private List<Outlink> toOutlinks(
String url, Metadata metadata, Map<String, List<String>> slinks) {
List<Outlink> outlinks = new LinkedList<>();
URL sourceUrl;
try {
sourceUrl = new URL(url);
} catch (MalformedURLException e) {
// we would have known by now as previous components check whether
// the URL is valid
LOG.error("MalformedURLException on {}", url);
eventCounter.scope("error_invalid_source_url").incrBy(1);
return outlinks;
}
Map<String, List<String>> linksKept = new HashMap<>();
for (Map.Entry<String, List<String>> linkEntry : slinks.entrySet()) {
String targetURL = linkEntry.getKey();
// filter the urls
if (urlFilters != null) {
targetURL = urlFilters.filter(sourceUrl, metadata, targetURL);
if (targetURL == null) {
eventCounter.scope("outlink_filtered").incr();
continue;
}
}
// the link has survived the various filters
if (targetURL != null) {
List<String> anchors = linkEntry.getValue();
linksKept.put(targetURL, anchors);
eventCounter.scope("outlink_kept").incr();
}
}
for (String outlink : linksKept.keySet()) {
// configure which metadata gets inherited from parent
Metadata linkMetadata = metadataTransfer.getMetaForOutlink(outlink, url, metadata);
Outlink ol = new Outlink(outlink);
// add the anchors to the metadata?
if (trackAnchors) {
List<String> anchors = linksKept.get(outlink);
if (anchors.size() > 0) {
linkMetadata.addValues(ANCHORS_KEY_NAME, anchors);
// sets the first anchor
ol.setAnchor(anchors.get(0));
}
}
ol.setMetadata(linkMetadata);
outlinks.add(ol);
}
return outlinks;
}
private void handleException(
String url,
Throwable e,
Metadata metadata,
Tuple tuple,
String errorSource,
String errorMessage) {
LOG.error(errorMessage);
// send to status stream in case another component wants to update
// its status
metadata.setValue(Constants.STATUS_ERROR_SOURCE, errorSource);
metadata.setValue(Constants.STATUS_ERROR_MESSAGE, errorMessage);
collector.emit(StatusStreamName, tuple, new Values(url, metadata, Status.ERROR));
collector.ack(tuple);
// Increment metric that is context specific
String s = "error_" + errorSource.replaceAll(" ", "_") + "_";
eventCounter.scope(s + e.getClass().getSimpleName()).incrBy(1);
// Increment general metric
eventCounter.scope("parse exception").incrBy(1);
}
@Override
public void execute(Tuple input) {
String urlString = input.getStringByField("url");
if (StringUtils.isBlank(urlString)) {
LOG.info("[Fetcher #{}] Missing value for field url in tuple {}", taskIndex, input);
// ignore silently
_collector.ack(input);
return;
}
Metadata metadata = null;
if (input.contains("metadata")) metadata = (Metadata) input.getValueByField("metadata");
if (metadata == null) metadata = Metadata.empty;
URL url;
try {
url = new URL(urlString);
} catch (MalformedURLException e) {
LOG.error("{} is a malformed URL", urlString);
// Report to status stream and ack
if (metadata == Metadata.empty) {
metadata = new Metadata();
}
metadata.setValue("error.cause", "malformed URL");
_collector.emit(
com.digitalpebble.storm.crawler.Constants.StatusStreamName,
input,
new Values(urlString, metadata, Status.ERROR));
_collector.ack(input);
return;
}
// check when we are allowed to process it
String key = getPolitenessKey(url);
Long timeAllowed = throttler.getIfPresent(key);
if (timeAllowed != null) {
long now = System.currentTimeMillis();
long timeToWait = timeAllowed - now;
if (timeToWait > 0) {
try {
Thread.sleep(timeToWait);
} catch (InterruptedException e) {
LOG.error("[Fetcher #{}] caught InterruptedException caught while waiting");
}
}
}
long delay = this.crawlDelay;
try {
Protocol protocol = protocolFactory.getProtocol(url);
BaseRobotRules rules = protocol.getRobotRules(urlString);
// autodiscovery of sitemaps
// the sitemaps will be sent down the topology
// as many times as there is a URL for a given host
// the status updater will certainly cache things
// but we could also have a simple cache mechanism here
// as well.
