[Jprogramming] crash

[David Lambert]

The crash took place while developing part 2 sea monster jigsaw puzzle 
code.  I didn't think there was any reasonable way to reconstruct the 
full j session (emacs probably stored it somewhere) and I thought it 
rather lucky to recreate the problem at all.  And by the way, rather 
than "access violation" the original message contained words something like

free invalid pointer


Conway's multidimensional life was easy with moving cuts of framed 
array.  For decisions I used a two level agenda, one based on the sum, 
the other on center value.


I'm stuck on part 2 of the monster message, regular expression tree.  
First I tried to limit the depth, after all there is a longest message 
length.  That was no help at all.  Next I've tried replacing the rules

8: 42 | 42 8    becomes

8: 42 | 42 42 | 42 42 42 |  ...

which doesn't work, maybe because of some sort of logic error for which 
I was lucky to pass part 1 yet fails in part 2.  Because of the 
symptoms, I'm pretty sure I can get the answer using about 25 rule 
variations UGH, but I won't know why.


|Date: Sun, 20 Dec 2020 18:08:03 +0000
|From: "'Mike Day' via Programming" <programming at jsoftware.com>
|To: programming at jsoftware.com
|Subject: Re: [Jprogramming] crash
|Message-ID: <2A3F7067-4C0E-492F-9551-9EE92ADB9EBF at tiscali.co.uk>
|Content-Type: text/plain;    charset=utf-8
|
|Your method looks classy!  No idea what’s wrong.
|
|FWIW,  I took the initial grid, adding enough empty planes so that no 
edge planes could be occupied in the required number of generations, set 
up a 26 * 3 array of offsets,  111,110,101, ... , -101, -110, -111,  
(omitting spaces) and then ravelled the data and changed the offsets to 
a vector using shape #: offsets. It was easy then to get the number of 
live neighbours for each point. Crude,  but fast enough.
|
|It was tempting to dig up Life in the J Wiki.
|
|Cheers,
|
|Mike