A TEST OF YOUR NEWLISP SKILLZ!
Something that spawned from
my-or is a macro, but not just any macro, it's a
Sounds simple? Try it, it's not. Here, I'll help you out (evil grin). Here is the solution as written in Scheme:
(lambda (x y)
(let ((temp (gensym)))
`(let ((,temp ,x))
(if ,temp ,temp ,y)))))
Once you've written 'my-or', run it against this program to test it:
(println "(my-or temp nil) = " (my-or temp nil))
(println "(my-or nil temp) = " (my-or nil temp))
(println "-----------")
(set 'value (my-or (begin (println "first arg") 1) (begin (println "second arg") 2)))
(println "should be 1: " value)
(println "-----------")
(set 'value (my-or (begin (println "first arg") nil) (begin (println "second arg") 2)))
(println "should be 2: " value)
IMPORTANT: In your implementation, make sure to name the symbol that you store the result of evaluating the first argument as 'temp'! Otherwise the variable-capture challenge is moot.
The winner is the first person to get the following output
(my-or nil temp) = 45
-----------
first arg
should be 1: 1
-----------
first arg
second arg
should be 2: 2
PRIZE
Hmm... since I'm also the developer of
Post your solution below if you've got it, I will give the solution on April 11th if no one solves it in that time.
(set 'my-or0 (lambda-macro(i j)
(let ((temp (eval i)))
(if temp
temp
(eval j)))))
(set 'my-or (protect2 'my-or 'my-or0 '(i j temp)))
(my-or nil temp) = 45
-----------
first arg
should be 1: 1
-----------
first arg
second arg
should be 2: 2
Described in article "don't fear dynamic scope (2)"
Nice, as I'm an avid reader of your blog, I should have anticipated your answer Kazimir. :-)
If you'd like a license to Espionage then send me a PM with your full name and email and I'll send it to you.
However, your solution, although very interesting (and a nice approach to these kinds of problems in newLISP), isn't quite the one that I was looking for, so I'm still leaving the challenge (and an additional license) open, but modifying the rules a little bit:
I'm looking for a single macro function, that gets the job done without any other user-defined functions or macros, and is much shorter than the length of the combined user-defined functions that Kazimir used to solve it. :-D
Edit: Kazimir, even though you've already submitted a working solution to the challenge as stated previously, you are free to try again with the new modifications. :-)
(lambda-macro (x y)
(eval (let ((temp (sym (string (inc counter)))))
(expand
'(let ((temp (eval x)))
(if temp ; Naive
temp ; version
(eval y)))
'temp)))))
I wish I could declare you the winner, as that is essentially the newLISP equivalent if the Scheme solution (ugly isn't it?), however, because the solution as specified in the contest revision does not allow user-defined functions, you were forced to do this:
Whereas the proper solution would have used (gensym) as in the Scheme solution. If newLISP had a gensym function I suppose I would be unable to debate and would be begrudgingly forced to accept this as the winner, but thankfully for the competition it does not. :-)
Because the contest rules forbid the use of a user-defined gensym function, this macro as presented is not safe from variable capture. If I were to replace (set 'temp 45) with (set 'counter 45), your function would not produce the correct output.
Kazimir, you are still a brilliant programmer in my eyes, but it is possible to solve this problem within the scope of the contest rules. That solution remains to be found. :-)
(lambda-macro (x y)
(first (list (eval (let ((temp (sym (append (string (last (symbols))) "+"))))
(expand
'(let ((temp (eval x)))
(if temp ; Naive
temp ; version
(eval y)))
'temp)))
(delete (last (symbols))) ))))
Wow.
I'm shocked and am laughing at the same time, you are incredible Kazimir!
That is a real surprise solution. At the very least this makes for a very intriguing topic. It fits within the contest rules, there's nothing that I can do to argue, so I begrudgingly accept this as the
HOWEVER! That doesn't mean that I can't offer the same prize for a second solution!! :-D
I have a problem with the above solution: it is too slow. I stipulate that it's still possible to find a second solution within the confines of the contest rules that is
Interesting note, of all the solutions that I have tested (which includes all of those presented here, and the yet-to-be discovered solution), the fastest one is the one using gensym:
(sym (string "gensym-" (inc gensym:counter))))
(define-macro (my-or)
(eval (let ((temp (gensym)))
(expand
'(let ((temp (eval (args 0))))
(if temp ; Naive
temp ; version
(eval (args 1))))
'temp))))
The yet-to-be discovered solution is about as fast as that one, but in my benchmarks it trails it just by a little bit (a couple milliseconds when executed 10000 times, but consistently).
