Asp Forum - Not quite getting it.

Roger Alsing

5/16/2008 6:46:00 PM

Hi,
I'm trying to get whats Ruby is all about.
I've been doing .NET since beta so I'm kind of stuck in the static typed
mindset.

The parts that I do get:

I love metaprogramming, I was able to make a port of my .NET AOP
framework that took about 3 months to do in .NET in a week in Ruby (w/o
ever touching ruby before.. )

I also understand that its nice with auto promotion of number types when
they would overflow in other languages.

What I don't get:

How do you deal with ducktyping?
How can anyone assume that just because a method exists that the
semantics is the same?

Brush.Draw( x , y )
vs
Cowboy.Draw( direction, speed )

The same name and signature, not even close to the same semantic
meaning.

Doesnt this put an extra burden on the developer, that he has to know
the exact intention of each and every method?
Where in a language that supports ifaces you simply know that the
methods associated with the interface have a certain meaning.

The backside of dynamic typing.
I got a mail from a newsgroup a few days ago with the title "we once
overflowed a long at google"

Ok ok, as I said, I get type promotion.
But what if you assign a "Giraffe" to a property called "Orders" , isnt
that just as bad or even worse than overflowing a number?

If someone say "that should be caught in a unit test" , well so should
the overflowed long (And Java and .NET can use big integers etc too if
you need)

And the thing that I have the hardest time with, how the heck do you
learn to use someone elses framework / api ?

in a typed language with decent intellisense I can see what types Im
supposed to pass in and what I get back.
One can often understand how to use an API by just looking at the
intellisense.

While I found myself browsing ruby docs back and forth just to see if
the result of a method would be a string or an object etc.

So can someone try to convert me or am I forever lost? :-P
--
Posted via http://www.ruby-....

33 Answers

Roger Alsing

5/16/2008 6:48:00 PM

"in a typed language"

Before I get killed, I ment static typed..
--
Posted via http://www.ruby-....

Dave Thomas

5/16/2008 6:58:00 PM

On May 16, 2008, at 1:45 PM, Roger Alsing wrote:

> Doesnt this put an extra burden on the developer, that he has to know
> the exact intention of each and every method?

I kinda think that this is the basis of programming in any language...

Here's the thing. I'm sure many people will try to convince you--this
may end up being a thread that runs and runs.

Ignore them. They'll never change your mind.

The only thing that will is for you to try it. Try Ruby on larger
programs. Try it for a month or too. And see what happens. If your
programs die because brushes end up shooting people, then Ruby isn't
the language for you. If instead you continue to enjoy a factor of 12
productivity improvement, maybe it is.

Everyone's mileage will vary.

Dave

Jason Roelofs

5/16/2008 7:03:00 PM

On 5/16/08, Roger Alsing <roger.alsing@precio.se> wrote:
> Hi,
> I'm trying to get whats Ruby is all about.
> I've been doing .NET since beta so I'm kind of stuck in the static typed
> mindset.
>
> The parts that I do get:
>
> I love metaprogramming, I was able to make a port of my .NET AOP
> framework that took about 3 months to do in .NET in a week in Ruby (w/o
> ever touching ruby before.. )
>
> I also understand that its nice with auto promotion of number types when
> they would overflow in other languages.
>
> What I don't get:
>
> How do you deal with ducktyping?
> How can anyone assume that just because a method exists that the
> semantics is the same?
>
> Brush.Draw( x , y )
> vs
> Cowboy.Draw( direction, speed )
>
> The same name and signature, not even close to the same semantic
> meaning.
>
> Doesnt this put an extra burden on the developer, that he has to know
> the exact intention of each and every method?
> Where in a language that supports ifaces you simply know that the
> methods associated with the interface have a certain meaning.
>
>
> The backside of dynamic typing.
> I got a mail from a newsgroup a few days ago with the title "we once
> overflowed a long at google"
>
> Ok ok, as I said, I get type promotion.
> But what if you assign a "Giraffe" to a property called "Orders" , isnt
> that just as bad or even worse than overflowing a number?
>
> If someone say "that should be caught in a unit test" , well so should
> the overflowed long (And Java and .NET can use big integers etc too if
> you need)
>
> And the thing that I have the hardest time with, how the heck do you
> learn to use someone elses framework / api ?
>
> in a typed language with decent intellisense I can see what types Im
> supposed to pass in and what I get back.
> One can often understand how to use an API by just looking at the
> intellisense.
>
> While I found myself browsing ruby docs back and forth just to see if
> the result of a method would be a string or an object etc.
>
>
> So can someone try to convert me or am I forever lost? :-P
>
> --
> Posted via http://www.ruby-....
>
>

Very oft brought up question, and valid in some points.

int someInterestingMethod(int x, int y, float spock) {
...
}

What do those parameters mean? What do you pass into this function and
what are you supposed to get out? Just having static types does not at
all help with understanding what a function does. You need well named
functions and arguments, with *documentation* on what the function
does and what parameters are supposed to be. This is no different in a
statically typed language vs dynamically typed language.

