Examples with Word6Extractor org.apache.poi.hwpf.extractor.Word6Extractor used on opensource projects

Example 1 with Word6Extractor

use of org.apache.poi.hwpf.extractor.Word6Extractor in project poi by apache.

the class TestExtractorFactory method testOPOIFS.

@Test
public void testOPOIFS() throws Exception {
    // Excel
    assertTrue(ExtractorFactory.createExtractor(new OPOIFSFileSystem(new FileInputStream(xls))) instanceof ExcelExtractor);
    assertTrue(ExtractorFactory.createExtractor(new OPOIFSFileSystem(new FileInputStream(xls))).getText().length() > 200);
    // Word
    assertTrue(ExtractorFactory.createExtractor(new OPOIFSFileSystem(new FileInputStream(doc))) instanceof WordExtractor);
    assertTrue(ExtractorFactory.createExtractor(new OPOIFSFileSystem(new FileInputStream(doc))).getText().length() > 120);
    assertTrue(ExtractorFactory.createExtractor(new OPOIFSFileSystem(new FileInputStream(doc6))) instanceof Word6Extractor);
    assertTrue(ExtractorFactory.createExtractor(new OPOIFSFileSystem(new FileInputStream(doc6))).getText().length() > 20);
    assertTrue(ExtractorFactory.createExtractor(new OPOIFSFileSystem(new FileInputStream(doc95))) instanceof Word6Extractor);
    assertTrue(ExtractorFactory.createExtractor(new OPOIFSFileSystem(new FileInputStream(doc95))).getText().length() > 120);
    // PowerPoint
    assertTrue(ExtractorFactory.createExtractor(new OPOIFSFileSystem(new FileInputStream(ppt))) instanceof PowerPointExtractor);
    assertTrue(ExtractorFactory.createExtractor(new OPOIFSFileSystem(new FileInputStream(ppt))).getText().length() > 120);
    // Visio
    assertTrue(ExtractorFactory.createExtractor(new OPOIFSFileSystem(new FileInputStream(vsd))) instanceof VisioTextExtractor);
    assertTrue(ExtractorFactory.createExtractor(new OPOIFSFileSystem(new FileInputStream(vsd))).getText().length() > 50);
    // Publisher
    assertTrue(ExtractorFactory.createExtractor(new OPOIFSFileSystem(new FileInputStream(pub))) instanceof PublisherTextExtractor);
    assertTrue(ExtractorFactory.createExtractor(new OPOIFSFileSystem(new FileInputStream(pub))).getText().length() > 50);
    // Outlook msg
    assertTrue(ExtractorFactory.createExtractor(new OPOIFSFileSystem(new FileInputStream(msg))) instanceof OutlookTextExtactor);
    assertTrue(ExtractorFactory.createExtractor(new OPOIFSFileSystem(new FileInputStream(msg))).getText().length() > 50);
    // Text
    try {
        ExtractorFactory.createExtractor(new OPOIFSFileSystem(new FileInputStream(txt)));
        fail();
    } catch (IOException e) {
    // Good
    }
}

Example 2 with Word6Extractor

use of org.apache.poi.hwpf.extractor.Word6Extractor in project poi by apache.

the class TestExtractorFactory method testInputStream.

