Conference noted::hackers_v1

$ type 'f$env("PROCEDURE")

	[RE .1]:  You beat me to it!
		<_Jym_>

You both ignored my statement in .0, saying:

	Of course, no reading of files allowed from within the procedure,
	lest the problem be trivial.

Then again, perhaps you're responses weren't claiming to be "better 
solutions", but rather they answer the question "what does this .COM file
do ?".  If so, you get credit.

By the way, from a theoretical computer language point of view, what 
sorts of languages allow such a program to be written and what do not ?

re: .1
------
and then there is 
	$ TYPE/OUT=TT TT
or even
	$ COPY TT TT

re: .3
------
I've seen self reproducing programs in Basic, Fortran and C.   I don't have it
anymore but the Basic version used the same idea as the DCL procedure in .0.  
Here are a few short ones in C:

* this one works almost everywhere:
    char*f="char*f=%c%s%c;main(){printf(f,34,f,34);}";main(){printf(f,34,f,34);}

* cheating from UNIX
    main(){system("cat self.c");}

- Paul

Re .3 last paragraph: I'd guess that any language that incorporates the
notion of string datatype and output statements can do this.  Then again,
I shudder at the thought of trying it in COBOL.  Interesting question:
which language allows the shortest one?  (APL freaks?)

	Paul

re .3
	Just answering the question.

	From a theoretical point of view, I believe that only the following
	features are needed:

	    String variables
	    String concatenation (if concatenation is only allowed on output,
		this should suffice)
	    Iteration
	    Ability to create/write a 'compilable' source

my 2 cents worth:

    I think that the requirement that "strings of characters" be
    manipulable is only true if we restrict "programs" to be things
    constructed of strings of characters.

    If, for instance, we allowed program images to be thought of
    as "programs", then the ability to work with bits and bytes
    could replace the ability to work with characters and strings.

    Generically:

    	It must be able to represent and manipulate data in the same
    	form as the program itself.

    	It must be able to execute a loop with a controlled exit.

    	It must be able to transcribe data externally to itself.

 
    Has anyone ever seen a similar program for a general Turing machine
    that not only replicates itself, but also invokes the replica?

    Now if you come up with a program that not only replicates itself,
    but also performs a useful function, you are beginning to have
    something like DNA.

    Dave

	$loop: if .not. f$verify(1) then goto loop

writes itself out...


/Bevin

Sorry, that's cheating, because f$verify is asking an external source
to generate your output for you, which is as bad as reading the source
file !

By the way, whoever sent in the on-line "C" program, congrads !  That
one is correct.  (I used DIFFERENCE command in DCL to verify it, and to
verify my original solution.  By the way, back on TOPS20 I wrote
a BLISS program that printed itself out.  Finally, when I thought I
had succeeded, I used the TOPS20 equivalent of DIFFERENCES to verify
my correctness.  It did.  THEN, just to be REALLY sure, I used
DIRECTORY,@@CHECKSUM, and it showed differences !!  So, I ended up
having to fix my program some more, which was covertly putting in
some extra nulls, which the DIFFERENCE utility on TOPS20 was missing.

So, perhaps my solution or the C solution are still not correct ?
Is there some equivalent to DIRECTORY,@@CHECKSUM on vms so we can
be more sure ?)

How about ...

A C program that writes a Pascal program that writes the C program? 

You get the idea.

Do a DUMP then a DIFF on the results....

/Bevin

The existence of self-reproducing programs is a very general result of the
theory of recursive functions.  In fact, if T(n) is the n'th Turing machine
in some recursive enumeration, it is a corollary of the Recursion Theorem
that there is an e with the property that, for all x, T(e)(x) = e; that is,
machine e, on any input, simply prints e.  All you have left then questions
of representation - if I consider the number e to represent the program for
its machine - i.e., I think of T as an interpreter - then, indeed, I have
a self-reproducing automaton, without worrying about strings, characters,
or anything of that sort - they are just syntactic gloss.

It is also interesting to create chains of functions Fi such that Fi prints
Fi+1 until finally some Fn prints F0.  Or things like pairs of functions F and
G such that F(G) = F and G(F) = G.  There has been a lot of work about these
kinds of questions in looking at how life began.  There was an article about
this in Scientific American a couple of years back.  At first, there are
single self-reproducing molecules.  But it turns out that past a certain
point, pairs, then triplets, gain an advantage in reproducing faster in
the resulting distributions of "food" molecules.  Beyond triplets, at
the time, no one could compute what happens...

