On Wed, Mar 21, 2007 at 10:36:05AM +0900, Daniel B. wrote:
Not true. Perl 6 will have optional static typing [1]. Sydney faked it
with a Behavior you could import (that used the parser to handle it, not
a compile time step, IIRC).
It’s possible. The question is whether or not it’s desirable. I vote
yes, others say no, and still others are in favor of some sort of type
inferencing. It’s been brought up before - you can search the archives.
I’m a fan of Ruby’s dynamic typing. When I want type inference, I use
OCaml instead. My thought is not that I should have one language that
does everything, but that each language I have available is optimally
useful for the tasks in which I choose to employ it.
That being said . . . I’m not saying we should necessarily have
overloading in Ruby. I just wonder if there might be another way to
solve the same problem without having to resort to development patterns
to “work around” a lack of easy functionality. I’m not entirely sold on
the idea that there’s enough “working around” to bother in this case,
but there might be.
On Wed, Mar 21, 2007 at 12:27:04PM +0900, David A. Black wrote:
things like attributes and class methods from the building blocks that
Ruby offers. The things that get conjured up tend to resemble things
from other languages – but maybe there are entirely new constructs
waiting to arise out of Ruby’s components.
Hmm. I like that explanation. It’ll bear some thought.
On Wed, Mar 21, 2007 at 10:30:56AM +0900, Nando S. wrote:
Hi Chad! I don’t think Ruby “lacks” anything, it’s just that it has a
different way of doing things. It’s a trade-off between the flexibility
of a dynamic language and some other things like methods with same name
and different parameters (how are you going to distinguish between one
method that uses a string and an integer and another one that uses an
integer and a hash, if you don’t have types defined at the parameter
level?). If you want to do things “like Java” or “like C++”, then I
think you should open your mind and try to do things in a different way,
in a rubbish way.
Trust me – I have no interest in shoehorning anything into a Java or
C++ mold. That wasn’t my point.
I’m a fan of Ruby’s dynamic typing. When I want type inference, I use
Thus, my question.
I think the availability of the allocate method is very characteristic
of Ruby, and not a workaround. It’s just Ruby being Ruby. One thing
I’ve always loved about Ruby is that it lets you do things through
composition of simple techniques that are virtually equivalent to
language-level constructs and facilities in other languages. The
classic examples are “attributes” and “class methods.” All you need
to create attributes in Ruby is:
instance variables
methods
and (almost) all you need for class methods (except a bit of
inheritance special-casing) is:
objects can have singleton methods
classes are objects.
It’s like Ruby is saying, “You want ‘attributes’? OK. Here. Now
you’ve got ‘attributes’.” It’s all built on the simplest possible
things in Ruby, and on not very many of them.
Multiple constructors seem to fall into that category. All you have
to do is use allocate, and you can have as many constructors as you
need.
I wonder what lies ahead. I mean, look how easy it is to conjure up
things like attributes and class methods from the building blocks that
Ruby offers. The things that get conjured up tend to resemble things
from other languages – but maybe there are entirely new constructs
waiting to arise out of Ruby’s components.
I wonder what lies ahead. I mean, look how easy it is to conjure up
things like attributes and class methods from the building blocks that
Ruby offers. The things that get conjured up tend to resemble things
from other languages – but maybe there are entirely new constructs
waiting to arise out of Ruby’s components.
This statement reminds me of something Alan Kay said to me after he
had given a talk on Smalltalk at IBM. He said that one of his
disappointments was that no one ever seemed to make new subclasses of
Behavior (which in Smalltalk is the abstract class which both Class
and Metaclass subclass).
I’ve heard it said – and I tend to agree – that regular use of
“patterns” is a sign that the language lacks something. I wonder if
that is the case with Ruby, as applies to multiple constructor behavior
implementation.
I don’t consider this to be a pattern, I consider overloading
constructors an anti-pattern.
One of the few good Java books (among all the mediocre or even worse
ones) is “Effective Java” by Joshua Bloch:
In the first chapter (actually chapter 2) in the first item Bloch
advises Java programmers to consider using static factory methods
instead of public and/or overloaded constructors. He lists the following
advantages:
: One advantage of static factory methods is that, unlike constructors,
they have names.
: If the parameters to a constructor do not, in and of themselves,
describe the object being
: returned, a static factory with a well-chosen name can make a class
easier to use and the
: resulting client code easier to read.
: […]
The same is true in Ruby, if you use class methods (methods of a class’s
eigenclass). The workarounds in the initialize method even look code
smelly. In Java overloading a constructor once or twice doesn’t seem to
be such a bad idea. But other programmers tend to just copy and paste
another constructor below the others. Once you have a class that comes
into the two digit range of constructors, it gets quite difficult to
remember all the different possible permutations of types. (Yes, I had
to maintain code like that.)
And I think, overloading in general isn’t such a good idea either. It
can lead to really funny semantic errors, if it teams up with Java’s
numeric type promotion for example. (Yeah, this happened to me, too.)
: A second advantage of static factory methods is that, unlike
constructors, they are not
: required to create a new object each time they’re invoked. This allows
immutable classes
: (Item 13) to use preconstructed instances or to cache instances as
they’re constructed and to
: dispense these instances repeatedly so as to avoid creating
unnecessary duplicate objects.
This is an advantage in Java only, in Ruby “constructors” already have
this advantage, the are actually factory methods - and classes are
nothing but object factories, objects that can make other objects. Ruby
doesn’t have “real” constructors like Java has. In Ruby a “constructor”
is a simple method, defined in class Class and thus can be shadowed by
defining a method in a class’s eigenclass, where new objects can be
allocated, configured, cached, and so on. These changes are also
transparent for client code, which doesn’t have to be changed to, e. g.,
profit from caching.
: A third advantage of static factory methods is that, unlike
constructors, they can return
: an object of any subtype of their return type. This gives you great
flexibility in choosing
: he class of the returned object.
This is a Java advantage only, because Ruby doesn’t have to work around
a rigid type system. You just have to take care not to screw up.
: One application of this flexibility is that an API can return objects
without making their
: classes public. Hiding implementation classes in this fashion can lead
to a very compact API.
: This technique lends itself to interface-based frameworks, where
interfaces provide natural
: return types for static factory methods.
Can be done in Ruby as well for both “new” factory methods or better
named factory methods.
: The main disadvantage of static factory methods is that classes
without public or
: protected constructors cannot be subclassed.
: […]
This isn’t a disadvantage in Ruby, because private methods can be
called from subclasses.
: A second disadvantage of static factory methods is that they are not
readily
: distinguishable from other static methods. They do not stand out in
API documentation in
: the way that constructors do. Furthermore, static factory methods
represent a deviation from
: the norm. Thus it can be difficult to figure out from the class
documentation how to instantiate
: a class that provides static factory methods instead of constructors.
This disadvantage can be
: reduced by adhering to standard naming conventions.
This might be a disadvantage, but you should choose meaningful names for
all methods and document accordingly anyway.
Why do people want to do this? Other than just to do it. What was it
about C letting you shoot yourself in the foot, and C++ letting you
blow your whole leg off…
and/or overloaded constructors. He lists the following advantages:
The same is true in Ruby, if you use class methods (methods of a class’s
type promotion for example. (Yeah, this happened to me, too.)
and this doesn’t even take into account the confusion that can be heaped
upon
the programmer when default parameters are thrown into the mix. if
parameter
overloading were allowed in ruby we could have this:
def m int
p int
p 42.0
end
def m int, float = 42.0
p int
p float
end
m 42
now, which one is called? in order to understand this kind of stuff in
c++
you have to become the compiler.
It’s all built on the simplest possible