Implement fetch_values for HashWithIndifferentAccess

[joshpencheon]

fetch_values was added to Hash in Ruby 2.3.0: https://bugs.ruby-lang.org/issues/10017

This patch adds an implemention for instances of HashWithIndifferentAccess, in line with the existing definitions of fetch and values_at.

Current behaviour, without patch:

hash = ActiveSupport::HashWithIndifferentAccess.new
hash[:a] = 'x'
hash[:b] = 'y'
hash.fetch_values('a', 'b') # => KeyError: key not found: "a"

New behaviour, with patch:

hash = ActiveSupport::HashWithIndifferentAccess.new
hash[:a] = 'x'
hash[:b] = 'y'
hash.fetch_values('a', 'b') # => ["x", "y"]
hash.fetch_values('a', 'c') { |key| 'z' } # => ["x", "z"]
hash.fetch_values('a', 'c') # => KeyError: key not found: "c"

Thanks, Josh

[matthewd]

Let's throw an if Hash.method_defined? :fetch_values onto the end here, so we don't get ahead of Hash.

(And a skip in the test to match)

[joshpencheon]

Thanks @matthewd, I hadn't considered that. I've made the changes you suggested; is there any formal way of documenting that these clauses can be removed once Ruby 2.3 is the minimum requirement?

I've rebased now development has moved to 5.2, and squashed my commits. Is there anything this PR needs?

[matthewd]

Thanks!

is there any formal way of documenting that these clauses can be removed once Ruby 2.3 is the minimum requirement

Nope. I don't think we have enough of them to warrant a formal process... we'll just look for version-relevant conditionals (like method_defined? calls) when the time comes.

[sandstrom]

I was just about to send a PR when I saw this! 😄 Great addition @joshpencheon!