You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
114 lines
3.7 KiB
114 lines
3.7 KiB
package net.yacy.document.parser.augment;
|
|
|
|
import java.io.IOException;
|
|
import java.io.InputStream;
|
|
import java.util.HashSet;
|
|
import java.util.Iterator;
|
|
import java.util.Set;
|
|
|
|
import net.yacy.data.ymark.YMarkUtil;
|
|
import net.yacy.document.AbstractParser;
|
|
import net.yacy.document.Document;
|
|
import net.yacy.document.Parser;
|
|
import net.yacy.document.parser.rdfa.impl.RDFaParser;
|
|
import net.yacy.kelondro.data.meta.DigestURI;
|
|
import net.yacy.kelondro.logging.Log;
|
|
import net.yacy.search.Switchboard;
|
|
|
|
|
|
public class AugmentParser extends AbstractParser implements Parser {
|
|
|
|
RDFaParser rdfaParser;
|
|
|
|
public AugmentParser() {
|
|
super("AugmentParser");
|
|
this.rdfaParser = new RDFaParser();
|
|
|
|
Log.logInfo("AugmentedParser", "augmented parser was initialized");
|
|
|
|
this.SUPPORTED_EXTENSIONS.add("html");
|
|
this.SUPPORTED_EXTENSIONS.add("php");
|
|
this.SUPPORTED_MIME_TYPES.add("text/html");
|
|
this.SUPPORTED_MIME_TYPES.add("text/xhtml+xml");
|
|
this.SUPPORTED_EXTENSIONS.add("html");
|
|
this.SUPPORTED_EXTENSIONS.add("htm");
|
|
}
|
|
|
|
@Override
|
|
public Document[] parse(DigestURI url, String mimeType, String charset, InputStream source) throws Failure, InterruptedException {
|
|
|
|
Document[] htmlDocs = this.rdfaParser.parse(url, mimeType, charset, source);
|
|
try {
|
|
source.reset();
|
|
} catch (IOException e) {
|
|
Log.logException(e);
|
|
}
|
|
|
|
Document alreadyParsedDocument = htmlDocs[0];
|
|
Document superDoc = analyze(alreadyParsedDocument, url, mimeType, charset);
|
|
Document augmentDoc = parseAndAugment(url, mimeType, charset);
|
|
Document[] retDocs = new Document[htmlDocs.length + 1];
|
|
for (int i = 1; i < htmlDocs.length; i++) {
|
|
retDocs[i - 1] = htmlDocs[i];
|
|
}
|
|
|
|
retDocs[retDocs.length - 1] = augmentDoc;
|
|
retDocs[retDocs.length - 2] = superDoc;
|
|
try { // merge additional result docs into the parse main document
|
|
alreadyParsedDocument.addSubDocuments(retDocs);
|
|
} catch (IOException ex) {
|
|
Log.logException(ex);
|
|
}
|
|
Document[] finalretDocs = new Document[1]; // return the merged document
|
|
finalretDocs[0] = alreadyParsedDocument;
|
|
return finalretDocs;
|
|
}
|
|
|
|
private static Document analyze (Document alreadyParsedDocument, DigestURI url,
|
|
String mimeType, String charset) {
|
|
|
|
Document newDoc = new Document(url, mimeType, charset, null, null, null, singleList(""), "",
|
|
"", null, "", 0, 0, null, null, null, null, false);
|
|
|
|
// if the magic word appears in the document, perform extra actions.
|
|
if (alreadyParsedDocument.getKeywords().contains("magicword")) {
|
|
String all = "";
|
|
all = "yacylatest";
|
|
newDoc = new Document(url, mimeType, charset, null, null, null, singleList(""), "",
|
|
"", null, "", 0, 0, all, null, null, null, false);
|
|
}
|
|
|
|
return newDoc;
|
|
}
|
|
|
|
private Document parseAndAugment(DigestURI url, String mimeType, String charset) {
|
|
|
|
String all = "";
|
|
Document newDoc = new Document(url, mimeType, charset, null, null, null, singleList(""), "",
|
|
"", null, "", 0, 0, all, null, null, null, false);
|
|
|
|
Iterator<net.yacy.kelondro.blob.Tables.Row> it;
|
|
try {
|
|
it = Switchboard.getSwitchboard().tables.iterator("aggregatedtags");
|
|
it = Switchboard.getSwitchboard().tables.orderBy(it, -1, "timestamp_creation").iterator();
|
|
while (it.hasNext()) {
|
|
net.yacy.kelondro.blob.Tables.Row r = it.next();
|
|
if (r.get("url", "").equals (url.toNormalform(false))) {
|
|
Set<String> tags = new HashSet<String>();
|
|
for (String s : YMarkUtil.keysStringToSet(r.get("scitag", ""))) {
|
|
tags.add(s);
|
|
}
|
|
newDoc.addTags(tags);
|
|
}
|
|
}
|
|
|
|
} catch (IOException e) {
|
|
Log.logException(e);
|
|
}
|
|
|
|
return newDoc;
|
|
}
|
|
|
|
|
|
}
|