Mongoid Search

Mongoid Search is a simple full text search implementation for Mongoid ORM. It supports Mongoid 3, 4, 5 and 6 and performs well for small data sets. If your searchable model is big (i.e. 1.000.000+ records), mongoid_fulltext, ElasticSearch, Solr or Sphinx may suit you better.

Installation

In your Gemfile:

gem 'mongoid_search'

Then:

bundle install

Examples

class Product
  include Mongoid::Document
  include Mongoid::Search
  field :brand
  field :name
  field :unit
  field :info, type: Hash

  has_many   :tags
  belongs_to :category

  search_in :brand, :name, tags: :name, category: :name, info: %i[summary description]
  search_in :unit, index: :_unit_keywords
end

class Tag
  include Mongoid::Document
  field :name

  belongs_to :product
end

class Category
  include Mongoid::Document
  field :name

  has_many :products
end

Now when you save a product, you get a _keywords field automatically:

p = Product.new brand: 'Apple', name: 'iPhone', unit: 'kilogram', info: { summary: 'Info-summary', description: 'Info-description' }
p.tags << Tag.new(name: 'Amazing')
p.tags << Tag.new(name: 'Awesome')
p.tags << Tag.new(name: 'Superb')
p.save
# => true
p._keywords
# => ["amazing", "apple", "awesome", "iphone", "superb", "Info-summary", "Info-description"]
p._unit_keywords
# => ["kilogram"]

Now you can run search, which will look in the _keywords field and return all matching results:

Product.full_text_search("apple iphone").size
# => 1

You can also search in "virtual" fields by defining them as methods. This can be useful when you have a method with dynamic fields (i.e. variable schema).

class ModelWithDynamicFields
  
  ...
  
  search_in :search_data

  def search_data
    # concatenate all String fields' values
    self.attributes.select{|k,v| v.is_a?(String) }.values.join(' ')
  end
end

Mongoid_search will run the method before save and use it's output to populate the _keywords field.

Of course, some models could have more than one index. For instance, two different searches with different fields, so you could even specify from which index should be searched:

Product.full_text_search("kilogram", index: :_unit_keywords).size
# => 1

Note that the search is case insensitive, and accept partial searching too:

Product.full_text_search('ipho').size
# => 1

Assuming you have a category with multiple products you can use the following code to search for 'iphone' in products cheaper than $499.

category.products.where(:price.lt => 499).full_text_search('iphone').asc(:price)

To index or reindex all existing records, run this rake task

$ rake mongoid_search:index

Options

match

:any - match any occurrence
:all - match all occurrences

Default is :any.

Product.full_text_search('apple motorola', match: :any).size
# => 1

Product.full_text_search('apple motorola', match: :all).size
# => 0

allow_empty_search

true - will return Model.all
false - will return []

Default is false.

Product.full_text_search('', allow_empty_search: true).size
# => 1

relevant_search

true - adds relevance information to the results
false - no relevance information

Default is false.

Product.full_text_search('amazing apple', relevant_search: true)
# => [#<Product _id: 5016e7d16af54efe1c000001, _type: nil, brand: "Apple", name: "iPhone", attrs: nil, info: nil, category_id: nil, _keywords: ["amazing", "apple", "awesome", "iphone", "superb"], relevance: 2.0>]

Please note that relevant_search will return an Array and not a Criteria object. The search method should always be called in the end of the method chain.

index

Default is _keywords.

Product.full_text_search('amazing apple', index: :_keywords)
# => [#<Product _id: 5016e7d16af54efe1c000001, _type: nil, brand: "Apple", name: "iPhone", unit: "l", attrs: nil, info: nil, category_id: nil, _keywords: ["amazing", "apple", "awesome", "iphone", "superb"], _unit_keywords: ["l"], relevance: 2.0>]

Product.full_text_search('kg', index: :_unit_keywords)
# => [#<Product _id: 5016e7d16af54efe1c000001, _type: nil, brand: "Apple", name: "iPhone", unit: "kg", attrs: nil, info: nil, category_id: nil, _keywords: ["amazing", "apple", "awesome", "iphone", "superb"], _unit_keywords: ["kg"], relevance: 2.0>]

index enables to have two or more different searches, with different or same fields. It should be noted that indexes are exclusive per each one.

Initializer

Alternatively, you can create an initializer to setup those options:

Mongoid::Search.setup do |config|
  ## Default matching type. Match :any or :all searched keywords
  config.match = :any

  ## If true, an empty search will return all objects
  config.allow_empty_search = false

  ## If true, will search with relevance information
  config.relevant_search = false

  ## Stem keywords
  config.stem_keywords = false

  ## Add a custom proc returning strings to replace the default stemmer
  # For example using ruby-stemmer:
  # config.stem_proc = Proc.new { |word| Lingua.stemmer(word, :language => 'nl') }

  ## Words to ignore
  config.ignore_list = []

  ## An array of words
  # config.ignore_list = %w{ a an to from as }

  ## Or from a file
  # config.ignore_list = YAML.load(File.open(File.dirname(__FILE__) + '/config/ignorelist.yml'))["ignorelist"]

  ## Search using regex (slower)
  config.regex_search = true

  ## Regex to search

  ## Match partial words on both sides (slower)
  config.regex = Proc.new { |query| /#{query}/ }

  ## Match partial words on the beginning or in the end (slightly faster)
  # config.regex = Proc.new { |query| /^#{query}/ }
  # config.regex = Proc.new { |query| /#{query}$/ }

  # Ligatures to be replaced
  # http://en.wikipedia.org/wiki/Typographic_ligature
  config.ligatures = { "œ"=>"oe", "æ"=>"ae" }

  # Strip symbols regex to be replaced. These symbols will be replaced by space
  config.strip_symbols = /[._:;'\"`,?|+={}()!@#%^&*<>~\$\-\\\/\[\]]/

  # Strip accents regex to be replaced. These sybols will be removed after strip_symbols replacing
  config.strip_accents = /[^\s\p{Alnum}]/

  # Minimum word size. Words smaller than it won't be indexed
  config.minimum_word_size = 2
end

mongoid_search

Development

Runtime