if (sitemapsAutoDiscovery) {
for (String sitemapURL : rules.getSitemaps()) {
handleOutlink(input, url, sitemapURL, metadata);
}
}
if (!rules.isAllowed(urlString)) {
LOG.info("Denied by robots.txt: {}", urlString);
metadata.setValue("error.cause", "robots.txt");
// Report to status stream and ack
_collector.emit(
com.digitalpebble.storm.crawler.Constants.StatusStreamName,
input,
new Values(urlString, metadata, Status.ERROR));
_collector.ack(input);
return;
}
// get the delay from robots
// value is negative when not set
long robotsDelay = rules.getCrawlDelay();
if (robotsDelay > 0) {
// cap the value to a maximum
// as some sites specify ridiculous values
if (robotsDelay > maxCrawlDelay) {
LOG.debug("Delay from robots capped at {} for {}", robotsDelay, url);
delay = maxCrawlDelay;
} else {
delay = robotsDelay;
}
}
long start = System.currentTimeMillis();
ProtocolResponse response = protocol.getProtocolOutput(urlString, metadata);
long timeFetching = System.currentTimeMillis() - start;
averagedMetrics.scope("fetch_time").update(timeFetching);
averagedMetrics.scope("bytes_fetched").update(response.getContent().length);
eventCounter.scope("fetched").incrBy(1);
perSecMetrics.scope("bytes_fetched_perSec").update(response.getContent().length);
perSecMetrics.scope("fetched_perSec").update(1);
LOG.info(
"[Fetcher #{}] Fetched {} with status {} in {}",
taskIndex,
urlString,
response.getStatusCode(),
timeFetching);
response
.getMetadata()
.setValue("fetch.statusCode", Integer.toString(response.getStatusCode()));
response.getMetadata().putAll(metadata);
// determine the status based on the status code
Status status = Status.fromHTTPCode(response.getStatusCode());
// if the status is OK emit on default stream
if (status.equals(Status.FETCHED)) {
_collector.emit(
Utils.DEFAULT_STREAM_ID,
input,
new Values(urlString, response.getContent(), response.getMetadata()));
} else if (status.equals(Status.REDIRECTION)) {
// Mark URL as redirected
_collector.emit(
com.digitalpebble.storm.crawler.Constants.StatusStreamName,
input,
new Values(urlString, response.getMetadata(), status));
// find the URL it redirects to
String redirection = response.getMetadata().getFirstValue(HttpHeaders.LOCATION);
if (allowRedirs && redirection != null && StringUtils.isNotBlank(redirection)) {
handleOutlink(input, url, redirection, response.getMetadata());
}
} else {
// Error
_collector.emit(
com.digitalpebble.storm.crawler.Constants.StatusStreamName,
input,
new Values(urlString, response.getMetadata(), status));
}
} catch (Exception exece) {
String message = exece.getMessage();
if (message == null) message = "";
// common exceptions for which we log only a short message
if (exece.getCause() instanceof java.util.concurrent.TimeoutException)
LOG.error("Socket timeout fetching {}", urlString);
else if (message.contains(" timed out")) LOG.error("Socket timeout fetching {}", urlString);
else if (exece.getCause() instanceof java.net.UnknownHostException)
LOG.error("Unknown host {}", urlString);
else if (message.contains(" timed out")) LOG.error("Socket timeout fetching {}", urlString);
// log the full stacktrace
else LOG.error("Exception while fetching {}", urlString, exece);
eventCounter.scope("exception").incrBy(1);
// could be an empty, immutable Metadata
if (metadata.size() == 0) {
metadata = new Metadata();
}
// add the reason of the failure in the metadata
metadata.setValue("fetch.exception", message);
_collector.emit(
com.digitalpebble.storm.crawler.Constants.StatusStreamName,
input,
new Values(urlString, metadata, Status.FETCH_ERROR));
}
// update the throttler
throttler.put(key, System.currentTimeMillis() + delay);
_collector.ack(input);
}
@Override
public void execute(Tuple tuple) {
byte[] content = tuple.getBinaryByField("content");
String url = tuple.getStringByField("url");
Metadata metadata = (Metadata) tuple.getValueByField("metadata");
// check that its content type is HTML
// look at value found in HTTP headers
boolean CT_OK = false;
String httpCT = metadata.getFirstValue(HttpHeaders.CONTENT_TYPE);
if (StringUtils.isNotBlank(httpCT)) {
if (httpCT.toLowerCase().contains("html")) {
CT_OK = true;
}
}
// simply ignore cases where the content type has not been set
// TODO sniff content with Tika?
