I'm in the U.S., and we usually format dates as "month/day/year". I'm trying to make sure that my Rails app, using Ruby 1.9, assumes this format everywhere, and works the way it did under Ruby 1.8.
I know that lots of people have this issue, so I'd like to create a definitive guide here.
Specifically:
- '04/01/2011' is April 1, 2011, not Jan 4, 2011.
- '4/1/2011' is also April 1, 2011 - the leading zeros should not be necessary.
How can I do this?
Here's what I have so far.
Controlling Date#to_s behavior
I have this line in application.rb
:
# Format our dates like "12/25/2011'
Date::DATE_FORMATS[:default] = '%m/%d/%Y'
This ensures that if I do the following:
d = Date.new(2011,4,1)
d.to_s
... I get "04/01/2011", not "2011-04-01".
Controlling String#to_date behavior
ActiveSupport's String#to_date
method currently looks like this (source):
def to_date
return nil if self.blank?
::Date.new(*::Date._parse(self, false).values_at(:year, :mon, :mday))
end
(In case you don't follow that, the second line creates a new date, passing in year, month and day, in that order. The way it gets the year, month and day values is by using Date._parse
, which parses a string and somehow decides what those values are, then returns a hash. .values_at
pulls the values out of that hash in the order Date.new
wants them.)
Since I know that I will normally pass in strings like "04/01/2011" or "4/1/2011", I can fix this by monkeypatching it like this:
class String
# Keep a pointer to ActiveSupport's String#to_date
alias_method :old_to_date, :to_date
# Redefine it as follows
def to_date
return nil if self.blank?
begin
# Start by assuming the values are in this order, separated by /
month, day, year = self.split('/').map(&:to_i)
::Date.new(year, month, day)
rescue
# If this fails - like for "April 4, 2011" - fall back to original behavior
begin
old_to_date
rescue NoMethodError => e
# Stupid, unhelpful error from the bowels of Ruby date-parsing code
if e.message == "undefined method `<' for nil:NilClass"
raise InvalidDateError.new("#{self} is not a valid date")
else
raise e
end
end
end
end
end
class InvalidDateError < StandardError; end;
This solution makes my tests pass, but is it crazy? Am I just missing a configuration option somewhere, or is there some other, easier solution?
Are there any other date-parsing cases I'm not covering?