lucene-custom-analyzer

(Micro)Library to build Lucene analyzers in a data-driven fashion.

Why Would You Want to Use lucene-custom-analyzer?

• Current Clojure Lucene libraries (e.g. jaju/lucene-clj, federkasten/clucie) doesn't provide a mechanism to build your custom Lucene Analyzers.
• Data-driven.
• Allows for extensibility using standard Lucene SPI, i.e. just put a JAR in the CLASSPATH.
• Allows to specify a directory from which resources will be loaded, e.g. synonyms dictionaries.
• Lucene 9+ supported.
• Already includes the most commonly used Lucene analysis components.

Quickstart

Dependencies:

lt.jocas/lucene-custom-analyzer {:mvn/version "1.0.34"}

Code:

(require '[lucene.custom.analyzer :as custom-analyzer])

(custom-analyzer/create
  {:tokenizer              {:standard {:maxTokenLength 4}}
   :char-filters           [{:patternReplace {:pattern     "foo"
                                              :replacement "foo"}}]
   :token-filters          [{:uppercase nil}
                            {:reverseString nil}]
   :offset-gap             2
   :position-increment-gap 3
   :config-dir             "."})
;; =>
;; #object[org.apache.lucene.analysis.custom.CustomAnalyzer
;;         0x4686f87d
;;         "CustomAnalyzer(org.apache.lucene.analysis.pattern.PatternReplaceCharFilterFactory@2f1300,org.apache.lucene.analysis.standard.StandardTokenizerFactory@7e71a244,org.apache.lucene.analysis.core.UpperCaseFilterFactory@54e9f0d6,org.apache.lucene.analysis.reverse.ReverseStringFilterFactory@3e494ba7)"]

Short notation for analysis components:

(custom-analyzer/create
  {:tokenizer :standard
   :char-filters [:htmlStrip]
   :token-filters [:uppercase]})
;; =>
;; #object[org.apache.lucene.analysis.custom.CustomAnalyzer
;;        0x16716eb1
;;        "CustomAnalyzer(org.apache.lucene.analysis.charfilter.HTMLStripCharFilterFactory@4c7f61fa,org.apache.lucene.analysis.standard.StandardTokenizerFactory@6fc69052,org.apache.lucene.analysis.core.UpperCaseFilterFactory@3944ccba)"]

If no options are provided then an Analyzer with just the standard tokenizer is created:

(custom-analyzer/create {})
;; =>
;; #object[org.apache.lucene.analysis.custom.CustomAnalyzer
;;         0x456fe86
;;         "CustomAnalyzer(org.apache.lucene.analysis.standard.StandardTokenizerFactory@5703f5b3)"]

If you want to check which analysis components are available run:

(lucene.custom.analyzer/char-filter-factories)
(lucene.custom.analyzer/tokenizer-factories)
(lucene.custom.analyzer/token-filter-factories)

Design

Under the hood this library uses the factory classes TokenizerFactory, TokenFilterFactory, and CharFilterFactory. The actual factories are loaded with java.util.ServiceLoader. All the available classes are automatically discovered.

If you want to include additional factory classes, e.g. your implementation of the TokenFilterFactory, you need to add it to the classpath 2 things:

1. The implementation class of one of the Factory classes
2. Under the META-INF/services add/change a file named org.apache.lucene.analysis.TokenFilterFactory that lists the classes from the step 1.

An example can be found here.

Future work

• [ ] Conditional token filters

License

Copyright © 2023 Dainius Jocas.

Distributed under The Apache License, Version 2.0.