/**
* Iterate over the map, and ensure that if the same string is in the cache, that the entry refers
* to the cached entry, rather than a new instance.
*
* @param stringMap
*/
public void mergeStrings(TIntObjectHashMap<String> stringMap) {
if (stringMap == null) return;
// FIXME: Investigate using String.intern() what's it's overhead if we
// do it on all strings.
// FIXME: Keep track of which integer keys (they should be 0-255) merge well and which don't,
// by tracking successes
for (TIntObjectIterator<String> it = stringMap.iterator(); it.hasNext(); ) {
it.advance();
int key = it.key();
assert key < 256;
// Eliminate poor merging strings
// FIXME: should also remove existing by having sep cache per key, which
// we can just remove once can see it's a problem. For now, we
// just stop after it seems that we've got no dupls
if (strFailuresByKey[key] > 5000 && strSuccessesByKey[key] < strFailuresByKey[key]) {
continue; // if after 5000 failures, we don't have more successes
}
String possDupl = it.value();
String cachedString = stringCache.get(possDupl);
if (cachedString != null) {
it.setValue(cachedString); // If have one cached, use it
strSuccessesByKey[key]++;
} else {
stringCache.put(possDupl, possDupl);
strFailuresByKey[key]++;
}
}
}
public static void save(final String filename) throws IOException {
final FileOutputStream fos = FileHelper.createFileOutputStream(filename);
final OutputBitStream ostream = new OutputBitStream(fos);
ostream.writeInt(ElementPropertiesLibrary.m_elements.size());
final TIntObjectIterator<ElementProperties> iter =
ElementPropertiesLibrary.m_elements.iterator();
for (int i = ElementPropertiesLibrary.m_elements.size(); i > 0; --i) {
iter.advance();
iter.value().save(ostream);
}
ostream.close();
}