// Input : the crawldb
platform.assertPathExists(crawlDbPath, "CrawlDb doesn't exist");
// Our crawl db is defined by the CrawlDbDatum
Tap inputSource = platform.makeTap(platform.makeBinaryScheme(CrawlDbDatum.FIELDS), crawlDbPath);
Pipe importPipe = new Pipe("import pipe");
// Split into tuples that are to be fetched and that have already been fetched
SplitterAssembly splitter = new SplitterAssembly(importPipe, new SplitFetchedUnfetchedCrawlDatums());
Pipe finishedDatumsFromDb = splitter.getRHSPipe();
Pipe urlsToFetchPipe = new Pipe("urls to Fetch", splitter.getLHSPipe());
// Convert the urlsToFetchPipe so that we now deal with UrlDatums.
urlsToFetchPipe = new Each(urlsToFetchPipe, new CreateUrlDatumFromCrawlDbFunction());
// A TupleLogger is a good way to follow the tuples around in a flow. You can enable the output
// of tuples by setting options.setDebugLogging() to true.
TupleLogger tupleLogger = new TupleLogger();
urlsToFetchPipe = TupleLogger.makePipe(urlsToFetchPipe, true);
// Create the output sinks :
// crawldb
// content
// parse
// status
BasePath outCrawlDbPath = platform.makePath(curWorkingDirPath, CrawlConfig.CRAWLDB_SUBDIR_NAME);
Tap loopCrawldbSink = platform.makeTap(platform.makeBinaryScheme(CrawlDbDatum.FIELDS), outCrawlDbPath, SinkMode.REPLACE);
BasePath contentDirPath = platform.makePath(curWorkingDirPath, CrawlConfig.CONTENT_SUBDIR_NAME);
Tap contentSink = platform.makeTap(platform.makeBinaryScheme(FetchedDatum.FIELDS), contentDirPath, SinkMode.REPLACE);
BasePath parseDirPath = platform.makePath(curWorkingDirPath, CrawlConfig.PARSE_SUBDIR_NAME);
Tap parseSink = platform.makeTap(platform.makeBinaryScheme(ParsedDatum.FIELDS), parseDirPath, SinkMode.REPLACE);
BasePath statusDirPath = platform.makePath(curWorkingDirPath, CrawlConfig.STATUS_SUBDIR_NAME);
Tap statusSink = platform.makeTap(platform.makeTextScheme(), statusDirPath, SinkMode.REPLACE);
// Create the sub-assembly that runs the fetch job
SimpleHttpFetcher fetcher = new SimpleHttpFetcher(options.getMaxThreads(), fetcherPolicy, userAgent);
fetcher.setMaxRetryCount(CrawlConfig.MAX_RETRIES);
fetcher.setSocketTimeout(CrawlConfig.SOCKET_TIMEOUT);
fetcher.setConnectionTimeout(CrawlConfig.CONNECTION_TIMEOUT);
// You can also provide a set of mime types you want to restrict what content type you
// want to deal with - for now keep it simple.
Set<String> validMimeTypes = new HashSet<String>();
validMimeTypes.add("text/plain");
validMimeTypes.add("text/html");
fetcherPolicy.setValidMimeTypes(validMimeTypes);
// The scorer is used by the FetchPipe to assign a score to every URL that passes the
// robots.txt processing. The score is used to sort URLs such that higher scoring URLs
// are fetched first. If URLs are skipped for any reason(s) lower scoring URLs are skipped.
BaseScoreGenerator scorer = new FixedScoreGenerator();
FetchPipe fetchPipe = new FetchPipe(urlsToFetchPipe, scorer, fetcher, platform.getNumReduceTasks());
Pipe statusPipe = new Pipe("status pipe", fetchPipe.getStatusTailPipe());
Pipe contentPipe = new Pipe("content pipe", fetchPipe.getContentTailPipe());
contentPipe = TupleLogger.makePipe(contentPipe, true);
// Take content and split it into content output plus parse to extract URLs.
SimpleParser parser;
if (options.isUseBoilerpipe()) {
parser = new SimpleParser(new BoilerpipeContentExtractor(), new SimpleLinkExtractor(), new ParserPolicy());
} else if (options.isGenerateHTML()) {
parser = new SimpleParser(new HtmlContentExtractor(), new SimpleLinkExtractor(), new ParserPolicy(), true);
} else {
parser = new SimpleParser();
}
parser.setExtractLanguage(false);
ParsePipe parsePipe = new ParsePipe(contentPipe, parser);
// Create the output map that connects each tail pipe to the appropriate sink, and the
// list of tail pipes.
Map<String, Tap> sinkMap = new HashMap<String, Tap>();
List<Pipe> tailPipes = new ArrayList<Pipe>();
if (options.isGenerateHTML()) {
// Let's write out the parse as text:
Pipe textParsePipe = new Pipe("text parse data", parsePipe.getTailPipe());
textParsePipe = new Each(textParsePipe, new Fields(ParsedDatum.PARSED_TEXT_FN), new RegexReplace(new Fields(ParsedDatum.PARSED_TEXT_FN), "[\\r\\n\\t]+", " ", true), Fields.REPLACE);
textParsePipe = new Each(textParsePipe, new Fields(ParsedDatum.URL_FN, ParsedDatum.PARSED_TEXT_FN), new Identity());
BasePath textParsePath = platform.makePath(curWorkingDirPath, CrawlConfig.HTML_SUBDIR_NAME);
Tap textParseTap = platform.makeTap(platform.makeTextScheme(), textParsePath, SinkMode.REPLACE);
sinkMap.put(textParsePipe.getName(), textParseTap);
tailPipes.add(textParsePipe);
}
// Let's output a WritableSequenceFile as an example - this file can
// then be used as input when working with Mahout.
// For now we only do it when we are running in Hadoop mode
Tap writableSeqFileSink = null;
Pipe writableSeqFileDataPipe = null;
if (!options.isLocalPlatformMode()) {
writableSeqFileDataPipe = new Pipe("writable seqfile data", new Each(parsePipe.getTailPipe(), new CreateWritableSeqFileData()));
BasePath writableSeqFileDataPath = platform.makePath(curWorkingDirPath, CrawlConfig.EXTRACTED_TEXT_SUBDIR_NAME);
WritableSequenceFile writableSeqScheme = new WritableSequenceFile(new Fields(CrawlConfig.WRITABLE_SEQ_FILE_KEY_FN, CrawlConfig.WRITABLE_SEQ_FILE_VALUE_FN), Text.class, Text.class);