							-- Jerry

It might be interesting to try this "the hard way", although probably
NOT in dcl.

First, notice that an EXTREMELY hard way is to write a decompiler, for instance
the Smalltalk-80 decompiler.  A decompiler could be probably written
such that when modified to decompile itself rather than ask the user
WHAT to decompile, it manages to print itself out.

This solution, although EXTREMELY hard (in terms of effort needed to write
a decompiler from scratch), is interesting because it isn't as likely to
provoke the "so what?" response as my .COM solution does, since when you list
it, you don't see many quoted strings in it.

But we don't need to solve it the EXTREMELY hard way in order to have
a program that prints itself out.  We merely need to write the portions
of a decompiler that know how to decompile the features of the language
used within the program !  And we can limit our local variables to names 
like "t0, t1, t2 etc." to make it even easier.

Perhaps such a self-printing program could be written in Bliss or
C ?  Any takers ?

One of the previous responses alluded to functions whose output is different
from the original function, such that executing the output produces
the original function (or functions that do this in more than two steps).

Can anyone write a .COM file whose output is NOT the original, but
when the output is executed it produces the original ?

Re:  Self-printing programs

The process of writing a self-printing program is trivial in DSM
(Digital Standard MUMPS).  The Xecute command can force a program to
load and print (or modify) itself, and then return control back to
itself.  Most MUMPS editors are variants of this idea (programs which
modify other programs, or themselves). 

Herewith is an small example:

SELF	;RSC ; Self-printing program
	XECUTE "ZLOAD SELF ZPRINT"
	QUIT

For those of you unfamiliar with MUMPS, the quoted string after Xecute
tells the interpreter to load the program named 'SELF' and print it to
the current output device. The QUIT terminates the execution of the
program. 

Something a little more useful along this line is a print utility I
wrote a while back which will print any MUMPS routine (including
itself) to any device.

(Just about any MUMPS hacker you meet will have a program like this
lying around somewhere.  I've added comments to help those of you who
don't know MUMPS, the comment marker is ";".  Some special characters
are "!" = linefeed, "?" = tab, "'" = NOT, "@" = indirection.  All
commands and functions (WRITE, SET, IF, $PIECE, $TEXT, etc.) are
spelled in full here, in normal use they would be abbreviated to the
first letter (W, S, I, $P, $T), which makes typing programs a lot
easier!) 

