From asheesh at asheesh.org  Thu Nov  3 11:20:07 2005
From: asheesh at asheesh.org (Asheesh Laroia)
Date: Thu, 3 Nov 2005 11:20:07 -0500 (EST)
Subject: [Patterns] Visitor pattern
Message-ID: <Pine.LNX.4.63.0511031110310.9576@supercore.laroia.net>

I'm looking at some basic Visitor design pattern code as part of this OOSE 
homework assignment.  I would like to share it, and then a comment I have.

Source: http://ootips.org/visitor-pattern.html

<Code>
package Transactions;

public class Transaction
{
   public abstract void accept(TransactionVisitor v);
}

interface TransactionVisitor
{
   public void visit(DepositTransaction dt);
   public void visit(WithdrawlTransaction wt);
}

class DepositTransaction extends Transaction
{
   public void accept(TransactionVisitor v)
   {v.visit(this);}
}

class WithdrawlTransaction extends Transaction
{
   public void accept(TransactionVisitor v)
   {v.visit(this);}
}

class TransactionUIFactory implements TransactionVisitor
{
   private UI itsUI;
   public UI getUI() {return itsUI;}
   public void visit(DepositTransaction dt) {
     itsUI = new DepositUI(dt);
   }
   public void visit(WithdrawlTransaction wt) {
     itsUI = new WithdrawlUI(wt);
   }
}
</Code>

My reaction: "How dumb!"

What this *should* be is a simple mapping of class type -> class 
constructor.  So screw this dumb overloading junk.  Just make a HashTable 
or Dictionary or whatever that maps class types to class constructor.

Then we turn TransactionUIFactory into:
class TransactionUIFactory implements TransactionVisitor
{
   private UI itsUI;
   public UI getUI() {return itsUI;}
   public Hash mapping = {DepositTransaction: DepositUI,
 			WithdrawalTransaction: WithdrawalUI};
   public void visit(Transaction t) {
     itsUI = new mapping[Transaction.type](t);
   }
}

Advantage of my code: The code is easier to read and has less duplication.

What's that, you can't do that in your dumb languages like Java or C#? 
You sure can in Python.  I see no reason you couldn't do it in LISP (or 
CLOS, Object-Oriented LISP, as Peter would remind me to mention and then 
cringe at the mention of).

And people tell me Java is object-oriented.  /me shakes his head

This is what I mean when I say the braindead uselessness of these 
languages makes you NEED many of our design patterns as a series of 
complex workarounds.  Surely there *are* interesting and worthwhile design 
patterns.  But when I see them used as crutches for missing features in 
these so-called object-oriented languages it makes me want to shoot 
someone.  This is hardly a visitor now; there's no explicit double 
dispatch.  It's just a regular-old boring function that takes some class 
in and returns some other class instance.

Moral of the story: There is a floor for how much duplication of code you 
have to have in every language.  One reason I prefer Python to Java 
because that floor is lower.  (And the kids these days seem to think Java 
is good.  Brendan, I'm looking straight at you.)

You are all free to take off your flame-retardant suits now.

-- Asheesh.

-- 
Some people have no respect for age unless it's bottled.

From phf at cs.jhu.edu  Fri Nov  4 02:36:25 2005
From: phf at cs.jhu.edu (=?ISO-8859-1?Q?Peter_Fr=F6hlich?=)
Date: Fri, 04 Nov 2005 02:36:25 -0500
Subject: [Patterns] Visitor pattern
In-Reply-To: <Pine.LNX.4.63.0511031110310.9576@supercore.laroia.net>
References: <Pine.LNX.4.63.0511031110310.9576@supercore.laroia.net>
Message-ID: <0062974055295b3f17572fd64eda2e78@cs.jhu.edu>

Hi all,

On Nov 3, 2005, at 11:20, Asheesh Laroia wrote:

> What's that, you can't do that in your dumb languages like Java or C#?
> You sure can in Python.  I see no reason you couldn't do it in LISP (or
> CLOS, Object-Oriented LISP, as Peter would remind me to mention and 
> then
> cringe at the mention of).

I hope I never really brought up CLOS? :-)

> And people tell me Java is object-oriented.  /me shakes his head

Well, the debate of what exactly "object-oriented" is has been raging 
for years. Actually it died down in the mid-to-late-nineties when most 
people settled on Peter Wegner's formula. I disagree with his, I don't 
think inheritance is a central feature. You seem to fall on the other 
side, thinking that lots of dynamism is essential. For me that's not an 
OO issue, just a language issue.

> This is what I mean when I say the braindead uselessness of these
> languages makes you NEED many of our design patterns as a series of
> complex workarounds.  Surely there *are* interesting and worthwhile 
> design
> patterns.  But when I see them used as crutches for missing features in
> these so-called object-oriented languages it makes me want to shoot
> someone.

Whoa... :-) The GoF is quite explicit in the book that some design 
patterns are "non-patterns" in certain languages, and they make the 
point by stating that in C even "inheritance" would be a pattern (if it 
were used widely that is). In fact, lots of people have gotten PhDs for 
the following research approach: (1) look at design patterns that are 
neat, (2) look at languages that need them, (3) extend those languages 
with features that make the design patterns go away (if the feature is 
used correctly). I'll try to include one of these discussions in the 
list of papers to look at for the last few meetings.

