Thread for basic questions

[dvgrn]

Does "merge multiple glider stream[s] onto the same lane" include dealing with possible collisions -- i.e., two signals trying to occupy the same place at the same time? Or can we assume that merging p46 signals won't conflict?

[PK22]

Does "merge multiple glider stream[s] onto the same lane" include dealing with possible collisions -- i.e., two signals trying to occupy the same place at the same time? Or can we assume that merging p46 signals won't conflict?

I meant that it should be possible to combine two separate gliders onto the same lane without them conflicting or interacting, although my original intent behind the question was to find if it would be possible to create single-channel recipes using only p46 devices, but after asking the question, I found a p46 G-to-H which should make it possible to both duplicate and merge glider streams.

[102564]

Are there any still lives where exactly one cell survives by S2? The block has no cells surviving by S2, and ship, long ship, very long ship, etc have two cells surviving by S2.

[WhiteHawk]

Kind of dumb, but what is the best way to feed this conduit back on itself ? (preferably without gliders intermediate steps and using both inputs)

Code: Select all

x = 73, y = 64, rule = B3/S23
4bo$4b3o$7bo$6b2o2$16b2o$16bobo$17b2o$2o$bo$bobo$2b2o13$26bo$26b3o$28b
o$28bo2$2bo$2bobo$2b3o$4bo2$27b2o$27bobo9b2o$29bo9bo19b2o$29b2o6bobo20b
o$37b2o21b3o3$45b2o23bo$46bo21b3o$46bobo18bo$11b2o11bo22b2o18b2o$6b2o
3b2o9b3o$6b2o14bo$22bo44bo$66bobo$5b2o58bo2bo2b2o$6bo4b2o53b2o3b2o$3b
3o5b2o34bo$3bo43bo$36b2o9b3o15b2o$23b2o11b2o11bo15b2o$24bo$21b3o$21bo
4$59b3o$59bo$58b2o!

Trying to make something that will either be a high-period oscillator or will self-destruct after a given amount of time.

[I6_I6]

What sort of pattern would you call this?

Code: Select all

x = 7, y = 7, rule = /2/3
2A$2BA$2.BA$.A.BA$.B2.BA$2.BA.BA$5.BA!

[hotdogPi]

What sort of pattern would you call this?

Code: Select all

x = 7, y = 7, rule = /2/3
2A$2BA$2.BA$.A.BA$.B2.BA$2.BA.BA$5.BA!

I think it falls under the replicator category. Sierpinski triangles are generally replicators; the main difference between this one and many others is that this one is two-dimensional. It's not a sawtooth like one-dimensional Sierpinski triangle replicators are.

What are the two axes, by the way? I initially wrote space on one axis and time on the other, but I don't think this is quite accurate.

[I6_I6]

What sort of pattern would you call this?

Code: Select all

x = 7, y = 7, rule = /2/3
2A$2BA$2.BA$.A.BA$.B2.BA$2.BA.BA$5.BA!

I think it falls under the replicator category. Sierpinski triangles are generally replicators; the main difference between this one and many others is that this one is two-dimensional. It's not a sawtooth like one-dimensional Sierpinski triangle replicators are.

What are the two axes, by the way? I initially wrote space on one axis and time on the other, but I don't think this is quite accurate.

What do you mean by the two axes?
Also, how is that a replicator? The definition of a replicator on the Wiki is this:

While there exist many conflicting definitions as to what a replicator is, many patterns classified as replicators tend to satisfy the following properties:

• Replicators are bound by a parity rule, in which the replicator will always revert to having a certain number of copies after a given integer multiple of a perfect power, usually 2, although some replicators based on higher powers exist. These are referred to as parity rule replicators. In effect, replicators must also be sawtooths. Replicators which do not revert in such a fashion are usually considered to be wickstretchers or spacefillers, or may belong to another class entirely outside of these three.

• In addition, replicators are also generally required to leave no debris, especially if any such debris would interfere with further replication. Patterns which do leave behind objects are referred to as "dirty" replicators, and many stop replicating after an amount of time due to the chaos created. Debris also prevents replicators from acting as a sawtooth.

There are further ways in which a replicator can be defined, such as strength, which measures its power to be a unit cell.

But to me it isn't even clear what a copy of the replicator would be. The pattern isn't a sawtooth, either.

[b-engine]

What sort of pattern would you call this?

Code: Select all

x = 7, y = 7, rule = /2/3
2A$2BA$2.BA$.A.BA$.B2.BA$2.BA.BA$5.BA!

I think that could fall under waves category.
Waves can have such trails and it's still called a wave.
Since such patterns aren't known in B3/S23, no one would define a new OCA-exclusive term just for such patterns.

[Resu]

How can I get the correct soups from Mateon1_Glider8_4_5_Test?

[PK22]

How can I get the correct soups from Mateon1_Glider8_4_5_Test?

