Java tutorial
/* * Licensed to the Apache Software Foundation (ASF) under one or more * contributor license agreements. See the NOTICE file distributed with * this work for additional information regarding copyright ownership. * The ASF licenses this file to You under the Apache License, Version 2.0 * (the "License"); you may not use this file except in compliance with * the License. You may obtain a copy of the License at * * http://www.apache.org/licenses/LICENSE-2.0 * * Unless required by applicable law or agreed to in writing, software * distributed under the License is distributed on an "AS IS" BASIS, * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. * See the License for the specific language governing permissions and * limitations under the License. */ package org.apache.tika.parser.microsoft; import java.io.FileNotFoundException; import java.io.IOException; import java.util.ArrayList; import java.util.HashMap; import java.util.HashSet; import java.util.List; import java.util.Locale; import java.util.Map; import java.util.Set; import org.apache.poi.hwpf.HWPFDocument; import org.apache.poi.hwpf.HWPFOldDocument; import org.apache.poi.hwpf.OldWordFileFormatException; import org.apache.poi.hwpf.extractor.Word6Extractor; import org.apache.poi.hwpf.model.FieldsDocumentPart; import org.apache.poi.hwpf.model.PicturesTable; import org.apache.poi.hwpf.model.StyleDescription; import org.apache.poi.hwpf.usermodel.CharacterRun; import org.apache.poi.hwpf.usermodel.Field; import org.apache.poi.hwpf.usermodel.HeaderStories; import org.apache.poi.hwpf.usermodel.Paragraph; import org.apache.poi.hwpf.usermodel.Picture; import org.apache.poi.hwpf.usermodel.Range; import org.apache.poi.hwpf.usermodel.Table; import org.apache.poi.hwpf.usermodel.TableCell; import org.apache.poi.hwpf.usermodel.TableRow; import org.apache.poi.poifs.filesystem.DirectoryEntry; import org.apache.poi.poifs.filesystem.DirectoryNode; import org.apache.poi.poifs.filesystem.Entry; import org.apache.poi.poifs.filesystem.NPOIFSFileSystem; import org.apache.tika.exception.TikaException; import org.apache.tika.io.TikaInputStream; import org.apache.tika.parser.ParseContext; import org.apache.tika.sax.XHTMLContentHandler; import org.xml.sax.SAXException; import org.xml.sax.helpers.AttributesImpl; import static java.nio.charset.StandardCharsets.UTF_8; public class WordExtractor extends AbstractPOIFSExtractor { private static final char UNICODECHAR_NONBREAKING_HYPHEN = '\u2011'; private static final char UNICODECHAR_ZERO_WIDTH_SPACE = '\u200b'; // could be improved by using the real delimiter in xchFollow [MS-DOC], v20140721, 2.4.6.3, Part 3, Step 3 private static final String LIST_DELIMITER = " "; private static final Map<String, TagAndStyle> fixedParagraphStyles = new HashMap<String, TagAndStyle>(); private static final TagAndStyle defaultParagraphStyle = new TagAndStyle("p", null); static { fixedParagraphStyles.put("Default", defaultParagraphStyle); fixedParagraphStyles.put("Normal", defaultParagraphStyle); fixedParagraphStyles.put("heading", new TagAndStyle("h1", null)); fixedParagraphStyles.put("Heading", new TagAndStyle("h1", null)); fixedParagraphStyles.put("Title", new TagAndStyle("h1", "title")); fixedParagraphStyles.put("Subtitle", new TagAndStyle("h2", "subtitle")); fixedParagraphStyles.put("HTML Preformatted", new TagAndStyle("pre", null)); } // True if we are currently in the named style tag: private boolean curStrikeThrough; private boolean curBold; private boolean curItalic; public WordExtractor(ParseContext context) { super(context); } private static int countParagraphs(Range... ranges) { int count = 0; for (Range r : ranges) { if (r != null) { count += r.numParagraphs(); } } return count; } /** * Given a style name, return what tag should be used, and * what style should be applied to it. */ public static TagAndStyle buildParagraphTagAndStyle(String styleName, boolean isTable) { TagAndStyle tagAndStyle = fixedParagraphStyles.get(styleName); if (tagAndStyle != null) { return tagAndStyle; } if (styleName.equals("Table Contents") && isTable) { return defaultParagraphStyle; } String tag = "p"; String styleClass = null; if (styleName.startsWith("heading") || styleName.startsWith("Heading")) { // "Heading 3" or "Heading2" or "heading 4" int num = 1; try { num = Integer.parseInt(styleName.substring(styleName.length() - 1)); } catch (NumberFormatException e) { } // Turn it into a H1 - H6 (H7+ isn't valid!) tag = "h" + Math.min(num, 6); } else { styleClass = styleName.replace(' ', '_'); styleClass = styleClass.substring(0, 1).toLowerCase(Locale.ROOT) + styleClass.substring(1); } return new TagAndStyle(tag, styleClass); } protected void parse(NPOIFSFileSystem filesystem, XHTMLContentHandler xhtml) throws IOException, SAXException, TikaException { parse(filesystem.getRoot(), xhtml); } protected void parse(DirectoryNode root, XHTMLContentHandler xhtml) throws IOException, SAXException, TikaException { HWPFDocument document; try { document = new HWPFDocument(root); } catch (OldWordFileFormatException e) { parseWord6(root, xhtml); return; } org.apache.poi.hwpf.extractor.WordExtractor wordExtractor = new org.apache.poi.hwpf.extractor.WordExtractor( document); HeaderStories headerFooter = new HeaderStories(document); // Grab the list of pictures. As far as we can tell, // the pictures should be in order, and may be directly // placed or referenced from an anchor PicturesTable pictureTable = document.getPicturesTable(); PicturesSource pictures = new PicturesSource(document); // Do any headers, if present Range[] headers = new Range[] { headerFooter.getFirstHeaderSubrange(), headerFooter.getEvenHeaderSubrange(), headerFooter.getOddHeaderSubrange() }; handleHeaderFooter(headers, "header", document, pictures, pictureTable, xhtml); // Do the main paragraph text Range r = document.getRange(); ListManager listManager = new ListManager(document); for (int i = 0; i < r.numParagraphs(); i++) { Paragraph p = r.getParagraph(i); i += handleParagraph(p, 0, r, document, FieldsDocumentPart.MAIN, pictures, pictureTable, listManager, xhtml); } // Do everything else for (String paragraph : wordExtractor.getMainTextboxText()) { xhtml.element("p", paragraph); } for (String paragraph : wordExtractor.getFootnoteText()) { xhtml.element("p", paragraph); } for (String paragraph : wordExtractor.getCommentsText()) { xhtml.element("p", paragraph); } for (String paragraph : wordExtractor.getEndnoteText()) { xhtml.element("p", paragraph); } // Do any footers, if present Range[] footers = new Range[] { headerFooter.getFirstFooterSubrange(), headerFooter.getEvenFooterSubrange(), headerFooter.getOddFooterSubrange() }; handleHeaderFooter(footers, "footer", document, pictures, pictureTable, xhtml); // Handle any pictures that we haven't output yet for (Picture p = pictures.nextUnclaimed(); p != null;) { handlePictureCharacterRun(null, p, pictures, xhtml); p = pictures.nextUnclaimed(); } // Handle any embeded office documents try { DirectoryEntry op = (DirectoryEntry) root.getEntry("ObjectPool"); for (Entry entry : op) { if (entry.getName().startsWith("_") && entry instanceof DirectoryEntry) { handleEmbeddedOfficeDoc((DirectoryEntry) entry, xhtml); } } } catch (FileNotFoundException e) { } } private void handleHeaderFooter(Range[] ranges, String type, HWPFDocument document, PicturesSource pictures, PicturesTable pictureTable, XHTMLContentHandler xhtml) throws SAXException, IOException, TikaException { if (countParagraphs(ranges) > 0) { xhtml.startElement("div", "class", type); ListManager listManager = new ListManager(document); for (Range r : ranges) { if (r != null) { for (int i = 0; i < r.numParagraphs(); i++) { Paragraph p = r.getParagraph(i); i += handleParagraph(p, 0, r, document, FieldsDocumentPart.HEADER, pictures, pictureTable, listManager, xhtml); } } } xhtml.endElement("div"); } } private int handleParagraph(Paragraph p, int parentTableLevel, Range r, HWPFDocument document, FieldsDocumentPart docPart, PicturesSource pictures, PicturesTable pictureTable, ListManager listManager, 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 if (p.isInTable() && p.getTableLevel() > parentTableLevel && parentTableLevel == 0) { Table t = r.getTable(p); xhtml.startElement("table"); xhtml.startElement("tbody"); for (int rn = 0; rn < t.numRows(); rn++) { TableRow row = t.getRow(rn); xhtml.startElement("tr"); for (int cn = 0; cn < row.numCells(); cn++) { TableCell cell = row.getCell(cn); xhtml.startElement("td"); for (int pn = 0; pn < cell.numParagraphs(); pn++) { Paragraph cellP = cell.getParagraph(pn); handleParagraph(cellP, p.getTableLevel(), cell, document, docPart, pictures, pictureTable, listManager, xhtml); } xhtml.endElement("td"); } xhtml.endElement("tr"); } xhtml.endElement("tbody"); xhtml.endElement("table"); return (t.numParagraphs() - 1); } String text = p.text(); if (text.replaceAll("[\\r\\n\\s]+", "").isEmpty()) { // Skip empty paragraphs return 0; } TagAndStyle tas; String numbering = null; if (document.getStyleSheet().numStyles() > p.getStyleIndex()) { StyleDescription style = document.getStyleSheet().getStyleDescription(p.getStyleIndex()); if (style != null && style.getName() != null && style.getName().length() > 0) { if (p.isInList()) { numbering = listManager.getFormattedNumber(p); } tas = buildParagraphTagAndStyle(style.getName(), (parentTableLevel > 0)); } else { tas = new TagAndStyle("p", null); } } else { tas = new TagAndStyle("p", null); } if (tas.getStyleClass() != null) { xhtml.startElement(tas.getTag(), "class", tas.getStyleClass()); } else { xhtml.startElement(tas.getTag()); } if (numbering != null) { xhtml.characters(numbering); } for (int j = 0; j < p.numCharacterRuns(); j++) { CharacterRun cr = p.getCharacterRun(j); // FIELD_BEGIN_MARK: if (cr.text().getBytes(UTF_8)[0] == 0x13) { Field field = document.getFields().getFieldByStartOffset(docPart, cr.getStartOffset()); // 58 is an embedded document // 56 is a document link if (field != null && (field.getType() == 58 || field.getType() == 56)) { // Embedded Object: add a <div // class="embedded" id="_X"/> so consumer can see where // in the main text each embedded document // occurred: String id = "_" + field.getMarkSeparatorCharacterRun(r).getPicOffset(); AttributesImpl attributes = new AttributesImpl(); attributes.addAttribute("", "class", "class", "CDATA", "embedded"); attributes.addAttribute("", "id", "id", "CDATA", id); xhtml.startElement("div", attributes); xhtml.endElement("div"); } } if (cr.text().equals("\u0013")) { 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.getFor(cr); handlePictureCharacterRun(cr, picture, pictures, xhtml); } else { handleCharacterRun(cr, tas.isHeading(), xhtml); } } // Close any still open style tags if (curStrikeThrough) { xhtml.endElement("s"); curStrikeThrough = false; } if (curItalic) { xhtml.endElement("i"); curItalic = false; } if (curBold) { xhtml.endElement("b"); curBold = false; } xhtml.endElement(tas.getTag()); return 0; } private void handleCharacterRun(CharacterRun cr, boolean skipStyling, XHTMLContentHandler xhtml) throws SAXException { // Skip trailing newlines if (!isRendered(cr) || cr.text().equals("\r")) return; if (!skipStyling) { if (cr.isBold() != curBold) { // Enforce nesting -- must close s and i tags if (curStrikeThrough) { xhtml.endElement("s"); curStrikeThrough = false; } if (curItalic) { xhtml.endElement("i"); curItalic = false; } if (cr.isBold()) { xhtml.startElement("b"); } else { xhtml.endElement("b"); } curBold = cr.isBold(); } if (cr.isItalic() != curItalic) { // Enforce nesting -- must close s tag if (curStrikeThrough) { xhtml.endElement("s"); curStrikeThrough = false; } if (cr.isItalic()) { xhtml.startElement("i"); } else { xhtml.endElement("i"); } curItalic = cr.isItalic(); } if (cr.isStrikeThrough() != curStrikeThrough) { if (cr.isStrikeThrough()) { xhtml.startElement("s"); } else { xhtml.endElement("s"); } curStrikeThrough = cr.isStrikeThrough(); } } // Clean up the text String text = cr.text(); text = text.replace('\r', '\n'); if (text.endsWith("\u0007")) { // Strip the table cell end marker text = text.substring(0, text.length() - 1); } // Copied from POI's org/apache/poi/hwpf/converter/AbstractWordConverter.processCharacters: // Non-breaking hyphens are returned as char 30 text = text.replace((char) 30, UNICODECHAR_NONBREAKING_HYPHEN); // Non-required hyphens to zero-width space text = text.replace((char) 31, UNICODECHAR_ZERO_WIDTH_SPACE); // Control characters as line break text = text.replaceAll("[\u0000-\u001f]", "\n"); xhtml.characters(text); } /** * Can be \13..text..\15 or \13..control..\14..text..