Thread for basic questions

[drc]

Can someone modify gsearch for non-totalistic use? Sorry for my lack of programming skills, but I feel like we could find lots of ultra-slow and odd-period gliders that way.

[drc]

Alternatively, if someone could teach me how to add support to things for nt-rules, I'd be glad to do so and stop clogging this thread.

[muzik]

How does one look for reactions that could theoretically be turned into spaceships, as in caterpillar, centipede and waterbear?

[dvgrn]

How does one look for reactions that could theoretically be turned into spaceships, as in caterpillar, centipede and waterbear?

Basically what you need is a reburnable fuse, preferably one that puts out some kind of surplus signal -- what people have lately been calling an "over-unity" reaction. If the fuse burns blocks, let's say, then it should leave behind exactly as many blocks as it burns -- or if there's any extra junk along with the chain of blocks, there will have to be some way to clean that up with surplus gliders.

The reburnable fuse used in the shield bug and centipede was found manually, by waving a block around in a Herschel's reaction envelope using Paul Chapman's Seeds of Destruction game. It only took about ten minutes for the key reaction to show up.

The space of possibilities is probably a bit better researched by now, but you could try something along the same lines, using LOMs or centuries or any other common recognizable active pattern. Try all combinations of active reaction R interacting with still life S, and see if you can find any instances of R' (a copy of R at a workable distance from the original) appearing somewhere where the same interaction can be repeated, with an offset S' left behind in the ash.

An output glider can be good, especially if they're optional somehow, and a small amount of cleanable junk near the edges of the reaction envelope is possibly okay. The best reactions are probably the ones that have dying sparks at their edges where multiple copies can be rubbed against each other, without affecting the creation of either R' or S'. That's how the Caterpillar and HBK work -- a signal can travel along the fuse trails and have no effect other than moving the trail objects, but if there are two trails and two signals in close proximity, then you get a rake or puffer effect or whatever.

... If you do find something that looks new, though, try to contain your excitement at first. There are far too many ways for this kind of reaction to have subtle flaws that mean that it turns out to be useless...!

[muzik]

The space of possibilities is probably a bit better researched by now, but you could try something along the same lines, using LOMs or centuries or any other common recognizable active pattern. Try all combinations of active reaction R interacting with still life S, and see if you can find any instances of R' (a copy of R at a workable distance from the original) appearing somewhere where the same interaction can be repeated, with an offset S' left behind in the ash.

...so something like this?

Is this known? This must be known.

Code: Select all

x = 30, y = 235, rule = B3/S23
obo$obo$3o3$4b2o$4b2o18$6b2o$6b2o18$8b2o$8b2o18$10b2o$10b2o18$12b2o$
12b2o18$14b2o$14b2o18$16b2o$16b2o18$18b2o$18b2o18$20b2o$20b2o18$22b2o$
22b2o18$24b2o$24b2o18$26b2o$26b2o18$28b2o$28b2o!

[dvgrn]

...so something like this?

Code: Select all

x = 30, y = 235, rule = B3/S23
obo$obo$3o3$4b2o$4b2o18$6b2o$6b2o18$8b2o$8b2o18$10b2o$10b2o18$12b2o$
12b2o18$14b2o$14b2o18$16b2o$16b2o18$18b2o$18b2o18$20b2o$20b2o18$22b2o$
22b2o18$24b2o$24b2o18$26b2o$26b2o18$28b2o$28b2o!

Yup, something like that, except probably with more accessible sparks to the sides, and either A) without the extra beehive or B) with a glider output that might help get rid of the extra beehive... in two or more copies of the fuse simultaneously, somehow.

Oblique fuses are harder to work with because even if you have a glider coming out in one direction or another, you can't just flip over the resulting rake and use the glider from each rake to clean up the other's debris, the way we could with the 31c/240 Herschel-pair climbers.

