Generation of alphanumeric person handle

author.rb

#
# This automatically creates a "handle" upon creation of person's record.
#
# "John Doe"'s handle shall look like "DO<x>J", where <x> is a minimal necessary number of digits,
# so we have to look up the existing handles first.  We exclude zeros (to avoid confusion with Os)
# and ones (to exclude jealousy "to be the first"), so we conveniently employ octal digits and add
# two subsequently.  We also have to do the best effort to normalize last names (drop diacritics,
# honorifics, Celtic and Dutch name prefixes, sr/jr suffixes, etc. to maximize variance.
#
# See this in action at http://dia-x.info/authors
#

class Author < ActiveRecord::Base
  has_and_belongs_to_many :diagrams

  validates_presence_of :name
  validates_format_of :name, with: /^(.+)\s+(\S+)$/, 
    message: 'must consist of given names and family name.'

  cattr_reader :per_page
  @@per_page = 99
  
  before_save :generate_code

  attr_accessible :name

  #########################################
  protected
  #########################################

  def generate_code
    
    return unless self.code.blank?
    
    nname = self.name.mb_chars.normalize(:kd).gsub(/[^\x20-\x7F]/,'').upcase.to_s
    nname.gsub! /\(.*\)/, '' # no parens
    nname.gsub! / MC/, ' '   # no McSomeone
    nname.gsub! / O\'/, ' '  # no O'Anybody
    nname.gsub! / VAN (DER )?/, ' '    # Vanderbilt is OK, van der Bilt is not
    nname.gsub! / SR\.?$| JR\.?$/, ''  # no sr or jr
    nname.strip!
    nname =~ /^(.+)\s+(\S+)$/
    names, family = $1[0,1], $2[0,2]
    
    @others = Author
      .where(:code ^ nil)
      .where(:code =~ family + '%' + names)


    while true
      trycode = family + random_code + names
      unless @others.find {|i| i.code == trycode}
        self.code = trycode
        # logger.info "*** #{trycode} ***"
        break
      end
    end
  end
  
  def random_code
    digit8 = ''
    limit = @others.size.to_s(8).size # cheap logarithm(8)
    limit.times do
      digit8 += (2 + rand(8)).to_s
    end
    return digit8
  end

end