writerpoolprocessor.java

Heritrix是一个开源,可扩展的web爬虫项目。Heritrix设计成严格按照robots.txt文件的排除指示和META robots标签。

    /**
     * Version of getAttributes that catches and logs exceptions
     * and returns null if failure to fetch the attribute.
     * @param name Attribute name.
     * @return Attribute or null.
     */
    public Object getAttributeUnchecked(String name) {
        Object result = null;
        try {
            result = super.getAttribute(name);
        } catch (AttributeNotFoundException e) {
            logger.warning(e.getLocalizedMessage());
        } catch (MBeanException e) {
            logger.warning(e.getLocalizedMessage());
        } catch (ReflectionException e) {
            logger.warning(e.getLocalizedMessage());
        }
        return result;
    }

   /**
    * Max size we want files to be (bytes).
    *
    * Default is ARCConstants.DEFAULT_MAX_ARC_FILE_SIZE.  Note that ARC
    * files will usually be bigger than maxSize; they'll be maxSize + length
    * to next boundary.
    * @return ARC maximum size.
    */
    public long getMaxSize() {
        Object obj = getAttributeUnchecked(ATTR_MAX_SIZE_BYTES);
        return (obj == null)? DEFAULT_MAX_FILE_SIZE: ((Long)obj).longValue();
    }

    public String getPrefix() {
        Object obj = getAttributeUnchecked(ATTR_PREFIX);
        return (obj == null)? WriterPoolMember.DEFAULT_PREFIX: (String)obj;
    }

    public List<File> getOutputDirs() {
        Object obj = getAttributeUnchecked(ATTR_PATH);
        List list = (obj == null)? Arrays.asList(DEFAULT_PATH): (StringList)obj;
        ArrayList<File> results = new ArrayList<File>();
        for (Iterator i = list.iterator(); i.hasNext();) {
            String path = (String)i.next();
            File f = new File(path);
            if (!f.isAbsolute()) {
                f = new File(getController().getDisk(), path);
            }
            if (!f.exists()) {
                try {
                    f.mkdirs();
                } catch (Exception e) {
                    e.printStackTrace();
                    continue;
                }
            }
            results.add(f);
        }
        return results;
    }
    
    public boolean isCompressed() {
        Object obj = getAttributeUnchecked(ATTR_COMPRESS);
        return (obj == null)? DEFAULT_COMPRESS:
            ((Boolean)obj).booleanValue();
    }

    /**
     * @return Returns the poolMaximumActive.
     */
    public int getPoolMaximumActive() {
        Object obj = getAttributeUnchecked(ATTR_POOL_MAX_ACTIVE);
        return (obj == null)? WriterPool.DEFAULT_MAX_ACTIVE:
            ((Integer)obj).intValue();
    }

    /**
     * @return Returns the poolMaximumWait.
     */
    public int getPoolMaximumWait() {
        Object obj = getAttributeUnchecked(ATTR_POOL_MAX_WAIT);
        return (obj == null)? WriterPool.DEFAULT_MAXIMUM_WAIT:
            ((Integer)obj).intValue();
    }

    public String getSuffix() {
        Object obj = getAttributeUnchecked(ATTR_SUFFIX);
        String sfx = (obj == null)?
            WriterPoolMember.DEFAULT_SUFFIX: (String)obj;
        if (sfx != null && sfx.trim().
                equals(WriterPoolMember.HOSTNAME_VARIABLE)) {
            String str = "localhost.localdomain";
            try {
                str = InetAddress.getLocalHost().getHostName();
            } catch (UnknownHostException ue) {
                logger.severe("Failed getHostAddress for this host: " + ue);
            }
            sfx = str;
        }
        return sfx;
    }
    
    public long getMaxToWrite() {
        Object obj = getAttributeUnchecked(ATTR_MAX_BYTES_WRITTEN);
        return (obj == null)? 0: ((Long)obj).longValue();
    }

	public void crawlEnding(String sExitMessage) {
		this.pool.close();
	}

	public void crawlEnded(String sExitMessage) {
        // sExitMessage is unused.
	}