Run the script below in Golly and enter the seed of the 'soup'. This one processes Mateon1_Glider8_4_5_Test, but you can change line 10 to fit any number of your choosing (although apgsearch currently only has official support for 6_5_6 and 8_4_5, but a little tinkering with lifelib lets it run with any integer values):

Code: Select all

# decode-Mateon-stdin-Golly4.py

import golly as g

import hashlib

def digest(bs):
    return hashlib.sha256(bs).digest()
# set this to the three values in the symmetry
(glis, width, length) = (8, 4, 5)

hashval = bytes("\0" * 32, "ascii")
randcnt = 0
randbyte = 0
randbit = 0

def getRLE(seed):
    global randbit, randbyte, randcnt, hashval

    cells = set()

    hashval = digest(seed)
    randcnt = 0
    randbyte = 0
    randbit = 0

    def bit():
        global randbit, randbyte, randcnt, hashval
        if randbyte >= 32:
            assert randbit == 0
            randbyte = 0
            randcnt += 1
            hashval = digest(seed + bytes(":%d" % randcnt, "ascii"))

        val = (hashval[randbyte] >> randbit) & 1
        randbit += 1
        if randbit >= 8:
            randbit = 0
            randbyte += 1
        return val

    def rand(bits):
        assert bits <= 64
        val = 0
        for b in range(bits):
            val |= (1 << b) if bit() else 0
        return val

    PHASES = [
        [[1, 1, 1],
         [1, 0, 0],
         [0, 1, 0]],
        [[0, 1, 1],
         [1, 1, 0],
         [0, 0, 1]]]

    for quadrant in [(-1, -1), (1, -1), (-1, 1), (1, 1)]:
        counter = 0
        while counter < glis:
            ori = bit()
            flip = bit()
            phase = bit()
            offs = rand(width)
            shift = rand(length) + 4
            (x, y) = (quadrant[0] * shift, quadrant[1] * shift)
            if ori:
                x += quadrant[0] * (offs + 1)
            else:
                y += quadrant[1] * offs
            r = 2
            bad = False
            for dx in range(-r, 3 + r):
                for dy in range(-r, 3 + r):
                    if (x + dx, y + dy) in cells:
                        bad = True
            if bad: continue
            for dy in range(3):
                for dx in range(3):
                    (gx, gy) = (2 - dx if quadrant[0] == -1 else dx, 2 - dy if quadrant[1] == -1 else dy)
                    (gx, gy) = (gy, gx) if flip else (gx, gy)
                    if PHASES[phase][gy][gx]:
                        cells.add((x + dx, y + dy))
            counter += 1

    xmin = min(x for (x, y) in cells)
    xmax = max(x for (x, y) in cells)
    ymin = min(y for (x, y) in cells)
    ymax = max(y for (x, y) in cells)

    rle = f"x = {xmax-xmin+1}, y = {ymax-ymin+1}, rule = {rule}\n"
    linerepeat = 0
    repeat = 0
    char = "b"
    for y in range(ymin, ymax + 1):
        for x in range(xmin, xmax + 1):
            c = "o" if (x, y) in cells else "b"
            if c == char:
                repeat += 1
            else:
                if repeat and linerepeat:
                    if linerepeat > 1:
                    	rle+=str(linerepeat) + "$"
                    else:
                    	rle += "$"
                    linerepeat = 0
                if repeat:
                    if repeat > 1:
                        rle += str(repeat) + char
                    else:
                    	rle += char
                char = c
                repeat = 1
        if repeat and char != "b":
            if repeat > 1:
                rle += str(repeat) + char
            else:
            	rle += char
        repeat = 0
        linerepeat += 1
    rle += "!"
    return rle

# and this to the rule
rule = "B3/S23"

#set this to the seed
seedstr = g.getclipstr()

# this was the previous sample, but it doesn't build anything interesting that I can recognize. (?)
# seedstr = "k_Nbke7qxLPRzS8441"

if seedstr[1] != "_":
  seedstr = g.getstring("Enter 18- or 22-character Catagolue seed for a Mateon stdin result: ","k_9HrHTiQjsT7V74016984")
seed = seedstr.encode()

g.new(seedstr)

# g.setclipstr(getRLE(seed))

rlestr = getRLE(seed)

pat = g.parse(rlestr.split("\n")[1])

g.putcells(pat)
g.fit()

[WhiteHawk]

Why were there so few major discoveries made in 2020, compared to 2021 and especially 2022, despite lockdown and whatnot?

No new oscillator periods, one new gun (28), and vastly fewer newer oscillators (to the point that the biggest oscillator discovered that year, metaphorically speaking, was only found in random soup). Even the 2020 POTY competition seems to just consist of improvements to existing speeds technology.

(That is not to say that stable L122, Anura/Soba, and Bandersnatch or any other discoveries made that year weren't notable, just that so few discoveries were made compared to the two following years)

[dvgrn]

Why were there so few major discoveries made in 2020, compared to 2021 and especially 2022, despite lockdown and whatnot?

When I look at 2020 I don't see any shortage of major discoveries at all. I can see why it might seem like that from a 2025 perspective, though.

What gets discovered in a given time chunk tends to depend a lot on what the community is focused on exploring at that time. 2020 was several years in to a focus on glider synthesis -- including a lot of work on slow-salvo synthesis via universal construction, of very large complicated objects up to and including the 0E0P metacell (2018), Slavic (2019), and the Speed Demonoid (2020). These show up as single discoveries, but they're a lot of hours of work each.

