--
Curt Welch                                            http://CurtWelch.Com/
c...@kcwc.com                                        http://NewsReader.Com/

-- I use @x to store the program
set @x = '''-- Since Comp.ai.philosophy is still dead
-- I thought I would introduce a more random
-- version.  Unfortunately, while I thought I
-- owned a developers edition of sql server
-- my home version does not work so this
-- code is untested.
declare @x varchar(max);
declare @x2 varchar(max);
declare @i int;
declare @q varchar(1);
declare @l int;
declare @r int;

-- I use @x to store the program
set @x = ;

-- @q is the quote character
set @q = LEFT(@x,1);

-- the second caracter is the pipe character
-- @i is set to the second instance of the pipe caracter
set @i = CHARINDEX(substring(@x,2,1),@x,3);

-- slice and dice
set @l = round(rand()*(len(@x)-2),0);
set @r = round(rand()*(len(@x)-2),0);
set @x2 = substring(@x,3,@l) + @q + @q + @x + @q + right(@x,@r);

-- does the new string execute
begin try
  exec(left(@x2,len(@x)-10)
end try
begin catch
  set @x2 = substring(@x,3,@i-3) + @q + @q + @x + @q + right(@x,len(@x)-@i);
end catch;

-- now print the random string if it executes
-- otherwise print the known string;
print @x2;';

-- @q is the quote character
set @q = LEFT(@x,1);

-- the second caracter is the pipe character
-- @i is set to the second instance of the pipe caracter
set @i = CHARINDEX(substring(@x,2,1),@x,3);

-- slice and dice
set @l = round(rand()*(len(@x)-2),0);
set @r = round(rand()*(len(@x)-2),0);
set @x2 = substring(@x,3,@l) + @q + @q + @x + @q + right(@x,@r);

-- does the new string execute
begin try
  exec(left(@x2,len(@x)-10)
end try
begin catch
  set @x2 = substring(@x,3,@i-3) + @q + @q + @x + @q + right(@x,len(@x)-@i);
end catch;

-- now print the random string if it executes
-- otherwise print the known string;
print @x2;

5  DEBATE  GOTO 5

--
Curt Welch                                            http://CurtWelch.Com/
c...@kcwc.com                                        http://NewsReader.Com/

My program is recursive and therefore more beautiful. I must find a way
to put a lambda in it.

declare @q varchar(1);
declare @p varchar(1);
declare @x varchar(max);
set @q='''';
set @p='|';
set @x='declare @q varchar(1);
declare @p varchar(1);
declare @x varchar(max);
set @q='''''''';
set @p=''|'';
set @x=''||'';
print REPLACE(@x,@p+@p,REPLACE(@x,@q,@q+@q));';
print REPLACE(@x,@p+@p,REPLACE(@x,@q,@q+@q));

I got a somewhat random version to work but it broke
on the fourth recursion.  I needed to make sure @q
was set or the routine failed.  This routine would
be quite brittle.  It turns out exec() doesn't allow
a function, at least on the version I'm running.
The easiest way to handle the try/catch was like
this except first setting @x2 to the randomized
version rather than the non-random version.

set @pr = '
print @x2+@pr;';
set @x2=REPLACE(@x,@p+@p,REPLACE(@x,@q,@q+@q));
begin try
  exec(@x2);
  print '--new routine';
end try
begin catch
  print '--old routine';
  set @x2=REPLACE(@x,@p+@p,REPLACE(@x,@q,@q+@q));
end catch
print @x2+@pr;

a less brittle routine would need to do some basic
checking for health such as replacing duplicate declares
with new names and making sure quotes were paired.

Maybe sql madlibs would be easier.

  create table #a {child varchar(max)};
  ...
  insert into #a
  select @x3;

while

  select @x3 into #a;

would be shorter, it would make hopeful monsters less likely
to be able to produce multiple children.  Letting the routine
create the temp table would also allow for the possibility
of creating two unique tables.  It seems unlikely the routine
would ever write to multiple unique tables who knows.

I think the simplest selection algorithm would be number
of unique children within a certain number of offspring
with population control starting at some point.  

In my first random reproducer the fourth generation produced
an empty string.  During development I found several cases
where it duplicated some of the set commands.  Most of the
cases just appended to comment lines.  I specifically included
lots of comment lines so this would be more probable.  I
think I need to have the routine inject new characters so
the pool isn't limited to the one's in the original routine.

It may take awhile for me to get my hands on a developer's
edition of sqlserver.  You won't hear again until I get
some results, positive or negative.  
Some milestones would be:
  surviving generations 10, 100, 1000, etc.  
  producing multiple offspring in a single run.
  creating multiple tables.
  creating a permanent table.

While I could make the production of multiple offsring
more likely by adding lots of comments just before and
after the insert statement I don't think my originator
will be that generous, at least to start.