The
OK, we'll unite these two of my prices for one, since the second one is just the version of the first one + that little gensym trick. When I'll buy Mac, I'll contact you.
But obviously you have something different (and faster) I cannot see now. I have to take some rest now, it is early morning in Croatia, it was fun session. Thanks for nice words and contribution to genlet library.
Good-night, and thank you for your creative solutions! :-)
The
(let (my-or:temp (eval (args 0)))
(if my-or:temp my-or:temp (eval (args 1)))))
(my-or 1 nil) => 1
(my-or nil 1) => 1
it is as fast as the
Many newLISP versions ago there was this utility function in the manual to define context-enclosed functions:
(def-new 'body (sym s s)))
the function takes a new function-name symbol in
(fn-macro (x y)
(let (temp (eval x)) (if temp temp (eval y)))))
(my-or 1 nil) => 1
(my-or nil 1) => 1
we also can omit the (args ...) trick now and name the variables for better readability.
A general comment:
Instead of trying to emulate Scheme or traditional LISPs we should emphasize typical newLISP solutions. Trying to do Scheme in newLISP only leads to in-efficient and often ugly code and newLISP is perceived as a second-class Lisp becuase of that.
Just as Scheme is designed around lexical scope and closures, newLISP is designed around dynamic scoping and namespaces. Both approaches are designed to avoid symbol-name clashes and maintain state. I believe that the newLISP approach is in the end easier to understand and more open to explore future programming paradigms.
Quote from: "Lutz"
As I mentioned
Perhaps I'm a lunatic, but I think that
This wouldn't be such a problem if newLISP supported nested contexts, but since it doesn't great care and caution must be taken when deciding upon the name of a context.
If you're going to recommend that as the "default" method of writing macros, then at the very least there should be a disclaimer somewhere in there about this, and a recommendation to use C-style naming conventions to attempt to avoid name conflicts.
I.e. If my company/organization name was "Pink Flowers" then I would write this as:
(let (PFmy-or:temp (eval PFmy-or:x))
(if PFmy-or:temp PFmy-or:temp (eval PFmy-or:y))))
IMO, if such naming conventions are ignored, then the
Quote
There is no such thing as "running out of namespaces". A namespace overhead consists of nothing more than one additional symbol, the context symbol. You can have as much namespaces has you can have symbols.
Quote
Yes, any non-trivial programming projects should employ naming conventions of functions and variables in any programming language.
You wouldn't write a bigger macro like this:
(let (PFmy-or:temp (eval PFmy-or:x))
(if PFmy-or:temp PFmy-or:temp (eval PFmy-or:y))))
But it is handy for smaller functions, to write it this way. For bigger functions you would use this:
(define-macro (PFmy-or x y)
(let (temp (eval x))
(if temp temp (eval y))))
(context MAIN)
Or use a function like 'def-static', as shown earlier. 'x' and 'y' are now part of 'PFmy-or' space. Of course naming conventions for functions and macros are still a good idea. But that has nothing to do with the fact, that a default functor is used to write the macro.
When writing many macros in a row, you could do this:
(define (Foo ...)
...
)
(context 'MAIN:Bar)
(defne (Bar ...)
...
)
; etc ;
This way saving one statement and closing the previous and opening the next namespace at the same time.
Quote from: "Lutz"
Sorry, that was a bad choice of words, I simply meant that it would be increasingly difficult to pick a name for a context without running into conflict with an existing one.
Quote
(let (PFmy-or:temp (eval PFmy-or:x))
(if PFmy-or:temp PFmy-or:temp (eval PFmy-or:y))))
But it is handy for smaller functions, to write it this way. For bigger functions you would use this:
(define-macro (PFmy-or x y)
(let (temp (eval x))
(if temp temp (eval y))))
(context MAIN)
Or use a function like 'def-static', as shown earlier. 'x' and 'y' are now part of 'PFmy-or' space. Of course naming conventions for functions and macros are still a good idea. But that has nothing to do with the fact, that a default functor is used to write the macro.
