public void testFunctionalFromTarHierarchy() throws Exception {
    log.debug3("in testFromTarHierarchy");
    // load the tarballs
    InputStream file_input = null;
    try {
      file_input = getResourceAsStream(realTARFile_A);
      // UrlCacher uc = au.makeUrlCacher(TAR_A_BASE);
      // uc.storeContent(file_input, tarHeader);
      UrlCacher uc = tarAu.makeUrlCacher(new UrlData(file_input, tarHeader, TAR_A_BASE));
      uc.storeContent();
      IOUtil.safeClose(file_input);

      file_input = getResourceAsStream(realTARFile_B);
      // uc = au.makeUrlCacher(TAR_B_BASE);
      // uc.storeContent(file_input, tarHeader);
      uc = tarAu.makeUrlCacher(new UrlData(file_input, tarHeader, TAR_B_BASE));
      uc.storeContent();
      IOUtil.safeClose(file_input);

    } catch (IOException e) {
      // TODO Auto-generated catch block
      e.printStackTrace();
    } finally {
      IOUtil.safeClose(file_input);
    }

    CachedUrlSet cus = tarAu.getAuCachedUrlSet();
    for (CachedUrl cu : cus.getCuIterable()) {
      log.debug3("AU - cu is: " + cu.getUrl());
      cu.release();
    }

    // We need to start from the level of the ArticleMetadataExtractor
    MyListEmitter emitter = new MyListEmitter();
    ArticleMetadataExtractor amEx =
        new ElsevierDeferredArticleMetadataExtractor(ArticleFiles.ROLE_ARTICLE_METADATA);

    Iterator<ArticleFiles> it = tarAu.getArticleIterator(MetadataTarget.Any());
    while (it.hasNext()) {
      ArticleFiles af = it.next();
      log.debug3("Metadata test - articlefiles " + af.toString());
      // CachedUrl cu = af.getFullTextCu();
      CachedUrl cu = af.getRoleCu(ArticleFiles.ROLE_ARTICLE_METADATA);
      log.debug3("metadata cu is " + cu.getUrl());
      // List<ArticleMetadata> mdlist = mle.extract(MetadataTarget.Any(), cu);
      amEx.extract(MetadataTarget.Any(), af, emitter);
      List<ArticleMetadata> returnList = emitter.getAmList();

      assertNotNull(returnList);
      log.debug3("size of returnList is " + returnList.size());
      Iterator<ArticleMetadata> mdIt = returnList.iterator();
      ArticleMetadata mdRecord = null;
      while (mdIt.hasNext()) {
        mdRecord = (ArticleMetadata) mdIt.next();
        validateCompleteMetadataRecord(mdRecord);
      }
    }
  }
Example #2
0
 protected void writeCu(CachedUrl cu) throws IOException {
   String url = cu.getUrl();
   long contentSize = cu.getContentSize();
   CIProperties props = cu.getProperties();
   long fetchTime = Long.parseLong(props.getProperty(CachedUrl.PROPERTY_FETCH_TIME));
   InputStream contentIn = cu.getUnfilteredInputStream();
   try {
     if (isResponse) {
       String hdrString = getHttpResponseString(cu);
       long size = contentSize + hdrString.length();
       InputStream headerIn = new ReaderInputStream(new StringReader(hdrString));
       InputStream concat = new SequenceInputStream(headerIn, contentIn);
       try {
         aw.write(xlateFilename(url), cu.getContentType(), getHostIp(), fetchTime, size, concat);
       } finally {
         IOUtil.safeClose(concat);
       }
     } else {
       aw.write(
           xlateFilename(url),
           cu.getContentType(),
           getHostIp(),
           fetchTime,
           cu.getContentSize(),
           contentIn);
     }
   } finally {
     AuUtil.safeRelease(cu);
   }
 }
 private static synchronized String getJavascript() {
   if (jstext == null) {
     InputStream istr = null;
     try {
       ClassLoader loader = Thread.currentThread().getContextClassLoader();
       istr = loader.getResourceAsStream(JAVASCRIPT_RESOURCE);
       jstext = StringUtil.fromInputStream(istr);
       istr.close();
     } catch (Exception e) {
       log.error("Can't load javascript", e);
     } finally {
       IOUtil.safeClose(istr);
     }
   }
   return jstext;
 }
  /** Explode the archive into its constituent elements */
  public void explode() throws CacheException {
    int goodEntries = 0;
    int badEntries = 0;
    int entriesBetweenSleep = 0;
    ArchiveReader arcReader = null;