\15 . * Nesting is allowed */ 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 + 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, pictures, xhtml); i += increment; } else if (cr.text().equals("\u0014")) { has14 = true; } else if (cr.text().equals("\u0015")) { if (!has14) { texts = controls; controls = new ArrayList<CharacterRun>(); } break; } else { if (has14) { texts.add(cr); } else { controls.add(cr); } } } // Do we need to do something special with this? if (controls.size() > 0) { String text = controls.get(0).text(); for (int j = 1; j < controls.size(); j++) { text += controls.get(j).text(); } if ((text.startsWith("HYPERLINK") || text.startsWith(" HYPERLINK")) && text.indexOf('"') > -1) { int start = text.indexOf('"') + 1; int end = findHyperlinkEnd(text, start); String url = ""; if (start >= 0 && start < end && end <= text.length()) { url = text.substring(start, end); } xhtml.startElement("a", "href", url); for (CharacterRun cr : texts) { handleCharacterRun(cr, skipStyling, xhtml); } xhtml.endElement("a"); } else { // Just output the text ones for (CharacterRun cr : texts) { if (pictures.hasPicture(cr)) { Picture picture = pictures.getFor(cr); handlePictureCharacterRun(cr, picture, pictures, xhtml); } else { handleCharacterRun(cr, skipStyling, xhtml); } } } } else { // We only had text // Output as-is for (CharacterRun cr : texts) { handleCharacterRun(cr, skipStyling, xhtml); } } // Tell them how many to skip over return i - index; } //temporary work around for TIKA-1512 private int findHyperlinkEnd(String text, int start) { int end = text.lastIndexOf('"'); if (end > start) { return end; } end = text.lastIndexOf('\u201D');//smart right double quote if (end > start) { return end; } end = text.lastIndexOf('\r'); if (end > start) { return end; } //if nothing so far, take the full length of the string //If the full string is > 256 characters, it appears //that the url is truncated in the .doc file. This //will return the value as it is in the file, which //may be incorrect; but it is the same behavior as opening //the link in MSWord. //This code does not currently check that length is actually >= 256. //we might want to add that? return text.length(); } private void handlePictureCharacterRun(CharacterRun cr, Picture picture, PicturesSource pictures, XHTMLContentHandler xhtml) throws SAXException, IOException, TikaException { if (!isRendered(cr) || 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 // the picture from the img tag and the embedded resource String filename = "image" + pictureNumber + (extension.length() > 0 ? "." + extension : ""); // Grab the mime type for the picture String mimeType = picture.getMimeType(); // Output the img tag AttributesImpl attr = new AttributesImpl(); attr.addAttribute("", "src", "src", "CDATA", "embedded:" + filename); attr.addAttribute("", "alt", "alt", "CDATA", filename); xhtml.startElement("img", attr); xhtml.endElement("img"); // 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, null, mimeType, xhtml, false); pictures.recordOutput(picture); } } /** * Outputs a section of text if the given text is non-empty. * * @param xhtml XHTML content handler * @param section the class of the <div/> section emitted * @param text text to be emitted, if any * @throws SAXException if an error occurs */ private void addTextIfAny(XHTMLContentHandler xhtml, String section, String text) throws SAXException { if (text != null && text.length() > 0) { xhtml.startElement("div", "class", section); xhtml.element("p", text); xhtml.endElement("div"); } } protected void parseWord6(NPOIFSFileSystem filesystem, XHTMLContentHandler xhtml) throws IOException, SAXException, TikaException { parseWord6(filesystem.getRoot(), xhtml); } protected void parseWord6(DirectoryNode root, XHTMLContentHandler xhtml) throws IOException, SAXException, TikaException { HWPFOldDocument doc = new HWPFOldDocument(root); Word6Extractor extractor = new Word6Extractor(doc); for (String p : extractor.getParagraphText()) { xhtml.element("p", p); } } /** * Determines if character run should be included in the extraction. * * @param cr character run. * @return true if character run should be included in extraction. */ private boolean isRendered(final CharacterRun cr) { return cr == null || !cr.isMarkedDeleted(); } public static class TagAndStyle { private String tag; private String styleClass; public TagAndStyle(String tag, String styleClass) { this.tag = tag; this.styleClass = styleClass; } public String getTag() { return tag; } public String getStyleClass() { return styleClass; } public boolean isHeading() { return tag.length() == 2 && tag.startsWith("h"); } } /** * 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(); // Build the Offset-Picture lookup map lookup = new HashMap<Integer, Picture>(); for (Picture p : all) { lookup.put(p.getStartOffset(), 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; } } }