Update boolean handling of false

[colinbruce]

Expand the logic around blank value checking.

If @allow_blank is false, make sure it's not a BooleanValidator before looking for value.blank? as false is a valid boolean response, but

false.blank? # => true

Create a test to replicate what we are seeing in use, when passing the value of false to a :boolean parameter (as opposed to "false" which is parsed as a string)

[colinbruce]

I couldn't find a collaboration guide, so I hope this is okay?
If there are changes needed for CI, release notes, etc feel free to shout

[dbussink]

Given the discussion in #750 I think there's probably a deeper underlying issue here and not something purely affecting only the boolean case. So I think a fix / change also would need to address things more generically and not only for boolean values.

[colinbruce]

Agreed - I was only fixing the issue I saw! 👍
Some analysis of #733 and regression testing of that may be useful @mathieujobin ?

[mathieujobin]

@colinbruce Thank you so much for this quick fix and added tests.
I didn't think of this case when reviewing the other PR.
After I get another 👍 I will merge and release as 0.7.1

[dbussink]

@colinbruce Thank you so much for this quick fix and added tests.

@mathieujobin I don't think this fix is complete then though? See also the comment earlier in #749 (comment). Would more changes be necessary here?

[mathieujobin]

@dbussink this still appears to be a good fix to add, the remaining issue can be fixed as a separate PR.
we need a test that reproduce it.

[dbussink]

@dbussink this still appears to be a good fix to add, the remaining issue can be fixed as a separate PR.
we need a test that reproduce it.

What I was trying to say is that the issue extends to anything where .blank? returns true, not only for booleans then. But I'll look at a separate change if that's required then. I'm mostly not sure what the intended behavior is here and if it really needs a code change or a change to the release notes to indicate the change.

[mathieujobin]

I am thinking we should actually question ActiveSupport as to why false.blank? Is true?

I don't know of any other objects that would have such a problem.

The allow_blank option was defaulting to false but wasn't enforced. So thing #733 and this PR is good.

I think I will add more tests with more object types.