private readonly Dictionary <int, List <Token> > materializedWords = new Dictionary <int, List <Token> >(); // Pooling? Configuration for expected unique words per document?
public void MergeOrAdd(TokenHash hash, StringBuilder word, WordLocation location)
{
if (this.materializedWords.TryGetValue(hash.HashValue, out var existingEntries))
{
foreach (var existingEntry in existingEntries)
{
if (word.SequenceEqual(existingEntry.Value))
{
existingEntry.AddLocation(location);
return;
}
}
existingEntries.Add(new Token(word.ToString(), location));
}
else
{
this.materializedWords.Add(
hash.HashValue,
new List <Token>()
{
new Token(word.ToString(), location)
});
}
}
/// <summary>
/// Captures a token at a location, merging the token with any locations
/// it previously matched at.
/// </summary>
public void MergeOrAdd(StringBuilder token, TokenLocation location)
{
var hash = new TokenHash(token);
if (this.materializedTokens.TryGetValue(hash.HashValue, out var existingEntries))
{
foreach (var existingEntry in existingEntries)
{
if (token.SequenceEqual(existingEntry.Value))
{
existingEntry.AddLocation(location);
return;
}
}
existingEntries.Add(new Token(token.ToString(), location));
}
else
{
this.materializedTokens.Add(
hash.HashValue,
new List <Token>()
{
new Token(token.ToString(), location)
});
}
}
public IEnumerable <Token> Process(ReadOnlySpan <char> input)
{
var processedWords = new TokenStore(); // TODO Pool?
var wordIndex = 0;
var start = 0;
var wordBuilder = new StringBuilder();
var hash = new TokenHash();
for (var i = 0; i < input.Length; i++)
{
var current = input[i];
if (this.IsWordSplitCharacter(current))
{
if (wordBuilder.Length > 0)
{
CaptureWord(processedWords, hash, wordIndex, start, i, wordBuilder);
wordIndex++;
wordBuilder.Length = 0;
hash = new TokenHash();
}
start = i + 1;
}
else
{
foreach (var processed in this.inputPreprocessorPipeline.Process(current))
{
wordBuilder.Append(processed);
hash = hash.Combine(processed);
}
}
}
if (wordBuilder.Length > 0)
{
CaptureWord(processedWords, hash, wordIndex, start, input.Length, wordBuilder);
}
return(processedWords.ToList());
}
private static void CaptureWord(TokenStore processedWords, TokenHash hash, int wordIndex, int start, int end, StringBuilder wordBuilder)
{
var length = end - start;
processedWords.MergeOrAdd(hash, wordBuilder, new WordLocation(wordIndex, start, length));
}