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

Example 1 with RequestData

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

the class FessXpathTransformer method getAnchorList.

protected List<String> getAnchorList(final Document document, final ResponseData responseData) {
    List<RequestData> anchorList = new ArrayList<>();
    final String baseHref = getBaseHref(document);
    try {
        final URL url = getBaseUrl(responseData.getUrl(), baseHref);
        for (final Map.Entry<String, String> entry : childUrlRuleMap.entrySet()) {
            for (final String u : getUrlFromTagAttribute(url, document, entry.getKey(), entry.getValue(), responseData.getCharSet())) {
                anchorList.add(RequestDataBuilder.newRequestData().get().url(u).build());
            }
        }
        anchorList = convertChildUrlList(anchorList);
    } catch (final Exception e) {
        logger.warn("Could not parse anchor tags.", e);
    }
    final List<String> urlList = new ArrayList<>(anchorList.size());
    for (final RequestData requestData : anchorList) {
        urlList.add(requestData.getUrl());
    }
    return urlList;
}

Example 2 with RequestData

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

the class FessXpathTransformer method putAdditionalData.

protected void putAdditionalData(final Map<String, Object> dataMap, final ResponseData responseData, final Document document) {
    // canonical
    if (StringUtil.isNotBlank(fessConfig.getCrawlerDocumentHtmlCanonicalXpath())) {
        final String canonicalUrl = getCanonicalUrl(responseData, document);
        if (canonicalUrl != null && !canonicalUrl.equals(responseData.getUrl())) {
            final Set<RequestData> childUrlSet = new HashSet<>();
            childUrlSet.add(RequestDataBuilder.newRequestData().get().url(canonicalUrl).build());
            throw new ChildUrlsException(childUrlSet, this.getClass().getName() + "#putAdditionalData(Map<String, Object>, ResponseData, Document)");
        }
    }
    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();
    final LabelTypeHelper labelTypeHelper = ComponentUtil.getLabelTypeHelper();
    String url = responseData.getUrl();
    final String indexingTarget = crawlingConfig.getIndexingTarget(url);
    url = pathMappingHelper.replaceUrl(sessionId, url);
    final String mimeType = responseData.getMimeType();
    final Map<String, String> fieldConfigMap = crawlingConfig.getConfigParameterMap(ConfigName.FIELD);
    final Map<String, String> xpathConfigMap = crawlingConfig.getConfigParameterMap(ConfigName.XPATH);
    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);
    }
    // lang
    final String lang = systemHelper.normalizeLang(getSingleNodeValue(document, getLangXpath(fessConfig, xpathConfigMap), true));
    if (lang != null) {
        putResultDataBody(dataMap, fessConfig.getIndexFieldLang(), lang);
    }
    // title
    // content
    final String body = getSingleNodeValue(document, getContentXpath(fessConfig, xpathConfigMap), prunedContent);
    putResultDataBody(dataMap, fessConfig.getIndexFieldContent(), documentHelper.getContent(responseData, body, dataMap));
    if ((Constants.TRUE.equalsIgnoreCase(fieldConfigMap.get(fessConfig.getIndexFieldCache())) || fessConfig.isCrawlerDocumentCacheEnabled()) && fessConfig.isSupportedDocumentCacheMimetypes(mimeType)) {
        if (responseData.getContentLength() > 0 && responseData.getContentLength() <= fessConfig.getCrawlerDocumentCacheMaxSizeAsInteger().longValue()) {
            String charSet = responseData.getCharSet();
            if (charSet == null) {
                charSet = Constants.UTF_8;
            }
            try (final BufferedInputStream is = new BufferedInputStream(responseData.getResponseBody())) {
                // cache
                putResultDataBody(dataMap, fessConfig.getIndexFieldCache(), new String(InputStreamUtil.getBytes(is), charSet));
                putResultDataBody(dataMap, fessConfig.getIndexFieldHasCache(), Constants.TRUE);
            } catch (final Exception e) {
                logger.warn("Failed to write a cache: " + sessionId + ":" + responseData, e);
            }
        } else {
            logger.debug("Content size is too large({} > {}): {}", responseData.getContentLength(), fessConfig.getCrawlerDocumentCacheMaxSizeAsInteger(), responseData.getUrl());
        }
    }
    // digest
    final String digest = getSingleNodeValue(document, getDigestXpath(fessConfig, xpathConfigMap), false);
    if (StringUtil.isNotBlank(digest)) {
        putResultDataBody(dataMap, fessConfig.getIndexFieldDigest(), digest);
    } else {
        putResultDataBody(dataMap, fessConfig.getIndexFieldDigest(), documentHelper.getDigest(responseData, body, dataMap, fessConfig.getCrawlerDocumentHtmlMaxDigestLengthAsInteger()));
    }
    // segment
    putResultDataBody(dataMap, fessConfig.getIndexFieldSegment(), sessionId);
    // host
    putResultDataBody(dataMap, fessConfig.getIndexFieldHost(), getHost(url));
    // site
    putResultDataBody(dataMap, fessConfig.getIndexFieldSite(), getSite(url, urlEncoding));
    // filename
    final String fileName = getFileName(url, urlEncoding);
    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);
    // anchor
    putResultDataBody(dataMap, fessConfig.getIndexFieldAnchor(), getAnchorList(document, responseData));
    // 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);
    }
    labelTypeSet.addAll(labelTypeHelper.getMatchedLabelValueSet(url));
    putResultDataBody(dataMap, fessConfig.getIndexFieldLabel(), labelTypeSet);
    // role: roleType
    final List<String> roleTypeList = new ArrayList<>();
    stream(crawlingConfig.getPermissions()).of(stream -> stream.forEach(p -> roleTypeList.add(p)));
    putResultDataBody(dataMap, fessConfig.getIndexFieldRole(), roleTypeList);
    // 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);
    }
    // thumbnail
    final String thumbnailUrl = getThumbnailUrl(responseData, document);
    if (StringUtil.isNotBlank(thumbnailUrl)) {
        putResultDataBody(dataMap, fessConfig.getIndexFieldThumbnail(), thumbnailUrl);
    }
    // from config
    final Map<String, String> scriptConfigMap = crawlingConfig.getConfigParameterMap(ConfigName.SCRIPT);
    xpathConfigMap.entrySet().stream().filter(e -> !e.getKey().startsWith("default.")).forEach(e -> {
        final String key = e.getKey();
        final String value = getSingleNodeValue(document, e.getValue(), true);
        putResultDataWithTemplate(dataMap, key, value, scriptConfigMap.get(key));
    });
    crawlingConfig.getConfigParameterMap(ConfigName.VALUE).entrySet().stream().forEach(e -> {
        final String key = e.getKey();
        final String value = e.getValue();
        putResultDataWithTemplate(dataMap, key, value, scriptConfigMap.get(key));
    });
}

