Encoding weirdness in .named_any scope

[Fodoj]

v. 3.2.1
ruby 2.0.0p353 (2013-11-22 revision 43784) [x86_64-darwin13.0.0]

There is some issue with encodings. In some cases encoding of tag is switched to ASCII, in other it's not and it gives an encoding exception explosion. Check it out:

# tag.rb
def self.named_any(list)
  if ActsAsTaggableOn.strict_case_match
    clause = list.map { |tag|
      sanitize_sql(["name = #{binary}?", as_8bit_ascii(tag)])
    }.join(' OR ')
    where(clause)
  else
    clause = list.map { |tag|
      puts as_8bit_ascii(unicode_downcase(tag))
      puts as_8bit_ascii(unicode_downcase(tag)).encoding
      puts sanitize_sql(['LOWER(name) = LOWER(?)', as_8bit_ascii(unicode_downcase(tag))])
      puts sanitize_sql(['SELECT (?)', as_8bit_ascii(unicode_downcase(tag))]).encoding
      puts "====="*10
      sanitize_sql(['LOWER(name) = LOWER(?)', as_8bit_ascii(unicode_downcase(tag))])
    }
    .join(' OR ')
    where(clause)
  end
end

=> str = ActsAsTaggableOn::Tag.named_any(["holä", "holä"])
=> holä
=> ASCII-8BIT
=> LOWER(name) = LOWER('holä')
=> ASCII-8BIT
=> ==================================================
=> holä
=> ASCII-8BIT
=> LOWER(name) = LOWER('holä')
=> ASCII-8BIT
### All is good, same encodings, but

=> str = ActsAsTaggableOn::Tag.named_any(["holä", "hol'ä"])
=> holä
=> ASCII-8BIT
=> LOWER(name) = LOWER('holä')
=> ASCII-8BIT
=> ==================================================
=> hol'ä
=> ASCII-8BIT
=> LOWER(name) = LOWER('hol\'ä')
=> UTF-8 # BAM! Different encoding! And as a result, when next part of code join(' OR ')
is executed..
=> ... incompatible character encodings: ASCII-8BIT and UTF-8

So one solution that works is to do instead:

sanitize_sql(['LOWER(name) = LOWER(?)', as_8bit_ascii(unicode_downcase(tag))]).force_encoding('BINARY')

then it's always ASCII and nothing explodes.

But looks weird to me. And I am not sure why ASCII is always needed here.

What could be another possible solution here? I can provide pull request with solution above though.

[jensljungblad]

Also noticed this problem. The above solution seems to work.

[Linuus]

Does any maintainer have any insights regarding this matter?

Ping @mbleigh @seuros

[seuros]

I think that affect mysql only.

[Linuus]

@seuros Ok. We are using Mysql so that seems right.

Any ideas on how to fix it? Perhaps merge the above proposal by @Fodoj

[Fodoj]

if you'd like I can prepare PR with this change

[seuros]

Submit a PR with tests, if it don't break other tests/adapter, i will merge it and release a new version.

[Linuus]

This PR fixed this issue, right?
#588

@seuros Any plans to release a new version soon?

[seuros]

Bundle update!