Page 10 of 16
Re: Hacking apgsearch
Posted: August 15th, 2017, 4:10 pm
by muzik
There's definitely some object separation issues father down the page, but I guess that was successful.
Why, exactly, is it so slow to search? I'd assume it would be down to the wickstretcher that forms from a single domino, but why does that cause so much lag if it's simple linear growth?
Re: Hacking apgsearch
Posted: August 15th, 2017, 6:28 pm
by Apple Bottom
Speaking of B2a3r4i/S, there's apparently another bug causing apgsearch 0.54+0.31i to error out when encountering oversized spaceships:
- no_ovqs.png (17.52 KiB) Viewed 34703 times
Re: Hacking apgsearch
Posted: August 15th, 2017, 10:05 pm
by praosylen
Apple Bottom wrote:Speaking of B2a3r4i/S, there's apparently another bug causing apgsearch 0.54+0.31i to error out when encountering oversized spaceships:
Sorry, everyone — replace
with
and it should work.
Re: Hacking apgsearch
Posted: August 17th, 2017, 4:49 pm
by Apple Bottom
Apple Bottom wrote:Some kind of bug appears to be left in 0.54+0.31i; soup-searching B2e3ai4a5j6n/S2-n3ijr4it finds, among other things, an empty object, supposedly a period-1001 oscillator:
xp1001_0.
Bump--
This also affects
B35y/S23, BTW -- and both this rule and the previous one have other objects that got classified as p1001 oscillators but aren't actually.
Re: Hacking apgsearch
Posted: August 18th, 2017, 2:15 pm
by drc
This is for catagolue, but can we have the hauls be colourized? e.g. xs=yellow, xp=blue, xq=green, and other=red.
Re: Hacking apgsearch
Posted: August 18th, 2017, 8:06 pm
by Saka
drc wrote:This is for catagolue, but can we have the hauls be colourized? e.g. xs=yellow, xp=blue, xq=green, and other=red.
I think we should also be able to customize our profile a bit, and it should also display hauls in other rules for "most recent hauls", not just B3/S23. Maybe we could have small avatars and say, custom username colors when we earn more than 2 badges? What about badges for people that seldom search Life but search other rules instead?
Re: Hacking apgsearch
Posted: August 21st, 2017, 6:05 am
by Apple Bottom
drc wrote:This is for catagolue, but can we have the hauls be colourized? e.g. xs=yellow, xp=blue, xq=green, and other=red.
You can probably already do this with client-side CSS (or alternatively Greasemonkey if you need a bit more power).
Saka wrote:[...] and it should also display hauls in other rules for "most recent hauls", not just B3/S23.
Definitely that!
Re: Hacking apgsearch
Posted: August 21st, 2017, 12:15 pm
by muzik
How about some more random pseudosymmetries? I wouldn't mind 32x32 (which seems to have been investigate but I can't get to work), 64x64, 1x512 and 1x1024.
Also, how about something like this? Searching one of those adjustable orthogonal speeds rules with a symmetry like this would be kind of interesting.
Code: Select all
x = 255, y = 3, rule = B3/S23
o2b3obo2b2obo6bo2b2obobo2bo3bo2bob4obobo2b2o5bo3b2ob3o4bo3b2ob3o3bob5o
bo2bo2bob4o5b4ob2o2b3ob3o3bob2o2b2obobob3ob3ob6obo2bo2bob7o2b3obo4bobo
8b2o2bo3b2o3b5obo2bob3ob2o2b3obo4bo3bo2b4ob2o2b3obo$2b2o3b2ob2o2bobob
2o3b3obo2b5o2bobo2bo3bo2bobob3o2b2obo4bo3bob4ob3o2bo2bo5bo3b2obo3bobob
2ob3obob2o4b2obo6b3ob3o2b2o2b2obobobob3o2b2o2bo4b2o3b2ob2o4bobo4b4o2bo
b2o2b2o2b2obobo4bob3o3bo2bobo5b2ob3o4bo3b2o4bo$o2b3obo2b2obo6bo2b2obob
o2bo3bo2bob4obobo2b2o5bo3b2ob3o4bo3b2ob3o3bob5obo2bo2bob4o5b4ob2o2b3ob
3o3bob2o2b2obobob3ob3ob6obo2bo2bob7o2b3obo4bobo8b2o2bo3b2o3b5obo2bob3o
b2o2b3obo4bo3bo2b4ob2o2b3obo!
Re: Hacking apgsearch
Posted: August 21st, 2017, 12:38 pm
by Apple Bottom
muzik wrote:How about some more random pseudosymmetries? I wouldn't mind 32x32 (which seems to have been investigate DHUT I can't get to work), 64x64, 1x512 and 1x1024.
