package werti.uima.ae;

import java.util.Iterator;
import java.util.regex.Pattern;

import org.apache.log4j.Logger;
import org.apache.uima.analysis_component.JCasAnnotator_ImplBase;
import org.apache.uima.analysis_engine.AnalysisEngineProcessException;
import org.apache.uima.cas.text.AnnotationIndex;
import org.apache.uima.jcas.JCas;
import org.apache.uima.cas.FSIterator;

import werti.uima.types.annot.PlainTextSentenceAnnotation;
import werti.uima.types.annot.RelevantText;
import werti.uima.types.annot.SentenceAnnotation;
import werti.uima.types.annot.Token;

/**
 * Wrapper for OpenNLP sentence detector.
 * 
 * Depends on {@link Token} Token annotation from {@link OpenNlpTokenizer}.
 * 
 * @author Adriane Boyd
 */
public class HTMLSentenceAnnotator extends JCasAnnotator_ImplBase {

	private static final Logger log =
		Logger.getLogger(HTMLSentenceAnnotator.class);
	
	// HTML tags that typically indicate sentence breaks, but not necessarily
	// a shift in content type
	private static Pattern htmlBreakPattern = Pattern.compile(".*(<li|</li>|<ul|</ul>|<ol|</ol>|<h[1..6]|</h[1-6]).*", Pattern.DOTALL); // heading1..6 added by Heli
	
	@SuppressWarnings("unchecked")
	@Override
	public void process(JCas jcas) throws AnalysisEngineProcessException {
		log.info("Starting HTML sentence detection");
		final AnnotationIndex sentIndex = jcas.getAnnotationIndex(PlainTextSentenceAnnotation.type);
		final AnnotationIndex rtIndex = jcas.getAnnotationIndex(RelevantText.type);
				
		final Iterator <PlainTextSentenceAnnotation> sit = sentIndex.iterator();
		
		while (sit.hasNext()) {
			final PlainTextSentenceAnnotation s = sit.next();
			final Iterator <RelevantText> rtit = rtIndex.subiterator(s, true, false);

			//String prevRTContentType = "";	// content of previous text span (to detect a change)
			int prevRTEnd = 0;				// end of previous text span in the loop
			int currentSentStart = -1;		// start of current new sentence under consideration 
			                                //   (may span multiple text segments)
			int lastAddedSentEnd = -1;		// end of last sentence added
			PlainTextSentenceAnnotation lastS = null;
			while (rtit.hasNext()) {
				final RelevantText t = rtit.next();
				//log.info("Relevant text "+t.getCoveredText());
				
				// initialize in first loop
				if (currentSentStart == -1) {
					currentSentStart = t.getBegin();
					//prevRTContentType = t.getHtmlContentType();
					prevRTEnd = s.getBegin();
				}
				
				// if the current type is different from the previous type,
				// add a sentence covering the last saved start to the end of
				// the previous added sentence
				/*final String currContentType = t.getHtmlContentType();
				if (!prevRTContentType.equals(currContentType)) {
					SentenceAnnotation sentence = new SentenceAnnotation(jcas, currentSentStart, prevRTEnd);
					sentence.addToIndexes();
					currentSentStart = t.getBegin();
					lastAddedSentEnd = prevRTEnd;
					prevRTContentType = currContentType;
				}*/
				
								
				// if a sentence boundary was not just added but any of the 
				// HTML tags in the pattern appear between the previous rt span and 
				// the current one, insert a sentence boundary
				if (currentSentStart != t.getBegin() && htmlBreakPattern.matcher(jcas.getDocumentText().substring(prevRTEnd, t.getBegin()).toLowerCase()).matches()) {
					//log.info("sentence boundary from HTML");
					SentenceAnnotation sentence = new SentenceAnnotation(jcas, currentSentStart, prevRTEnd);
					sentence.addToIndexes();
					currentSentStart = t.getBegin();
					lastAddedSentEnd = prevRTEnd;
					//prevRTContentType = currContentType;
				}
				
				prevRTEnd = t.getEnd();
				lastS = s;
			}

			// if no sentences were added (because the whole sentence 
			// corresponded to a single RelevantText span or all spans were 
			// of the same type), add the whole original plain text sentence 
			// as a sentence
			if (lastAddedSentEnd == -1) {
				SentenceAnnotation sentence = new SentenceAnnotation(jcas, s.getBegin(), s.getEnd());
				sentence.addToIndexes();
			// add the last sentence if needed
			} else if (lastS != null && lastAddedSentEnd != lastS.getEnd()) {
				SentenceAnnotation sentence = new SentenceAnnotation(jcas, currentSentStart, lastS.getEnd());
				sentence.addToIndexes();
			}
		}

		log.info("Finished HTML sentence detection");
	}	
}