Examples with ResponseData org.codelibs.fess.crawler.entity.ResponseData used on opensource projects

Example 11 with ResponseData

use of org.codelibs.fess.crawler.entity.ResponseData in project fess by codelibs.

the class FessCrawlerThread method isContentUpdated.

@Override
protected boolean isContentUpdated(final CrawlerClient client, final UrlQueue<?> urlQueue) {
    if (ComponentUtil.getFessConfig().isIncrementalCrawling()) {
        final long startTime = System.currentTimeMillis();
        final FessConfig fessConfig = ComponentUtil.getFessConfig();
        final CrawlingConfigHelper crawlingConfigHelper = ComponentUtil.getCrawlingConfigHelper();
        final CrawlingInfoHelper crawlingInfoHelper = ComponentUtil.getCrawlingInfoHelper();
        final SambaHelper sambaHelper = ComponentUtil.getSambaHelper();
        final IndexingHelper indexingHelper = ComponentUtil.getIndexingHelper();
        final FessEsClient fessEsClient = ComponentUtil.getFessEsClient();
        final String url = urlQueue.getUrl();
        ResponseData responseData = null;
        try {
            final CrawlingConfig crawlingConfig = crawlingConfigHelper.get(crawlerContext.getSessionId());
            final Map<String, Object> dataMap = new HashMap<>();
            dataMap.put(fessConfig.getIndexFieldUrl(), url);
            final List<String> roleTypeList = new ArrayList<>();
            stream(crawlingConfig.getPermissions()).of(stream -> stream.forEach(p -> roleTypeList.add(p)));
            if (url.startsWith("smb://")) {
                if (url.endsWith("/")) {
                    // directory
                    return true;
                }
                if (fessConfig.isSmbRoleFromFile()) {
                    // head method
                    responseData = client.execute(RequestDataBuilder.newRequestData().head().url(url).build());
                    if (responseData == null) {
                        return true;
                    }
                    final ACE[] aces = (ACE[]) responseData.getMetaDataMap().get(SmbClient.SMB_ACCESS_CONTROL_ENTRIES);
                    if (aces != null) {
                        for (final ACE item : aces) {
                            final SID sid = item.getSID();
                            final String accountId = sambaHelper.getAccountId(sid);
                            if (accountId != null) {
                                roleTypeList.add(accountId);
                            }
                        }
                        if (logger.isDebugEnabled()) {
                            logger.debug("smbUrl:" + responseData.getUrl() + " roleType:" + roleTypeList.toString());
                        }
                    }
                }
            }
            dataMap.put(fessConfig.getIndexFieldRole(), roleTypeList);
            final String id = crawlingInfoHelper.generateId(dataMap);
            if (logger.isDebugEnabled()) {
                logger.debug("Searching indexed document: " + id);
            }
            final Map<String, Object> document = indexingHelper.getDocument(fessEsClient, id, new String[] { fessConfig.getIndexFieldId(), fessConfig.getIndexFieldLastModified(), fessConfig.getIndexFieldAnchor(), fessConfig.getIndexFieldSegment(), fessConfig.getIndexFieldExpires(), fessConfig.getIndexFieldClickCount(), fessConfig.getIndexFieldFavoriteCount() });
            if (document == null) {
                storeChildUrlsToQueue(urlQueue, getChildUrlSet(fessEsClient, id));
                return true;
            }
            final Date expires = DocumentUtil.getValue(document, fessConfig.getIndexFieldExpires(), Date.class);
            if (expires != null && expires.getTime() < System.currentTimeMillis()) {
                final Object idValue = document.get(fessConfig.getIndexFieldId());
                if (idValue != null && !indexingHelper.deleteDocument(fessEsClient, idValue.toString())) {
                    logger.debug("Failed to delete expired document: " + url);
                }
                return true;
            }
            final Date lastModified = DocumentUtil.getValue(document, fessConfig.getIndexFieldLastModified(), Date.class);
            if (lastModified == null) {
                return true;
            }
            urlQueue.setLastModified(lastModified.getTime());
            log(logHelper, LogType.CHECK_LAST_MODIFIED, crawlerContext, urlQueue);
            if (responseData == null) {
                // head method
                responseData = client.execute(RequestDataBuilder.newRequestData().head().url(url).build());
                if (responseData == null) {
                    return true;
                }
            }
            final int httpStatusCode = responseData.getHttpStatusCode();
            if (logger.isDebugEnabled()) {
                logger.debug("Accessing document: " + url + ", status: " + httpStatusCode);
            }
            if (httpStatusCode == 404) {
                storeChildUrlsToQueue(urlQueue, getAnchorSet(document.get(fessConfig.getIndexFieldAnchor())));
                if (!indexingHelper.deleteDocument(fessEsClient, id)) {
                    logger.debug("Failed to delete 404 document: " + url);
                }
                return false;
            } else if (responseData.getLastModified() == null) {
                return true;
            } else if (responseData.getLastModified().getTime() <= lastModified.getTime() && httpStatusCode == 200) {
                log(logHelper, LogType.NOT_MODIFIED, crawlerContext, urlQueue);
                responseData.setExecutionTime(System.currentTimeMillis() - startTime);
                responseData.setParentUrl(urlQueue.getParentUrl());
                responseData.setSessionId(crawlerContext.getSessionId());
                responseData.setHttpStatusCode(org.codelibs.fess.crawler.Constants.NOT_MODIFIED_STATUS);
                processResponse(urlQueue, responseData);
                storeChildUrlsToQueue(urlQueue, getAnchorSet(document.get(fessConfig.getIndexFieldAnchor())));
                final Date documentExpires = crawlingInfoHelper.getDocumentExpires(crawlingConfig);
                if (documentExpires != null && !indexingHelper.updateDocument(fessEsClient, id, fessConfig.getIndexFieldExpires(), documentExpires)) {
                    logger.debug("Failed to update " + fessConfig.getIndexFieldExpires() + " at " + url);
                }
                return false;
            }
        } finally {
            if (responseData != null) {
                IOUtils.closeQuietly(responseData);
            }
        }
    }
    return true;
}

