Thanks for running my puzzle solver “444”. 8% success stinks but every one
timed out. A typical puzzle which the results state took 39 seconds
grinding away until a timeout, solves just fine interactively in 0.2 seconds.
If someone can head me in the direction of the problem I would appreciate it,
especially if it requires changing my program.
Interactively, using my home grown driver I get around 80% success ignoring
all requirements other than getting a stable sequence so I am motivated in
finding what the issue is with the timeout so I can continue to move forward.
So I did put all code that required array processing variables into functions
so the newest program calls a series of functions. Lets’ see if that works.
Well… Thanks for running my newest script. 44% is better than 8% but still
something is not right. All I did was encapsulate blocks of code into functions.
But still all the failures are timeouts that solve interactively within a few seconds, not the 10 -30 seconds that the batch processor states.
I will ecapsulate EVERYTHING in functions is all I can thinks of and will submit again.
Thanks for running script.
Looks better, now I can see what is going on.
Lesson: put everything into functions.
Noticed you have shiftables in the list, Brourd like puzzles will be near impossible,
the shiftables I would not even attempt, might want to take them out of there.
Are the results accessible by script?
So far I found these problems I can fix:
1 flip of middle gc in 3 stack to stabilize puzzle
sav_seqloop error
bad boost error
add 1 gc in 4 stack with loop with adjacent stacks req. to stabilize puzzle
flip adjacent gc pair in loop next to error stack on other side of bulge req to stabilize puzzle
these improvements should keep me off the streets and out of trouble for at least a week I would think.
Latest success rate 18%. Timeouts again.
Comparing runs:
To get to 70% from 44% all I did was put all code with indexes into functions
and put “return;” at the end of each function. I changed very little to drop the
score to 11% then 18% so I am at a loss.
How could this fail: ( … ) puzzle id 3956246. but it did.
structures run Interactively or 20 at a time run fine.
So I will keep poking around with my code.
If I do a :
hxxp://eterna.cmu.edu/get/?type=puzzles"
I get around 5400 puzzle. The total is around 7400.
puzzle 3712900 is missing from list but if I do:
hxxp://eterna.cmu.edu/get/?type=puzzle&nid=3712900
I find the puzzle.
All the difficult puzzles seem to be missing from the 5400 list.
I ran 3000 puzzles from the easy list interactively and batted 90% success ,
JL’s script reduced my error list 20% so his script should bat around 93%.
Putting all the difficult puzzles into the run would decrease the success rate to
70% which is what we are seeing from the batch run.
So how do I access the missing puzzles?
And is there a list of Ajaxmanager fields I am able to access?