This reverts commit 867dc1700f32aae6f98c4651bd501597e6b52bc0, reversing
changes made to 9a421aaa8285cf2a7ecb1af370748b0337818930.
This breaks anyone who's using ForceSSL: https://travis-ci.org/rails-api/rails-api/jobs/5556065
Please see comments on #8156 for some discussion.
1. When comparing the directory to delete against the top level
cache_path, use File.realpath to make sure we aren't comparing two
unequal strings that point to the same path. This occurs, for
example, when cache_path has a trailing slash, which it does in the
default Rails configuration. Since the input to
delete_empty_directories never has a trailing slash, the comparison
will never be true and the top level cache directory (and above) may
be deleted. However…
2. File.delete raises EPERM when trying to delete a directory, so no
directories have ever been deleted. Changing the code to Dir.delete
fixes that.
ActiveSupport::NumberHelper#number_to_human now returns the number unaltered when
the units hash does not contain the needed key, e.g. when the number provided is less
than the largest key provided.
This commit standardises the return value of `to_time` to an instance
of `Time` in the local system timezone, matching the Ruby core and
standard library behavior.
The default form for `String#to_time` has been changed from :utc to
:local but research seems to suggest the latter is the more common form.
Also fix an edge condition with `String#to_time` where the string has
a timezone offset in it and the mode is :local. e.g:
# Before:
>> "2000-01-01 00:00:00 -0500".to_time(:local)
=> 2000-01-01 05:00:00 -0500
# After:
>> "2000-01-01 00:00:00 -0500".to_time(:local)
=> 2000-01-01 00:00:00 -0500
Closes#2453
I did this because to_date gives a very unhelpful error message if you
do not pass in a correct date. In the process I think this cleans up the
code nicely and even better it tends to be slightly faster than the
current implementation.
Benchmark
https://gist.github.com/4440875
Date, DateTime, Time and TimeWithZone can now be compared to infinity,
so it's now possible to create ranges with one infinite bound and
date/time object as another bound.
Ex.: @range = Range.new(Date.today, Float::INFINITY)
Also it's possible to check inclusion of date/time in range with
conversion.
Ex.: @range.include?(Time.now + 1.year) # => true
@range.include?(DateTime.now + 1.year) # => true
Ability to create date/time ranges with infinite bound is required
for handling postgresql range types.
When you add one callack in two separate `set_callback` calls - it is
only called once.
When you do it in one `set_callback` call - it is called twice.
This violates the principle of least astonishment for me. Duplicating
callback is usually an error. There is a correct and obvious way to do
anything without this "feature".
If you want to do
before_save :clear_balance, :calculate_tax, :clear_balance
or whatever, you should better do
before_save :carefully_calculate_tax
def carefully_calculate_tax
clear_balance
calculate_tax
clear_balance
end
And this even opens gates for some advanced refactorings, unlike the
first approach.
My assumptions are:
- Principle of least astonishment is violated, when callbacks are either
prevented from duplication, or not.
- Duplicating callbacks is usually an error. When it is intentional -
it's a smell of a bad design and can be approached without abusing
this "feature".
My suggestion is: do not allow duplicating callbacks in one callback
call, like it is not allowed in separate callbacks call.
The encoding scheme (e.g. ☠ -> "\u2620") was broken for characters
not in the Basic Multilingual Plane. It is possible to escape them
for json using the weird encoding scheme of a twelve-character
sequence representing the UTF-16 surrogate pair (e.g. '𠜎' ->
"\u270e\u263a") but this wasn't properly handled in the escaping code.
Since raw UTF-8 is allowed in json, it was decided to simply pass
through the raw bytes rather than attempt to escape them.
The Time.time_with_datetime_fallback, Time.utc_time and Time.local_time
methods were added to handle the limitations of Ruby's native Time
implementation. Those limitations no longer apply so we are deprecating
them in 4.0 and they will be removed in 4.1.
The to_time_in_current_zone method doesn't match the naming of the methods
for converting to ActiveSupport::TimeWithZone on Time and DateTime. Since
DateTime inherits from Date that has led to confusion with some users
using the to_time_in_current_zone method with DateTime instances and having
the time part dropped and the UTC offset lost.
This commit fixes this by deprecating the old method and adding a new
in_time_zone method which matches the naming for DateTime and Time. This
should prevent accidently dropping times and UTC offsets when converting
DateTime instances to ActiveSupport::TimeWithZone.
This commit adds a convenience method for converting a string to an
ActiveSupport::TimeWithZone instance using the configured Time.zone or
another passed as an argument.
Chrome, Safari and Firefox serialize Date objects to strings such
as 'Mon May 28 2012 00:00:00 GMT-0700 (PDT)'. When these strings
are parsed the zone is interpreted as 'GMT-0700' which doesn't
exist in the TzInfo list of timezones.
By taking advantage of the improved date/time handling in 1.9.3
we can use `Date._parse` and the `:offset` value which is parsed
correctly.
Three tests were amended to make them pass:
1. test_parse_with_old_date
This needed changing to a different value because the original
value was before EST was adopted so was being changed to a
LMT (Local Mean Time) value after the change. It didn't before
because `DateTime` just has offsets from UTC not timezones.
2. test_parse_should_not_black_out_system_timezone_dst_jump
Changed the implementation of this test as the stubs were
dependent on internal implementation details of the test.
Confirmed that the modified test still failed when the
implementation of `parse` was restored to pre-#5571.
3. test_parse_should_black_out_app_timezone_dst_jump
Ditto.
Closes#5770.
This behavior mattered under Ruby 1.8, but that doesn't matter now
that we don't support it.
In addition, we don't want to proxy the #class method. A test was added
to prevent against regressions.