    logger.debug(
        (storeArchive ? "Storing" : "Fetching") + " WARC file: " + origUrl + " will explode");
    try {
      // Wrap it in an ArchiveReader
      logger.debug3("About to wrap stream");
      arcReader = wrapStream(fetchUrl, arcStream);
      logger.debug3("wrapStream() returns " + (arcReader == null ? "null" : "non-null"));
      // Explode it
      if (arcReader == null) {
        throw new CacheException.ExploderException("no WarcReader for " + origUrl);
      }
      ArchivalUnit au = crawlFacade.getAu();
      logger.debug("Exploding " + fetchUrl);
      // Iterate through the elements in the WARC file, except the first
      Iterator<ArchiveRecord> iter = arcReader.iterator();
      // Skip first record
      if (iter.hasNext()) iter.next();
      while (iter.hasNext()) {
        helper.pokeWDog();
        // check need to pause
        handlePause(++entriesBetweenSleep);
        // handle each element in the archive
        ArchiveRecord element = iter.next();
        // Each element is a URL to be cached in our AU
        ArchiveRecordHeader elementHeader = element.getHeader();
        String elementUrl = elementHeader.getUrl();
        String elementMimeType = elementHeader.getMimetype();
        long elementLength = elementHeader.getLength();
        long elementDate;
        try {
          elementDate = ArchiveUtils.parse14DigitDate(elementHeader.getDate()).getTime();
        } catch (ParseException e) {
          elementDate = 0;
        }
        logger.debug2("WARC url " + elementUrl + " mime " + elementMimeType);
        // add check to determine if this is a url which should be cached
        if (au.shouldBeCached(elementUrl) && elementUrl.startsWith("http:")) {
          ArchiveEntry ae =
              new ArchiveEntry(
                  elementUrl,
                  elementLength,
                  elementDate,
                  element, // ArchiveRecord extends InputStream
                  this,
                  fetchUrl);
          ae.setHeaderFields(makeCIProperties(elementHeader));
          long bytesStored = elementLength;
          logger.debug3("ArchiveEntry: " + ae.getName() + " bytes " + bytesStored);
          try {
            helper.process(ae);
          } catch (PluginException ex) {
            throw new CacheException.ExploderException("helper.process() threw", ex);
          }
          if (ae.getBaseUrl() != null) {
            if (ae.getRestOfUrl() != null && ae.getHeaderFields() != null) {
              storeEntry(ae);
              handleAddText(ae);
              goodEntries++;
              // this needs to use the correct depth ? how
              CrawlUrlData cud = new CrawlUrlData(elementUrl, 0);
              crawlFacade.addToParseQueue(cud);
              crawlFacade.getCrawlerStatus().addContentBytesFetched(bytesStored);
            }
          } else {
            badEntries++;
            logger.debug2("Can't map " + elementUrl + " from " + archiveUrl);
          }
        }
      }
    } catch (IOException ex) {
      throw new CacheException.ExploderException(ex);
    } finally {
      if (arcReader != null) {
        try {
          arcReader.close();
        } catch (IOException ex) {
          throw new CacheException.ExploderException(ex);
        }
      }
      IOUtil.safeClose(arcStream);
    }
    // report failed fetches
    if (badEntries != 0) {
      String msg = archiveUrl + ": " + badEntries + "/" + goodEntries + " bad entries";
      throw new CacheException.UnretryableException(msg);
    }
  }
Example #5
0
  /** Explode the archive into its constituent elements */
  public void explode() throws CacheException {
    CachedUrl cachedUrl = null;
    int goodEntries = 0;
    int badEntries = 0;
    int ignoredEntries = 0;
    int entriesBetweenSleep = 0;
    ArchiveReader arcReader = null;