OK, I have to rescind my earlier statement:
Quote from: "itistoday"
I now agree with you, that the proper solution to this problem is to use a default function, while making sure to use naming conventions in a large project. The reason is that I realized that even when not using the default function and going the gensym route, you are still affecting the global namespace, just as you would be if you were to place the macro in its own context. Edit: The gensym approach might still be something to keep in mind if your macro lives in some large context and you don't want it polluting the global namespace, but even then, you'd be better off simply using good naming conventions as the context approach is much faster.
Question: in light of this, I think it would be very handy to have the def-static function brought back into the standard library, except could it be changed so that it can be used in the same way that one would use define-macro? Perhaps name it 'define-smacro' for "safe macro":
(let (temp (eval x))
(if temp temp (eval y))))
I mean, look at that! That would have the Scheme people jealous! :-)
There's also the more radical option of making *all* macros safe, by making the define-macro behavior create a new context by default, although that might be a bad idea since newLISP doesn't support nested contexts...
.aaahhh...I think i just missed this posting... Nice topic!
Does this work for you?
(let (
temp (eval (args 0))
temp1 (eval (args 1))
)
(if temp
temp
temp1
)
)
)
results:
(my-or nil temp) = 45
-----------
first arg
should be 1: 1
-----------
first arg
second arg
should be 2: 2
Here's a minimal version:
(or (eval (args 0)) (eval (args 1)))
)
m i c h a e l
Quote from: "DrDave"
results:
How did you get those results? With that code the second argument is always evaluated, therefore producing this result:
(my-or nil temp) = 45
-----------
first arg
second arg
should be 1: 1
-----------
first arg
second arg
should be 2: 2
Quote from: "m i c h a e l"
(or (eval (args 0)) (eval (args 1)))
)
That's cheating. :-p
I guess I didn't state it, but part of the idea/challenge is that you're supposed to define your own 'or', if you could use the built-in 'or' then it wouldn't really be a challenge..
Quote from: "itistoday"
How did you get those results? With that code the second argument is always evaluated, therefore producing this result:
(my-or nil temp) = 45
-----------
first arg
second arg
should be 1: 1
-----------
first arg
second arg
should be 2: 2
Ah, my bad. You're right. I didn't look carefully at the results.
Quote from: "itistoday"
Quote from: "m i c h a e l"
(or (eval (args 0)) (eval (args 1)))
)
That's cheating. :-p
I guess I didn't state it, but part of the idea/challenge is that
Inheritance is a powerful (and in 2009), a fundamental programming concept... For OOP, or for LISP, which at it's core "invented" inheritance by it's "functional" programming language nature... Every new function "inherits" code from existing functions... "or" is the base function... "my-or" is the derived functionality that reuses the "or" code and extends it's functionality... Using the "if" function is no more, or no less a valid idiom than using the "or" function...
Michael's version is clear, concise, and I can understand at a glance what it is doing... It is "pure"... It is "beautiful"... The only potential fault with such "beauty" is that it must also work... In real world programming, functionality must always trump beautiful form... When both functional form and beauty come together, it is rare, and profound...
"OR" is this to be an obfuscated newLISP contest? ;-)
-- xytroxon
Not an obfuscated contest, no, just a challenge to see if you can figure out a way to do it. While working on this problem myself I came up with what I thought was a clever little solution and I thought it would make for a fun topic/challenge. :-)
(define-macro (my-or)
(not (and (eval (args 0)) (eval (args 1))))
)
but probably thats a sidekick ;-)
Quote from: "newdep"
Yeah... but in addition that doesn't produce the correct results:
(my-or nil temp) = true
-----------
first arg
second arg
should be 1: nil
-----------
first arg
should be 2: true
yeah Im still looking into it because I want small code result ;-)
So i like micheal solution but im aming for a more scheme lookalike..
I cant get closer then this to scheme..
(define-macro (my-or)
(let (temp
(unless (eval (args 0))
(eval (args 1))))
temp))
(my-or nil temp) = 45
-----------
first arg
should be 1: 1
-----------
first arg
second arg
should be 2: 2
At least its short (shorter then the if version)
returns the values correctly and doesnt drop variables..
But then again it uses (args) and it uses 'unless (which is an 'or).
PS: Dropin another quizzzzzzzzzz ;-)
Quote from: "newdep"
(not (and (eval (args 0)) (eval (args 1))))
)
but probably thats a sidekick ;-)
LOL!!!
I love these code challenges :-)
But the quest should always be for the shortest path...
