boolean skipRecord(GoogleURL urlObject, Reporter reporter) {
if (_skipDomain) {
reporter.incrCounter(Counters.SKIPPING_BAD_DOMAIN_URL, 1);
return true;
}
if (!urlObject.isValid()) {
reporter.incrCounter(Counters.SKIPPING_INVALID_URL, 1);
return true;
} else if (urlObject.has_query()) {
reporter.incrCounter(Counters.HIT_QUERY_CHECK_CONDITION, 1);
if ((_flags & (HAS_HOMEPAGE_URLDATA | HAS_FEED_URLDATA)) == 0) {
reporter.incrCounter(Counters.SKIPPING_QUERY_URL, 1);
return true;
}
} else {
// if redirect ... skip
if ((_flags & HAS_REDIRECT_DATA) != 0) {
reporter.incrCounter(Counters.SKIPPING_REDIRECTED_URL, 1);
return true;
}
if ((_flags & (HAS_HOMEPAGE_URLDATA | HAS_FEED_URLDATA)) != 0) {
if (!_skipEverythingButHomepage || ((_flags & HAS_HOMEPAGE_URLDATA) != 0)) {
reporter.incrCounter(Counters.ALLOWING_HOMEPAGE_OR_FEEDURL, 1);
return false;
}
}
if (_skipEverythingButHomepage) {
reporter.incrCounter(Counters.SKIPPING_EVERYTHING_BUT_HOMEPAGE_URL, 1);
return true;
}
if (_crawlStatus != null) {
if (_crawlStatus.has("crawl_status")) {
JsonObject realCrawlStatus = _crawlStatus.get("crawl_status").getAsJsonObject();
if (realCrawlStatus.has("http_result")) {
int httpResult = realCrawlStatus.get("http_result").getAsInt();
if (httpResult == 200 || httpResult == 404) {
if ((_flags & HAS_BLOGPROBE_URLDATA) != 0) {
if (_blogURLSkipFlag.get()) {
reporter.incrCounter(Counters.SKIPPING_BLOGPROBE_URL, 1);
return true;
} else {
reporter.incrCounter(Counters.RECRAWLING_BLOGPROBE_URL, 1);
return false;
}
} else {
reporter.incrCounter(Counters.SKIPPING_ALREADY_FETCHED, 1);
return true;
}
}
}
}
}
}
return false;
}
void emitLastRecord(Reporter reporter) throws IOException {
if (_flags != 0) {
if (_domainStats == null) {
reporter.incrCounter(Counters.EMITTING_URL_RECORD_WITH_NULL_DOMAINSTATS, 1);
} else {
reporter.incrCounter(Counters.EMITTING_URL_RECORD_WITH_DOMINSTATS, 1);
}
if (_crawlStatus != null) {
reporter.incrCounter(Counters.EMITTED_RECORD_HAD_CRAWLSTATUS, 1);
} else {
reporter.incrCounter(Counters.EMITTED_RECORD_HAD_NULL_CRAWLSTATUS, 1);
}
}
if (_contextURLBytes.getLength() >= 4097) {
reporter.incrCounter(Counters.SKIPPING_INVALID_LENGTH_URL, 1);
} else {
GoogleURL urlObject = new GoogleURL(_contextURLBytes.toString());
if (!skipRecord(urlObject, reporter)) {
if (urlObject.has_query()) {
reporter.incrCounter(Counters.LET_THROUGH_QUERY_URL, 1);
}
URLFPV2 fp = URLUtils.getURLFPV2FromURLObject(urlObject);
if (fp != null) {
if (_emittedURLSFilter.isPresent(fp)) {
reporter.incrCounter(Counters.SKIPPING_ALREADY_EMITTED_URL, 1);
} else {
_emittedURLSFilter.add(fp);
_emittedURLSInFilter++;
SegmentGeneratorItem itemValue = new SegmentGeneratorItem();
itemValue.setDomainFP(fp.getDomainHash());
itemValue.setRootDomainFP(fp.getRootDomainHash());
itemValue.setUrlFP(fp.getUrlHash());
itemValue.setUrl(urlObject.getCanonicalURL());
itemValue.setPageRank(0);
itemValue.setModifiedStatus((byte) 0);
items.add(itemValue);
if (items.size() >= SPILL_THRESHOLD) spillItems(reporter);
}
} else {
reporter.incrCounter(Counters.NULL_FP_FOR_URL, 1);
}
}
}
// reset stuff
_flags = 0;
_crawlStatus = null;
_contextURLBytes.clear();
_blogURLSkipFlag.set(true);
}
