Attribute setters for instance variables


#1

Hello,

The Pickaxe book presents the following example:

class Incorrect
attr_accessor :one, :two
def initialize
one = 1
self.two = 2
end
end

obj = Incorrect.new

p obj.one
p obj.two

The first thing printed is ‘nil’ and not ‘1’ since “one = 1” assigns to
a local variable of method initialize instead of the instance variable.
So far it’s clear.
Now, they propose a solution: to use self.one instead of just one.
Won’t be using @one simpler and also correct ? What am I missing ?

Thanks


#2

removed_email_address@domain.invalid wrote:

end
Won’t be using @one simpler and also correct ? What am I missing ?

Thanks

@one = 1 is setting the instance variable itself, and that’s fine for
accessors. If you want to do something special on =() then you will want
to use the self.one style. e.g.

def Foo
def one=(n)
@one = n+42
end
def initialize(n)
self.one = 0 # @one == 42
@one = 0 # @one == 0
end
end


#3

Hans F. wrote:

end
Now, they propose a solution: to use self.one instead of just one.
def one=(n)
@one = n+42
end
def initialize(n)
self.one = 0 # @one == 42
@one = 0 # @one == 0
end
end

Thanks for your swift and insightful reply.


#4

On Sun, 12 Mar 2006 removed_email_address@domain.invalid wrote:

end
Won’t be using @one simpler and also correct ? What am I missing ?
def one= arg
log{ “somthing about one” }
validate{ arg }
notify_observers :one, arg
@one = arg
end

just a few examples.

this become even more true when the number attribues increases since you
can

class C
ATTRIBUTES = %w[ one two three four five six seven ]
def initialize kws = {}
kws.each{|kw, val| send “#{ kw }=”, val}
end
end

in one line even if the number of attributes is huge.

i’d also argue that the simplest thing is to simply make the getter take
and
optional argument so one can do

def initialize
one 1
two 2
end

but that’s a matter of taste.

hth - cheers.

-a


#5

Mike A. wrote:

self.two = 2

So far it’s clear.
def Foo
advantages to this, including fix of the problem above, hard to debug
spelling mistakes, clearer code, etc. A simple ‘var’ declaration would
be simple:

What problem? I don’t see any problem, I quite like that behavior. I
also like not declaring variables, the inconvenience from spelling
errors is minimal, especially when the telltale nil problems are almost
as useful as a compiler error most of the time.


#6

Hans F. wrote:

end
Now, they propose a solution: to use self.one instead of just one.
def one=(n)
@one = n+42
end
def initialize(n)
self.one = 0 # @one == 42
@one = 0 # @one == 0
end
end

I really wish Ruby required explicit locals declaration. There are many
advantages to this, including fix of the problem above, hard to debug
spelling
mistakes, clearer code, etc. A simple ‘var’ declaration would be
simple:

def Foo
def one=(n)
@one = n+42
end
def initialize(n)
var x, y = n * n, 1 / n
one = 0 # @one == 42
@one = 0 # @one == 0
z = 10 # error
end
end

Regards,
Mike


#7

Hans F. wrote: [full message below]

I really wish Ruby required explicit locals declaration. There are
many advantages to this, including fix of the problem above, hard to
debug spelling mistakes, clearer code, etc. A simple ‘var’
declaration would be simple:

What problem? I don’t see any problem, I quite like that behavior. I
also like not declaring variables, the inconvenience from spelling
errors is minimal, especially when the telltale nil problems are
almost
as useful as a compiler error most of the time.

I can’t imagine coding thousands of lines of code without variable
declarations. If you catch the nil errors caused by spelling mistakes
right
away, sure, you can easily track it down. But I’d rather have the
compiler
tell me that a variable doesn’t exist than to give me an nil error who
knows
where. ‘var’ is only 4 more letters, and I’m not saying limit them to
the
beginning of the method or block. The only change would be prepending
‘var’.

Mike