Question

I’ve got a gem that’s used a bunch of people using a bunch of different Ruby interpreters, and it includes what boils down to this code:

res = RestClient.post(...)
doc = REXML::Document.new(res).root

The content of res is always UTF-8, and this works fine in Ruby 1.8, but it blows up under Ruby 1.9 if the response is not pure ASCII and the user’s default encoding is not UTF-8.

Now, if I wanted to make this work on Ruby 1.9 alone, I’d just stick res.force_encoding('utf-8') in there and be done with it, but that method is 1.9-only and then breaks under Ruby 1.8:

NoMethodError: undefined method `force_encoding' for #<String:0x101318178>

The best solution can come up with is this, which forces the systemwide default encoding to UTF-8:

Encoding.default_external = 'UTF-8' if defined? Encoding

Better ideas, or is this as good as it gets? Will there be any negative impact on library users who are trying to use different encodings?

Answer 1

I’m with Mike Lewis in using respond_to, but don’t do it on the variable res everywhere throughout your code.

I took a look at your code in gateway.rb and it looks like everywhere you are using res, it gets set by a call to make_api_request so you could add this before your return statement in that method:

doc = doc.force_encoding("UTF-8") if doc.respond_to?(:force_encoding) 

Even if it’s other places but it’s not literally with every string you encounter, I’m sure you can find a way to refactor the code that makes sense and solves the problems in one place instead of everywhere you encounter it.

Are you having a problem with other places?