    /* (non-Javadoc)
     * @see org.archive.crawler.event.CrawlStatusListener#crawlStarted(java.lang.String)
     */
    public void crawlStarted(String message) {
        // TODO Auto-generated method stub
    }
    
    protected String getCheckpointStateFile() {
    	return this.getClass().getName() + ".state";
    }
    
    public void crawlCheckpoint(File checkpointDir) throws IOException {
        int serial = getSerialNo().get();
        if (this.pool.getNumActive() > 0) {
            // If we have open active Archive files, up the serial number
            // so after checkpoint, we start at one past current number and
            // so the number we serialize, is one past current serialNo.
            // All this serial number manipulation should be fine in here since
            // we're paused checkpointing (Revisit if this assumption changes).
            serial = getSerialNo().incrementAndGet();
        }
        saveCheckpointSerialNumber(checkpointDir, serial);
        // Close all ARCs on checkpoint.
        try {
            this.pool.close();
        } finally {
            // Reopen on checkpoint.
            setupPool(new AtomicInteger(serial));
        }
    }
    
	public void crawlPausing(String statusMessage) {
        // sExitMessage is unused.
	}

	public void crawlPaused(String statusMessage) {
        // sExitMessage is unused.
	}

	public void crawlResuming(String statusMessage) {
        // sExitMessage is unused.
	}
	
    private void readObject(ObjectInputStream stream)
    throws IOException, ClassNotFoundException {
        stream.defaultReadObject();
        ObjectPlusFilesInputStream coistream =
            (ObjectPlusFilesInputStream)stream;
        coistream.registerFinishTask( new Runnable() {
            public void run() {
            	setupPool(new AtomicInteger());
            }
        });
    }

	protected WriterPool getPool() {
		return pool;
	}

	protected void setPool(WriterPool pool) {
		this.pool = pool;
	}

	protected long getTotalBytesWritten() {
		return totalBytesWritten;
	}

	protected void setTotalBytesWritten(long totalBytesWritten) {
        this.totalBytesWritten = totalBytesWritten;
    }
	
    /**
     * Called out of {@link #initialTasks()} when recovering a checkpoint.
     * Restore state.
     */
    protected void checkpointRecover() {
        int serialNo = loadCheckpointSerialNumber();
        if (serialNo != -1) {
            getSerialNo().set(serialNo);
        }
    }

    /**
     * @return Serial number from checkpoint state file or if unreadable, -1
     * (Client should check for -1).
     */
    protected int loadCheckpointSerialNumber() {
        int result = -1;
        
        // If in recover mode, read in the Writer serial number saved
        // off when we checkpointed.
        File stateFile = new File(getSettingsHandler().getOrder()
                .getController().getCheckpointRecover().getDirectory(),
                getCheckpointStateFile());
        if (!stateFile.exists()) {
            logger.info(stateFile.getAbsolutePath()
                    + " doesn't exist so cannot restore Writer serial number.");
        } else {
            DataInputStream dis = null;
            try {
                dis = new DataInputStream(new FileInputStream(stateFile));
                result = dis.readShort();
            } catch (FileNotFoundException e) {
                e.printStackTrace();
            } catch (IOException e) {
                e.printStackTrace();
            } finally {
                try {
                    if (dis != null) {
                        dis.close();
                    }
                } catch (IOException e) {
                    e.printStackTrace();
                }
            }
        }
        return result;
    }
    
    protected void saveCheckpointSerialNumber(final File checkpointDir,
            final int serialNo)
    throws IOException {
        // Write out the current state of the ARCWriter serial number.
        File f = new File(checkpointDir, getCheckpointStateFile());
        DataOutputStream dos = new DataOutputStream(new FileOutputStream(f));
        try {
            dos.writeShort(serialNo);
        } finally {
            dos.close();
        }
    }
    