Example 12 with ResponseData

use of org.codelibs.fess.crawler.entity.ResponseData in project fess by codelibs.

the class AbstractFessFileTransformer method generateData.

protected Map<String, Object> generateData(final ResponseData responseData) {
    final Extractor extractor = getExtractor(responseData);
    final Map<String, String> params = new HashMap<>();
    params.put(TikaMetadataKeys.RESOURCE_NAME_KEY, getResourceName(responseData));
    final String mimeType = responseData.getMimeType();
    params.put(HttpHeaders.CONTENT_TYPE, mimeType);
    params.put(HttpHeaders.CONTENT_ENCODING, responseData.getCharSet());
    final StringBuilder contentMetaBuf = new StringBuilder(1000);
    final Map<String, Object> dataMap = new HashMap<>();
    final Map<String, Object> metaDataMap = new HashMap<>();
    String content;
    try (final InputStream in = responseData.getResponseBody()) {
        final ExtractData extractData = getExtractData(extractor, in, params);
        content = extractData.getContent();
        if (fessConfig.isCrawlerDocumentFileIgnoreEmptyContent() && StringUtil.isBlank(content)) {
            return null;
        }
        if (getLogger().isDebugEnabled()) {
            getLogger().debug("ExtractData: " + extractData);
        }
        // meta
        //
        extractData.getKeySet().stream().filter(//
        k -> extractData.getValues(k) != null).forEach(key -> {
            final String[] values = extractData.getValues(key);
            metaDataMap.put(key, values);
            if (fessConfig.isCrawlerMetadataContentIncluded(key)) {
                final String joinedValue = StringUtils.join(values, ' ');
                if (StringUtil.isNotBlank(joinedValue)) {
                    if (contentMetaBuf.length() > 0) {
                        contentMetaBuf.append(' ');
                    }
                    contentMetaBuf.append(joinedValue.trim());
                }
            }
            final Pair<String, String> mapping = fessConfig.getCrawlerMetadataNameMapping(key);
            if (mapping != null) {
                if (Constants.MAPPING_TYPE_ARRAY.equalsIgnoreCase(mapping.getSecond())) {
                    dataMap.put(mapping.getFirst(), values);
                } else if (Constants.MAPPING_TYPE_STRING.equalsIgnoreCase(mapping.getSecond())) {
                    final String joinedValue = StringUtils.join(values, ' ');
                    dataMap.put(mapping.getFirst(), joinedValue.trim());
                } else if (values.length == 1) {
                    try {
                        if (Constants.MAPPING_TYPE_LONG.equalsIgnoreCase(mapping.getSecond())) {
                            dataMap.put(mapping.getFirst(), Long.parseLong(values[0]));
                        } else if (Constants.MAPPING_TYPE_DOUBLE.equalsIgnoreCase(mapping.getSecond())) {
                            dataMap.put(mapping.getFirst(), Double.parseDouble(values[0]));
                        } else {
                            logger.warn("Unknown mapping type: {}={}", key, mapping);
                        }
                    } catch (final NumberFormatException e) {
                        logger.warn("Failed to parse " + values[0], e);
                    }
                }
            }
        });
    } catch (final Exception e) {
        final CrawlingAccessException rcae = new CrawlingAccessException("Could not get a text from " + responseData.getUrl(), e);
        rcae.setLogLevel(CrawlingAccessException.WARN);
        throw rcae;
    }
    if (content == null) {
        content = StringUtil.EMPTY;
    }
    final String contentMeta = contentMetaBuf.toString().trim();
    final FessConfig fessConfig = ComponentUtil.getFessConfig();
    final CrawlingInfoHelper crawlingInfoHelper = ComponentUtil.getCrawlingInfoHelper();
    final String sessionId = crawlingInfoHelper.getCanonicalSessionId(responseData.getSessionId());
    final PathMappingHelper pathMappingHelper = ComponentUtil.getPathMappingHelper();
