use of org.corpus_tools.annis.gui.widgets.grid.Row in project ANNIS by korpling.
the class EventExtractor method mergeAllRowsIfPossible.
/**
* Merges the rows. This function uses a heuristical approach that guarantiess to merge all rows
* into one if there is no conflict at all. If there are conflicts the heuristic will be
* best-efford but with linear runtime (given a a number of rows).
*
* @param rows Will be altered, if no conflicts occcured this wil have only one element.
*/
private static void mergeAllRowsIfPossible(ArrayList<Row> rows) {
// use fixed seed in order to get consistent results (with random properties)
Random rand = new Random(5711l);
int tries = 0;
// this should be enough to be quite sure we don't miss any optimization
// possibility
final int maxTries = rows.size() * 2;
// or we give up until too much tries
while (rows.size() > 1 && tries < maxTries) {
// choose two random entries
int oneIdx = rand.nextInt(rows.size());
int secondIdx = rand.nextInt(rows.size());
if (oneIdx == secondIdx) {
// try again if we choose the same rows by accident
continue;
}
Row one = rows.get(oneIdx);
Row second = rows.get(secondIdx);
if (one.merge(second)) {
// remove the second one since it is merged into the first
rows.remove(secondIdx);
// success: reset counter
tries = 0;
} else {
// increase counter to avoid endless loops if no improvement is possible
tries++;
}
}
}
use of org.corpus_tools.annis.gui.widgets.grid.Row in project ANNIS by korpling.
the class SingleGridComponent method createAnnotationGrid.
private boolean createAnnotationGrid() {
String resultID = input.getId();
grid = new AnnotationGrid(mediaController, pdfController, resultID);
grid.addStyleName(getMainStyle());
grid.addStyleName(Helper.CORPUS_FONT_FORCE);
grid.setEscapeHTML(Boolean.parseBoolean(input.getMappings().getOrDefault(MAPPING_ESCAPE_HTML, "true")));
LinkedList<Class<? extends SNode>> types = new LinkedList<>();
if (isShowingSpanAnnotations()) {
types.add(SSpan.class);
}
if (isShowingTokenAnnotations()) {
types.add(SToken.class);
}
grid.setAnnosWithNamespace(EventExtractor.computeDisplayedNamespace(input, types));
layout.addComponent(grid);
SDocumentGraph graph = input.getDocument().getDocumentGraph();
List<SNode> sortedSegmentationNodes = Helper.getSortedSegmentationNodes(this.segmentationName, graph);
Map<SToken, Integer> token2index = Helper.createToken2IndexMap(graph, this.enforcedText);
Preconditions.checkArgument(!token2index.isEmpty(), "Token list must be non-empty");
LinkedHashMap<String, ArrayList<Row>> rowsByAnnotation = computeAnnotationRows(token2index);
// Get Mappings
String gridTemplates = input.getMappings().getOrDefault(MAPPING_GRID_TEMPLATES, "");
// Parse Mappings
if (!gridTemplates.equals("")) {
String[] split = gridTemplates.split("\\|\\|");
for (String s : split) {
// example of s: entity="person"==>:), or infstat==><b>%%value%%</b>
String[] unit_split = s.split("==>");
Set<Map.Entry<String, ArrayList<Row>>> set = rowsByAnnotation.entrySet();
// Displaying elements of LinkedHashMap
Iterator<Map.Entry<String, ArrayList<Row>>> iterator = set.iterator();
while (iterator.hasNext()) {
// iterate over rows
Map.Entry<String, ArrayList<Row>> me = iterator.next();
String rowKey = me.getKey();
ArrayList<Row> rowValue = me.getValue();
for (Row rowValue1 : rowValue) {
ArrayList<GridEvent> rowEvents = rowValue1.getEvents();
if (unit_split[0].indexOf('=') < 0) {
// unit_split[0] is a single instruction, e.g., infstat
// check if the key of a row in rowsByAnnotation is unit_split[0]
// if it is, we need to change every value of this row, else we dont do anything
String rowName = rowKey.split("::")[1];
if (rowName.equals(unit_split[0])) {
// iterate over all values and replace the value with the unit_split[1]
for (GridEvent ev : rowEvents) {
String origValue = ev.getValue();
String newValue = unit_split[1].replaceAll("%%value%%", origValue);
ev.setValue(newValue);
}
}
} else {
// its a instruction like entity='person'
// first break this split into entity and person
// check if rowKey is entity, then when iterating over events, check if value is
// person
String rowName = rowKey.split("::")[1];
String targetRow = unit_split[0].split("=")[0];
String targetValue = unit_split[0].split("=")[1].replaceAll("\"", "");
if (rowName.equals(targetRow)) {
