Re: Recursion-Loop Macro?

[Lindsay John Lawrence]

There is lots of discussion on the mail archive on this topic over the
years.
http://www.mail-archive.com/picolisp@software-lab.de/msg03860.html
There are others. Try searching tail recursion, macros, fexpr.

The mail archive has been an excellent source of information on picolisp.

I haven't found significant advantage to having tail recursion. Either in
picolisp, or other languages I generally work with.
Not recursing saves 'stack' but by the time that is an issue I am looking
for a different algorithm.

I am discovering picolisp has some great general purpose functions for just
that.

Here's an example I worked out recursively last week (set intersection),
then tried to do the same thing without recursion, then with (idx).
(randl, concurret, concurreti, concurretix are defined at the end of this
message)

# Make two lists of random numbers with random range large enough for small
collision set
: (nil (setq M (randl 1 999) N (randl 1 999)))
-> NIL

# Recurse
: (bench (sort (uniq (concurret M N
4.103 sec
-> (1144218 1809896 2048032 2258715 2819894 5571078 9413422 9726100 9936515)

# loop
: (bench (sort (uniq (concurreti M N
2.566 sec
-> (1144218 1809896 2048032 2258715 2819894 5571078 9413422 9726100 9936515)

# idx
: (bench (sort (uniq (concurretix M N
0.018 sec
-> (1144218 1809896 2048032 2258715 2819894 5571078 9413422 9726100 9936515)

Idx 'wins'.


/Lindsay


# ==
# Set intersection: recursive solution.
(de membro (X Y)
   (cond
  ((not Y) NIL)
  ((= X (car Y)) T)
  (T (membro X (cdr Y))) ) )

(de concurret (X Y)
   (cond
  ((not X) NIL)
  ((membro (car X) Y)
 (cons (car X) (concurret (cdr X) Y)) )
  (T (concurret (cdr X) Y)) ) )

# Set intersection: no recursion
(de membri (X Y)
   (until (or (not Y) (= X (car Y)))
  (setq Y (cdr Y)) )
   (= X (car Y)) )

(de concurreti (X Y)
   (let (V NIL)
  (until (not X)
 (when (membri (car X) Y)
(setq V (cons (car X) V)) )
 (setq X (cdr X)) )
  V ) )
# Set intersection: Filter X into (idx); filter Y on that.
(de concurretix (X Y)
   (let (R NIL)
  (filter
 '((V)
(not (idx 'R (cons (hash V) V) T)) )
 X )
  (uniq
 (filter
'((V) (idx 'R (cons (hash V) V)))
Y ) ) ) )

# Generate list of random numbers in given range
(de randl (L R) (make (do L (link (rand 0 R)

Re: Recursion-Loop Macro?

[Joh-Tob Schäg]

I will think about doing a f-expr for that.
Am 12.02.2017 11:50 schrieb "Mike Pechkin" :

> write yourself recursion call and you will see you don't need macro.
>
>
> On Sun, Feb 12, 2017 at 5:40 AM, Christopher Howard <
> christopher.how...@qlfiles.net> wrote:
>
>> Hi list. I got the impression from some of the documentation that we
>> aren't planning to implement TCO in picoLisp, and that the preferred
>> approach is to just use a loop. I don't want to argue against that, but
>> I was just wondering if perhaps somebody had a macro that made a loop
>> look like a recursive tail call?
>>
>> --
>> https://qlfiles.net
>> --
>> UNSUBSCRIBE: mailto:picolisp@software-lab.de?subject=Unsubscribe
>>
>
>

Re: Recursion-Loop Macro?

[Mike Pechkin]

write yourself recursion call and you will see you don't need macro.


On Sun, Feb 12, 2017 at 5:40 AM, Christopher Howard <
christopher.how...@qlfiles.net> wrote:

> Hi list. I got the impression from some of the documentation that we
> aren't planning to implement TCO in picoLisp, and that the preferred
> approach is to just use a loop. I don't want to argue against that, but
> I was just wondering if perhaps somebody had a macro that made a loop
> look like a recursive tail call?
>
> --
> https://qlfiles.net
> --
> UNSUBSCRIBE: mailto:picolisp@software-lab.de?subject=Unsubscribe
>

Recursion-Loop Macro?

[Christopher Howard]

Hi list. I got the impression from some of the documentation that we
aren't planning to implement TCO in picoLisp, and that the preferred
approach is to just use a loop. I don't want to argue against that, but
I was just wondering if perhaps somebody had a macro that made a loop
look like a recursive tail call?

-- 
https://qlfiles.net
-- 
UNSUBSCRIBE: mailto:picolisp@software-lab.de?subject=Unsubscribe