    final CrawlingConfigHelper crawlingConfigHelper = ComponentUtil.getCrawlingConfigHelper();
    final CrawlingConfig crawlingConfig = crawlingConfigHelper.get(responseData.getSessionId());
    final Date documentExpires = crawlingInfoHelper.getDocumentExpires(crawlingConfig);
    final SystemHelper systemHelper = ComponentUtil.getSystemHelper();
    final FileTypeHelper fileTypeHelper = ComponentUtil.getFileTypeHelper();
    final DocumentHelper documentHelper = ComponentUtil.getDocumentHelper();
    String url = responseData.getUrl();
    final String indexingTarget = crawlingConfig.getIndexingTarget(url);
    url = pathMappingHelper.replaceUrl(sessionId, url);
    final Map<String, String> fieldConfigMap = crawlingConfig.getConfigParameterMap(ConfigName.FIELD);
    String urlEncoding;
    final UrlQueue<?> urlQueue = CrawlingParameterUtil.getUrlQueue();
    if (urlQueue != null && urlQueue.getEncoding() != null) {
        urlEncoding = urlQueue.getEncoding();
    } else {
        urlEncoding = responseData.getCharSet();
    }
    // cid
    final String configId = crawlingConfig.getConfigId();
    if (configId != null) {
        putResultDataBody(dataMap, fessConfig.getIndexFieldConfigId(), configId);
    }
    //  expires
    if (documentExpires != null) {
        putResultDataBody(dataMap, fessConfig.getIndexFieldExpires(), documentExpires);
    }
    // segment
    putResultDataBody(dataMap, fessConfig.getIndexFieldSegment(), sessionId);
    // content
    final StringBuilder buf = new StringBuilder(content.length() + 1000);
    if (fessConfig.isCrawlerDocumentFileAppendBodyContent()) {
        buf.append(content);
    }
    if (fessConfig.isCrawlerDocumentFileAppendMetaContent()) {
        if (buf.length() > 0) {
            buf.append(' ');
        }
        buf.append(contentMeta);
    }
    final String bodyBase = buf.toString().trim();
    final String body = documentHelper.getContent(responseData, bodyBase, dataMap);
    putResultDataBody(dataMap, fessConfig.getIndexFieldContent(), body);
    if ((Constants.TRUE.equalsIgnoreCase(fieldConfigMap.get(fessConfig.getIndexFieldCache())) || fessConfig.isCrawlerDocumentCacheEnabled()) && fessConfig.isSupportedDocumentCacheMimetypes(mimeType)) {
        if (responseData.getContentLength() > 0 && responseData.getContentLength() <= fessConfig.getCrawlerDocumentCacheMaxSizeAsInteger().longValue()) {
            final String cache = content.trim().replaceAll("[ \\t\\x0B\\f]+", " ");
            // text cache
            putResultDataBody(dataMap, fessConfig.getIndexFieldCache(), cache);
            putResultDataBody(dataMap, fessConfig.getIndexFieldHasCache(), Constants.TRUE);
        }
    }
    // digest
    putResultDataBody(dataMap, fessConfig.getIndexFieldDigest(), documentHelper.getDigest(responseData, bodyBase, dataMap, fessConfig.getCrawlerDocumentFileMaxDigestLengthAsInteger()));
    // title
    final String fileName = getFileName(url, urlEncoding);
    if (!dataMap.containsKey(fessConfig.getIndexFieldTitle())) {
        if (url.endsWith("/")) {
            if (StringUtil.isNotBlank(content)) {
                putResultDataBody(dataMap, fessConfig.getIndexFieldTitle(), documentHelper.getDigest(responseData, body, dataMap, fessConfig.getCrawlerDocumentFileMaxTitleLengthAsInteger()));
            } else {
                putResultDataBody(dataMap, fessConfig.getIndexFieldTitle(), fessConfig.getCrawlerDocumentFileNoTitleLabel());
            }
        } else {
            if (StringUtil.isBlank(fileName)) {
                putResultDataBody(dataMap, fessConfig.getIndexFieldTitle(), decodeUrlAsName(url, url.startsWith("file:")));
            } else {
                putResultDataBody(dataMap, fessConfig.getIndexFieldTitle(), fileName);
            }
        }
    }
    // host
    putResultDataBody(dataMap, fessConfig.getIndexFieldHost(), getHostOnFile(url));
    // site
