From: Daniel Brockman <daniel@...>
Date: 2005-07-27T10:31:40+09:00
Subject: Re: defining to_s works oddly?

Hi Martin,

> Am I the only one that finds it odd that if I define a
> to_s function for a class, that if I do
>
>    puts obj
>
> I won't get errors, but if I try something like
>
>    s = String.new
>    s += obj
>
> I get a [TypeError]

You're probably not the only one.  The reason why ‘puts’
converts its argument using ‘to_s’ but String#+ doesn't is
far from obvious.

> It's no prob to write obj.to_s, it's just I expected the
> first version to call to_s automatically...

As James Edward Gray II quoted Why saying, “to_str is an
implicit cast, whereas to_s is an explicit cast,” and you
could probably redefine String#+ to perform the explicit
cast instead of the implicit with no problems resulting.

But I wouldn't do that.  There is a significant difference
between ‘"#{foo}"’ (or ‘puts foo’) and ‘"bar" + foo’:
The former *must* result in a string, whereas the latter
could reasonably result in something other than a string.

Consider these two expressions:

         "1" + 2        and         1 + "2"

Current Ruby behavior is to reject both with type errors.
See, Fixnum#+ implicitly casts its argument into a number,
and String#+ implicitly casts its argument into a string,
but neither argument can be implicitly casted that way.

[Again, the term “implicit cast” is not an exact description
of what happens, but rather a shorter way of saying “convert
if possible and fail otherwise.”]

You could make either method use an explicit cast instead of
an implicit, but if you make both explicit you would be in
the following weird situation:

   "1" + 2  #=> "12"

   1 + "2"  #=> 3

Assuming you agree that this asymmetry is unacceptable,
the question is, which cast should take precedence?

Perl avoids this ambiguity by letting "1" == 1, so that

   "1" + 2 == 1 + "2" == "3" == 3.

Ruby avoids it by requiring an explicit cast:

   "1" + 2.to_s  #=> "12"

   1.to_s + "2"  #=> "12"

   1 + "2".to_i  #=> 3

   "1".to_i + 2  #=> 3


I know this message doesn't provide any solutions, but I
hope it can help you accept Ruby's behavior as reasonable
(something I consider very important for trust reasons).

-- 
Daniel Brockman <daniel@brockman.se>

    So really, we all have to ask ourselves:
    Am I waiting for RMS to do this?   --TTN.