In other LISPs "eval" can be a speed hog (major slow down)...
-- xytroxon
Quote from: "newdep"
(define-macro (my-or)
(let (temp
(unless (eval (args 0))
(eval (args 1))))
temp))
(my-or nil temp) = 45
-----------
first arg
should be 1: 1
-----------
first arg
second arg
should be 2: 2
At least its short (shorter then the if version)
returns the values correctly and doesnt drop variables..
But then again it uses (args) and it uses 'unless (which is an 'or).
PS: Dropin another quizzzzzzzzzz ;-)
It is similar in principle to the one I came up with, but actually yours is better because it is much faster. In fact it's almost as fast as placing the macro in its own namespace. Send me a PM with your full name and email if you'd like a license to Espionage.
Here is the one I came up with, as you can see the principle of avoiding the evaluation is sortof the same, whereas you do it with 'unless', I did it with the exception mechanism. ](define-macro (my-or)
(catch
(begin
(let (temp (eval (args 0)))
(if temp (throw temp))
)
(throw (eval (args 1)))
)
)
)[/code]
Yes, the 'unless' and 'or' are similar, but the 'or' solution as was presented was equivalent to doing this (and therefore not an acceptable solution):
I was mainly looking for something similar to what I had, which I feel your solution was, as it used the same principle of avoiding the extra evaluation that was in DrDave's code while avoiding variable capture as well.
I hope though that nobody actually writes their macros this way, as to an outsider it would be completely non-obvious why the extra hoop-jumping was taking place.
This just demonstrates the need for a
Aha...I win finaly something here ;-)
Thanks for the price.. I dont have a Mac with OSX only
a MacClassic and an IMac.. But If you dont mind I would Like to store
you kind generosity for the "yearly newlisp Contest".. Where a winner
then also could get your nice software package.. (If thats oke with you, it would be a nice addon for the price shelf...)
Coming back to the Scheme example, I think its somehow unfair too what
Scheme does with GenSym. So personaly I prefer the newlisp (arg) way
of solving this. Where GenSym is a nice workaround its also not transparent for the function..
..I liked the quizzzz.. nice brain trainers these are...
Quote from: "newdep"
you kind generosity for the "yearly newlisp Contest".. Where a winner
then also could get your nice software package.. (If thats oke with you, it would be a nice addon for the price shelf...)
Sure! :-)
(define-macro (my-or)
(
(lambda ()
(if (args 0)
(args 0)
(eval (args 1))
)
)
(eval (args 0))
(args 1)
)
)
This solution uses a lambda to store the eval'ed first arg to the macro and defer evaluation of the second. It's not as fast as the solution presented by lutz, which has an execution time about 65% of mine, probably because mine entails an extra function call. But it does, at least, avoid the need for creating a new context just for one simple macro. This was a fun and surprisingly challenging problem. Thanks, itistoday!
So you used nested function call to transfer arguments in nested function so you do not have to use any variable at all. As (arg n) are always local, accidental overshadowing is impossible. Very interesting idea.
Welcome to the forum.
Safe OR Eval
or maybe
Pronounced "safari".
In honor of the safe journey through the newLISP namespace jungle ;)
-- xytroxon
Quote from: "Kazimir Majorinc"
Welcome to the forum.
Thanks! While this particular problem is unlikely to arise in any real-life work (after all, we have 'or), I think the basic underlying idea is potentially quite useful.
Nice full functional solution cgs1019!
Here a rewrite easier to understand for beginners:
(let ( func (lambda () (if (args 0) (args 0) (eval (args 1)) )))
(func (eval (args 0)) (args 1))
)
)
cgs1019 passes the user arguments to an inner function, with only the first argument evaluated. The inner function then decides if the second must be evaluated too. cgs1019 just uses the anonymous version of func. This is like doing:
ps: apropos "namespace jungle ;)": basic usage patterns of namespaces in newLISP are quite simple, like partitioning code into modules. Beyond that it gets more complicated and requires some manual studying to understand that namespaces in newLISP are not dynamic closures like in Scheme, but can be just as (and more) powerful when using them in their own way ;-)
Quote from: "cgs1019"
Your solution is beautiful Chris, I'd give you an Espionage license for it if I didn't know that you already had one. :-)
( (lambda (x) (+ x x)) 1) => 2
...efficient thinking is not always efficient for the thinking...
I forget about these powerfull solutions too often..
Nice nice..!