How to get a Token from a Lucene TokenStream?

I'm trying to use Apache Lucene for tokenizing, and I am baffled at the process to obtain Tokens from a TokenStream.

The worst part is that I'm looking at the comments in the JavaDocs that address my question.

http://lucene.apache.org/java/3_0_1/api/core/org/apache/lucene/analysis/TokenStream.html#incrementToken%28%29

Somehow, an AttributeSource is supposed to be used, rather than Tokens. I'm totally at a loss.

Can anyone explain how to get token-like information from a TokenStream?


Yeah, it's a little convoluted (compared to the good ol' way), but this should do it:

TokenStream tokenStream = analyzer.tokenStream(fieldName, reader);
OffsetAttribute offsetAttribute = tokenStream.getAttribute(OffsetAttribute.class);
TermAttribute termAttribute = tokenStream.getAttribute(TermAttribute.class);

while (tokenStream.incrementToken()) {
    int startOffset = offsetAttribute.startOffset();
    int endOffset = offsetAttribute.endOffset();
    String term = termAttribute.term();
}

Edit: The new way

According to Donotello, TermAttribute has been deprecated in favor of CharTermAttribute. According to jpountz (and Lucene's documentation), addAttribute is more desirable than getAttribute.

TokenStream tokenStream = analyzer.tokenStream(fieldName, reader);
OffsetAttribute offsetAttribute = tokenStream.addAttribute(OffsetAttribute.class);
CharTermAttribute charTermAttribute = tokenStream.addAttribute(CharTermAttribute.class);

tokenStream.reset();
while (tokenStream.incrementToken()) {
    int startOffset = offsetAttribute.startOffset();
    int endOffset = offsetAttribute.endOffset();
    String term = charTermAttribute.toString();
}

This is how it should be (a clean version of Adam's answer):

TokenStream stream = analyzer.tokenStream(null, new StringReader(text));
CharTermAttribute cattr = stream.addAttribute(CharTermAttribute.class);
stream.reset();
while (stream.incrementToken()) {
  System.out.println(cattr.toString());
}
stream.end();
stream.close();

For the latest version of lucene 7.3.1

    // Test the tokenizer
    Analyzer testAnalyzer = new CJKAnalyzer();
    String testText = "Test Tokenizer";
    TokenStream ts = testAnalyzer.tokenStream("context", new StringReader(testText));
    OffsetAttribute offsetAtt = ts.addAttribute(OffsetAttribute.class);
    try {
        ts.reset(); // Resets this stream to the beginning. (Required)
        while (ts.incrementToken()) {
            // Use AttributeSource.reflectAsString(boolean)
            // for token stream debugging.
            System.out.println("token: " + ts.reflectAsString(true));

            System.out.println("token start offset: " + offsetAtt.startOffset());
            System.out.println("  token end offset: " + offsetAtt.endOffset());
        }
        ts.end();   // Perform end-of-stream operations, e.g. set the final offset.
    } finally {
        ts.close(); // Release resources associated with this stream.
    }

Reference: https://lucene.apache.org/core/7_3_1/core/org/apache/lucene/analysis/package-summary.html


There are two variations in the OP question:

  1. What is "the process to obtain Tokens from a TokenStream"?
  2. "Can anyone explain how to get token-like information from a TokenStream?"

Recent versions of the Lucene documentation for Token say (emphasis added):

NOTE: As of 2.9 ... it is not necessary to use Token anymore, with the new TokenStream API it can be used as convenience class that implements all Attributes, which is especially useful to easily switch from the old to the new TokenStream API.

And TokenStream says its API:

... has moved from being Token-based to Attribute-based ... the preferred way to store the information of a Token is to use AttributeImpls.

The other answers to this question cover #2 above: how to get token-like information from a TokenStream in the "new" recommended way using attributes. Reading through the documentation, the Lucene developers suggest that this change was made, in part, to reduce the number of individual objects created at a time.

But as some people have pointed out in the comments of those answers, they don't directly answer #1: how do you get a Token if you really want/need that type?

With the same API change that makes TokenStream an AttributeSource, Token now implements Attribute and can be used with TokenStream.addAttribute just like the other answers show for CharTermAttribute and OffsetAttribute. So they really did answer that part of the original question, they simply didn't show it.

It is important that while this approach will allow you to access Token while you're looping, it is still only a single object no matter how many logical tokens are in the stream. Every call to incrementToken() will change the state of the Token returned from addAttribute; So if your goal is to build a collection of different Token objects to be used outside the loop then you will need to do extra work to make a new Token object as a (deep?) copy.