void setConfig(Configuration config) {
log.debug("config: " + config);
proxyHost = config.get(PARAM_PROXY_HOST);
proxyPort = config.getInt(PARAM_PROXY_PORT, DEFAULT_PROXY_PORT);
if (StringUtil.isNullString(proxyHost) || proxyPort <= 0) {
String http_proxy = System.getenv("http_proxy");
if (!StringUtil.isNullString(http_proxy)) {
try {
HostPortParser hpp = new HostPortParser(http_proxy);
proxyHost = hpp.getHost();
proxyPort = hpp.getPort();
} catch (HostPortParser.InvalidSpec e) {
log.warning("Can't parse http_proxy environment var, ignoring: " + http_proxy + ": " + e);
}
}
}
if (StringUtil.isNullString(proxyHost) || proxyPort <= 0) {
proxyHost = null;
} else {
log.info("Proxying through " + proxyHost + ":" + proxyPort);
}
userAgent = config.get(PARAM_USER_AGENT);
if (StringUtil.isNullString(userAgent)) {
userAgent = null;
} else {
log.debug("Setting User-Agent to " + userAgent);
}
}
public void testArticleCountAndType() throws Exception {
int expCount = 28;
PluginTestUtil.crawlSimAu(sau);
String pat1 = "branch(\\d+)/(\\d+file\\.html)";
String rep1 = "aps/journal/v123/n$1/full/$2";
PluginTestUtil.copyAu(sau, nau, ".*[^.][^p][^d][^f]$", pat1, rep1);
String pat2 = "branch(\\d+)/(\\d+file\\.pdf)";
String rep2 = "aps/journal/v123/n$1/pdf/$2";
PluginTestUtil.copyAu(sau, nau, ".*\\.pdf$", pat2, rep2);
// Remove some URLs
int deleted = 0;
for (Iterator it = nau.getAuCachedUrlSet().contentHashIterator(); it.hasNext(); ) {
CachedUrlSetNode cusn = (CachedUrlSetNode) it.next();
if (cusn instanceof CachedUrl) {
CachedUrl cu = (CachedUrl) cusn;
String url = cu.getUrl();
if (url.contains("/journal/")
&& (url.endsWith("1file.html") || url.endsWith("2file.pdf"))) {
deleteBlock(cu);
++deleted;
}
}
}
assertEquals(8, deleted);
Iterator<ArticleFiles> it = nau.getArticleIterator();
int count = 0;
int countHtmlOnly = 0;
int countPdfOnly = 0;
while (it.hasNext()) {
ArticleFiles af = it.next();
log.info(af.toString());
CachedUrl cu = af.getFullTextCu();
String url = cu.getUrl();
assertNotNull(cu);
String contentType = cu.getContentType();
log.debug("count " + count + " url " + url + " " + contentType);
count++;
if (af.getRoleUrl(ArticleFiles.ROLE_FULL_TEXT_PDF) == null) {
++countHtmlOnly;
}
if (af.getRoleUrl(ArticleFiles.ROLE_FULL_TEXT_PDF) == url) {
++countPdfOnly;
}
}
log.debug("Article count is " + count);
assertEquals(expCount, count);
assertEquals(4, countHtmlOnly);
assertEquals(4, countPdfOnly);
}
protected FilterRule constructFilterRule(String contentType) {
String mimeType = HeaderUtil.getMimeTypeFromContentType(contentType);
Object filter_el =
definitionMap.getMapElement(mimeType + DefinableArchivalUnit.SUFFIX_FILTER_RULE);
if (filter_el instanceof String) {
log.debug("Loading filter " + filter_el);
return (FilterRule) newAuxClass((String) filter_el, FilterRule.class);
} else if (filter_el instanceof List) {
if (((List) filter_el).size() > 0) {
return new DefinableFilterRule((List) filter_el);
}
}
return super.constructFilterRule(mimeType);
}
/** If in testing mode FOO, copy values from FOO_override map, if any, to main map */
void processOverrides(TypedEntryMap map) {
String testMode = getTestingMode();
if (StringUtil.isNullString(testMode)) {
return;
}
Object o = map.getMapElement(testMode + DefinableArchivalUnit.SUFFIX_OVERRIDE);
if (o == null) {
return;
}
if (o instanceof Map) {
Map overrideMap = (Map) o;
for (Map.Entry entry : (Set<Map.Entry>) overrideMap.entrySet()) {
String key = (String) entry.getKey();
Object val = entry.getValue();
