Syck does not like Noémie. aka: Issues with UTF8 strings interpreted as Boolean values

[Bertg]

TLDR:

• It seems certain UTF8 strings will be parsed into Boolean values.
• It's not clear (by the tests or other examples) if this is intended
• It's very surprising
• Have no idea how to fix it

Bertg@acf9778#diff-4037826ea7c5075dd50bbcd81e29d17fR36

Longer story:

So, I was tracing down an issue in our test suite, where the faker gem was generating a false value for something that should be a string.

After a lot of digging I found that the Translation hash it loaded (using I18n), actually contained a false value. However, the YAML file does not.

This lead me to this awesome orb session:

2.3.1 :001 > $LOADED_FEATURES.grep(/YAML/).first
 => nil
2.3.1 :002 > require 'yaml'
 => true
2.3.1 :003 > $LOADED_FEATURES.grep(/YAML/).first
 => nil
2.3.1 :004 > YAML.load("Mélissa")
 => "Mélissa"
2.3.1 :005 > YAML.load("Noémie")
 => "Noémie"
2.3.1 :006 > require 'syck'
 => true
2.3.1 :007 > $LOADED_FEATURES.grep(/syck/).first
 => "/Users/bertg/.rvm/gems/ruby-2.3.1/gems/syck-1.3.0/lib/syck.bundle"
2.3.1 :008 > YAML.load("Mélissa")
 => "Mélissa"
2.3.1 :009 > YAML.load("Noémie")
 => false

And that's not all

2.3.1 :010 > YAML.load("--- 'Noémie'")
 => "Noémie"
2.3.1 :011 > YAML.load("--- ['Noémie']")
 => ["Noémie"]
2.3.1 :012 > YAML.load("--- [Noémie]")
 => [false]

So, let's see where the fail is!

2.3.1 :013 > YAML.load("Noé")
 => false
2.3.1 :014 > YAML.load("No")
 => false
2.3.1 :016 > YAML.load("Nom")
 => "Nom"
2.3.1 :017 > YAML.load("Noå")
 => false
2.3.1 :018 > YAML.load("Yeså")
 => true

So basically it seams a non ASCII character following a Boolean like value, will interpret that value as boolean.