datalevin.search-utils — datalevin 0.9.14

Some useful utility functions that can be passed as options to search engine to customize search.

create-analyzer clj

(create-analyzer opts)

Creates an analyzer fn ready for use in search.

opts have the following keys:

• :tokenizer is a tokenizing fn that takes a string and returns a seq of [term, position, offset], where term is a word, position is the sequence number of the term, and offset is the character offset of this term. e.g. create-regexp-tokenizer produces such fn.

• :token-filters is an ordered list of token filters. A token filter receives a [term, position, offset] and returns a transformed list of tokens to replace it with.

create-max-length-token-filter clj

(create-max-length-token-filter max-length)

Filters tokens that are strictly longer than max-length.

create-min-length-token-filter clj

(create-min-length-token-filter min-length)

Filters tokens that are strictly shorter than min-length.

create-ngram-token-filter clj

(create-ngram-token-filter gram-size)

(create-ngram-token-filter min-gram-size max-gram-size)

Produces character ngrams between min and max size from the token and returns everything as tokens. This is useful for producing efficient fuzzy search.

create-regexp-tokenizer clj

(create-regexp-tokenizer pat)

Creates a tokenizer that splits text on the given regular expression pat.

create-stemming-token-filter clj

(create-stemming-token-filter language)

Create a token filter that replaces tokens with their stems.

The stemming algorithm is Snowball https://snowballstem.org/

language is a string, its value can be one of the following:

arabic, armenian, basque, catalan, danish, dutch, english, french, finnish, german, greek, hindi, hungarian, indonesian, irish, italian, lithuanian, nepali, norwegian, portuguese, romanian, russian, serbian, swedish, tamil, turkish, spanish, yiddish, and porter

create-stop-words-token-filter clj

(create-stop-words-token-filter stop-word-pred)

Takes a stop words predicate that returns true when the given token is a stop word

en-stop-words-token-filter clj

This token filter removes "empty" tokens (for english language).

lower-case-token-filter clj

This token filter converts tokens to lower case.

prefix-token-filter clj

Produces a series of every possible prefixes in a token and replace it with them. For example: vault -> v, va, vau, vaul, vault

Takes a vector [word position start-offset].

This is useful for producing efficient autocomplete engines, provided this filter is NOT applied at query time.

unaccent-token-filter clj

This token filter removes accents and diacritics from tokens.