UPRT	;R.S. CRAIG - Print selected routines
	WRITE !!,"Print Routines Utility",!!
	;  Get the output device (default = LPA0).
DEV	SET %DEF="LPA0:",%QTY=2 DO ^%IOS
	;  No device (%ZIOD) selected, stop routine.
	IF '$DATA(%ZIOD) SET %ZIOD="" GOTO EXIT
	;  Select right margin of 80 or 132, depending on device in %ZIOD.
	SET ROU="",RM=$SELECT(%ZIOD["LPA0":130,1:80)
	;  Call utility to select routine(s).
	WRITE ! DO ^%RSEL
	IF '$DATA(%UTILITY) WRITE !?2,"No routines selected" GOTO EXIT
COPY	WRITE !!,"How many copies?  1",$CHAR(8) READ COPY IF COPY="" SET COPY=1
	IF COPY="?" WRITE !?3,"Enter the number of copies to print of each program" GOTO COPY
	;  Set copy to 1 if less then/equal to zero.
	SET COPY=+COPY IF COPY'>0 SET COPY=1
	;  Create executable variables LINE and HDR using MUMPS indirection.
INIT	FOR VARIABLE="LINE","HDR" SET @VARIABLE=$PIECE($TEXT(@VARIABLE),";;",2)
	; Functions to get system date and time.
	SET DATE=$ZCALL(%CDATASC,+$HOROLOG,2),TIME=$ZCALL(%CTIMASC,$PIECE($HOROLOG,",",2),2)
	;  Format the time to civilized format.
	SET TIME=$PIECE(TIME,":",1,2)_" "_$PIECE(TIME," ",2)
	;  Function to get default VMS device and directory.
	SET SOURCE=$PIECE($PIECE($ZCALL(%PGMSHOW),",",1),"]",1)_"]"
	;  Some VT100 escape sequences for showing off.
	SET BOLD=$CHAR(27)_"[1m",NORMAL=$CHAR(27)_"[0m"
	WRITE !!,"Printing routines on ",%ZIOD,!!
	;  Loop through the array in alpha order, return routine names 
	;  in ROU, call PRINT subroutine from FOR loop, COUNT tells us
	;  how many times we did the loop.
	FOR COUNT=0:1 SET ROU=$ORDER(%UTILITY(ROU)) Q:ROU=""  DO PRINT
	;  Done with FOR loop, let us know how many we printed.
	USE 0 WRITE !!?3,COUNT," routine" WRITE:COUNT'=1 "s" WRITE " printed"
	IF COPY>1 WRITE " (",COPY," copies each)"
	;  All done, clean up and quit.
EXIT	CLOSE %ZIOD K I,DATE,TIME,%DTY,%IOD,%UTILITY,%DTY,%ZIOD
	K COPY,COUNT,CT,HDR,LINE,LN,PG,RM,ROU,SOURCE,BOLD,NORMAL
	WRITE ! QUIT
	;
PRINT	;  Write routine names to terminal (device 0).
	USE 0 WRITE:$X>70 ! WRITE ROU
	;  eXecute string to load routine, initialize PG (page count),
	;  redirect output to device and eXecute code in LINE.
	XECUTE "ZL @ROU FOR COPY=1:1:COPY SET PG=0 USE %ZIOD XECUTE LINE"
	;  Redirect output back to terminal, write some spaces,
	;  and quit subroutine.
	USE 0 WRITE ?$X+(10-$LENGTH(ROU)) Q
	;
	; Executable code
	;  This line eXecutes HDR, loops through program text, writes the
	;  line tag (1st $PIECE), and line code (2nd $PIECE), keeps track
	;  of the line count and does header again if > 62.
LINE	;;XECUTE HDR FOR I=1:1 SET LN=$TEXT(+I) QUIT:LN=""  WRITE !,$PIECE(LN," ",1),?8,$PIECE(LN," ",2,99) SET CT=CT+1+($X\80) IF CT>62 XECUTE HDR
	;  This line increments page count, writes a formfeed, default
	;  device and directory, date and time, page count and bumps up line
	;  count.
HDR	;;SET PG=PG+1 WRITE #,BOLD,ROU,?12,SOURCE,"  ",DATE," ",TIME,?RM-(7+$LENGTH(PG)),"Pg. ",PG,NORMAL,! SET CT=4

Its gotta be trivial to have a .COM file start of "$ I = 1", generate "$ I = 2"
and for this second form to generate "$ I = 1"...

/Bevin

re: .15

That's no fair.  You just loaded your source program and printed it 
out.  Could you write a DSM program that will display itself without 
using ZLOAD?  The code should be able to reproduce itself without 
access to the source.

tom p

My apologies.  I haven't looked at this file in a long time.  Here is a very
small DCL .COM file which prints itself. Have fun.

------------------------------Cut Here------------------------------------------
$ A[0,8] = 34
$ B = "$ A[0,8] = 34!/$ B = !AS!AS!AS!/$ C = F$FAO(B,A,B,A)!/$ WRITE SYS$OUTPUT C!/"
$ C = F$FAO(B,A,B,A)
$ WRITE SYS$OUTPUT C

It doesn't print itself out - the resulting file has one record instead
of three.

/Bevin

PS: I still think its really cute...

Hey Barton, I like that one.

I'll let the one-record business pass.

One small gripe though.  When I "@" it, I see an extra blank line that "TYPE"
doesn't display.  Can you fix this flaw ?

/Eric

A slightly shorter version of Paul Wong's C program in reply .4 is:

p="p=%c%s%c;main(){printf(p,34,p,34);}";main(){printf(p,34,p,34);}

This version is 66 characters versus Paul's, which is 76 characters.

	Mike

    What does this do. I found it in my LOGIN.COM after I SET H 0/LOG=MORE.LOG  
    
    
    