> You are all free to take off your flame-retardant suits now.

I think we should rather put them on and make sure that they're sealed 
tightly. :-)

Peter
--
Peter H. Froehlich <><><><><><> http://www.cs.jhu.edu/~phf/
OpenPGP: ABC2 9BCC 1445 86E9 4D59  F532 A8B2 BFAE 342B E9D9


From phf at cs.jhu.edu  Wed Nov 16 16:51:30 2005
From: phf at cs.jhu.edu (Peter Froehlich)
Date: Wed, 16 Nov 2005 16:51:30 -0500
Subject: [Patterns] Oops...
Message-ID: <6ec6fb2cf611a16923e16197fd291d05@cs.jhu.edu>

Hi all,

The reminder just popped up and I realized that I never posted a paper 
to read. Darn! So I'll come over to the room to tell you all the sad 
story and then that's that. :-/

Peter
--
Peter H. Froehlich <><><><><><> http://www.cs.jhu.edu/~phf/
OpenPGP: ABC2 9BCC 1445 86E9 4D59  F532 A8B2 BFAE 342B E9D9


From asheesh at asheesh.org  Wed Nov 30 01:35:57 2005
From: asheesh at asheesh.org (Asheesh Laroia)
Date: Wed, 30 Nov 2005 01:35:57 -0500 (EST)
Subject: [Patterns] Hey now, everybody now
Message-ID: <Pine.LNX.4.63.0511300135140.16998@supercore.laroia.net>

Peter, want to give us some reading?

-- Asheesh.

-- 
If we were meant to get up early, God would have created us with alarm clocks.

From phf at cs.jhu.edu  Wed Nov 30 11:30:13 2005
From: phf at cs.jhu.edu (=?ISO-8859-1?Q?Peter_Fr=F6hlich?=)
Date: Wed, 30 Nov 2005 11:30:13 -0500
Subject: [Patterns] Hey now, everybody now
In-Reply-To: <Pine.LNX.4.63.0511300135140.16998@supercore.laroia.net>
References: <Pine.LNX.4.63.0511300135140.16998@supercore.laroia.net>
Message-ID: <86d502c308d16f332a1f49c89250519c@cs.jhu.edu>

Hi all,

On Nov 30, 2005, at 01:35, Asheesh Laroia wrote:

> Peter, want to give us some reading?

I suck! Yesterday I ran into Promit and he reminded me as well, but I 
can't keep my wits about me as I am trying to write a simple 3D engine 
for the foundations course. :-/

Alright, so here's the first paper I thought of:

   http://portal.acm.org/citation.cfm?doid=286936.286952

It's somewhat "controversial" and maybe not written in the best English 
possible, but still interesting. Another interesting one is here:

   http://www.ccs.neu.edu/home/lorenz/papers/oopsla97/

Then there's the whole business about "meta patterns" that Wolfgang 
Pree started in 1994 or so, but no "official" copy of his papers seems 
to survive online right now. This one is just as good however:

   http://www.exciton.cs.rice.edu/comp410/frameworks/Pree/C010.pdf

I suggest we focus on the first paper and try to dissect/understand 
what they are saying and how much impact that has, but feel free to 
read the other two as well (if they strike you as more interesting).

Peter
--
Peter H. Froehlich <><><><><><> http://www.cs.jhu.edu/~phf/
OpenPGP: ABC2 9BCC 1445 86E9 4D59  F532 A8B2 BFAE 342B E9D9


From ussjoin at acm.jhu.edu  Wed Nov 30 18:33:23 2005
From: ussjoin at acm.jhu.edu (Brendan O'Connor)
Date: Wed, 30 Nov 2005 18:33:23 -0500
Subject: [Patterns] (no subject)
Message-ID: <438E36C3.8090600@acm.jhu.edu>

http://www.daimi.au.dk/~apaipi/workshop/master.pdf

Is the thesis.
---BFO

From phf at cs.jhu.edu  Wed Nov 30 18:58:08 2005
From: phf at cs.jhu.edu (Peter Froehlich)
Date: Wed, 30 Nov 2005 18:58:08 -0500
Subject: [Patterns] (no subject)
In-Reply-To: <438E36C3.8090600@acm.jhu.edu>
References: <438E36C3.8090600@acm.jhu.edu>
Message-ID: <6a63aaf14c6bebeab70852eb410cc388@cs.jhu.edu>

Hi all,

On Nov 30, 2005, at 18:33, Brendan O'Connor wrote:

> http://www.daimi.au.dk/~apaipi/workshop/master.pdf
> Is the thesis.

Thanks, material for a whole semester... :-) Asheesh probably likes 
this one, although I don't know where to find more information: 
http://dx.doi.org/10.1145/289423.289478

Peter
--
Peter H. Froehlich <><><><><><> http://www.cs.jhu.edu/~phf/
OpenPGP: ABC2 9BCC 1445 86E9 4D59  F532 A8B2 BFAE 342B E9D9