    putResultDataBody(dataMap, fessConfig.getIndexFieldSite(), getSiteOnFile(url, urlEncoding));
    // filename
    if (StringUtil.isNotBlank(fileName)) {
        putResultDataBody(dataMap, fessConfig.getIndexFieldFilename(), fileName);
    }
    // url
    putResultDataBody(dataMap, fessConfig.getIndexFieldUrl(), url);
    // created
    final Date now = systemHelper.getCurrentTime();
    putResultDataBody(dataMap, fessConfig.getIndexFieldCreated(), now);
    // TODO anchor
    putResultDataBody(dataMap, fessConfig.getIndexFieldAnchor(), StringUtil.EMPTY);
    // mimetype
    putResultDataBody(dataMap, fessConfig.getIndexFieldMimetype(), mimeType);
    if (fileTypeHelper != null) {
        // filetype
        putResultDataBody(dataMap, fessConfig.getIndexFieldFiletype(), fileTypeHelper.get(mimeType));
    }
    // content_length
    putResultDataBody(dataMap, fessConfig.getIndexFieldContentLength(), Long.toString(responseData.getContentLength()));
    // last_modified
    final Date lastModified = responseData.getLastModified();
    if (lastModified != null) {
        putResultDataBody(dataMap, fessConfig.getIndexFieldLastModified(), lastModified);
        // timestamp
        putResultDataBody(dataMap, fessConfig.getIndexFieldTimestamp(), lastModified);
    } else {
        // timestamp
        putResultDataBody(dataMap, fessConfig.getIndexFieldTimestamp(), now);
    }
    // indexingTarget
    putResultDataBody(dataMap, Constants.INDEXING_TARGET, indexingTarget);
    //  boost
    putResultDataBody(dataMap, fessConfig.getIndexFieldBoost(), crawlingConfig.getDocumentBoost());
    // label: labelType
    final Set<String> labelTypeSet = new HashSet<>();
    for (final String labelType : crawlingConfig.getLabelTypeValues()) {
        labelTypeSet.add(labelType);
    }
    final LabelTypeHelper labelTypeHelper = ComponentUtil.getLabelTypeHelper();
    labelTypeSet.addAll(labelTypeHelper.getMatchedLabelValueSet(url));
    putResultDataBody(dataMap, fessConfig.getIndexFieldLabel(), labelTypeSet);
    // role: roleType
    final List<String> roleTypeList = getRoleTypes(responseData);
    stream(crawlingConfig.getPermissions()).of(stream -> stream.forEach(p -> roleTypeList.add(p)));
    putResultDataBody(dataMap, fessConfig.getIndexFieldRole(), roleTypeList);
    // lang
    if (StringUtil.isNotBlank(fessConfig.getCrawlerDocumentFileDefaultLang())) {
        putResultDataBody(dataMap, fessConfig.getIndexFieldLang(), fessConfig.getCrawlerDocumentFileDefaultLang());
    }
    // id
    putResultDataBody(dataMap, fessConfig.getIndexFieldId(), crawlingInfoHelper.generateId(dataMap));
    // parentId
    String parentUrl = responseData.getParentUrl();
    if (StringUtil.isNotBlank(parentUrl)) {
        parentUrl = pathMappingHelper.replaceUrl(sessionId, parentUrl);
        putResultDataBody(dataMap, fessConfig.getIndexFieldUrl(), parentUrl);
        putResultDataBody(dataMap, fessConfig.getIndexFieldParentId(), crawlingInfoHelper.generateId(dataMap));
        // set again
        putResultDataBody(dataMap, fessConfig.getIndexFieldUrl(), url);
    }
    // from config
    final Map<String, String> scriptConfigMap = crawlingConfig.getConfigParameterMap(ConfigName.SCRIPT);
    final Map<String, String> metaConfigMap = crawlingConfig.getConfigParameterMap(ConfigName.META);
    for (final Map.Entry<String, String> entry : metaConfigMap.entrySet()) {
        final String key = entry.getKey();
        final String[] values = entry.getValue().split(",");
        for (final String value : values) {
            putResultDataWithTemplate(dataMap, key, metaDataMap.get(value), scriptConfigMap.get(key));
        }
    }
    final Map<String, String> valueConfigMap = crawlingConfig.getConfigParameterMap(ConfigName.VALUE);
    for (final Map.Entry<String, String> entry : valueConfigMap.entrySet()) {
        final String key = entry.getKey();
        putResultDataWithTemplate(dataMap, key, entry.getValue(), scriptConfigMap.get(key));
    }
    return dataMap;
}

