Search in sources :

Example 1 with SitemapsHelper

use of org.codelibs.fess.crawler.helper.SitemapsHelper in project fess-crawler by codelibs.

the class SitemapsRuleTest method setUp.

@Override
protected void setUp() throws Exception {
    super.setUp();
    StandardCrawlerContainer container = new StandardCrawlerContainer().singleton("sitemapsHelper", // 
    SitemapsHelper.class).singleton("sitemapsRule", SitemapsRule.class);
    sitemapsRule = container.getComponent("sitemapsRule");
}
Also used : StandardCrawlerContainer(org.codelibs.fess.crawler.container.StandardCrawlerContainer) SitemapsHelper(org.codelibs.fess.crawler.helper.SitemapsHelper)

Example 2 with SitemapsHelper

use of org.codelibs.fess.crawler.helper.SitemapsHelper in project fess-crawler by codelibs.

the class CrawlerTest method setUp.

@Override
protected void setUp() throws Exception {
    super.setUp();
    final Map<String, String> featureMap = newHashMap();
    featureMap.put("http://xml.org/sax/features/namespaces", "false");
    final Map<String, String> propertyMap = newHashMap();
    final Map<String, String> childUrlRuleMap = newHashMap();
    childUrlRuleMap.put("//A", "href");
    childUrlRuleMap.put("//AREA", "href");
    childUrlRuleMap.put("//FRAME", "src");
    childUrlRuleMap.put("//IFRAME", "src");
    childUrlRuleMap.put("//IMG", "src");
    childUrlRuleMap.put("//LINK", "href");
    childUrlRuleMap.put("//SCRIPT", "src");
    container = new StandardCrawlerContainer();
    container.<HcHttpClient>prototype("internalHttpClient", HcHttpClient.class, client -> {
        client.setCookieSpec(CookieSpecs.BEST_MATCH);
        client.setClientConnectionManager(container.getComponent("clientConnectionManager"));
    }).prototype("httpClient", FaultTolerantClient.class, client -> {
        client.setCrawlerClient(container.getComponent("internalHttpClient"));
        client.setMaxRetryCount(5);
        client.setRetryInterval(500);
    }).prototype("fsClient", FileSystemClient.class).prototype("ruleManager", RuleManagerImpl.class, manager -> {
        manager.addRule(container.getComponent("sitemapsRule"));
        manager.addRule(container.getComponent("fileRule"));
    }).prototype("accessResult", AccessResultImpl.class).prototype("urlQueue", UrlQueueImpl.class).prototype("crawlerThread", CrawlerThread.class).prototype("crawler", Crawler.class).prototype("urlFilterService", UrlFilterServiceImpl.class).prototype("urlQueueService", UrlQueueServiceImpl.class).prototype("dataService", DataServiceImpl.class).prototype("urlFilter", UrlFilterImpl.class).singleton("urlConvertHelper", UrlConvertHelper.class).singleton("intervalController", DefaultIntervalController.class).singleton("sitemapsHelper", SitemapsHelper.class).singleton("logHelper", LogHelperImpl.class).singleton("encodingHelper", EncodingHelper.class).singleton("contentLengthHelper", ContentLengthHelper.class).singleton("mimeTypeHelper", MimeTypeHelperImpl.class).<FileTransformer>singleton("fileTransformer", FileTransformer.class, transformer -> {
        transformer.setName("fileTransformer");
        transformer.setFeatureMap(featureMap);
        transformer.setPropertyMap(propertyMap);
        transformer.setChildUrlRuleMap(childUrlRuleMap);
    }).singleton("dataHelper", MemoryDataHelper.class).singleton("robotsTxtHelper", RobotsTxtHelper.class).<CrawlerClientFactory>singleton("clientFactory", CrawlerClientFactory.class, factory -> {
        factory.addClient("http:.*", container.getComponent("httpClient"));
        factory.addClient("file:.*", container.getComponent("fsClient"));
    }).singleton("tikaExtractor", TikaExtractor.class).<ExtractorFactory>singleton("extractorFactory", ExtractorFactory.class, factory -> {
        TikaExtractor tikaExtractor = container.getComponent("tikaExtractor");
        factory.addExtractor("text/plain", tikaExtractor);
        factory.addExtractor("text/html", tikaExtractor);
    }).singleton("httpClient", // 
    HcHttpClient.class).singleton("sitemapsResponseProcessor", // 
    SitemapsResponseProcessor.class).<SitemapsRule>singleton("sitemapsRule", SitemapsRule.class, rule -> {
        rule.setResponseProcessor(container.getComponent("sitemapsResponseProcessor"));
        rule.setRuleId("sitemapsRule");
        rule.addRule("url", ".*sitemap.*");
    }).<// 
    DefaultResponseProcessor>singleton("defaultResponseProcessor", DefaultResponseProcessor.class, processor -> {
        processor.setTransformer(container.getComponent("fileTransformer"));
        processor.setSuccessfulHttpCodes(new int[] { 200 });
        processor.setNotModifiedHttpCodes(new int[] { 304 });
    }).<// 
    RegexRule>singleton("fileRule", RegexRule.class, rule -> {