// iterate over all values and replace the value with the unit_split[1]
for (GridEvent ev : rowEvents) {
String origValue = ev.getValue();
if (origValue.equals(targetValue)) {
ev.setValue(unit_split[1]);
}
// String newValue = unit_split[1].replaceAll("%%value%%",origValue);
}
}
}
}
}
}
}
// add tokens as row
Row tokenRow = computeTokenRow(sortedSegmentationNodes, graph, rowsByAnnotation, token2index);
String tokenRowCaption = "tok";
if (isHidingToken()) {
// We have to add the invisible token row avoid issues with the layout
// (see https://github.com/korpling/ANNIS/issues/524)
// but we don't want the invisible token layer to override an actual "tok"
// annotation layer (see https://github.com/korpling/ANNIS/issues/596)
tokenRow.setStyle("invisible_token");
tokenRowCaption = "";
grid.setTokRowKey("");
}
if (isTokenFirst()) {
// copy original list but add token row at the beginning
LinkedHashMap<String, ArrayList<Row>> newList = new LinkedHashMap<>();
newList.put(tokenRowCaption, Lists.newArrayList(tokenRow));
newList.putAll(rowsByAnnotation);
rowsByAnnotation = newList;
} else {
// just add the token row to the end of the list
rowsByAnnotation.put(tokenRowCaption, Lists.newArrayList(tokenRow));
}
EventExtractor.removeEmptySpace(rowsByAnnotation, tokenRow);
// check if the token row only contains empty values
boolean tokenRowIsEmpty = true;
for (GridEvent tokenEvent : tokenRow.getEvents()) {
if (tokenEvent.getValue() != null && !tokenEvent.getValue().trim().isEmpty()) {
tokenRowIsEmpty = false;
break;
}
}
if (!isHidingToken() && canShowEmptyTokenWarning()) {
lblEmptyToken.setVisible(tokenRowIsEmpty);
}
grid.setRowsByAnnotation(rowsByAnnotation);
return !tokenRowIsEmpty;
}
use of org.corpus_tools.annis.gui.widgets.grid.Row in project ANNIS by korpling.
the class SingleGridComponent method computeTokenRow.
private Row computeTokenRow(List<SNode> tokens, SDocumentGraph graph, LinkedHashMap<String, ArrayList<Row>> rowsByAnnotation, Map<SToken, Integer> token2index) {
/*
* we will only add tokens of one texts which is mentioned by any included annotation.
*/
Set<String> validTextIDs = new HashSet<>();
if (enforcedText == null) {
Iterator<ArrayList<Row>> itAllRows = rowsByAnnotation.values().iterator();
while (itAllRows.hasNext()) {
ArrayList<Row> rowsForAnnotation = itAllRows.next();
for (Row r : rowsForAnnotation) {
validTextIDs.addAll(r.getTextIDs());
}
}
/**
* we want to show all token if no valid text was found and we have only one text and the
* first one if there is more than one text.
*/
List<STextualDS> allTexts = graph.getTextualDSs();
if (validTextIDs.isEmpty() && allTexts != null && (allTexts.size() == 1 || allTexts.size() == 2)) {
validTextIDs.add(allTexts.get(0).getId());
}
} else {
validTextIDs.add(enforcedText.getId());
}
Row tokenRow = new Row();
for (SNode t : tokens) {
// get the Salt ID of the STextualDS of this token
STextualDS tokenText = Helper.getTextualDSForNode(t, graph);
// only add token if text ID matches the valid one
if (tokenText != null && validTextIDs.contains(tokenText.getId()) && hasSegmentation(t, this.segmentationName)) {
Range<Integer> coveredRange = Helper.getLeftRightSpan(t, graph, token2index);
String text = extractTextForToken(t, segmentationName);
GridEvent event = new GridEvent(t.getId(), coveredRange.lowerEndpoint(), coveredRange.upperEndpoint(), text);
event.setTextID(tokenText.getId());
// check if the token is a matched node
Long match = isCoveredTokenMarked() ? markCoveredTokens(input.getMarkedAndCovered(), t) : tokenMatch(t);
event.setMatch(match);
tokenRow.addEvent(event);
}
}
return tokenRow;
}
use of org.corpus_tools.annis.gui.widgets.grid.Row in project ANNIS by korpling.
the class EventExtractor method parseSalt.
/**
* Converts Salt document graph to rows.
*
* @param input
* @param showSpanAnnos
* @param showTokenAnnos
* @param mediaLayer A set of all annotation layers which should be treated as special media
* layer.
* @param annotationNames
* @param replaceValueWithMediaIcon If true the actual value is removed and an icon for playing
* the media file is shown instead.
* @param token2index
* @param pdfController makes status of all pdfviewer available for the events.
* @param text If non-null only include annotations for nodes of the specified text.
