Remove picture from data

[mensfeld]

Hey, is there a way to remove once indexed file?

[mrkamel]

Hi, not yet, but i'll add the feature.
Thx for proposing it.

[mensfeld]

Hey - well I "fixed" that on my side by replacing (same ID, same keywords) proper image with 1x1 small gif ;) That way the original file is removed. I also wrote a simple wrapper around RestClient for Similarity. You can take a look ;)

require "base64"
require 'ostruct'

# A simple wrapper for Similarity library
# @see https://github.com/mrkamel/similarity
class Similarity

  # We should raise it when we get response different than expected
  class InvalidResponse < Exception; end
  # We should raise it when we try to process non-existing file
  class NonExistingFile < Exception; end

  # We can't remove data directly, from Lire, so instead we replace
  # given record, with 1x1 GIF file. This is its's content
  NON_EXISTING_FILE = 'R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBRAA7'

  HOST      = Settings.similarity.host
  PORT      = Settings.similarity.port
  LIMIT     = Settings.similarity.limit
  NAMESPACE = Settings.similarity.namespace
  SOURCE    = "http://#{HOST}:#{PORT}"

  # Adds a new record about image to similarity
  # @param file [String, Tempfile] path to a file or a ready tempfile
  # @param file_id [Integer] id of this file
  # @param namespace [String] a namespace/keywords describing image - thanks
  #   to this value, we can narrow search to a given scope/namespace instead
  #   of querying all the images
  # @return [true]
  # @raise [InvalidResponse] InvalidResponse if something bad happened
  # @example Create a new record in Similarity
  #   Similarity.create('/path/to/file.jpg', 123, 'Artworks')
  def self.create(file, file_id, namespace = NAMESPACE)
    if file.is_a?(String)
      self.ensure(file)

      file = File.new(file)
    end

    response = self.post(
      "#{SOURCE}/uploads", 
      file: file, 
      id: file_id.to_s, 
      text: namespace
    )

    raise InvalidResponse, response unless response == ""

    true
  end

  # Removes information 
  # @param file_id [Integer] id of this file
  # @param namespace [String] a namespace/keywords of image that we want to remove
  # @return [true]
  # @raise [InvalidResponse] InvalidResponse if something bad happened
  # @example Remove information about given file
  #   Similarity.destroy(123, 'Artworks')
  def self.destroy(file_id, namespace = NAMESPACE)
    self.create(blank_file, file_id, namespace)
  end

  # Searches for similar images
  # @param file [String] path to a file that we want to check fior similarity
  # @param namespace [String] a namespace/keywords of image that we want to remove
  # @param limit [Integer] maximum number of similar images
  # @return [Array<OpenStruct>] Array of similar images
  # @example Search for similar images
  #   Similarity.search('path/to/file.jpg') #=> [#<Openstruct id="12">]
  def self.search(file, namespace = NAMESPACE, limit = LIMIT)
    self.ensure(file)
    file = File.new(file)

    response = self.post(
      "#{SOURCE}/search", 
      file: file,
      q: namespace, 
      start: 0, 
      limit: limit
    )

    data = [Hash.from_xml(response)['response']['result']].flatten
    data.collect{ |r| OpenStruct.new(r) }
  end

  private

  # Performs a RestClient post request and handles errors
  # @param url [String] url where we should send post
  # @param params [Hash] request params
  # @example
  #   post('http://localhost:8984/search', id: 1)
  def self.post(url, params = {})
    RestClient.post(url, params)
  rescue RestClient::InternalServerError => e
    raise InvalidResponse, e.to_s
  end

  # Ensures that we won't work on non-existing file
  # @param [String] file path
  # @raise [NonExistingFile] raised when given file doesn't exist
  def self.ensure(file)
    raise NonExistingFile, file unless File.exists?(file)
  end

  # @return [Tempfile] tempfile with 1x1 GIF image
  def self.blank_file
    @blank_file ||= Tempfile.new('').tap do |tmp|
      tmp.binmode
      tmp << Base64.decode64(NON_EXISTING_FILE)
      tmp.rewind
    end
  end

end

Although in the end I won't be using Similarity (I don't have much record and pure hamming distance is enough) - I love your solution!

Ah and one more thing - there's a new Lire version that works great with transparent PNG files (the one that you're using is quite old) - in a spare time would be awesome if you could update it :)
Great job btw!

[mrkamel]

Thanks a lot! I'll update Lire asap as well.