JarCastinglucene4s
Light-weight convenience wrapper around Lucene to simplify complex tasks and add Scala sugar.
Setup
lucene4s is published to Sonatype OSS and Maven Central currently supporting Scala 2.11, 2.12, and 2.13.
Configuring the dependency in SBT simply requires:
libraryDependencies += "com.outr" %% "lucene4s" % "1.10.0"
Using
Imports
You may find yourself needing other imports depending on what you're doing, but the majority of functionality can be achieved simply importing com.outr.lucene4s._:
import com.outr.lucene4s._
Creating a Lucene Instance
Lucene is the object utilized for doing anything with Lucene, so you first need to instantiate it:
val directory = Paths.get("index")
val lucene = new DirectLucene(Nil, directory = Option(directory))
NOTE: If you leave directory blank or set it to None (the default) it will use an in-memory index.
Creating Fields
For type-safety and convenience we can create the fields we'll be using in the document ahead of time:
val name = lucene.create.field[String]("name")
val address = lucene.create.field[String]("address")
Inserting Documents
Inserting is quite easy using the document builder:
lucene.doc().fields(name("John Doe"), address("123 Somewhere Rd.")).index()
lucene.doc().fields(name("Jane Doe"), address("123 Somewhere Rd.")).index()
Querying Documents
Querying documents is just as easy with the query builder:
val paged = lucene.query().sort(Sort(name)).search()
paged.results.foreach { searchResult =>
println(s"Name: ${searchResult(name)}, Address: ${searchResult(address)}")
}
This will return a PagedResults instance with the page size set to the limit. There are convenience methods for navigating the pagination and accessing the results.
The above code will output:
Name: John Doe, Address: 123 Somewhere Rd.
Name: Jane Doe, Address: 123 Somewhere Rd.
Highlighting Results
Though querying is nice, we may want to stylize the output to show the matched results. This is pretty simple:
val paged = lucene.query().sort(Sort(name)).filter(fuzzy(name("jhn"))).highlight().search()
paged.results.foreach { searchResult =>
val highlighting = searchResult.highlighting(name).head
println(s"Fragment: ${highlighting.fragment}, Word: ${highlighting.word}")
}
The above code will output:
Fragment: <em>John</em> Doe, Word: John
Fragment: <em>Jane</em> Doe, Word: Jane
Faceted Searching
See https://github.com/outr/lucene4s/blob/master/implementation/src/test/scala/tests/FacetsSpec.scala
Full-Text Searching
In lucene4s the Lucene instance holds a fullText Field that contains a concatenation of all the fields that are configured as fullTextSearchable. This defaults to Lucene.defaultFullTextSearchable which defaults to false.
The fullText field is the default field used for searches if it's not specified in the SearchTerm. Let's see an example:
val paged = lucene.query().filter(wildcard("doe*")).search()
paged.total should be(4)
paged.results(0)(firstName) should be("John")
paged.results(1)(firstName) should be("Jane")
paged.results(2)(firstName) should be("Baby")
paged.results(3)(firstName) should be("James")
For a complete example, see: https://github.com/outr/lucene4s/blob/master/implementation/src/test/scala/tests/FullTextSpec.scala
Keyword Searching
As we saw previously, the fullText field provides us with a concatenation of all fields configured to be fullTextSearchable. In addition, if you create an instance of KeywordIndexing you can query against a no-duplicates index of keywords for the fullText (although you can override defaults to apply keyword indexing to any field). All we have to do is create and instance referencing the Lucene instance and the name (used for storage purposes):
val keywordIndexing = KeywordIndexing(lucene, "keywords")
val keywords = keywordIndexing.search("do*")
println("Keywords: ${keywords.results.map(_.word).mkString(", ")}")
The above code would output:
Keywords: Doe
For the complete example see: https://github.com/outr/lucene4s/blob/master/implementation/src/test/scala/tests/SimpleSpec.scala
Case Class Support
lucene4s provides a powerful Macro-based system to generate two-way mappings between case classes and Lucene fields at compile-time. This is accomplished through the use of Searchable. The setup is pretty simple.
Setup
First we need to define a case class to model the data in the index:
case class Person(id: Int, firstName: String, lastName: String, age: Int, address: String, city: String, state: String, zip: String)
As you can see, this is a bare-bones case class with nothing special about it.
Next we need to define a Searchable trait the defines the unique identification for update and delete:
trait SearchablePerson extends Searchable[Person] {
// This is necessary for update and delete to reference the correct document.
override def idSearchTerms(person: Person): List[SearchTerm] = List(exact(id(person.id)))
/*
Though at compile-time all fields will be generated from the params in `Person`, for code-completion we can define
an unimplemented method in order to properly reference the field. This will still compile without this definition,
but most IDEs will complain.
*/
def id: Field[Int]
}
As the last part of our set up we simply need to generate it from our Lucene instance:
val people = lucene.create.searchable[SearchablePerson]
Inserting
Now that we've configured everything inserting a person is trivial:
people.insert(Person(1, "John", "Doe", 23, "123 Somewhere Rd.", "Lalaland", "California", "12345")).index()
Notice that we still have to call index() at the end for it to actually invoke. This allows us to do more advanced tasks like adding facets, adding non-Searchable fields, etc. before actually inserting.
Updating
Now lets try updating our Person:
people.update(Person(1, "John", "Doe", 23, "321 Nowhere St.", "Lalaland", "California", "12345")).index()
As you can see here, the signature is quite similar to insert. Internally this will utilize idSearchTerms as we declared previously to apply the update. In this case that means as long as we don't change the id (1) then calls to update will replace an existing record if one exists.
Querying
Querying works very much the same as in the previous examples, except we get our QueryBuilder from our people instance:
val paged = people.query().search()
paged.entries.foreach { person =>
println(s"Person: $person")
}
Note that instead of calling paged.results we call paged.entries as it represents the conversion to Person. We can still use paged.results if we want access to the SearchResult like before.
Deleting
Deleting is just as easy as inserting and updating:
people.delete(Person(1, "John", "Doe", 23, "321 Nowhere St.", "Lalaland", "California", "12345"))
Additional Information
All Searchable implementations automatically define a docType field that is used to uniquely separate different Searchable instances so you don't have to worry about multiple different instances overlapping.
For more examples see https://github.com/outr/lucene4s/blob/master/implementation/src/test/scala/tests/SearchableSpec.scala
Geospatial Support
One of the great features of Lucene is geospatial querying and what Lucene wrapper would be complete without it?
Creating a Spatial Field
In order to create a stored, queryable, filterable, and sortable latitude and longitude you need only create a SpatialPoint field:
val location: Field[SpatialPoint] = lucene.create.field[SpatialPoint]("location")
Sorting Nearest a Point
Most of the time it's most useful to take an existing latitude and longitude and sort your results returning the nearest documents to that location:
val paged = lucene.query().sort(Sort.nearest(location, SpatialPoint(40.7142, -74.0119))).search()
Filtering by Distance
If you want to filter your results to only include entries within a certain range of a location:
val newYorkCity = SpatialPoint(40.7142, -74.0119)
val paged = lucene
.query()
.sort(Sort.nearest(location, newYorkCity))
.filter(spatialDistance(location, newYorkCity, 50.miles))
.search()
