Hi, I know 1 is truthy and nil is falsy, so it still does as described in the method documentation, but in previous versions it returned true/false (as I expect by its signature).
Is there a reason this has changed?
Thanks!
Ruby methods don't have signatures. 
Hi, I know 1 is truthy and nil is falsy, so it still does as described in
the method documentation, but in previous versions it returned true/false (as I expect by its signature).
In Ruby you should not expect true/false from a method whose name ends in ? In Java they are the only allowed values, but in Ruby (as in Perl, C, and many others) *any* value has an interpretation in boolean context. Therefore you exploit that feature of the language.
From a method you have to expect what its contract guarantees.
Is there a reason this has changed?
The change happened here:
and made into 4.0.0 (3-2-stable still has the ternary).
The reason Rails in general does not commit to singletons in predicates is that we do not want to write artificial code like the one in 3-2-stable. It is ugly an unnecessary. Doing select_value we are done, so return right there.
Documenting a generic predicate interface allows you that freedom. So you use the predicate this way:
do_foo if exists?
and you test it this way
assert exists?
or
exists?.should be_true
You never do
do_foo if exists? == true
or
exists?.should == true
If the caller needs a singleton, given our contracts, it is its responsibility to get one out of the predicate:
(!!exists?).to_json
or whatever.