mediaTypes.get(1)); // Icon of embedded office doc?
- assertEquals(MediaType.parse("image/unknown"), handler.mediaTypes.get(2)); // Icon of embedded office doc?
- assertEquals(TYPE_PNG, handler.mediaTypes.get(3)); // Embedded image
- assertEquals(TYPE_JPG, handler.mediaTypes.get(4)); // Embedded image
- assertEquals(TYPE_PNG, handler.mediaTypes.get(5)); // Embedded image
+ assertEquals(TYPE_PNG, handler.mediaTypes.get(1)); // Embedded image - logo
+ assertEquals(TYPE_JPG, handler.mediaTypes.get(2)); // Embedded image - safe
+ assertEquals(TYPE_PNG, handler.mediaTypes.get(3)); // Embedded image - try
+ assertEquals(MediaType.parse("image/unknown"), handler.mediaTypes.get(4)); // Icon of embedded office doc?
+ assertEquals(MediaType.parse("image/unknown"), handler.mediaTypes.get(5)); // Icon of embedded office doc?
assertEquals(TYPE_DOCX, handler.mediaTypes.get(6)); // Embedded office doc
assertEquals(TYPE_PNG, handler.mediaTypes.get(7)); // PNG inside .docx
assertEquals(TYPE_JPG, handler.mediaTypes.get(8)); // JPG inside .docx