Example 13 with ResponseData

use of org.codelibs.fess.crawler.entity.ResponseData in project fess by codelibs.

the class DocumentHelper method processRequest.

public Map<String, Object> processRequest(final CrawlingConfig crawlingConfig, final String crawlingInfoId, final String url) {
    if (StringUtil.isBlank(crawlingInfoId)) {
        throw new CrawlingAccessException("sessionId is null.");
    }
    final CrawlerClientFactory crawlerClientFactory = ComponentUtil.getCrawlerClientFactory();
    crawlingConfig.initializeClientFactory(crawlerClientFactory);
    final CrawlerClient client = crawlerClientFactory.getClient(url);
    if (client == null) {
        throw new CrawlingAccessException("CrawlerClient is null for " + url);
    }
    final long startTime = System.currentTimeMillis();
    try (final ResponseData responseData = client.execute(RequestDataBuilder.newRequestData().get().url(url).build())) {
        if (responseData.getRedirectLocation() != null) {
            final Set<RequestData> childUrlList = new HashSet<>();
            childUrlList.add(RequestDataBuilder.newRequestData().get().url(responseData.getRedirectLocation()).build());
            throw new ChildUrlsException(childUrlList, "Redirected from " + url);
        }
        responseData.setExecutionTime(System.currentTimeMillis() - startTime);
        responseData.setSessionId(crawlingInfoId);
        final RuleManager ruleManager = SingletonLaContainer.getComponent(RuleManager.class);
        final Rule rule = ruleManager.getRule(responseData);
        if (rule == null) {
            throw new CrawlingAccessException("No url rule for " + url);
        } else {
            responseData.setRuleId(rule.getRuleId());
            final ResponseProcessor responseProcessor = rule.getResponseProcessor();
            if (responseProcessor instanceof DefaultResponseProcessor) {
                final Transformer transformer = ((DefaultResponseProcessor) responseProcessor).getTransformer();
                final ResultData resultData = transformer.transform(responseData);
                final byte[] data = resultData.getData();
                if (data != null) {
                    try {
                        @SuppressWarnings("unchecked") final Map<String, Object> result = (Map<String, Object>) SerializeUtil.fromBinaryToObject(data);
                        return result;
                    } catch (final Exception e) {
                        throw new CrawlerSystemException("Could not create an instance from bytes.", e);
                    }
                }
            } else {
                throw new CrawlingAccessException("The response processor is not DefaultResponseProcessor. responseProcessor: " + responseProcessor + ", url: " + url);
            }
        }
        return null;
    } catch (final Exception e) {
        throw new CrawlingAccessException("Failed to parse " + url, e);
    }
}