... Speaking of which, one line of investigation that I'm not sure was ever followed through on, was the idea of cleaning up a different block in the 31c/240 reaction:

Code: Select all

x = 55, y = 252, rule = B3/S23
53b2o$53b2o8$8b2o$8b2o21$53b2o$53b2o8$8b2o$8b2o21$53b2o$53b2o8$8b2o$8b
2o21$53b2o$53b2o8$8b2o$8b2o21$53b2o$53b2o8$8b2o$8b2o21$53b2o$53b2o8$8b
2o$8b2o21$53b2o$53b2o8$8b2o$8b2o21$53b2o$53b2o8$8b2o$8b2o13$45b3o$46bo
$46b3o7$3o$bo$b3o!
#C [[ STEP 50 ]]

There are several relative placements that might shift the fuses over in an interesting way, while producing lots of gliders, which could be maybe used somehow to clean up Herschels at the top and re-insert them at the bottom (?). This would produce something more like an HBK than like a Caterpillar -- probably it would take too much synchronized construction to be practical, but maybe somebody can think of another idea.

(You could maybe do a centipede-like trick and use the gliders to build blocks at the top, but then you'd just end up with an oblique-looking 31c/240 orthogonal spaceship.)

[wildmyron]

Can someone modify gsearch for non-totalistic use? Sorry for my lack of programming skills, but I feel like we could find lots of ultra-slow and odd-period gliders that way.

Alternatively, if someone could teach me how to add support to things for nt-rules, I'd be glad to do so and stop clogging this thread.

There's a fair bit of work involved in this kind of modification - and similarly for your request on the Script Request thread regarding Eppstein's glider-db program. At a minimum, here is a list of required modifications:

• Modify data structure for rule specification to support isotropic rules
• Support specifying isotropic rules on the command line
• Update any File I/O functions to support saving and reading isotropic rule specifications
• Review the assumptions of any optimisations in the program which might break under isotropic rules
• In the case of a search program - implement an iterator which supports isotropic rules
• Where pattern symmetry is important - review all tests related to symmetry and update for isotropic rules
• Review and update any other parts of the code which depend on the rule

For some programs (like the glider-db program) it might be possible to utilise an existing implementation of the first and second points and not have too much more work to do. In other cases starting over will be a better option. I'm sure that more programs supporting isotropic rules on the Moore neighbourhood will be published, but probably not as a quick response to a request here.

[muzik]

Would a glide symmetric knightship be impossible?


Would a replicator in a 3D CA assume the shape of a cube, octahedron, or either depending on the rule, assuming the grid was made of cubes? Would such a rule be applied to CA of higher dimensions (tesseract/16-cell, 5cube/5orthoplex, etc.)?

[BlinkerSpawn]

Would a glide symmetric knightship be impossible?


Would a replicator in a 3D CA assume the shape of a cube, octahedron, or either depending on the rule, assuming the grid was made of cubes? Would such a rule be applied to CA of higher dimensions (tesseract/16-cell, 5cube/5orthoplex, etc.)?

1) Yes. Run a knightship for one cycle, flip it, run it for another cycle, and flip again in the same direction through the same cell. It should be displaced either orthogonally or diagonally from its starting position depending on your choice of axis.

2) Depends on the rule, just like the patterns in the Quadratic Replicators thread, regardless of dimension.

[muzik]

Is it possible for a still life, oscillator, spaceship, replicator, etc. to be an essential Garden of Eden to itself - i.e. having no predecessors except from its own evolution sequence?

[blah]

Is it possible for a still life, oscillator, spaceship, replicator, etc. to be an essential Garden of Eden to itself - i.e. having no predecessors except from its own evolution sequence?

This is every pattern in B/S012345678 (still lifes), or B012345678/S (p2 oscillators).
Also every oscillator in a reversible CA.

[muzik]

Could such patterns exist in life?

[Saka]

What's the best programming language to write a CA program in?

