Currently, as far as I know, `if' and `unless' merely check if an object is of either types NilClass or FalseClass. This seems to contradict the duck-typing paradigm. I propose that we implement a method Object#to_b that is called by `if' and `unless'. It should return a boolean value. That way, a developer can decide whether or not an object should evaluate to true or false.
class Connection
def open?; end
def closed?; end
def to_b; open?; end
end
connection = Connection.new(...)
if connection
# ...
end
Implementation:
class Object
def to_b
true
end
end
class FalseClass
def to_b
false
end
end
class NilClass
def to_b
false
end
end
One problem I see with this proposal is that it seems like a lot of people check if a variable has been set by writing `if var ...'. This change would require that people wrote `unless var.nil?', which I personally find more correct as well.
This would break a lot of my code, personally. I used to use "unless
var.nil?" but it is easier to say:
if var and var.foo
than:
if (not var.nil?) and var.foo
I'm not *quite* sure what the "unless" version of the test would be:
unless var.nil? or var.foo.nil?
Not quite what I want. I think that #to_b would be problematic,
especially in the example that you gave, since it's more expressive to
say:
if conn.open?
than:
if conn
which implies you're checking to see if conn is really there.
-austin
···
On 11/16/05, Daniel Schierbeck <daniel.schierbeck@gmail.com> wrote:
Currently, as far as I know, `if' and `unless' merely check if an object
is of either types NilClass or FalseClass. This seems to contradict the
duck-typing paradigm. I propose that we implement a method Object#to_b
that is called by `if' and `unless'. It should return a boolean value.
That way, a developer can decide whether or not an object should
evaluate to true or false.
One problem I see with this proposal is that it seems like a lot of
people check if a variable has been set by writing `if var ...'. This
change would require that people wrote `unless var.nil?', which I
personally find more correct as well.
You could also ask that "and", "or", "not", "!", "||", "&&" be
operator methods. I think the reason all of this is built-in is
performance. All of this stuff doesn't have to go through the normal
method-call mechanism. If you do a benchmark on "unless x" vs. "if
x.nil?" you'll see the difference.
···
On 11/16/05, Daniel Schierbeck <daniel.schierbeck@gmail.com> wrote:
Currently, as far as I know, `if' and `unless' merely check if an object
is of either types NilClass or FalseClass. This seems to contradict the
duck-typing paradigm. I propose that we implement a method Object#to_b
that is called by `if' and `unless'.It should return a boolean value.
That way, a developer can decide whether or not an object should
evaluate to true or false.
I used to use "unless
var.nil?" but it is easier to say:
if var and var.foo
than:
if (not var.nil?) and var.foo
`if var and var.foo' would still work. If `var' is undefined (nil), its #to_b method will return false. So unless var is an object that overrides the #to_b method, your code will still work.
I can see the problem with the performance Matz brought up, but this was more meant as a philosophical discussion (though I can see now that I was quite concrete.) It actually came in response to the whole "subclassing FalseClass and NilClass" discussion, where I believe the #to_b method would be more true to the idea of duck-typing.