Example 14 with ResponseData

use of org.codelibs.fess.crawler.entity.ResponseData in project fess by codelibs.

the class FessXpathTransformerTest method test_transform.

public void test_transform() throws Exception {
    String data = "<html><head><title>Test</title></head><body><h1>Header1</h1><p>This is a pen.</p></body></html>";
    final FessXpathTransformer fessXpathTransformer = new FessXpathTransformer();
    fessXpathTransformer.init();
    SingletonLaContainerFactory.getContainer().register(CrawlingInfoHelper.class, "crawlingInfoHelper");
    SingletonLaContainerFactory.getContainer().register(PathMappingHelper.class, "pathMappingHelper");
    SingletonLaContainerFactory.getContainer().register(CrawlingConfigHelper.class, "crawlingConfigHelper");
    SingletonLaContainerFactory.getContainer().register(SystemHelper.class, "systemHelper");
    SingletonLaContainerFactory.getContainer().register(FileTypeHelper.class, "fileTypeHelper");
    SingletonLaContainerFactory.getContainer().register(DocumentHelper.class, "documentHelper");
    SingletonLaContainerFactory.getContainer().register(LabelTypeHelper.class, "labelTypeHelper");
    WebConfig webConfig = new WebConfig();
    setValueToObject(webConfig, "labelTypeList", new ArrayList<LabelType>());
    ComponentUtil.getCrawlingConfigHelper().store("test", webConfig);
    setValueToObject(ComponentUtil.getLabelTypeHelper(), "labelTypePatternList", new ArrayList<LabelTypePattern>());
    for (int i = 0; i < 10000; i++) {
        if (i % 1000 == 0) {
            logger.info(MemoryUtil.getMemoryUsageLog() + ":" + i);
            System.gc();
        }
        ResponseData responseData = new ResponseData();
        responseData.setCharSet("UTF-8");
        responseData.setContentLength(data.length());
        responseData.setExecutionTime(1000L);
        responseData.setHttpStatusCode(200);
        responseData.setLastModified(new Date());
        responseData.setMethod("GET");
        responseData.setMimeType("text/html");
        responseData.setParentUrl("http://fess.codelibs.org/");
        responseData.setResponseBody(data.getBytes());
        responseData.setSessionId("test-1");
        responseData.setStatus(0);
        responseData.setUrl("http://fess.codelibs.org/test.html");
        ResultData resultData = fessXpathTransformer.transform(responseData);
    // System.out.println(resultData.toString());
    }
    System.gc();
    Thread.sleep(1000L);
    logger.info(MemoryUtil.getMemoryUsageLog());
    assertTrue(MemoryUtil.getUsedMemory() < 100000000L);
}

Example 15 with ResponseData

use of org.codelibs.fess.crawler.entity.ResponseData in project fess by codelibs.

the class FessXpathTransformerTest method test_processMetaRobots_nofollow.

public void test_processMetaRobots_nofollow() throws Exception {
    final String data = "<meta name=\"robots\" content=\"nofollow\" />";
    final Document document = getDocument(data);
    final FessXpathTransformer transformer = new FessXpathTransformer();
    final ResponseData responseData = new ResponseData();
    responseData.setUrl("http://example.com/");
    transformer.processMetaRobots(responseData, new ResultData(), document);
    assertTrue(responseData.isNoFollow());
}