Parse guard statements like regular conditionals
I just learned that these two pieces of code are not the same:
def test if result = calculate_result return result end ... end
def test return result if result = calculate_result ... end
The second one will give this unhelpful error message:
NameError: undefined local variable or method `result' for main:Object Did you mean? result
I think this is confusing, especially since tools like RuboCop sometimes encourage users to change conditionals to inline ones ("guard statements"). I'd like to propose that Ruby be changed to treat these as the same.
#1 [ruby-core:87245] Updated by shevegen (Robert A. Heiler) about 1 month ago
Interesting. I personally very rarely do assignment of variables
followed by if-conditionals in my ruby code.
Perhaps there may be some parsing limitation? Otherwise it does
indeed looks like the same code to me (from the semantics),
just spread out onto more than one line in the first case).
As for tools such as rubocop - while this can be ignored for the
suggestion here (since your suggestion is just as valid without
any tools making recommendation), ruby hackers should never
become autobots working for the tools; it should be the other
way around, including making decisions. But that is just my
personal opinion - rubocop can be customized and adjusted as
far as I know anyway, so it's not a big deal. :)
#4 [ruby-core:87255] Updated by jeremyevans0 (Jeremy Evans) about 1 month ago
jacobevelyn (Jacob Evelyn) wrote:
I agree the
did_you_meanmessage is not good, but I would rather change the behavior so that line of code works and doesn't produce an error at all.
Doing that would change the behavior of code that currently works:
def calculate_result :calculate_result end def result :result end def test return result if result = calculate_result :test end
:result. With the change you are proposing,
test would return
#5 [ruby-core:87256] Updated by jacobevelyn (Jacob Evelyn) about 1 month ago
Ah you're right Jeremy, I hadn't thought about that case. I still think it's unexpected though that in your example the behavior would change if we split the conditional into an
if...end, but I don't know if others agree, and I don't know how a proposal like this that could possibly break existing code should move forward.