log.debug(getDefaultPluginName() + ": Overriding " + key + " with " + val);
map.setMapElement(key, val);
}
}
}
protected void initMimeMap() throws PluginException.InvalidDefinition {
for (Iterator iter = definitionMap.entrySet().iterator(); iter.hasNext(); ) {
Map.Entry ent = (Map.Entry) iter.next();
String key = (String) ent.getKey();
Object val = ent.getValue();
if (key.endsWith(DefinableArchivalUnit.SUFFIX_LINK_EXTRACTOR_FACTORY)) {
String mime = stripSuffix(key, DefinableArchivalUnit.SUFFIX_LINK_EXTRACTOR_FACTORY);
if (val instanceof String) {
String factName = (String) val;
log.debug(mime + " link extractor: " + factName);
MimeTypeInfo.Mutable mti = mimeMap.modifyMimeTypeInfo(mime);
LinkExtractorFactory fact =
(LinkExtractorFactory) newAuxClass(factName, LinkExtractorFactory.class);
mti.setLinkExtractorFactory(fact);
}
} else if (key.endsWith(DefinableArchivalUnit.SUFFIX_CRAWL_FILTER_FACTORY)) {
// XXX This clause must precede the one for SUFFIX_HASH_FILTER_FACTORY
// XXX unless/until that key is changed to not be a terminal substring
// XXX of this one
String mime = stripSuffix(key, DefinableArchivalUnit.SUFFIX_CRAWL_FILTER_FACTORY);
if (val instanceof String) {
String factName = (String) val;
log.debug(mime + " crawl filter: " + factName);
MimeTypeInfo.Mutable mti = mimeMap.modifyMimeTypeInfo(mime);
FilterFactory fact = (FilterFactory) newAuxClass(factName, FilterFactory.class);
mti.setCrawlFilterFactory(fact);
}
} else if (key.endsWith(DefinableArchivalUnit.SUFFIX_HASH_FILTER_FACTORY)) {
String mime = stripSuffix(key, DefinableArchivalUnit.SUFFIX_HASH_FILTER_FACTORY);
if (val instanceof String) {
String factName = (String) val;
log.debug(mime + " filter: " + factName);
MimeTypeInfo.Mutable mti = mimeMap.modifyMimeTypeInfo(mime);
FilterFactory fact = (FilterFactory) newAuxClass(factName, FilterFactory.class);
mti.setHashFilterFactory(fact);
}
} else if (key.endsWith(DefinableArchivalUnit.SUFFIX_FETCH_RATE_LIMIT)) {
String mime = stripSuffix(key, DefinableArchivalUnit.SUFFIX_FETCH_RATE_LIMIT);
if (val instanceof String) {
String rate = (String) val;
log.debug(mime + " fetch rate: " + rate);
MimeTypeInfo.Mutable mti = mimeMap.modifyMimeTypeInfo(mime);
RateLimiter limit = mti.getFetchRateLimiter();
if (limit != null) {
limit.setRate(rate);
} else {
mti.setFetchRateLimiter(new RateLimiter(rate));
}
}
} else if (key.endsWith(DefinableArchivalUnit.SUFFIX_LINK_REWRITER_FACTORY)) {
String mime = stripSuffix(key, DefinableArchivalUnit.SUFFIX_LINK_REWRITER_FACTORY);
String factName = (String) val;
log.debug(mime + " link rewriter: " + factName);
MimeTypeInfo.Mutable mti = mimeMap.modifyMimeTypeInfo(mime);
LinkRewriterFactory fact =
(LinkRewriterFactory) newAuxClass(factName, LinkRewriterFactory.class);
mti.setLinkRewriterFactory(fact);
} else if (key.endsWith(DefinableArchivalUnit.SUFFIX_METADATA_EXTRACTOR_FACTORY_MAP)) {
String mime = stripSuffix(key, DefinableArchivalUnit.SUFFIX_METADATA_EXTRACTOR_FACTORY_MAP);
Map factNameMap = (Map) val;
Map factClassMap = new HashMap();
MimeTypeInfo.Mutable mti = mimeMap.modifyMimeTypeInfo(mime);
for (Iterator it = factNameMap.keySet().iterator(); it.hasNext(); ) {
String mdTypes = (String) it.next();
String factName = (String) factNameMap.get(mdTypes);
log.debug(mime + " (" + mdTypes + ") metadata extractor: " + factName);
for (String mdType : (List<String>) StringUtil.breakAt(mdTypes, ";")) {
setMdTypeFact(factClassMap, mdType, factName);
}
}
mti.setFileMetadataExtractorFactoryMap(factClassMap);
}
}
}