Examples with ContentHandler org.xml.sax.ContentHandler used on opensource projects

Example 96 with ContentHandler

use of org.xml.sax.ContentHandler in project tika by apache.

the class PowerPointParserTest method testMasterFooter.

@Test
public void testMasterFooter() throws Exception {
    ContentHandler handler = new BodyContentHandler();
    Metadata metadata = new Metadata();
    try (InputStream stream = PowerPointParserTest.class.getResourceAsStream("/test-documents/testPPT_masterFooter.ppt")) {
        new OfficeParser().parse(stream, handler, metadata, new ParseContext());
    }
    String content = handler.toString();
    assertContains("Master footer is here", content);
    // Make sure boilerplate text didn't come through:
    assertEquals(-1, content.indexOf("Click to edit Master"));
    //TIKA-1171
    assertEquals(-1, content.indexOf("*"));
}

Example 97 with ContentHandler

use of org.xml.sax.ContentHandler in project tika by apache.

the class PublisherParserTest method testPublisherParser.

@Test
public void testPublisherParser() throws Exception {
    try (InputStream input = PublisherParserTest.class.getResourceAsStream("/test-documents/testPUBLISHER.pub")) {
        Metadata metadata = new Metadata();
        ContentHandler handler = new BodyContentHandler();
        new OfficeParser().parse(input, handler, metadata, new ParseContext());
        assertEquals("application/x-mspublisher", metadata.get(Metadata.CONTENT_TYPE));
        assertEquals(null, metadata.get(TikaCoreProperties.TITLE));
        assertEquals("Nick Burch", metadata.get(TikaCoreProperties.CREATOR));
        assertEquals("Nick Burch", metadata.get(Metadata.AUTHOR));
        String content = handler.toString();
        assertContains("0123456789", content);
        assertContains("abcdef", content);
    }
}

Example 98 with ContentHandler

use of org.xml.sax.ContentHandler in project tika by apache.

the class ForkParserTest method testParallelParsing.

@Test
public void testParallelParsing() throws Exception {
    final ForkParser parser = new ForkParser(ForkParserTest.class.getClassLoader(), new ForkTestParser());
    try {
        final ParseContext context = new ParseContext();
        Thread[] threads = new Thread[10];
        ContentHandler[] output = new ContentHandler[threads.length];
        for (int i = 0; i < threads.length; i++) {
            final ContentHandler o = new BodyContentHandler();
            output[i] = o;
            threads[i] = new Thread() {

                public void run() {
                    try {
                        InputStream stream = new ByteArrayInputStream(new byte[0]);
                        parser.parse(stream, o, new Metadata(), context);
                    } catch (Exception e) {
                        e.printStackTrace();
                    }
                }
            };
            threads[i].start();
        }
        for (int i = 0; i < threads.length; i++) {
            threads[i].join();
            assertEquals("Hello, World!", output[i].toString().trim());
        }
    } finally {
        parser.close();
    }
}

Example 99 with ContentHandler

use of org.xml.sax.ContentHandler in project tika by apache.

the class ForkParserTest method testPoolSizeReached.

@Test
public void testPoolSizeReached() throws Exception {
    final ForkParser parser = new ForkParser(ForkParserTest.class.getClassLoader(), new ForkTestParser());
    try {
        final Semaphore barrier = new Semaphore(0);
        Thread[] threads = new Thread[parser.getPoolSize()];
        PipedOutputStream[] pipes = new PipedOutputStream[threads.length];
        final ParseContext context = new ParseContext();
        for (int i = 0; i < threads.length; i++) {
            final PipedInputStream input = new PipedInputStream() {

                @Override
                public synchronized int read() throws IOException {
                    barrier.release();
                    return super.read();
                }
            };
            pipes[i] = new PipedOutputStream(input);
            threads[i] = new Thread() {

                public void run() {
                    try {
                        ContentHandler o = new DefaultHandler();
                        parser.parse(input, o, new Metadata(), context);
                    } catch (Exception e) {
                        e.printStackTrace();
                    }
                }
            };
            threads[i].start();
        }
        // Wait until all the background parsers have been started
        barrier.acquire(parser.getPoolSize());
        final ContentHandler o = new BodyContentHandler();
        Thread blocked = new Thread() {

            public void run() {
                try {
                    barrier.release();
                    InputStream stream = new ByteArrayInputStream(new byte[0]);
                    parser.parse(stream, o, new Metadata(), context);
                } catch (Exception e) {
                    e.printStackTrace();
                }
            }
        };
        blocked.start();
        // Wait until the last thread is started, and then some to
        // make sure that it would have had a chance to start processing
        // data had it not been blocked.
        barrier.acquire();
        Thread.sleep(1000);
        assertEquals("", o.toString());
        for (int i = 0; i < threads.length; i++) {
            pipes[i].close();
            threads[i].join();
        }
        blocked.join();
        assertEquals("Hello, World!", o.toString().trim());
    } finally {
        parser.close();
    }
}

Example 100 with ContentHandler

use of org.xml.sax.ContentHandler in project tika by apache.

the class BasicContentHandlerFactoryTest method testBody.

@Test
public void testBody() throws Exception {
    Parser p = new MockParser(OVER_DEFAULT);
    BasicContentHandlerFactory.HANDLER_TYPE type = BasicContentHandlerFactory.HANDLER_TYPE.BODY;
    ContentHandler handler = new BasicContentHandlerFactory(type, -1).getNewContentHandler();
    assertTrue(handler instanceof BodyContentHandler);
    p.parse(null, handler, null, null);
    String extracted = handler.toString();
    assertNotContains("title", extracted);
    assertContains("aaaaaaaaaa", extracted);
    assertTrue(extracted.length() > 110000);
    //now test write limit
    p = new MockParser(10);
    handler = new BasicContentHandlerFactory(type, 5).getNewContentHandler();
    assertTrue(handler instanceof BodyContentHandler);
    assertWriteLimitReached(p, (BodyContentHandler) handler);
    extracted = handler.toString();
    assertNotContains("This ", extracted);
    assertContains("aaaa", extracted);
    //now test outputstream call
    p = new MockParser(OVER_DEFAULT);
    ByteArrayOutputStream os = new ByteArrayOutputStream();
    handler = new BasicContentHandlerFactory(type, -1).getNewContentHandler(os, ENCODING);
    assertTrue(handler instanceof BodyContentHandler);
    p.parse(null, handler, null, null);
    assertNotContains("title", os.toByteArray());
    assertContains("aaaaaaaaaa", os.toByteArray());
    assertNotContains("<body", os.toByteArray());
    assertNotContains("<html", os.toByteArray());
    assertTrue(os.toByteArray().length > 110000);
    p = new MockParser(10);
    os = new ByteArrayOutputStream();
    handler = new BasicContentHandlerFactory(type, 5).getNewContentHandler(os, ENCODING);
    assertTrue(handler instanceof WriteOutContentHandler);
    assertWriteLimitReached(p, (WriteOutContentHandler) handler);
    assertEquals(0, os.toByteArray().length);
}