@Test
public void testInputStream() throws Exception {
    // Excel
    POITextExtractor extractor = ExtractorFactory.createExtractor(new FileInputStream(xls));
    assertTrue(extractor instanceof ExcelExtractor);
    assertTrue(extractor.getText().length() > 200);
    extractor.close();
    extractor = ExtractorFactory.createExtractor(new FileInputStream(xlsx));
    assertTrue(extractor.getClass().getName(), extractor instanceof XSSFExcelExtractor);
    assertTrue(extractor.getText().length() > 200);
    // TODO Support OOXML-Strict, see bug #57699
    //        assertTrue(
    //                ExtractorFactory.createExtractor(new FileInputStream(xlsxStrict))
    //                instanceof XSSFExcelExtractor
    //        );
    //        assertTrue(
    //                ExtractorFactory.createExtractor(new FileInputStream(xlsxStrict)).getText().length() > 200
    //        );
    extractor.close();
    // Word
    extractor = ExtractorFactory.createExtractor(new FileInputStream(doc));
    assertTrue(extractor.getClass().getName(), extractor instanceof WordExtractor);
    assertTrue(extractor.getText().length() > 120);
    extractor.close();
    extractor = ExtractorFactory.createExtractor(new FileInputStream(doc6));
    assertTrue(extractor.getClass().getName(), extractor instanceof Word6Extractor);
    assertTrue(extractor.getText().length() > 20);
    extractor.close();
    extractor = ExtractorFactory.createExtractor(new FileInputStream(doc95));
    assertTrue(extractor.getClass().getName(), extractor instanceof Word6Extractor);
    assertTrue(extractor.getText().length() > 120);
    extractor.close();
    extractor = ExtractorFactory.createExtractor(new FileInputStream(docx));
    assertTrue(extractor instanceof XWPFWordExtractor);
    assertTrue(extractor.getText().length() > 120);
    extractor.close();
    // PowerPoint
    extractor = ExtractorFactory.createExtractor(new FileInputStream(ppt));
    assertTrue(extractor instanceof PowerPointExtractor);
    assertTrue(extractor.getText().length() > 120);
    extractor.close();
    extractor = ExtractorFactory.createExtractor(new FileInputStream(pptx));
    assertTrue(extractor instanceof XSLFPowerPointExtractor);
    assertTrue(extractor.getText().length() > 120);
    extractor.close();
    // Visio
    extractor = ExtractorFactory.createExtractor(new FileInputStream(vsd));
    assertTrue(extractor instanceof VisioTextExtractor);
    assertTrue(extractor.getText().length() > 50);
    extractor.close();
    // Visio - vsdx
    extractor = ExtractorFactory.createExtractor(new FileInputStream(vsdx));
    assertTrue(extractor instanceof XDGFVisioExtractor);
    assertTrue(extractor.getText().length() > 20);
    extractor.close();
    // Publisher
    extractor = ExtractorFactory.createExtractor(new FileInputStream(pub));
    assertTrue(extractor instanceof PublisherTextExtractor);
    assertTrue(extractor.getText().length() > 50);
    extractor.close();
    // Outlook msg
    extractor = ExtractorFactory.createExtractor(new FileInputStream(msg));
    assertTrue(extractor instanceof OutlookTextExtactor);
    assertTrue(extractor.getText().length() > 50);
    extractor.close();
    // Text
    try {
        FileInputStream stream = new FileInputStream(txt);
        try {
            ExtractorFactory.createExtractor(stream);
            fail();
        } finally {
            IOUtils.closeQuietly(stream);
        }
    } catch (IllegalArgumentException e) {
    // Good
    }
}

Example 3 with Word6Extractor

use of org.apache.poi.hwpf.extractor.Word6Extractor in project poi by apache.

the class TestHWPFOldDocument method testCodePageBug50955.

@Test
public void testCodePageBug50955() throws IOException {
    //windows 1251
    HWPFOldDocument doc = HWPFTestDataSamples.openOldSampleFile("Bug50955.doc");
    Word6Extractor ex = new Word6Extractor(doc);
    StringBuilder sb = new StringBuilder();
    for (String p : ex.getParagraphText()) {
        sb.append(p);
    }
    //Greetings!
    assertContains(sb.toString(), "привет");
    ex.close();
    doc.close();
}

Example 4 with Word6Extractor

use of org.apache.poi.hwpf.extractor.Word6Extractor in project poi by apache.

the class TestHWPFOldDocument method testDefaultCodePageEncoding.

@Test
public void testDefaultCodePageEncoding() throws IOException {
    HWPFOldDocument doc = HWPFTestDataSamples.openOldSampleFile("Bug60942.doc");
    Word6Extractor ex = new Word6Extractor(doc);
    String txt = ex.getText();
    assertContains(txt, "BERTHOD");
    assertContains(txt, "APPLICOLOR");
    assertContains(txt, "les meilleurs");
    assertContains(txt, "GUY LECOLE");
    ex.close();
    doc.close();
}

Example 5 with Word6Extractor

use of org.apache.poi.hwpf.extractor.Word6Extractor in project tika by apache.

the class WordExtractor method parseWord6.

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);
    }
}