/**
* Add a scored ID without boxing. The default value will be used for each channel.
*
* @param id The ID to add.
* @param score The score for the ID.
* @return The builder (for chaining).
*/
public ScoredIdListBuilder add(long id, double score) {
Preconditions.checkState(ids != null, "builder has been finished");
ids.add(id);
scores.add(score);
for (ChannelStorage chan : channels.values()) {
assert chan.values.size() == ids.size() - 1;
chan.values.add(chan.defaultValue);
}
for (TypedChannelStorage chan : typedChannels.values()) {
assert chan.values.size() == ids.size() - 1;
chan.values.add(chan.defaultValue);
}
return this;
}
/**
* Add a scored ID. The ID is copied into the builder, not referenced. All side channels on the ID
* must have already been added with one of the {@code addChannel} methods.
*
* @param id The ID.
* @return The builder (for chaining).
*/
public ScoredIdListBuilder add(ScoredId id) {
Preconditions.checkState(ids != null, "builder has been finished");
// check whether all symbols are valid
Collection<SymbolValue<?>> chans = id.getChannels();
if (!ignoreUnknown) {
for (SymbolValue<?> chan : chans) {
TypedSymbol<?> sym = chan.getSymbol();
boolean good =
sym.getType().equals(Double.class)
? channels.containsKey(sym.getRawSymbol())
: typedChannels.containsKey(sym);
if (!good) {
throw new IllegalArgumentException("channel " + sym + " not known");
}
}
}
// now we're ready to add
int idx = ids.size();
add(id.getId(), id.getScore());
assert ids.size() == idx + 1;
assert scores.size() == idx + 1;
for (SymbolValue<?> sv : chans) {
TypedSymbol<?> sym = sv.getSymbol();
if (sym.getType().equals(Double.class) && channels.containsKey(sym.getRawSymbol())) {
ChannelStorage chan = channels.get(sym.getRawSymbol());
assert chan.values.size() == idx + 1;
if (sv instanceof DoubleSymbolValue) {
chan.values.set(idx, ((DoubleSymbolValue) sv).getDoubleValue());
} else {
Object v = sv.getValue();
chan.values.set(idx, (Double) v);
}
} else {
TypedChannelStorage chan = typedChannels.get(sv.getSymbol());
if (chan != null) {
assert chan.values.size() == idx + 1;
chan.values.set(idx, sv.getValue());
}
}
}
return this;
}
/**
* Build a sparse vector directly from the list of IDs. This allows a scored ID list builder to be
* used to efficiently accumulate a sparse vector. If the same ID is added multiple times, the
* first instance is used.
*
* @return A sparse vector containing the data accumulated.
*/
public ImmutableSparseVector buildVector() {
MutableSparseVector msv = MutableSparseVector.create(ids);
final int size = size();
for (int i = 0; i < size; i++) {
msv.set(ids.get(i), scores.get(i));
}
for (ChannelStorage chan : channels.values()) {
MutableSparseVector vchan = msv.getOrAddChannelVector(chan.symbol);
for (int i = 0; i < size; i++) {
vchan.set(ids.get(i), chan.values.get(i));
}
}
for (TypedChannelStorage<?> chan : typedChannels.values()) {
Long2ObjectMap vchan = msv.getOrAddChannel(chan.symbol);
for (int i = 0; i < size; i++) {
vchan.put(ids.get(i), chan.values.get(i));
}
}
return msv.freeze();
}
/**
* Implementation of {@link #build()} and {@link #finish()}.
*
* @param reuse Whether we should try to reuse the builder's storage for the packed list. If
* {@code true}, the builder will be invalid after finishing and the packed list will use the
* same arrays as the builder if they are full.
* @return The packed ID list.
*/
private PackedScoredIdList finish(boolean reuse) {
Preconditions.checkState(ids != null, "builder has been finished");
Map<Symbol, DoubleList> chans;
Map<TypedSymbol<?>, List<?>> typedChans;
if (size() > 0) {
ImmutableMap.Builder<Symbol, DoubleList> cbld = ImmutableMap.builder();
for (ChannelStorage chan : channels.values()) {
DoubleArrayList built;
if (reuse) {
built = chan.values;
built.trim();
} else {
built = new DoubleArrayList(chan.values);
}
cbld.put(chan.symbol, built);
}
chans = cbld.build();
ImmutableMap.Builder<TypedSymbol<?>, List<?>> tcbld = ImmutableMap.builder();
for (TypedChannelStorage<?> chan : typedChannels.values()) {
List<?> built;
if (reuse) {
chan.values.trimToSize();
built = chan.values;
} else {
built = new ArrayList<Object>(chan.values);
}
tcbld.put(chan.symbol, built);
}
typedChans = tcbld.build();
} else {
chans = Collections.emptyMap();
typedChans = Collections.emptyMap();
}
LongList builtIds;
DoubleList builtScores;
if (reuse) {
ids.trim();
builtIds = ids;
scores.trim();
builtScores = scores;
clear();
} else {
builtIds = new CompactableLongArrayList(ids);
builtScores = new DoubleArrayList(scores);
}
return new PackedScoredIdList(builtIds, builtScores, typedChans, chans);
}
/**
* Get the number of items currently in the builder.
*
* @return The number of items in the builder.
*/
public int size() {
assert ids.size() == scores.size();
return ids.size();
}