* @return
*/
public static LinkedHashMap<String, ArrayList<Row>> parseSalt(VisualizerInput input, boolean showSpanAnnos, boolean showTokenAnnos, List<String> annotationNames, Set<String> mediaLayer, boolean replaceValueWithMediaIcon, Map<SToken, Integer> token2index, PDFController pdfController, STextualDS text) {
SDocumentGraph graph = input.getDocument().getDocumentGraph();
// only look at annotations which were defined by the user
LinkedHashMap<String, ArrayList<Row>> rowsByAnnotation = new LinkedHashMap<>();
for (String anno : annotationNames) {
rowsByAnnotation.put(anno, new ArrayList<Row>());
}
AtomicInteger eventCounter = new AtomicInteger();
PDFPageHelper pageNumberHelper = new PDFPageHelper(input);
if (showSpanAnnos) {
for (SSpan span : graph.getSpans()) {
if (text == null || text == Helper.getTextualDSForNode(span, graph)) {
addAnnotationsForNode(span, graph, token2index, pdfController, pageNumberHelper, eventCounter, rowsByAnnotation, true, mediaLayer, replaceValueWithMediaIcon);
}
}
// end for each span
}
if (showTokenAnnos) {
for (SToken tok : graph.getTokens()) {
if (text == null || text == Helper.getTextualDSForNode(tok, graph)) {
addAnnotationsForNode(tok, graph, token2index, pdfController, pageNumberHelper, eventCounter, rowsByAnnotation, false, mediaLayer, replaceValueWithMediaIcon);
}
}
}
// 2. merge rows when possible
for (Map.Entry<String, ArrayList<Row>> e : rowsByAnnotation.entrySet()) {
mergeAllRowsIfPossible(e.getValue());
}
// 3. sort events on one row by left token index
for (Map.Entry<String, ArrayList<Row>> e : rowsByAnnotation.entrySet()) {
for (Row r : e.getValue()) {
sortEventsByTokenIndex(r);
}
}
// 4. split up events if they cover islands
for (Map.Entry<String, ArrayList<Row>> e : rowsByAnnotation.entrySet()) {
for (Row r : e.getValue()) {
splitRowsOnIslands(r, graph, text, token2index);
}
}
// 5. split up events if they have gaps
for (Map.Entry<String, ArrayList<Row>> e : rowsByAnnotation.entrySet()) {
for (Row r : e.getValue()) {
splitRowsOnGaps(r, graph, token2index);
}
}
return rowsByAnnotation;
}
use of org.corpus_tools.annis.gui.widgets.grid.Row in project ANNIS by korpling.
the class EventExtractor method removeEmptySpace.
public static void removeEmptySpace(LinkedHashMap<String, ArrayList<Row>> rowsByAnnotation, Row tokenRow) {
List<Range<Integer>> gaps = new LinkedList<>();
BitSet totalOccupancyGrid = new BitSet();
for (Map.Entry<String, ArrayList<Row>> layer : rowsByAnnotation.entrySet()) {
for (Row r : layer.getValue()) {
totalOccupancyGrid.or(r.getOccupancyGridCopy());
}
}
// bug report.
if (tokenRow != null) {
totalOccupancyGrid.or(tokenRow.getOccupancyGridCopy());
}
// The Range class can give us the next bit that is not set. Use this
// to detect gaps. A gap starts from the next non-set bit and goes to
// the next set bit.
Range<Integer> gap = Range.closed(-1, totalOccupancyGrid.nextSetBit(0));
while (true) {
int gapStart = totalOccupancyGrid.nextClearBit(gap.upperEndpoint() + 1);
int gapEnd = totalOccupancyGrid.nextSetBit(gapStart);
if (gapEnd <= 0) {
break;
}
gap = Range.closed(gapStart, gapEnd - 1);
gaps.add(gap);
}
int gapID = 0;
int totalOffset = 0;
for (Range<Integer> gRaw : gaps) {
// adjust the space range itself
Range<Integer> g = Range.closed(gRaw.lowerEndpoint() - totalOffset, gRaw.upperEndpoint() - totalOffset);
int offset = g.upperEndpoint() - g.lowerEndpoint();
totalOffset += offset;
for (Entry<String, ArrayList<Row>> rowEntry : rowsByAnnotation.entrySet()) {
ArrayList<Row> rows = rowEntry.getValue();
for (Row r : rows) {
List<GridEvent> eventsCopy = new LinkedList<>(r.getEvents());
for (GridEvent e : eventsCopy) {
if (e.getLeft() >= g.upperEndpoint()) {
r.removeEvent(e);
e.setLeft(e.getLeft() - offset);
e.setRight(e.getRight() - offset);
r.addEvent(e);
}
}
// add a special space event
String spaceCaption = "";
if ("tok".equalsIgnoreCase(rowEntry.getKey())) {
spaceCaption = "(...)";
}
GridEvent spaceEvent = new GridEvent("gap-" + gapID, g.lowerEndpoint(), g.lowerEndpoint(), spaceCaption);
spaceEvent.setSpace(true);
r.addEvent(spaceEvent);
gapID++;
}
}
}
}
Aggregations