Also, how about something like this? Searching one of those adjustable orthogonal speeds rules with a symmetry like this would be kind of interesting.
Why not just implement it yourself? It's fairly easy to do in apgsearch 1.x (or Aidan's hacked version). Be sure to submit results to a _Test symmetry first, and share your efforts.
Take a look at how others have implemented additional new (pseudo-)symmetries before, and if you have further (specific) questions or if you get stuck, post here again.
Re: Hacking apgsearch
Posted: August 23rd, 2017, 4:38 am
by Apple Bottom
Apologies for double-posting, but here's another bug--
Wickstretchers in B3-c4i/S1c2-ck34a sometimes get
misclassified as oversized still lifes in 0.54+0.3i.
Re: Hacking apgsearch
Posted: August 23rd, 2017, 12:17 pm
by wwei23
This is misclassified as yl120, zz_LINEAR, zz_REPLICATOR, and zz_EXPLOSIVE. When will it realize that it's an x+log(x)?
Code: Select all
x = 3, y = 6, rule = B3ai6i/S2ai3aei5i
bo$bo$bo2$bo$3o!
Re: Hacking apgsearch
Posted: August 23rd, 2017, 1:19 pm
by muzik
Apple Bottom wrote:muzik wrote:How about some more random pseudosymmetries? I wouldn't mind 32x32 (which seems to have been investigated but I can't get to work), 64x64, 1x512 and 1x1024.
Also, how about something like this? Searching one of those adjustable orthogonal speeds rules with a symmetry like this would be kind of interesting.
Why not just implement it yourself?
...I knew I shouldn't have dropped computing science.
Re: Hacking apgsearch
Posted: August 23rd, 2017, 7:45 pm
by toroidalet
wwei23 wrote:When will it realize that it's an x+log(x)?
Never. Also it's not x+log(x). It's the addition of a sawtooth function and x.
Re: Hacking apgsearch
Posted: August 24th, 2017, 1:36 pm
by muzik
Could apgsearch be modified as to support certain explosive rules, by ignoring any patterns that still show any activity after a set number of generations?
Re: Hacking apgsearch
Posted: August 24th, 2017, 8:07 pm
by toroidalet
muzik wrote:Could apgsearch be modified as to support certain explosive rules, by ignoring any patterns that still show any activity after a set number of generations?
That would be problematic in rules that are very nearly explosive (patterns in rules like B2c3ajk46-ak/S1c237e8
can last more than 900000 generations). It would also skew the selection.
But
how would this skew the selection? Let's find out! (Yay science)
In B37/S23 (Drylife) there exists a highly dirty puffer:
Code: Select all
x = 11, y = 3, rule = B37/S23
b3o3b3o$o2bo3bo2bo$4bobo!
This puffer's trail doesn't explode, but the base does (this can be stabilized with some junk, but that's not important.). Upon excluding patterns which are somehow determined to be explosive (This hypothetical method is 100% correct, never labelling an explosive pattern as stable or a stable pattern as explosive (Which is pretty much impossible to achieve)) we arrive at the conclusion that this puffer, yl160_[rest of apgcode] (I don't know how to label this) is much rarer than it actually is, because a soup that produces it is less likely to produce ash which stabilizes the exploding back than it is for a soup to produce this puffer in a way that makes the junk at the base explode (even discounting the other part of the soup already exploding). And the statistical algorithm reports incorrect values for any rule containing long-lasting patterns.
