The TranslatedString#to_s now returns nil some cases

[raxoft]

The commit 2e54394 introduced some bugs.

First of all, I believe TranslatedString#to_s should check if the @value responds to html_safe, not self.

More important thing however is that now to_s doesn't work at all in some cases. Before, code like this was working fine for translated strings:

puts text.split("\n").first.to_s.encoding

Now it fails (with undefined method encoding for nil:NilClass error). The reason is that split creates array of TranslatedStrings, none of which however have the @value set. Cloning the class of the actual string itself seems to be a feature of split, but it obviously causes mismatch between the value of the string itself and the @value variable, which is now lacking (not to mention the lack of other instance variables, but it's not much of an problem for me).

I believe other String operations might suffer from the same problem, so it might be best to drop the @value idea entirely and base the value of the string solely on the string itself.

[raxoft]

BTW, the split behaviour can be easily tested with this snippet:

class C < String
end

p C.new("test").split("t").map{ |x| [x.class, x] }

Other string methods really seem to share this class cloning behaviour, such as scan.

[raxoft]

BTW, testing the mentioned commit a bit further, it also seems that the conversion of value to string added to the beginning of TranslatedString#initialize handles only TranslatedString, but not Untranslated strings. As a consequence, calling to_s on Untranslated string now returns TranslatedString, rather than String as expected.

[raxoft]

And while we are at it, the change at translation.rb:96 from @filters to @filter looks just like a typo - there is no @filter instance variable in the Translation class.

[ai]

Thanks for really good investigation. I will look closer on this week.

[raxoft]

Hello, sorry for the bump, but is there any chance to get this fixed? Alternatively, I can try to create a patch myself. Just one question - do you remember what was the idea behind introducing @value instead of using self as before? Thanks.

[ai]

Yeap, PR will be very welcome

[raxoft]

OK, I will have a look. Is support for 2.2+ enough or shall I aim for 2.1 as well?

[ai]

2.1 support will be good, because I need a major release to drop 2.1 support.

[raxoft]

OK, the PR is ready in #147.