When using a validation matcher in the negative, i.e.:
should_not validate_*(...)
as opposed to:
should validate_*(...)
...it's common to receive the following error:
undefined method `attribute_setter' for nil:NilClass
This happens particularly when using a matcher that makes use of
AllowValueMatcher or DisallowValueMatcher internally (which all of the
validation matchers do).
Whenever you make an assertion by using a matcher in the negative as
opposed to the positive, RSpec still calls the `matches?` method for
that matcher; however, the assertion will pass if that returns *false*
as opposed to true. In other words, it just inverts the result.
However, whenever we are using AllowValueMatcher or
DisallowValueMatcher, it doesn't really work to invert the result. like
this. This is because AllowValueMatcher and DisallowValueMatcher,
despite their name, aren't truly opposites of each other.
AllowValueMatcher performs these steps:
1. Set the attribute on the record to some value
2. Run validations on the record
3. Ask whether validations pass or fail
4. If validations fail, store the value that caused the failure along
with the validation errors and return false
5. Otherwise, return true
However, DisallowValueMatcher performs these steps:
1. Set the attribute on the record to some value
2. Run validations on the record
3. Ask whether validations pass or fail
4. If validations *pass*, store the value that caused the failure along
with some metadata and return false
5. Otherwise, return true
This difference in logic is achieved by having AllowValueMatcher
implement `does_not_match?` and then having DisallowValueMatcher use
this for its positive case and use `matches?` for its negative case.
It's easy to see because of this that `does_not_match?` is not the same
as `!matches?` and vice versa.
So a matcher that makes use of these submatchers internally needs to use
their opposite versions whenever that matcher is used in the negative
case. In other words, all of the matchers need a `does_not_match?` which
is like `matches?`, except that all of the logic is inverted, and in all
the cases in which AllowValueMatcher is used, DisallowValueMatcher needs
to be used.
Doing this ensures that when `failure_message` is called on
AllowValueMatcher or DisallowValueMatcher, step 4 in the list of steps
above stores a proper value that can then be referenced in the failure
message for the validation matcher itself.
Forcing people to add ignoring_interference_by_writer for each and every
case in which an attribute changes incoming values is pretty obnoxious
on our part (for instance, when using the numericality matcher against
an integer column + `only_integer`). So now, it's enabled by default.
This effectively means that people should never get
AtttributeChangedValueErrors again.
`allow_value` matcher is, of course, concerned with setting values on a
particular attribute on a particular record, and then checking that the
record is valid after doing so. That comes with a caveat: if the
attribute is overridden in such a way so that the same value going into
the attribute isn't the same value coming out of it, then `allow_value`
will balk -- it'll say, "I can't do that because that changes how I
work."
That's all well and good, but what the attribute intentionally changes
incoming values? ActiveRecord's typecasting behavior, for instance,
would trigger such an exception. What if the developer needs a way to
get around this? This is where `ignoring_interference_by_writer` comes
into play. You can tack it on to the end of the matcher, and you're free
to go on your way.
So, prior to this commit you could already apply it to `allow_value`,
but now in this commit it also works on any other matcher.
But, one little thing: sometimes using this qualifier isn't going to
work. Perhaps you or something else actually *is* overriding the
attribute to change incoming values in a specific way, and perhaps the
value that comes out makes the record fail validation, and there's
nothing you can do about it. So in this case, even if you're using
`ignoring_interference_by_writer`, we want to inform you about what the
attribute is doing -- what the input and output was. And so we do.
Why:
* Previously, `allow_value` would raise a CouldNotSetAttributeError
if the value being set didn't match the value the attribute had after
being set, but only if the attribute was being changed from nil to
non-nil or non-nil to nil.
* It turns out it doesn't matter which value you're trying to set the
attribute to -- if the attribute rejects that change it's confusing
either way. (In fact, I was recently bit by a case in which I was
trying to validate numericality of an attribute, where the writer
method for that attribute was overridden to ensure that the attribute
always stored a number and never contained non-number characters.
This ended up making the numericality validation useless, of
course -- but it caused confusion because the test acted in a way
I didn't expect.)
To satisfy the above:
* `allow_value` now raises a CouldNotSetAttributeError if the attribute
rejects the value being set in *any* way.
* However, add a `ignoring_interference_by_writer` qualifier so that
it is possible to manually override this behavior.
* Fix tests that are failing now because of this new change:
* Fix tests for allow_value matcher
* Fix tests for numericality matcher
* Remove tests for numericality matcher + integer column
* An integer column will typecast any non-integer value to an
integer.
* Because of the typecasting, our tests for the numericality matcher
against an integer column don't quite work, because we can't
really test what happens when the attribute is set to a
non-integer value. Now that `allow_value` is more strict, we're
getting a CouldNotSetAttributeError when attempting to do so.
* The tests mentioned were originally added to ensure that we are
handling RangeErrors that ActiveRecord used to emit. This doesn't
happen anymore, so the tests aren't necessary anymore either.
* Fix tests for acceptance matcher
* Fix tests for absence matcher
When running unit tests we need to ensure that we are simulating how
other people will use the matchers as best we can. This means, for
instance, tagging any example groups that test controller matchers as
controller example groups.
* Change 'spec' Rake task to 'spec:unit'
* Require unit_spec_helper.rb in unit tests, not spec_helper.rb
* Re-namespace files in spec/support/unit under UnitTests
* Files in spec/support/unit/helpers no longer automatically add
themselves to RSpec - this happens in unit_spec_helper.rb
* Extract RecordWithDifferentErrorAttributeBuilder and
RecordValidatingConfirmationBuilder to separate files
* Move spec/shoulda to spec/unit_tests/shoulda
* Move spec/support/*.rb to spec/support/unit_tests/{helpers,matchers}
* Move spec_helper.rb to unit_spec_helper.rb
Elliot Winkler