void PopulateSymbolTable(int[] frequencies)
{
Symbols = new HuffmanTableEntry[NumSymbols];
for (int i = 0; i < NumSymbols; i++)
{
Symbols[i] = new HuffmanTableEntry {
Symbol = i, Count = frequencies[i]
};
}
}
// TODO: This completely fails if there is only 1 observed symbol. We should fix this.
// However, in practice there should always be at least one literal symbol + EOF symbol (shouldn't be compressing 0 byte files).
public void ConstructTree()
{
var symbolsToProcess = new List <HuffmanTableEntry>(Symbols.Where(s => s.Count > 0));
while (symbolsToProcess.Count > 1)
{
// Pop the lowest count symbols off the list, and insert a tree node each time you loop.
// Loop until you only have one symbol, the top of the tree.
var lowest1 = symbolsToProcess[0];
var lowest2 = symbolsToProcess[1];
if (lowest1.Count > lowest2.Count)
{
lowest1 = symbolsToProcess[1];
lowest2 = symbolsToProcess[0];
}
for (int i = 2; i < symbolsToProcess.Count; i++)
{
if (symbolsToProcess[i].Count < lowest1.Count)
{
lowest2 = lowest1;
lowest1 = symbolsToProcess[i];
}
else if (symbolsToProcess[i].Count < lowest2.Count)
{
lowest2 = symbolsToProcess[i];
}
}
// Now, lowest1 and lowest2 should point to the lowest two probability items in the list.
// So, create a merged node with lowest two probability items.
var merged = new HuffmanTableEntry {
Symbol = -1,
Count = lowest1.Count + lowest2.Count
};
lowest1.Parent = merged;
lowest2.Parent = merged;
// Insert merged node and pop off lowest
symbolsToProcess.Remove(lowest1);
symbolsToProcess.Remove(lowest2);
symbolsToProcess.Add(merged);
}
}