The only thing a static typed language guarantees is that proper types
were passed into the function. That's no guarantee that you're using
the function correctly or that the function itself works at all. Only
a comprehensive test suite will catch these kind of problems.

Things like overflowing a float will happen to anyone in any language.

So in short, all code should have:

1) documentation of what the method does, what the parameters are and
what the return is.
2) properly named functions and parameters.

neither of which has anything to do with the typing system of the
language. There will be libraries who's maintainers don't document
anything, leaving you to experiment and ask around. This is true for
all programming and is something we just have to live with (and
hopefully not propogate).

Hope that helps you understand why dynamic vs static typing really
isn't a issue.

Jason

Phillip Gawlowski

5/16/2008 7:21:00 PM

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Roger Alsing wrote:

| How do you deal with ducktyping?
| How can anyone assume that just because a method exists that the
| semantics is the same?

Well, if it walks like a duck, and it talks like a duck, it probably is
a duck.

I.e. if a Class or variable reacts to certain methods, it probably is of
the correct type.

| Brush.Draw( x , y )
| vs
| Cowboy.Draw( direction, speed )
|
| The same name and signature, not even close to the same semantic
| meaning.
|
| Doesnt this put an extra burden on the developer, that he has to know
| the exact intention of each and every method?

That's true of any language. .NET baffles *me*, for example. :P

| Where in a language that supports ifaces you simply know that the
| methods associated with the interface have a certain meaning.

It's similar in Ruby (but not quite). Unlike more 'blue-collar'
languages (not my term! And it is usually used in conjunction with Java,
but it probably applies to .NET as well), Ruby doesn't try to protect
you. It has some 'safety' measures, though.

'a string' + 2 Won't work, as the string is, well, a string, and the 2
is an integer. And you can't implicitly cast types in Ruby, for example.

| The backside of dynamic typing.
| I got a mail from a newsgroup a few days ago with the title "we once
| overflowed a long at google"
|
| Ok ok, as I said, I get type promotion.
| But what if you assign a "Giraffe" to a property called "Orders" , isnt
| that just as bad or even worse than overflowing a number?

Depends on the scenario, I guess, and what 'Orders' is supposed to do.
If it is, for example, an array holding all items in a shopping cart,
there is no issue. If 'Orders' is only supposed to contain objects of
the class 'Fruit', you are of course in a bit of a pickle. However, the
class 'Orders' is supposed to handle that.

Note: I don't think that data structures like arrays or hashes can
actually overflow. You can create them with arbitrary lengths (subject
to limitations like available memory, obviously).

Ruby's Array's aren't like (I think this is .NET syntax, but feel free
to correct me, it's been a while) Array ary = Array[9];, where ary can
hold only 10 members.

| If someone say "that should be caught in a unit test" , well so should
| the overflowed long (And Java and .NET can use big integers etc too if
| you need)

Well, this probably more an issue with the developer, than the language,
from what I can see.

| And the thing that I have the hardest time with, how the heck do you
| learn to use someone elses framework / api ?

How did you learn .NET? Same thing, really: Reading documentation, using
tutorials, experimentation.

| in a typed language with decent intellisense I can see what types Im
| supposed to pass in and what I get back.
| One can often understand how to use an API by just looking at the
| intellisense.

NetBeans, Eclipse, and Ruby in Steal (which is built upon visual Studio,
but isn't free, but has actual IntelliSense(tm)) can do that for you.

It hinges on the documentation provided by the gem / application you use
to tell you what the method expects.

| While I found myself browsing ruby docs back and forth just to see if
| the result of a method would be a string or an object etc.

Well, usually a Ruby object returns what you pass in. So, something that
handles arrays, returns an array. I think. If that is what you mean. :)

| So can someone try to convert me or am I forever lost? :-P

Hey, doing something that took you 3 months in .NET in 1 week in a
foreign language is a rather strong testimonial, isn't it?

