Incorrect code detection

[whizzzkid]

I was wondering with any possibilities to use this awesome library to detect code snippets. But I think there is some errors with even the test cases mentioned. I tried using many code snippets shown here http://highlightjs.org/static/test.html let's have a look at the ruby code. I added the following code to the highlight.pack.js:

var code='class A < B; def self.create(object = User) object end end' +
'class Zebra; def inspect; "X#{2 + self.object_id}" end end' +
'' +
'module ABC::DEF' +
'  include Comparable' +
'' +
'  # @param test' +
'  # @return [String] nothing' +
'  def foo(test)' +
'    Thread.new do |blockvar|' +
'      ABC::DEF.reverse(:a_symbol, :\'a symbol\', :<=>, \'test\' + ?\012)' +
'      answer = valid?4 && valid?CONST && ?A && ?A.ord' +
'    end.join' +
'  end' +
'' +
'  def [](index) self[index] end' +
'  def ==(other) other == self end' +
'end' +
'' +
'anIdentifier = an_identifier' +
'Constant = 1' +
'render action: :new' +
'' +
'str =~ /^(?:foo)$/' +
'str =~ %r{foo|bar|buz$}' +
'str =~ %r!foo|bar$!' +
'str =~ %r[foo|bar$]' +
'str =~ %r(\(foo|bar\)$)';
console.log(hljs.highlightAuto(code).language);
console.log(hljs.highlightAuto(code).second_best.language);

On executing:

node highlight.pack.js

I was expecting the language to be ruby, but I got the output:

livecodeserver
ruby

In case of java I get

coffeescript
java

These were cases from the test, I even tried using rather some random snippet (let's say: http://attachment.fbsbx.com/hackercup_source.php?sid=1402033520037632) I got:

d
ocaml

I might be wrong with the usage, please correct me, or does this qualifies as a bug?
Thanks!

[isagalaev]

You're concatenating all the lines without line breaks between them, it becomes completely mangled as a result.

[whizzzkid]

You're awesome!... It worked... I built http://code.nishantarora.in/langdetect.js/ over this... now I can detect code languages of multiple files I needed to profile!...

Thanks!

[whizzzkid]

closing this!

[isagalaev]

I should note that using highlight.js specifically for detecting languages might not be a good idea as our goal is to produce a useful highlighting, not detect a language. I.e., we don't consider it a bug if a snippet language is detected incorrectly but resulting highlighting looks okay.

[whizzzkid]

Makes sense, but I just needed a crude mechanism to profile thousands of source code files (withoutfile extensions or meta comments). I think I can live with a percentage of errors. Though as you say it might not detect the language correctly, as of now this is the closest thing I have... Thanks!

[isagalaev]

Glad it worked for you :-)