        rule.setRuleId("fileRule");
        rule.setDefaultRule(true);
        rule.setResponseProcessor(container.getComponent("defaultResponseProcessor"));
    }).<// 
    PoolingHttpClientConnectionManager>singleton("clientConnectionManager", new PoolingHttpClientConnectionManager(5, TimeUnit.MINUTES), manager -> {
        manager.setMaxTotal(200);
        manager.setDefaultMaxPerRoute(20);
    });
    crawler = container.getComponent("crawler");
    dataService = container.getComponent("dataService");
    urlQueueService = container.getComponent("urlQueueService");
    fileTransformer = container.getComponent("fileTransformer");
}
Also used : StandardCrawlerContainer(org.codelibs.fess.crawler.container.StandardCrawlerContainer) MimeTypeHelperImpl(org.codelibs.fess.crawler.helper.impl.MimeTypeHelperImpl) HcHttpClient(org.codelibs.fess.crawler.client.http.HcHttpClient) UrlQueueImpl(org.codelibs.fess.crawler.entity.UrlQueueImpl) PlainTestCase(org.dbflute.utflute.core.PlainTestCase) ExtractorFactory(org.codelibs.fess.crawler.extractor.ExtractorFactory) FileSystemClient(org.codelibs.fess.crawler.client.fs.FileSystemClient) DataService(org.codelibs.fess.crawler.service.DataService) CookieSpecs(org.apache.http.client.config.CookieSpecs) SitemapsHelper(org.codelibs.fess.crawler.helper.SitemapsHelper) DefaultResponseProcessor(org.codelibs.fess.crawler.processor.impl.DefaultResponseProcessor) RuleManagerImpl(org.codelibs.fess.crawler.rule.impl.RuleManagerImpl) DataServiceImpl(org.codelibs.fess.crawler.service.impl.DataServiceImpl) RegexRule(org.codelibs.fess.crawler.rule.impl.RegexRule) FaultTolerantClient(org.codelibs.fess.crawler.client.FaultTolerantClient) LogHelperImpl(org.codelibs.fess.crawler.helper.impl.LogHelperImpl) SitemapsRule(org.codelibs.fess.crawler.rule.impl.SitemapsRule) Map(java.util.Map) UrlFilterServiceImpl(org.codelibs.fess.crawler.service.impl.UrlFilterServiceImpl) PoolingHttpClientConnectionManager(org.apache.http.impl.conn.PoolingHttpClientConnectionManager) CrawlerWebServer(org.codelibs.fess.crawler.util.CrawlerWebServer) ContentLengthHelper(org.codelibs.fess.crawler.helper.ContentLengthHelper) EncodingHelper(org.codelibs.fess.crawler.helper.EncodingHelper) RobotsTxtHelper(org.codelibs.fess.crawler.helper.RobotsTxtHelper) AccessResultImpl(org.codelibs.fess.crawler.entity.AccessResultImpl) SitemapsResponseProcessor(org.codelibs.fess.crawler.processor.impl.SitemapsResponseProcessor) UrlFilterImpl(org.codelibs.fess.crawler.filter.impl.UrlFilterImpl) MemoryDataHelper(org.codelibs.fess.crawler.helper.MemoryDataHelper) DefaultIntervalController(org.codelibs.fess.crawler.interval.impl.DefaultIntervalController) ResourceUtil(org.codelibs.core.io.ResourceUtil) TikaExtractor(org.codelibs.fess.crawler.extractor.impl.TikaExtractor) File(java.io.File) FileTransformer(org.codelibs.fess.crawler.transformer.impl.FileTransformer) TimeUnit(java.util.concurrent.TimeUnit) UrlQueueService(org.codelibs.fess.crawler.service.UrlQueueService) UrlConvertHelper(org.codelibs.fess.crawler.helper.UrlConvertHelper) UrlQueueServiceImpl(org.codelibs.fess.crawler.service.impl.UrlQueueServiceImpl) CrawlerClientFactory(org.codelibs.fess.crawler.client.CrawlerClientFactory) UrlQueue(org.codelibs.fess.crawler.entity.UrlQueue) LogHelperImpl(org.codelibs.fess.crawler.helper.impl.LogHelperImpl) ContentLengthHelper(org.codelibs.fess.crawler.helper.ContentLengthHelper) HcHttpClient(org.codelibs.fess.crawler.client.http.HcHttpClient) DefaultIntervalController(org.codelibs.fess.crawler.interval.impl.DefaultIntervalController) RobotsTxtHelper(org.codelibs.fess.crawler.helper.RobotsTxtHelper) CrawlerClientFactory(org.codelibs.fess.crawler.client.CrawlerClientFactory) ExtractorFactory(org.codelibs.fess.crawler.extractor.ExtractorFactory) StandardCrawlerContainer(org.codelibs.fess.crawler.container.StandardCrawlerContainer) UrlQueueImpl(org.codelibs.fess.crawler.entity.UrlQueueImpl) RuleManagerImpl(org.codelibs.fess.crawler.rule.impl.RuleManagerImpl) PoolingHttpClientConnectionManager(org.apache.http.impl.conn.PoolingHttpClientConnectionManager) FaultTolerantClient(org.codelibs.fess.crawler.client.FaultTolerantClient) FileTransformer(org.codelibs.fess.crawler.transformer.impl.FileTransformer) UrlFilterImpl(org.codelibs.fess.crawler.filter.impl.UrlFilterImpl) DefaultResponseProcessor(org.codelibs.fess.crawler.processor.impl.DefaultResponseProcessor) SitemapsRule(org.codelibs.fess.crawler.rule.impl.SitemapsRule) TikaExtractor(org.codelibs.fess.crawler.extractor.impl.TikaExtractor) RegexRule(org.codelibs.fess.crawler.rule.impl.RegexRule) UrlQueueServiceImpl(org.codelibs.fess.crawler.service.impl.UrlQueueServiceImpl)