    /**
     * Return list of metadatas to add to first arc file metadata record.
     * 
     * Default is to stylesheet the order file.  To specify stylesheet,
     * override {@link #getFirstrecordStylesheet()}.
     *
     * Get xml files from settingshandler.  Currently order file is the
     * only xml file.  We're NOT adding seeds to meta data.
     *
     * @return List of strings and/or files to add to arc file as metadata or
     * null.
     */
    public synchronized List<String> getMetadata() {
        if (this.cachedMetadata != null) {
            return this.cachedMetadata;
        }
        return cacheMetadata();
    }
    
    protected synchronized List<String> cacheMetadata() {
        if (this.cachedMetadata != null) {
            return this.cachedMetadata;
        }
        
        // If no stylesheet, return empty metadata.
        if (getFirstrecordStylesheet() == null ||
                getFirstrecordStylesheet().length() == 0) {
            this.cachedMetadata = new ArrayList<String>(1);
            this.cachedMetadata.add("");
            return this.cachedMetadata;
        }
        
        List<String> result = null;
        if (!XMLSettingsHandler.class.isInstance(getSettingsHandler())) {
            logger.warning("Expected xml settings handler (No warcinfo).");
            // Early return
            return result;
        }
        
        XMLSettingsHandler xsh = (XMLSettingsHandler)getSettingsHandler();
        File orderFile = xsh.getOrderFile();
        if (!orderFile.exists() || !orderFile.canRead()) {
                logger.severe("File " + orderFile.getAbsolutePath() +
                    " is does not exist or is not readable.");
        } else {
            result = new ArrayList<String>(1);
            result.add(getFirstrecordBody(orderFile));
        }
        this.cachedMetadata = result;
        return this.cachedMetadata;
    }
    
    /**
     * @preturn Full path to stylesheet (Its read off the CLASSPATH
     * as resource).
     */
    protected String getFirstrecordStylesheet() {
        return null;
    }

    /**
     * Write the arc metadata body content.
     *
     * Its based on the order xml file but into this base we'll add other info
     * such as machine ip.
     *
     * @param orderFile Order file.

     *
     * @return String that holds the arc metaheader body.
     */
    protected String getFirstrecordBody(File orderFile) {
        String result = null;
        TransformerFactory factory = TransformerFactory.newInstance();
        Templates templates = null;
        Transformer xformer = null;
        try {
            templates = factory.newTemplates(new StreamSource(
                this.getClass().getResourceAsStream(getFirstrecordStylesheet())));
            xformer = templates.newTransformer();
            // Below parameter names must match what is in the stylesheet.
            xformer.setParameter("software", "Heritrix " +
                Heritrix.getVersion() + " http://crawler.archive.org");
            xformer.setParameter("ip",
                InetAddress.getLocalHost().getHostAddress());
            xformer.setParameter("hostname",
                InetAddress.getLocalHost().getHostName());
            StreamSource source = new StreamSource(
                new FileInputStream(orderFile));
            StringWriter writer = new StringWriter();
            StreamResult target = new StreamResult(writer);
            xformer.transform(source, target);
            result= writer.toString();
        } catch (TransformerConfigurationException e) {
            logger.severe("Failed transform " + e);
        } catch (FileNotFoundException e) {
            logger.severe("Failed transform, file not found " + e);
        } catch (UnknownHostException e) {
            logger.severe("Failed transform, unknown host " + e);
        } catch(TransformerException e) {
            SourceLocator locator = e.getLocator();
            int col = locator.getColumnNumber();
            int line = locator.getLineNumber();
            String publicId = locator.getPublicId();
            String systemId = locator.getSystemId();
            logger.severe("Transform error " + e + ", col " + col + ", line " +
                line + ", publicId " + publicId + ", systemId " + systemId);
        }

        return result;
    }
}