void setDomainStats(JsonObject domainStats, Reporter reporter) throws IOException {
_domainStats = domainStats;
if (_domainStats.has("dR")) {
_domainRank = _domainStats.get("dR").getAsDouble();
} else {
_domainRank = 0.0;
}
if (_domainStats.has("urls")) {
int urlCount = _domainStats.get("urls").getAsInt();
int crawledCount = _domainStats.get("crawled").getAsInt();
int Http200Count = (_domainStats.has("200")) ? _domainStats.get("200").getAsInt() : 0;
if (urlCount != 0 && crawledCount != 0 && Http200Count == 0) {
reporter.incrCounter(Counters.SKIPPING_BAD_DOMAIN_BASED_ON_CRAWL_HISTORY, 1);
LOG.info(
"Skipping Everything But Homepage for Domain:"
+ _newDomainBytes.toString()
+ " CrawledCount:"
+ crawledCount
+ " HTTP200Count:"
+ Http200Count
+ " URLCount:"
+ urlCount);
_skipEverythingButHomepage = true;
} else if (urlCount > 25000 && urlCount < 100000) {
if (!_domainStats.has("dR") || _domainStats.get("dR").getAsDouble() < 3.0) {
LOG.info("Skipping Domain:" + _newDomainBytes.toString());
reporter.incrCounter(Counters.SKIPPING_DOMAIN_EXCEEDED_URL_COUNT_AND_LOW_DR, 1);
_skipDomain = true;
}
} else if (urlCount > 250000 && urlCount < 1000000) {
if (!_domainStats.has("dR") || _domainStats.get("dR").getAsDouble() < 4.0) {
LOG.info("Skipping Domain:" + _newDomainBytes.toString());
reporter.incrCounter(Counters.SKIPPING_DOMAIN_EXCEEDED_URL_COUNT_AND_LOW_DR, 1);
_skipDomain = true;
}
} else if (urlCount > 1000000) {
if (!_domainStats.has("dR") || _domainStats.get("dR").getAsDouble() < 5.0) {
LOG.info("Skipping Domain:" + _newDomainBytes.toString());
reporter.incrCounter(Counters.SKIPPING_DOMAIN_EXCEEDED_URL_COUNT_AND_LOW_DR, 1);
_skipDomain = true;
}
}
}
if (_emittedURLSInFilter >= FLUSH_THRESHOLD) {
_emittedURLSFilter.clear();
_emittedURLSInFilter = 0;
reporter.incrCounter(Counters.FLUSHED_BLOOMFILTER, 1);
}
}
