/**
* Splits events of a row if they overlap an island. Islands are areas between the token which are
* included in the result.
*
* @param row
* @param graph
* @param text
* @param startTokenIndex token index of the first token in the match
* @param endTokenIndex token index of the last token in the match
*/
private static void splitRowsOnIslands(
Row row,
final SDocumentGraph graph,
STextualDS text,
long startTokenIndex,
long endTokenIndex) {
BitSet tokenCoverage = new BitSet();
// get the sorted token
List<SToken> sortedTokenList = graph.getSortedTokenByText();
// add all token belonging to the right text to the bit set
ListIterator<SToken> itToken = sortedTokenList.listIterator();
while (itToken.hasNext()) {
SToken t = itToken.next();
if (text == null || text == CommonHelper.getTextualDSForNode(t, graph)) {
RelannisNodeFeature feat =
(RelannisNodeFeature) t.getFeature(ANNIS_NS, FEAT_RELANNIS_NODE).getValue();
long tokenIndexRaw = feat.getTokenIndex();
tokenIndexRaw = clip(tokenIndexRaw, startTokenIndex, endTokenIndex);
int tokenIndex = (int) (tokenIndexRaw - startTokenIndex);
tokenCoverage.set(tokenIndex);
}
}
ListIterator<GridEvent> itEvents = row.getEvents().listIterator();
while (itEvents.hasNext()) {
GridEvent event = itEvents.next();
BitSet eventBitSet = new BitSet();
eventBitSet.set(event.getLeft(), event.getRight() + 1);
// restrict event bitset on the locations where token are present
eventBitSet.and(tokenCoverage);
// if there is is any 0 bit before the right border there is a break in the event
// and we need to split it
if (eventBitSet.nextClearBit(event.getLeft()) <= event.getRight()) {
// remove the original event
row.removeEvent(itEvents);
// The event bitset now marks all the locations which the event should
// cover.
// Make a list of new events for each connected range in the bitset
int subElement = 0;
int offset = eventBitSet.nextSetBit(0);
while (offset >= 0) {
int end = eventBitSet.nextClearBit(offset) - 1;
if (offset < end) {
GridEvent newEvent = new GridEvent(event);
newEvent.setId(event.getId() + "_islandsplit_" + subElement++);
newEvent.setLeft(offset);
newEvent.setRight(end);
row.addEvent(itEvents, newEvent);
}
offset = eventBitSet.nextSetBit(end + 1);
}
} // end if we need to split
}
}
private static String getTextID(SToken tok, SDocumentGraph graph) {
List<? extends SRelation<? extends SNode, ? extends SNode>> tokenOutEdges =
graph.getOutRelations(tok.getId());
if (tokenOutEdges != null) {
for (SRelation<? extends SNode, ? extends SNode> tokEdge : tokenOutEdges) {
if (tokEdge instanceof STextualRelation) {
return ((STextualRelation) tokEdge).getTarget().getId();
}
}
}
return null;
}
private static void addAnnotationsForNode(
SNode node,
SDocumentGraph graph,
long startTokenIndex,
long endTokenIndex,
PDFController pdfController,
PDFPageHelper pageNumberHelper,
AtomicInteger eventCounter,
LinkedHashMap<String, ArrayList<Row>> rowsByAnnotation,
boolean addMatch,
Set<String> mediaLayer,
boolean replaceValueWithMediaIcon) {
List<String> matchedAnnos = new ArrayList<>();
SFeature featMatchedAnnos = graph.getFeature(ANNIS_NS, FEAT_MATCHEDANNOS);
if (featMatchedAnnos != null) {
matchedAnnos = Splitter.on(',').trimResults().splitToList(featMatchedAnnos.getValue_STEXT());
}
// check if the span is a matched node
SFeature featMatched = node.getFeature(ANNIS_NS, FEAT_MATCHEDNODE);
Long matchRaw = featMatched == null ? null : featMatched.getValue_SNUMERIC();
String matchedQualifiedAnnoName = "";
if (matchRaw != null && matchRaw <= matchedAnnos.size()) {
matchedQualifiedAnnoName = matchedAnnos.get((int) ((long) matchRaw) - 1);
}
// calculate the left and right values of a span
// TODO: howto get these numbers with Salt?
RelannisNodeFeature feat =
(RelannisNodeFeature) node.getFeature(ANNIS_NS, FEAT_RELANNIS_NODE).getValue();
long leftLong = feat.getLeftToken();
long rightLong = feat.getRightToken();
leftLong = clip(leftLong, startTokenIndex, endTokenIndex);
rightLong = clip(rightLong, startTokenIndex, endTokenIndex);
int left = (int) (leftLong - startTokenIndex);
int right = (int) (rightLong - startTokenIndex);
for (SAnnotation anno : node.getAnnotations()) {
ArrayList<Row> rows = rowsByAnnotation.get(anno.getQName());
if (rows == null) {
// try again with only the name
rows = rowsByAnnotation.get(anno.getName());
}
if (rows != null) {
// only do something if the annotation was defined before
// 1. give each annotation of each span an own row
Row r = new Row();
String id = "event_" + eventCounter.incrementAndGet();
GridEvent event = new GridEvent(id, left, right, anno.getValue_STEXT());
event.setTooltip(Helper.getQualifiedName(anno));
if (addMatch && matchRaw != null) {
long match = matchRaw;
if (matchedQualifiedAnnoName.isEmpty()) {
// always set the match when there is no matched annotation at all
event.setMatch(match);
}
// check if the annotation also matches
else if (matchedQualifiedAnnoName.equals(anno.getQName())) {
event.setMatch(match);
}
}
if (node instanceof SSpan) {
// calculate overlapped SToken
List<? extends SRelation<? extends SNode, ? extends SNode>> outEdges =
graph.getOutRelations(node.getId());
if (outEdges != null) {
for (SRelation<? extends SNode, ? extends SNode> e : outEdges) {
if (e instanceof SSpanningRelation) {
SSpanningRelation spanRel = (SSpanningRelation) e;
SToken tok = spanRel.getTarget();
event.getCoveredIDs().add(tok.getId());
// get the STextualDS of this token and add it to the event
String textID = getTextID(tok, graph);
if (textID != null) {
event.setTextID(textID);
}
}
}
} // end if span has out edges
} else if (node instanceof SToken) {
event.getCoveredIDs().add(node.getId());
// get the STextualDS of this token and add it to the event
String textID = getTextID((SToken) node, graph);
if (textID != null) {
event.setTextID(textID);
}
}
// try to get time annotations
if (mediaLayer == null || mediaLayer.contains(anno.getQName())) {
double[] startEndTime = TimeHelper.getOverlappedTime(node);
if (startEndTime.length == 1) {
if (replaceValueWithMediaIcon) {
event.setValue(" ");
event.setTooltip("play excerpt " + event.getStartTime());
}
event.setStartTime(startEndTime[0]);
} else if (startEndTime.length == 2) {
event.setStartTime(startEndTime[0]);
event.setEndTime(startEndTime[1]);
if (replaceValueWithMediaIcon) {
event.setValue(" ");
event.setTooltip("play excerpt " + event.getStartTime() + "-" + event.getEndTime());
}
}
}
r.addEvent(event);
rows.add(r);
if (pdfController != null && pdfController.sizeOfRegisterdPDFViewer() > 0) {
String page = pageNumberHelper.getPageFromAnnotation(node);
if (page != null) {
event.setPage(page);
}
}
}
} // end for each annotation of span
}