Though, as Dave Black suggested, you'll have to try for yourself. Work
with Ruby for a while, and if you like it, you like it, and if you don't
like it, then you don't like it. :)

In my, admittedly limited experience in using statically typed language
versus Ruby, I don't need the additional safety or benefits the static
types provide. How ever, you mileage will vary.

- --
Phillip Gawlowski
Twitter: twitter.com/cynicalryan
Blog: http://justarubyist.bl...

~ - You know you've been hacking too long when...
...you're trying to get to sleep but can't because you don't know the
right VMS logical--you keep typing SHOW LOGICAL *SLEEP* and never get
anything but: SHOW-S-NOTRAN, no translation for logical name *SLEEP*
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.8 (MingW32)
Comment: Using GnuPG with Mozilla - http://enigmail....

iEYEARECAAYFAkgt3pwACgkQbtAgaoJTgL+kfQCfWxaHIIUz65IWBAnzfuJar9vH
pk0AnjFjlIbpgW6F11fblZgMHmf0tOJg
=8an1
-----END PGP SIGNATURE-----

Florian Gilcher

5/16/2008 8:17:00 PM

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On May 16, 2008, at 8:45 PM, Roger Alsing wrote:

> How do you deal with ducktyping?
> How can anyone assume that just because a method exists that the
> semantics is the same?
>
> Brush.Draw( x , y )
> vs
> Cowboy.Draw( direction, speed )
>
> The same name and signature, not even close to the same semantic
> meaning.

Actually, how do you guarantee those semantics in a statically typed
language?

Consider this (pseudo-java, sorry ;) ):

interface Drawable
{
abstract public void draw(int a, int b);
}

class Point implements Drawable
{
public void draw(int x, int y)
{
Canvas.drawPoint(x,y);
}
}

class CarThief implements Drawable
{
public void draw(int x, int y)
{
Car c = steal(Cars.random());
c.drive_to(x,y);
}
}

Both Classes implement Drawable but there is no way to check on two
"Drawable" whether
they share the same semantics.

Regards
Skade
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.8 (Darwin)

iEYEARECAAYFAkgt67EACgkQJA/zY0IIRZZ1sACgmVHN1laF+Xqr72t887Pd+n3J
TwwAniR9ESpQ0gwVGcOA5r4opjn1hLJ+
=GhYL
-----END PGP SIGNATURE-----

Keith Bennett

5/16/2008 10:51:00 PM

Florian, your example is not a fair one; it illustrates an intentional
violation of the semantics, whereas Roger is referring to an
accidental one. In fact, it illustrates the lengths to which one must
go to trigger the problem in Java.

Roger, your question is one that I have shared. I think it's
mathematically impossible to deny that the less restrictive nature of
duck typing makes it more likely to encounter an inappropriate
object. The question, though, is how *much* more likely? As in
premature optimization, do we really know that this will be a
problem? Or are we incorrectly assuming it?

People with much more experience with Ruby than I (such as Dave)
report that this is not a problem in practice. I think the minimal
risk here needs to be balanced with the formidable benefits of coding
in such a highly productive language as Ruby.

- Keith

ara.t.howard

5/16/2008 11:02:00 PM

On May 16, 2008, at 12:45 PM, Roger Alsing wrote:

> Ok ok, as I said, I get type promotion.
> But what if you assign a "Giraffe" to a property called "Orders" ,
> isnt
> that just as bad or even worse than overflowing a number?

then you will need a tall box.

a @ http://codeforp...
--
we can deny everything, except that we have the possibility of being
better. simply reflect on that.
h.h. the 14th dalai lama

Todd Benson

5/16/2008 11:49:00 PM

On Fri, May 16, 2008 at 5:55 PM, Keith Bennett <keithrbennett@gmail.com> wrote:
> Florian, your example is not a fair one; it illustrates an intentional
> violation of the semantics, whereas Roger is referring to an
> accidental one. In fact, it illustrates the lengths to which one must
> go to trigger the problem in Java.
>
> Roger, your question is one that I have shared. I think it's
> mathematically impossible to deny that the less restrictive nature of
> duck typing makes it more likely to encounter an inappropriate
> object. The question, though, is how *much* more likely? As in
> premature optimization, do we really know that this will be a
> problem? Or are we incorrectly assuming it?
>
> People with much more experience with Ruby than I (such as Dave)
> report that this is not a problem in practice. I think the minimal
> risk here needs to be balanced with the formidable benefits of coding
> in such a highly productive language as Ruby.
>
> - Keith

I tend to think that using a RDBMS for data integrity with an overlay
of Ruby is a nice compromise. I'm still anti-unit_test, but then, I'm
not a seasoned programmer and am probably missing the good stuff.

