Generate data files with frozen_string_literal: true

[casperisfine]

It's not a lot, but some of the strings in these generated files are duplicated and a bit of memory could be saved by enabling frozen_string_literal.

There's another source of duplication that could be removed, but not sure if you'd be open to the fix.

For instance:

# encoding: UTF-8

# This file contains data derived from the IANA Time Zone Database
# (https://www.iana.org/time-zones).

module TZInfo
  module Data
    module Definitions
      module Cuba
        include TimezoneDefinition
        
        linked_timezone 'Cuba', 'America/Havana'
      end
    end
  end
end

Here "Cuba" exist twice, one is the backing string for :Cuba symbol, because they are automatically casted as US-ASCII when applicable:

>> Cuba.name.encoding
=> #<Encoding:US-ASCII>

So a trick to reduce the memory further could be to set # encoding: US-ASCII.