Example 3 with SitemapsHelper

use of org.codelibs.fess.crawler.helper.SitemapsHelper in project fess-crawler by codelibs.

the class SitemapsResponseProcessor method process.

@Override
public void process(final ResponseData responseData) {
    final SitemapsHelper sitemapsHelper = crawlerContainer.getComponent("sitemapsHelper");
    try (final InputStream responseBody = responseData.getResponseBody()) {
        final SitemapSet sitemapSet = sitemapsHelper.parse(responseBody);
        final Set<RequestData> requestDataSet = new LinkedHashSet<>();
        for (final Sitemap sitemap : sitemapSet.getSitemaps()) {
            if (sitemap != null) {
                requestDataSet.add(RequestDataBuilder.newRequestData().get().url(sitemap.getLoc()).build());
            }
        }
        throw new ChildUrlsException(requestDataSet, this.getClass().getName() + "#process");
    } catch (final IOException e) {
        throw new IORuntimeException(e);
    }
}
Also used : LinkedHashSet(java.util.LinkedHashSet) ChildUrlsException(org.codelibs.fess.crawler.exception.ChildUrlsException) Sitemap(org.codelibs.fess.crawler.entity.Sitemap) IORuntimeException(org.codelibs.core.exception.IORuntimeException) InputStream(java.io.InputStream) RequestData(org.codelibs.fess.crawler.entity.RequestData) SitemapSet(org.codelibs.fess.crawler.entity.SitemapSet) IOException(java.io.IOException) SitemapsHelper(org.codelibs.fess.crawler.helper.SitemapsHelper)

Example 4 with SitemapsHelper

use of org.codelibs.fess.crawler.helper.SitemapsHelper in project fess-crawler by codelibs.

the class RuleManagerImplTest method setUp.

@Override
protected void setUp() throws Exception {
    super.setUp();
    StandardCrawlerContainer container = new StandardCrawlerContainer().singleton("sitemapsHelper", // 
    SitemapsHelper.class).singleton("sitemapsRule", // 
    SitemapsRule.class).singleton("fileRule", // 
    RegexRule.class).singleton("ruleManager", RuleManagerImpl.class);
    ruleManager = container.getComponent("ruleManager");
    SitemapsRule sitemapsRule = container.getComponent("sitemapsRule");
    sitemapsRule.setRuleId("sitemapsRule");
    sitemapsRule.addRule("url", ".*sitemap.*");
    ruleManager.addRule(sitemapsRule);
    RegexRule fileRule = container.getComponent("fileRule");
    fileRule.setRuleId("fileRule");
    fileRule.setDefaultRule(true);
    ruleManager.addRule(fileRule);
}
Also used : StandardCrawlerContainer(org.codelibs.fess.crawler.container.StandardCrawlerContainer) SitemapsHelper(org.codelibs.fess.crawler.helper.SitemapsHelper)

Aggregations

SitemapsHelper (org.codelibs.fess.crawler.helper.SitemapsHelper)4 StandardCrawlerContainer (org.codelibs.fess.crawler.container.StandardCrawlerContainer)3 File (java.io.File)1 IOException (java.io.IOException)1 InputStream (java.io.InputStream)1 LinkedHashSet (java.util.LinkedHashSet)1 Map (java.util.Map)1 TimeUnit (java.util.concurrent.TimeUnit)1 CookieSpecs (org.apache.http.client.config.CookieSpecs)1 PoolingHttpClientConnectionManager (org.apache.http.impl.conn.PoolingHttpClientConnectionManager)1 IORuntimeException (org.codelibs.core.exception.IORuntimeException)1 ResourceUtil (org.codelibs.core.io.ResourceUtil)1 CrawlerClientFactory (org.codelibs.fess.crawler.client.CrawlerClientFactory)1 FaultTolerantClient (org.codelibs.fess.crawler.client.FaultTolerantClient)1 FileSystemClient (org.codelibs.fess.crawler.client.fs.FileSystemClient)1 HcHttpClient (org.codelibs.fess.crawler.client.http.HcHttpClient)1 AccessResultImpl (org.codelibs.fess.crawler.entity.AccessResultImpl)1 RequestData (org.codelibs.fess.crawler.entity.RequestData)1 Sitemap (org.codelibs.fess.crawler.entity.Sitemap)1 SitemapSet (org.codelibs.fess.crawler.entity.SitemapSet)1