    logger.info(
        (storeArchive ? "Storing" : "Fetching") + " WARC file: " + origUrl + " will explode");
    try {
      if (storeArchive) {
        UrlCacher uc = au.makeUrlCacher(new UrlData(arcStream, arcProps, fetchUrl));
        BitSet bs = new BitSet();
        bs.set(UrlCacher.DONT_CLOSE_INPUT_STREAM_FLAG);
        uc.setFetchFlags(bs);
        uc.storeContent();
        archiveData.resetInputStream();
        arcStream = archiveData.input;
      }
      // Wrap it in an ArchiveReader
      logger.debug3("About to wrap stream");
      arcReader = wrapStream(fetchUrl, arcStream);
      logger.debug3("wrapStream() returns " + (arcReader == null ? "null" : "non-null"));
      // Explode it
      if (arcReader == null) {
        throw new CacheException.ExploderException("no WarcReader for " + origUrl);
      }
      ArchivalUnit au = crawlFacade.getAu();
      Set stemSet = new HashSet();
      logger.debug("Exploding " + fetchUrl);
      // Iterate through the elements in the WARC file, except the first
      Iterator i = arcReader.iterator();
      // Skip first record
      for (i.next(); i.hasNext(); ) {
        // XXX probably not necessary
        helper.pokeWDog();
        if ((++entriesBetweenSleep % sleepAfter) == 0) {
          long pauseTime =
              CurrentConfig.getTimeIntervalParam(PARAM_RETRY_PAUSE, DEFAULT_RETRY_PAUSE);
          Deadline pause = Deadline.in(pauseTime);
          logger.debug3("Sleeping for " + StringUtil.timeIntervalToString(pauseTime));
          while (!pause.expired()) {
            try {
              pause.sleep();
            } catch (InterruptedException ie) {
              // no action
            }
          }
        }
        ArchiveRecord element = (ArchiveRecord) i.next();
        // Each element is a URL to be cached in a suitable AU
        ArchiveRecordHeader elementHeader = element.getHeader();
        String elementUrl = elementHeader.getUrl();
        String elementMimeType = elementHeader.getMimetype();
        long elementLength = elementHeader.getLength();
        logger.debug2("WARC url " + elementUrl + " mime " + elementMimeType);
        if (elementUrl.startsWith("http:")) {
          ArchiveEntry ae =
              new ArchiveEntry(
                  elementUrl,
                  elementLength,
                  0, // XXX need to convert getDate string to long
                  element, // ArchiveRecord extends InputStream
                  this,
                  fetchUrl);
          ae.setHeaderFields(makeCIProperties(elementHeader));
          long bytesStored = elementLength;
          logger.debug3("ArchiveEntry: " + ae.getName() + " bytes " + bytesStored);
          try {
            helper.process(ae);
          } catch (PluginException ex) {
            throw new CacheException.ExploderException("helper.process() threw", ex);
          }
          if (ae.getBaseUrl() != null) {
            if (ae.getRestOfUrl() != null && ae.getHeaderFields() != null) {
              storeEntry(ae);
              handleAddText(ae);
              goodEntries++;
              crawlFacade.getCrawlerStatus().addContentBytesFetched(bytesStored);
            } else {
              ignoredEntries++;
            }
          } else {
            badEntries++;
            logger.debug2("Can't map " + elementUrl + " from " + archiveUrl);
          }
        }
      }
    } catch (IOException ex) {
      throw new CacheException.ExploderException(ex);
    } finally {
      if (arcReader != null)
        try {
          arcReader.close();
          arcReader = null;
        } catch (IOException ex) {
          throw new CacheException.ExploderException(ex);
        }
      if (cachedUrl != null) {
        cachedUrl.release();
      }
      IOUtil.safeClose(arcStream);
    }
    if (badEntries == 0 && goodEntries > 0) {
      // Make it look like a new crawl finished on each AU to which
      // URLs were added.
      for (Iterator it = touchedAus.iterator(); it.hasNext(); ) {
        ArchivalUnit au = (ArchivalUnit) it.next();
        logger.debug3(archiveUrl + " touching " + au.toString());
        AuUtil.getDaemon(au).getNodeManager(au).newContentCrawlFinished();
      }
    } else {
      ArchivalUnit au = crawlFacade.getAu();
      String msg = archiveUrl + ": " + badEntries + "/" + goodEntries + " bad entries";
      throw new CacheException.UnretryableException(msg);
    }
  }