Please? I want to make one. The programs key features will be:
User friendly (GUI similar to Golly)
Support for nt rules
Support for Larger Than Life rules
Support for Moore radius 2, vonNeumann radius 2
A new ruletable format, human and computer readable (example is GoL. Please review and suggest improvements!):

Code: Select all

Life
(Parentheses denote comment. First line is rule name)
1 Live
(# of states, the background is not counted)
Moore
(Neighborhood)
Format BC
(The format of the rule, there is BC for birth/change and Adv for "advanced")
1:B3-1,C:0145678-1>0
(State 1: Birth on 3 neighbors of state 1, Change on 0,1,4,5,6,7,8 neighbors of state 1, and change into 0)

Adv form: (not yet, will make)

[dvgrn]

Could such patterns exist in life?

See Lifeline Volume 6, October 1972. John Conway offered a $50 prize for a still life with no predecessors other than itself (fading junk around the edges not counting, of course) -- and this is still an open question as far as I know. There's been some discussion of this in other threads and even a little bit in this thread in the last few months.

I should probably resist speculating about oscillators/spaceships with unique parents, which would therefore be provably unsynthesizable. I think all we can say at present is that every oscillator or spaceship that has been discovered so far -- and there are a lot of them -- can be shown via a quick lifesrc/WLS/JLS search to have at least two distinct parents. Those parents may themselves not be constructible -- for a larger pattern, it's easy to backtrack to a Garden of Eden parent whether you want to or not -- but the fact remains that there are multiple parents.

That doesn't tell us much about huge 100x100 or 1000x1000 oscillators or spaceships, purposely designed to have a unique parent. But nobody seems to know how to go about designing such a thing without an Earth-sized supercomputer. @mtve...?

[blah]

What's the best programming language to write a CA program in?

Please? I want to make one.

Maybe HTML5 (with js and css) would be a good idea. In terms of speed it wouldn't be the best, but if that's not important it might be worth considering. I created Reasoning Realm with HTML5. The point is that it's simply an easier language to use, especially when it comes to the GUI.

It depends on your experience and needs. The final decision should be yours to make. By the way, would you be interested in me posting a newer version of Reasoning Realm? I've wanted to for a while, but I'm not sure anyone would care.

[Saka]

Maybe HTML5 (with js and css) would be a good idea. In terms of speed it wouldn't be the best, but if that's not important it might be worth considering. I created Reasoning Realm with HTML5. The point is that it's simply an easier language to use, especially when it comes to the GUI.

It depends on your experience and needs. The final decision should be yours to make. By the way, would you be interested in me posting a newer version of Reasoning Realm? I've wanted to for a while, but I'm not sure anyone would care.

Alright. I will look into HTML5. But what about a language for an "offline" software (Like Golly)?

[blah]

Alright. I will look into HTML5. But what about a language for an "offline" software (Like Golly)?

Reasoning Realm requires no internet connection to run, it's just an HTML file. But in terms of standalone applications, maybe Java or something would be more appropriate and powerful, though harder to get into. It depends on how much effort you're willing to put in.

[muzik]

Is it possible for a Garden of Eden to exist where, if every cell within its bounding box were to be inverted from its original state, it would remain a GoE?

[BlinkerSpawn]

Is it possible for a Garden of Eden to exist where, if every cell within its bounding box were to be inverted from its original state, it would remain a GoE?

Most likely.
I'll see if I can find it but I know at one point that somebody did a search for GoE agars: subpatterns that weren't GoEs in their own right but were impossible to create a predecessor to when copied several times. I think it's rather likely that one of these could be found that is invertible or even self-inverse.
Edit: Here are the results.

[dvgrn]

Is it possible for a Garden of Eden to exist where, if every cell within its bounding box were to be inverted from its original state, it would remain a GoE?

Most likely.
I'll see if I can find it but I know at one point that somebody did a search for GoE agars: subpatterns that weren't GoEs in their own right but were impossible to create a predecessor to when copied several times. I think it's rather likely that one of these could be found that is invertible or even self-inverse.