With meta: there's a very subtle geeky wet dream underlying the whole
concept of meta-programming. I'll stick around and see what happens,
but I'll use it sparingly.

Todd

Florian Gilcher

5/17/2008 12:12:00 AM

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On May 17, 2008, at 12:55 AM, Keith Bennett wrote:

> Florian, your example is not a fair one; it illustrates an intentional
> violation of the semantics, whereas Roger is referring to an
> accidental one. In fact, it illustrates the lengths to which one must
> go to trigger the problem in Java.

Actually, you don't have to go that far :). I know, the example is a
artificial.

Actually, i did only add the interface for illustration. A more common
case
would be the assumption that methods of subclasses share
common semantics with methods of superclasses. I had such
a case some time ago, when a fellow student assumed that a method
name was not taken and implemented it on a subclass, causing havoc
in the code calling it :). Granted, this is an indicator that the naming
was too generic, but the language didn't help a bit.

But the question was:

>How can anyone assume that just because a method exists that the
>semantics is the same?

This problem is present in both Java and Ruby and none of the
languages helps me much in this. It is much more a question of
programmers discipline, good naming and knowledge of the
ecosystem the software lives in :). Thats perhaps why you don't
experience it as a big problem. Considering that i assume that
ruby is not the first OOP-Language for many people, they've
already got a headstart when it comes to this.

But actually, i can produce a real-world sample for an interesting
semantic
changing behaviour that is much harder to reproduce in Java/C# and
would really be abuse in such languages.

If you know you way around ActiveRecord, calling #find on associated set
of models (e.g. @some_person.friends.find) will not call Array#find but
Friend.find. _But_ if you try inspecting the Object, it will happily
tell you that
it is an Array. (and not some subclass ...)
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.8 (Darwin)

iEYEARECAAYFAkguItkACgkQJA/zY0IIRZbmsACgujIPlOK7pEM5AeqEEWLf1aYI
mAsAnAjlb3EKVHGFGF/Jn5pBDBVul0AS
=aT3y
-----END PGP SIGNATURE-----

Florian Gilcher

5/17/2008 12:25:00 AM

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On May 17, 2008, at 1:02 AM, ara.t.howard wrote:

> On May 16, 2008, at 12:45 PM, Roger Alsing wrote:
>
>> Ok ok, as I said, I get type promotion.
>> But what if you assign a "Giraffe" to a property called "Orders" ,
>> isnt
>> that just as bad or even worse than overflowing a number?
>
>
>
> then you will need a tall box.

Pragmatic Programming ;).

Roger, just because variables are not typed doesn't mean that you can't
check for Types. Actually, Rubys type-system is quite strong, it doesn't
allow casts. You can very easily build objekt properties that check for
type or a certain set of methods. Something like:

class Test
attr_accessor :order

def order=(order)
if order.kind_of?(Giraffe)
raise Exception.new("We package nearly everything, but
Giraffes????")
end
end
end

So Ruby does typing, but on a different level. It assumes that all
objects are sane and
then gives you the possiblity to check all the information that the
object propagates and
use that for reasoning. See it as a service-based approach. Wouldn't
be the first time
a service does not do what the name implies ;).

The numbers problem: Conceptually, there is also no such thing as an
"overflowing number",
in Ruby because Ruby abstracts high enough that you as a Programmer do
not have to to
care about those technical implications. This is a completely
different ballpark ;).

Regards,
Florian Gilcher
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.8 (Darwin)

iEYEARECAAYFAkguJcMACgkQJA/zY0IIRZbWPQCfWGQIQGaBh1LkaHHqcUCtUbpZ
GiwAn3pWUIjEvwTubAwgodxuyK1oRLZX
=95BN
-----END PGP SIGNATURE-----

comp.lang.ruby

Not quite getting it.

Roger Alsing

Roger Alsing

Dave Thomas

Jason Roelofs

Phillip Gawlowski

Florian Gilcher

Keith Bennett

ara.t.howard

Todd Benson

Florian Gilcher

Florian Gilcher

x Login to ForumsZone