Detection improvements and documentation

.gitignore

@@ -0,0 +1 @@
+.DS_Store

README.md

@@ -5,7 +5,7 @@ Languages.jl
 [![Languages](http://pkg.julialang.org/badges/Languages_0.6.svg)](http://pkg.julialang.org/?pkg=Languages)
 
 
-# Introduction
+## Introduction
 
 Languages.jl is a Julia package for working with human languages. It provides:
 
@@ -17,11 +17,45 @@ Languages.jl is a Julia package for working with human languages. It provides:
 	* Pronouns
 	* Stopwords
 
-# Usage
+	These methods are supported only for English and German currently.
+
+	This package also detects the script and language for written text in a wide variety of languages.  
+
+## Usage
 
 	using Languages
 
-	articles(EnglishLanguage)
-	stopwords(EnglishLanguage)
+	articles(Languages.English())
+	stopwords(Languages.English())
 
 All word lists are returned as vectors of UTF-8 strings.
+
+## Script detection
+
+Script detection model works by checking the unicode character ranges present within
+the input text
+
+  Languages.detect_script("To be or not to be") # => Languages.LatinScript()
+
+## Language Detection
+
+A trigram based model is used to detect the language for the text. The model is
+filtered based on the detected script.
+
+We detect 84 of the most common languages spoken around the world. This usually
+covers most languages with more than 10 million native speakers.
+
+  Languages.detect("To be or not to be")
+  # (Languages.English(), Languages.LatinScript(), 1.0)
+
+The `detect` function returns the language, the script, and the confidence.
+
+The language and script detection code in this package is ported from  the rust package [whatlang-rs](https://github.com/greyblake/whatlang-rs). That package is in turn derived from [franc](https://github.com/wooorm/franc). See `LICENSE.whatlang-rs` for details. 
+
+## Deprecations
+
+The API of this package has been refurbished recently. If you have used this package earlier,
+please be aware of these changes.
+
+  * The language names have been shortened. So `English` instead of `EnglishLanguage`. However, the language names are no longer exported. So they should be referred to with the package name: `Languages.English`
+	* Every language is a type. However all functions now accept and return instances of these types, rather than the types themselves.

src/whatlang.jl

@@ -62,9 +62,9 @@ function detect_script(text::AbstractString)
         end
     end
 
-    sort(script_counters, lt=(x,y)->x[2]<y[2])
-    if script_counters[2] > 0
-        return script_counters[1]
+    sort!(script_counters, lt=(x,y)->x[2]<y[2])
+    if script_counters[1][2] > 0
+        return script_counters[1][2]
     else
         return nothing
     end
@@ -410,7 +410,9 @@ function calculate_distance(lang_trigrams,  text_trigrams)
 end
 
 function detect(text::AbstractString, options=default_options())
+    if text==""; throw(ArgumentError("Cannot detect language for empty text")); end
     script = detect_script(text)
+    if script == nothing; return (nothing, nothing, 0); end
     lang, conf = detect_lang_based_on_script(text, script, options)
     return (from_code(lang), script, conf)
 end

src/word_lists.jl

@@ -5,17 +5,12 @@ for f in ("articles", "indefinite_articles", "definite_articles", "prepositions"
       return fetch_word_list(filename)
     end
   end
-end
-
 
-for f in ("articles", "indefinite_articles", "definite_articles", "prepositions", "pronouns", "stopwords")
+  # Deprecations
   @eval begin
     function $(Symbol(f)){T <: Language}(l::Type{T})
-      filename = Pkg.dir("Languages", "data", $f, string(string(l.name.name), ".txt"))
       Base.depwarn("Use of Languages as types is deprecated. Use instances.", Symbol(T))
-      return fetch_word_list(filename)
+      $(Symbol(f))(l())
     end
-
-    $(Symbol(f))(l::T) where T <: Language = $(Symbol(f))(T)
   end
 end