Self-inverse might be an interesting case. Otherwise, doesn't the question have a trivial answer? Patterns matching that description are going to be all over the place.

Any large enough random 50% fill pattern -- 100x100 should be safe, and probably a lot smaller -- will have a probability very close to 1 of being both GoE and inverse-GoE.

Take the current record-holder, GoE #11 (hmm, somebody needs to tell LifeWiki about the last four records... In Achim's numbering, the LifeWiki's Garden of Eden #6 is actually #7. Probably it's a good idea to go with Achim's numbering, though he kind of messes it up with "Garden of Eden 8x12" which isn't in the numbered series.)

Anyway: take any rectangular-ish Garden of Eden, and put its inverse pattern next to it. The whole thing is still a Garden of Eden, because any pattern that contains a subpattern with no predecessor, also has no predecessor. So we have

Code: Select all

x = 22, y = 9, rule = B3/S23
2b2ob3obob2o2bo3bobo$bo2bobobo2bob2obobob2o$ob3o2b3o2bo3b2o3bo$bobobo
2bob2obobob2obo$b3o2b2o2b2o3b2o2b2o$2o3b2o2bo3b3o2b2obo$obobob3o3bobob
o3b2o$o2bobo2bo3b2obob2ob2o$b3o2b4obo3b2o4bo!

or even

Code: Select all

x = 21, y = 9, rule = B3/S23
2b2ob3ob3o2bo3bobo$bo2bobobobob2obobob2o$ob3o2b3obo3b2o3bo$bobobo2bobo
bobob2obo$b3o2b2o2bo3b2o2b2o$2o3b2o2bo2b3o2b2obo$obobob3o2bobobo3b2o$o
2bobo2bo2b2obob2ob2o$b3o2b5o3b2o4bo!

as a guaranteed GoE, no matter how you assign ON and OFF to the colors... Right? Am I missing something about the problem statement?

[Saka]

If I draw a random viable 3 glider collision (that works when rewinded), what are the chances it is "new"? What about a 4 glider collision?

[dvgrn]

If I draw a random viable 3 glider collision (that works when rewinded), what are the chances it is "new"? What about a 4 glider collision?

Good questions. I guess nobody really knows the answer yet.

Eventually I think the answer to the first question will be "zero". A complete database of all three-glider collisions should be just about within reach, with some careful work.

Right now the odds are maybe about one in a hundred thousand -- I just totally made that up -- that a randomly drawn three-glider collision happens to fall outside of the range of the various searches that Bob Shemyakin and others have done over the years. Does it count if gencols was run at some point and collided those gliders, but the filters were set to look for something else and so that combination was discarded? That probably happened lots of times to the three-glider switch engine recipe... Does it count as "new" if the collision has never been tried before, but it just makes some pointless junk that other three-glider collisions also make?

For four-glider collisions, well... a lot of gencols searches have been run with four gliders, too, but they're bound to have missed 99.9% of the collisions out there. The great majority of four-glider collisions are the weird useless cases, like delayed kickbacks from R-pentominoes, and 2-glider messes interacting with each other, and a glider hitting an active switch engine after a few hundred ticks, or a thousand ticks... and so on and so forth.

Maybe the question there is whether you'd actually be likely to pick one of the weird cases if you're just drawing glider collisions at random. Probably not... there are trillions of big slow messy four-glider explosions, but they're relatively hard to find because one or two of the gliders will probably be way off in the distance somewhere. If you drew that kind of thing randomly, the gliders would just miss each other almost all the time.

[bprentice]

By the way, would you be interested in me posting a newer version of Reasoning Realm?

Yes! Please do so.

Brian Prentice

[muzik]

Could exponential (or anything greater than quadratic) growth be possible in a 2D universe if the cell's neighbourhood is also allowed to increase infinitely?

[BlinkerSpawn]

Could exponential (or anything greater than quadratic) growth be possible in a 2D universe if the cell's neighbourhood is also allowed to increase infinitely?

Sure, just use B(all but 0)/S(all).