else {
CT_OK = true;
}
if (!CT_OK) {
String errorMessage = "Exception content-type " + httpCT + " for " + url;
RuntimeException e = new RuntimeException(errorMessage);
handleException(url, e, metadata, tuple, "content-type checking", errorMessage);
return;
}
LOG.info("Parsing : starting {}", url);
long start = System.currentTimeMillis();
String charset = getContentCharset(content, metadata);
// get the robots tags from the fetch metadata
RobotsTags robotsTags = new RobotsTags(metadata);
Map<String, List<String>> slinks;
String text;
DocumentFragment fragment;
try (ByteArrayInputStream bais = new ByteArrayInputStream(content)) {
org.jsoup.nodes.Document jsoupDoc = Jsoup.parse(bais, charset, url);
fragment = JSoupDOMBuilder.jsoup2HTML(jsoupDoc);
// extracts the robots directives from the meta tags
robotsTags.extractMetaTags(fragment);
// store a normalised representation in metadata
// so that the indexer is aware of it
robotsTags.normaliseToMetadata(metadata);
// do not extract the links if no follow has been set
// and we are in strict mode
if (robotsTags.isNoFollow() && robots_noFollow_strict) {
slinks = new HashMap<>(0);
} else {
Elements links = jsoupDoc.select("a[href]");
slinks = new HashMap<>(links.size());
for (Element link : links) {
// abs:href tells jsoup to return fully qualified domains
// for
// relative urls.
// e.g.: /foo will resolve to http://shopstyle.com/foo
String targetURL = link.attr("abs:href");
// nofollow
boolean noFollow = "nofollow".equalsIgnoreCase(link.attr("rel"));
// remove altogether
if (noFollow && robots_noFollow_strict) {
continue;
}
// link not specifically marked as no follow
// but whole page is
if (!noFollow && robotsTags.isNoFollow()) {
noFollow = true;
}
String anchor = link.text();
if (StringUtils.isNotBlank(targetURL)) {
// any existing anchors for the same target?
List<String> anchors = slinks.get(targetURL);
if (anchors == null) {
anchors = new LinkedList<>();
slinks.put(targetURL, anchors);
}
// track the anchors only if no follow is false
if (!noFollow && StringUtils.isNotBlank(anchor)) {
anchors.add(anchor);
}
}
}
}
text = jsoupDoc.body().text();
} catch (Throwable e) {
String errorMessage = "Exception while parsing " + url + ": " + e;
handleException(url, e, metadata, tuple, "content parsing", errorMessage);
return;
}
// store identified charset in md
metadata.setValue("parse.Content-Encoding", charset);
long duration = System.currentTimeMillis() - start;
LOG.info("Parsed {} in {} msec", url, duration);
List<Outlink> outlinks = toOutlinks(url, metadata, slinks);
ParseResult parse = new ParseResult();
parse.setOutlinks(outlinks);
// parse data of the parent URL
ParseData parseData = parse.get(url);
parseData.setMetadata(metadata);
parseData.setText(text);
parseData.setContent(content);
// apply the parse filters if any
try {
parseFilters.filter(url, content, fragment, parse);
} catch (RuntimeException e) {
String errorMessage = "Exception while running parse filters on " + url + ": " + e;
handleException(url, e, metadata, tuple, "content filtering", errorMessage);
return;
}
if (emitOutlinks) {
for (Outlink outlink : parse.getOutlinks()) {
collector.emit(
StatusStreamName,
tuple,
new Values(outlink.getTargetURL(), outlink.getMetadata(), Status.DISCOVERED));
}
}
// emit each document/subdocument in the ParseResult object
// there should be at least one ParseData item for the "parent" URL
for (Map.Entry<String, ParseData> doc : parse) {
ParseData parseDoc = doc.getValue();
collector.emit(
tuple,
new Values(
doc.getKey(), parseDoc.getContent(), parseDoc.getMetadata(), parseDoc.getText()));
}
collector.ack(tuple);
eventCounter.scope("tuple_success").incr();
}