Example 3 with RequestData

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

the class FessXpathTransformer method convertChildUrlList.

@Override
protected List<RequestData> convertChildUrlList(final List<RequestData> urlList) {
    if (urlList != null) {
        for (final RequestData requestData : urlList) {
            String url = requestData.getUrl();
            for (final Map.Entry<String, String> entry : convertUrlMap.entrySet()) {
                url = url.replaceAll(entry.getKey(), entry.getValue());
            }
            requestData.setUrl(replaceDuplicateHost(url));
        }
    }
    return urlList;
}

Example 4 with RequestData

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

the class FessCrawlerThread method getChildUrlSet.

protected Set<RequestData> getChildUrlSet(final FessEsClient fessEsClient, final String id) {
    final FessConfig fessConfig = ComponentUtil.getFessConfig();
    final IndexingHelper indexingHelper = ComponentUtil.getIndexingHelper();
    final List<Map<String, Object>> docList = indexingHelper.getChildDocumentList(fessEsClient, id, new String[] { fessConfig.getIndexFieldUrl() });
    if (docList.isEmpty()) {
        return null;
    }
    if (logger.isDebugEnabled()) {
        logger.debug("Found documents: " + docList);
    }
    final Set<RequestData> urlSet = new HashSet<>(docList.size());
    for (final Map<String, Object> doc : docList) {
        final String url = DocumentUtil.getValue(doc, fessConfig.getIndexFieldUrl(), String.class);
        if (StringUtil.isNotBlank(url)) {
            urlSet.add(RequestDataBuilder.newRequestData().get().url(url).build());
        }
    }
    return urlSet;
}

Example 5 with RequestData

use of org.codelibs.fess.crawler.entity.RequestData 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);
    }
}