* Add failing specs for hashtag and username extraction in language detector * Remove usernames and hashtags from text before language detection * Handle multiple instances of special case, and reduce whitespace
		
			
				
	
	
		
			49 lines
		
	
	
		
			953 B
		
	
	
	
		
			Ruby
		
	
	
	
	
	
			
		
		
	
	
			49 lines
		
	
	
		
			953 B
		
	
	
	
		
			Ruby
		
	
	
	
	
	
| # frozen_string_literal: true
 | |
| 
 | |
| class LanguageDetector
 | |
|   attr_reader :text, :account
 | |
| 
 | |
|   def initialize(text, account = nil)
 | |
|     @text = text
 | |
|     @account = account
 | |
|     @identifier = CLD3::NNetLanguageIdentifier.new(1, 2048)
 | |
|   end
 | |
| 
 | |
|   def to_iso_s
 | |
|     detected_language_code || default_locale.to_sym
 | |
|   end
 | |
| 
 | |
|   def prepared_text
 | |
|     simplified_text.strip
 | |
|   end
 | |
| 
 | |
|   private
 | |
| 
 | |
|   def detected_language_code
 | |
|     result.language.to_sym if detected_language_reliable?
 | |
|   end
 | |
| 
 | |
|   def result
 | |
|     @result ||= @identifier.find_language(prepared_text)
 | |
|   end
 | |
| 
 | |
|   def detected_language_reliable?
 | |
|     result.reliable?
 | |
|   end
 | |
| 
 | |
|   def simplified_text
 | |
|     text.dup.tap do |new_text|
 | |
|       URI.extract(new_text).each do |url|
 | |
|         new_text.gsub!(url, '')
 | |
|       end
 | |
|       new_text.gsub!(Account::MENTION_RE, '')
 | |
|       new_text.gsub!(Tag::HASHTAG_RE, '')
 | |
|       new_text.gsub!(/\s+/, ' ')
 | |
|     end
 | |
|   end
 | |
| 
 | |
|   def default_locale
 | |
|     account&.user_locale || I18n.default_locale
 | |
|   end
 | |
| end
 |