EDIT: typos
Re: Hacking apgsearch
Posted: August 25th, 2017, 3:21 am
by muzik
Then maybe said search should be dedicated to a certain pseudo symmetry,
Re: Hacking apgsearch
Posted: August 28th, 2017, 2:38 am
by muzik
3 pathologicals:
Code: Select all
x = 60, y = 31, rule = B2c3ae4ai5iy6c/S2-kn3ijry4cit
4bo16b4obob2obob3o8bobob2o3b5o$bo2bo2bo15bo4bob2o4bo7bobob4ob2ob2o$3b
2o2bob4ob2o9b4ob2o2bo10b2o4bo3bobobo$2bo2b2o2b2ob3o6b2obob3obobo2b2o7b
2o4bob2o3bobo$4bobobo2b2o2bo8bobobo3bob2o8b5ob3o3b2obo$obobobob4obo7b
3obo4bobob2o9bobob2o2bo2bo$ob2o2b3obobobo7b5o2b2o2bo2bo10bo2b2o3b5o$2o
b2obob5obo8bo3bo3bobob2o7b3o2b2ob4ob3o$2bo2bobobob2o2bo5bo4b3ob3o2b2o
7bo2b2obobobo4bo$5obo2bo3b2o7b3obo2bo2b2ob2o11b3obo4b3o$obob3ob5o9bo2b
4o5bobo7b2o3bobobo2bo2bo$3b2obobo2bo10bo2b2o2b2o3b3o9b2obo2bo2b2ob2o$b
3o2b2obo2bo2bo5b2ob3o3b3o2b2o10b3o4b4obo$4o2bobo4b2o8bo2b3obo4b2o7bobo
3b2o6b2o$bob6o2b3obo5bo3bob2ob4o10b2obo2bobobob3o$o3bo2bo6bo6b2obo4b2o
2b3o8bo4bo3b3o2bo$bob6o2b3obo5bo3bob2ob4o10b2obo2bobobob3o$4o2bobo4b2o
8bo2b3obo4b2o7bobo3b2o6b2o$b3o2b2obo2bo2bo5b2ob3o3b3o2b2o10b3o4b4obo$
3b2obobo2bo10bo2b2o2b2o3b3o9b2obo2bo2b2ob2o$obob3ob5o9bo2b4o5bobo7b2o
3bobobo2bo2bo$5obo2bo3b2o7b3obo2bo2b2ob2o11b3obo4b3o$2bo2bobobob2o2bo
5bo4b3ob3o2b2o7bo2b2obobobo4bo$2ob2obob5obo8bo3bo3bobob2o7b3o2b2ob4ob
3o$ob2o2b3obobobo7b5o2b2o2bo2bo10bo2b2o3b5o$obobobob4obo7b3obo4bobob2o
9bobob2o2bo2bo$4bobobo2b2o2bo8bobobo3bob2o8b5ob3o3b2obo$2bo2b2o2b2ob3o
6b2obob3obobo2b2o7b2o4bob2o3bobo$3b2o2bob4ob2o9b4ob2o2bo10b2o4bo3bobob
o$bo2bo2bo15bo4bob2o4bo7bobob4ob2ob2o$4bo16b4obob2obob3o8bobob2o3b5o!
I think I know what's happening on the right two, but don't know how to explain it.
Re: Hacking apgsearch
Posted: August 28th, 2017, 7:48 am
by Apple Bottom
This is a question for wwei23, primarily -- you've searched the "25p" pseudo-symmetry earlier this summer, which I
understand is based on taking the Kronecker product of a C1 soup with itself:
wwei23 wrote:
Basically, take a 16x16 soup, and replace each cell with a 16x16 block of cells. Then replace each 16x16 block of cells with the original soup. It was the only way I could get 25p to work.
Can you share the code you used for this, either the entire modified apgsearch script or the hashsoup() routine it uses? Thanks.
Re: Hacking apgsearch
Posted: August 29th, 2017, 9:50 pm
by drc
Re: Hacking apgsearch
Posted: August 29th, 2017, 9:55 pm
by praosylen
Thanks for pointing that out! I'll try to see what's happening there — although I'm really busy right now in general, I should have a bit of time tonight to look at that.
Re: Hacking apgsearch
Posted: August 30th, 2017, 10:06 am
by praosylen
Several bugfixes, including fixing the bug that was causing the xp1001s, one that made the script error out on actual oversized spaceships, and another that made the script inoperable and very difficult to exit when the rule was entered in all lowercase.
Hopefully coming soon: speed improvements.
P.S. It turned out that the cause of the xp1001 bug was a quite insidious one where I was decrementing a non-existing variable, which Python didn't error on for some reason. It only affected patterns oriented a certain way, causing some of the strange observed behavior.
Re: Hacking apgsearch
Posted: August 30th, 2017, 1:32 pm
by muzik
Will B0 non-totalistic ever be supported?
Re: Hacking apgsearch
Posted: August 30th, 2017, 4:04 pm
by praosylen
muzik wrote:Will B0 non-totalistic ever be supported?
Probably not, unless it gets added to apgluxe (or apgsearch v5).
Re: Hacking apgsearch
Posted: August 30th, 2017, 4:04 pm
by Apple Bottom
muzik wrote:Will B0 non-totalistic ever be supported?
Will you ever stop posting questions like this every time some piece of software updates?
Honestly, the signal-to-noise ratio of these posts is not very high, and it's starting to get just a little bit annoying to constantly receive topic reply notifications for posts that are frankly fairly pointless.
Re: Hacking apgsearch
Posted: August 31st, 2017, 2:43 pm
by muzik
You should probably make the script simplify rulestrings, since I've managed to submit
this and
this, which should be simplified to b2-ce3-i45678s012345678 and b2cs, respectively.