public void endDocument() throws SAXException {
// time to actually set up the edges
IntIterator rows = m_edges.rows();
while (rows.hasNext()) {
int r = rows.nextInt();
String src = m_edges.getString(r, SRCID);
if (!m_nodeMap.containsKey(src)) {
throw new SAXException(
"Tried to create edge with source node id=" + src + " which does not exist.");
}
int s = ((Integer) m_nodeMap.get(src)).intValue();
m_edges.setInt(r, SRC, s);
String trg = m_edges.getString(r, TRGID);
if (!m_nodeMap.containsKey(trg)) {
throw new SAXException(
"Tried to create edge with target node id=" + trg + " which does not exist.");
}
int t = ((Integer) m_nodeMap.get(trg)).intValue();
m_edges.setInt(r, TRG, t);
}
m_edges.removeColumn(SRCID);
m_edges.removeColumn(TRGID);
// now create the graph
m_graph = new Graph(m_nodes, m_edges, m_directed);
if (m_graphid != null) m_graph.putClientProperty(ID, m_graphid);
}
private float compareNodesCaseSensitiveBy(
String attributeColumn, Integer nodeOneIndex, Integer nodeTwoIndex, Table nodeTable) {
String nodeOneAttribute =
(String) nodeTable.getString(nodeOneIndex.intValue(), attributeColumn);
String nodeTwoAttribute =
(String) nodeTable.getString(nodeTwoIndex.intValue(), attributeColumn);
float similarity = this.similarityChecker.getSimilarity(nodeOneAttribute, nodeTwoAttribute);
return similarity;
}
/** @see prefuse.data.io.TableWriter#writeTable(prefuse.data.Table, java.io.OutputStream) */
public void writeTable(Table table, OutputStream os) throws DataIOException {
try {
// get print stream
PrintWriter out = new PrintWriter(new OutputStreamWriter(os, "UTF-8"));
// write out header row
if (m_printHeader) {
for (int i = 0; i < table.getColumnCount(); ++i) {
if (i > 0) out.print(',');
out.print(makeCSVSafe(table.getColumnName(i)));
}
out.println();
}
// write out data
for (IntIterator rows = table.rows(); rows.hasNext(); ) {
int row = rows.nextInt();
for (int i = 0; i < table.getColumnCount(); ++i) {
if (i > 0) out.print(',');
String str = table.getString(row, table.getColumnName(i));
out.print(makeCSVSafe(str));
}
out.println();
}
// finish up
out.flush();
} catch (Exception e) {
throw new DataIOException(e);
}
}
// this mutates nodeTable
private Table makeMergeTable(Table nodeTable, Graph mergeGraph, StringBuffer mergeLog) {
// get all the clusters in our merge graph (where each cluster is a group of nodes to be merged)
List clusters = extractWeakComponentClusters(mergeGraph);
// for each cluster...
for (Iterator clusterIt = clusters.iterator(); clusterIt.hasNext(); ) {
LinkedHashSet cluster = (LinkedHashSet) clusterIt.next();
// mark that we will merge every node into a single node
// (this step could be made smarter, but is ok for now. Will need user intervention in some
// cases)
Integer[] eachNodeInCluster = (Integer[]) cluster.toArray(new Integer[cluster.size()]);
Integer firstNode = null;
if (eachNodeInCluster.length <= 1) continue;
for (int ii = 0; ii < eachNodeInCluster.length; ii++) {
Integer node = eachNodeInCluster[ii];
if (firstNode == null) {
// (we arbitrarily choose the first node as the node to merge other nodes into)
firstNode = node;
String nodeOneAttribute =
(String) nodeTable.getString(firstNode.intValue(), this.compareAttributeName);
mergeLog.append("for \"" + nodeOneAttribute + "\"..." + "\r\n");
} else {
Integer nodeBeyondFirst = node;
// (we merge nodes beyond the first into the first node)
// (last value is off by one, because unique indices are 1-based instead of 0-based, but
// otherwise correlate with row number.
nodeTable.setInt(
nodeBeyondFirst.intValue(), UNIQUE_INDEX_COLUMN_NAME, firstNode.intValue() + 1);
String iAmNotThePrimaryNodeInCluster = "";
nodeTable.setString(
nodeBeyondFirst.intValue(),
COMBINE_VALUES_COLUMN_NAME,
iAmNotThePrimaryNodeInCluster);
String nodeOneAttribute =
(String) nodeTable.getString(nodeBeyondFirst.intValue(), this.compareAttributeName);
mergeLog.append(" merging in \"" + nodeOneAttribute + "\"" + "\r\n");
}
}
}
return nodeTable;
}
private Graph makeMergeGraph(Table nodeTable, StringBuffer noteLog) {
Graph mergeGraph = makeEmptyMergeGraph(nodeTable);
// for each group of nodes with a common attribute prefix...
ListMap groupedNodes =
sortNodesByAttributePrefix(nodeTable, this.compareAttributeName, this.numPrefixLetters);
for (Iterator groupIt = groupedNodes.values().iterator(); groupIt.hasNext(); ) {
List nodeGroup = (List) groupIt.next();
// for each pair of nodes in the group...
for (int i = 0; i < nodeGroup.size(); i++) {
Integer firstNodeIndex = (Integer) nodeGroup.get(i);
for (int j = i; j < nodeGroup.size(); j++) {
Integer secondNodeIndex = (Integer) nodeGroup.get(j);
// test how similar the two nodes are
float similarity =
compareNodesCaseInsensitiveBy(
this.compareAttributeName, firstNodeIndex, secondNodeIndex, nodeTable);
// if their similarity is high enough to merge...
if (similarity >= this.mergeOnSimilarity) {
// link the nodes in the merge graph
mergeGraph.addEdge(firstNodeIndex.intValue(), secondNodeIndex.intValue());
}
// else if their similarity is noteworthy...
else if (similarity >= this.makeNoteOnSimilarity) {
// record it in the log
String nodeOneAttribute =
(String) nodeTable.getString(firstNodeIndex.intValue(), this.compareAttributeName);
String nodeTwoAttribute =
(String) nodeTable.getString(secondNodeIndex.intValue(), this.compareAttributeName);
noteLog.append("" + similarity + " similar:" + "\r\n");
noteLog.append(" \"" + nodeOneAttribute + "\"" + "\r\n");
noteLog.append(" \"" + nodeTwoAttribute + "\"" + "\r\n");
}
}
}
}
return mergeGraph;
}