Z
[62"p F
>
>
[62;1"pP1;1;1;4{>
__?~????/??@B????;
???oOOOO/???N????;
OOOOOOOO/????????;
???}@@A?/???N????;
???~????/CGGF????;
???~????/???N????;
???}AAA?/???N????;
???~????/???FCCC?;
AAA}????/???N????;
???~????/CCCF????;
???{A@??/???N????;
???~????/???BCG??;
?@A{????/???N????;
???~????/?GCB????;
?_ON????/??@M????;
???NO_??/???M@???;
???OoOOO/?????@AC;
???OWSQP/????????;
@ACGO_??/??????@?;
???_OGC?/CA@?????;
OOOOOOo?/??????@?;
OOOOOO[?/????????;
@ACGo???/GCA@????;
?}AAA???/?BAAA???;
??WWW???/????????;
??AAA}??/??AAAB??;
GGG}GGG?/AAAAAAA?;
?OWSQOO?/?@@@@@@?;
?ggwkig?/?A@?????;
?OOQSWO?/?@@@@@@?;
???}@@A?/CGGF????;
???KK???/?BB??BB?;
wCCwCCC?/?@@?@@@?;
wCCwCCw?/?@@?@@??;
OOOSOOO?/???@????;
?oKAKo??/BAAAAAB?;
KScCcSK?/???@????;
Wcd~dcW?/??ABA???;
?@~@@B??/?ABA????;
OGGO__O?/????????;
GCCGOOG?/@@@@@@@?;
wCQQQCw?/?@AAA@??;
?CgOgC??/?@???@??;
?oKAKo??/BA???AB?;
gkigikg?/?@A?A@??;
?gggikg?/????A@??;
?gggggg?/?AAAAAA?;
A}AAA}A?/?B???B??;
MO_}_OM?/??ABA???;
?CIIIC??/????????;
AEiQAAA?/ABAAAAA?;
?SiiiO??/??AAA@??;
?KQQ}Q}?/????B?B?;
_??_WEA?/?@A@????;
wCAAACw?/AB???BA?;
?AQQQQA?/?AAAAAA?;
CAA{AAC?/???B????;
oGCCCCC?/?@AAAAA?;
CCCCCGo?/AAAAA@??;
wCAAACw?/B?????B?;
}?????}?/?@AAA@??;
?oKAKo??/B?????B?;
EW_?_WE?/??@A@???;
?GGGGGW?/????????;
_OGGoOG?/?@AA@@A?;
?_WSQQk?/E@AAAA@?;
?GGO__W?/GCA@?@M?;
?_O[QaC?/?@AAA@??;
?ogggG??/?@AAAA??;
_OOOoOg?/HEABAA@?;
GCGo?_O?/??KN@???;
GGoOGGo?/??B???N?;
???w????/???@AA@?;
oggggo??/@AAAA@??;
C{?_OG??/?B@@AA??;
?CCgO_??/AA@???B?;
?w????w?/E@AAAA@?;
CcW??_[?/ABA@@???;
CCcSGGo?/??@AAA@?;
?GwGGwG?/?A@??B??;
_??w??_?/?@ANA@??;
??_OGGo?/KB??@@??;
_OGGWgG?/@AAAA@??;
OGGwGGC?/???BA???;
????????/????????;
???}PPA?/EGGF????;
oG?_?Go?/@AA@AA@?;
OgicCC??/?BAIIE??;
Cw????w?/??@AA@??;
cQIEEAA?/?@@DDB??;
OwSOOOO?/??@?????;
?GC}CG??/???B????;
OOOOSwO?/????@???;
?_?}?_??/??@B@???;
\P1;1|23/1B;24/08;25/0A\
<
    

Does your account get enabled for WPS when you log in? If so, and you
are on a VT2%%, it's probably loading up your function keys...

Al

    That MUST be it.  When I log in under a VT100, it doesn't do it.
     Thanks for the quick reply as not to keepp me in suspense.
    -=*>Jason<*=-

re: .23	> it's probably loading up your function keys...

	Some more details:

	these escape sequences first identify your terminal type ( <esc>Z ),
	sets   it   to   VT200   mode,   7  bit  controls  ( <ESC>[62;1"p ),
	requests that it sends only 7-bit codes ( <esc><space>F ), downline-
	loads  a  character set ( <esc>P1;1;1;4{> and all the sixel patterns
	that  follow ), then defines the function keys F11,F12 and F13 to be
	<ESC>, <BS> and <LF> , respectively .

	The character set is quite funny, mostly composed of various math
	and other signs (sigma, omega, and a lot whose name and existence
	i never heard of before...) - might be greek alphabet , any greek
	out there care to check it ?

	You can activate that character set after you've loaded it (by typing
	the note .22) by sending <ESC>(> to your terminal.

		Hope that helps...

			Marc.

Judging from some of the symbols used, it's a math character set of some sort.
Or maybe APL.

-Mike

Most likely to be the Technical Character Set.

    Most definitely the TCS. WPS+ loads the TCS into a terminal with soft
    fonts. 
    
    	Jon