The same goes for another active topic that had a lot of momentum at the time. Working through and finding a synthesis of all N-bit still lifes shows up as just one discovery, but it's a lot of synthesizing... and for example the 20-bit still life synthesis project started in mid-2020, but didn't show up as a CurrentNews line item until 2021.

Long story short, the "average discovery" in 2020 probably needed a much bigger time investment than the average discovery in the years after that.

The Great Oscillator Discovery Project
One big reason for 2021 being an inflection point is that 2021 is when hotdogPi showed up with a newly written search program. A lot of people suddenly realized that there was an enormous amount of unexplored search space within easy reach, that could produce new mechanisms for oscillators and guns. It just hadn't occurred to anyone to work through that search space in an organized way.

Like many lines of Life investigation, the more people looked, the more they found, and the reachable search space kept getting bigger as new catalysts were invented and new search code was written. The community suddenly had a new focus... and within a few years of repeatedly rolling the dice to try to hit missing periods, we had completed the huge job of case-bashing and proved omniperiodicity.

There were few enough missing periods for oscillators and guns (and, yes, there are still a few missing periods for guns) that each time a gap got filled it seemed reasonable to think of it as a separate discovery.

Conversely, we don't tend to think of the several hundred 20-bit still lifes that were synthesized in 2020 as separate discoveries. There were several other 2020 mega-projects that are hiding a lot of highly non-trivial subproblem solutions under one big label like "17-bit RCT" or "barge crosser" or "crabstretcher seed".

[WhiteHawk]

When I look at 2020 I don't see any shortage of major discoveries at all. I can see why it might seem like that from a 2025 perspective, though.

What gets discovered in a given time chunk tends to depend a lot on what the community is focused on exploring at that time. 2020 was several years in to a focus on glider synthesis -- including a lot of work on slow-salvo synthesis via universal construction, of very large complicated objects up to and including the 0E0P metacell (2018), Slavic (2019), and the Speed Demonoid (2020). These show up as single discoveries, but they're a lot of hours of work each.

The same goes for another active topic that had a lot of momentum at the time. Working through and finding a synthesis of all N-bit still lifes shows up as just one discovery, but it's a lot of synthesizing... and for example the 20-bit still life synthesis project started in mid-2020, but didn't show up as a CurrentNews line item until 2021.

Long story short, the "average discovery" in 2020 probably needed a much bigger time investment than the average discovery in the years after that.

The Great Oscillator Discovery Project
One big reason for 2021 being an inflection point is that 2021 is when hotdogPi showed up with a newly written search program. A lot of people suddenly realized that there was an enormous amount of unexplored search space within easy reach, that could produce new mechanisms for oscillators and guns. It just hadn't occurred to anyone to work through that search space in an organized way.

Like many lines of Life investigation, the more people looked, the more they found, and the reachable search space kept getting bigger as new catalysts were invented and new search code was written. The community suddenly had a new focus... and within a few years of repeatedly rolling the dice to try to hit missing periods, we had completed the huge job of case-bashing and proved omniperiodicity.

There were few enough missing periods for oscillators and guns (and, yes, there are still a few missing periods for guns) that each time a gap got filled it seemed reasonable to think of it as a separate discovery.

Conversely, we don't tend to think of the several hundred 20-bit still lifes that were synthesized in 2020 as separate discoveries. There were several other 2020 mega-projects that are hiding a lot of highly non-trivial subproblem solutions under one big label like "17-bit RCT" or "barge crosser" or "crabstretcher seed".

Thanks for the clarification.

What would you say was the most productive year for life discoveries? Also where would you say 2025 fits on the "productivity" scale

[dvgrn]

What would you say was the most productive year for life discoveries? Also where would you say 2025 fits on the "productivity" scale...

I don't think I'd dare to pick a single year -- there were a bunch of local maxima, with productivity values highly dependent on the metrics you use -- number of discoveries? Amount of research work put in? Number of people collaborating on discoveries?

I hear 1970 was pretty epic.
1996 was when people suddenly went from only knowing about one weird exceptional odd-period glider gun, to being able to build an unbounded number of odd-period guns on demand.
But I wasn't in the community for either of those years.

2003 was when Karel Suhajda and a bunch of other people filled in the last holes in the p14-p999 gun collection -- that was a _lot_ of activity for the time period.

In 2006, Brice Due's OTCA metapixel impressed a lot of people -- it was way bigger than anything that seemed like it ought to run on the PCs of the time, and yet Golly could run it... same with the p1 megapixel in 2008.

Several of APG's mega-projects showed up in 2010 -- e.g., the pi and phi calculators and Osqrtlogt -- and so did one of the biggest surprises of all time, Andrew Wade's Gemini. So somewhere in that half-decade was where Lifenthusiasts' collective imagination suddenly started playing in a much bigger space.

2015 started a productive time for spaceship syntheses that had been almost completely beyond everyone's abilities up to then -- with Martin Grant/Extrementhusiast leading the way, and Goldtiger997 and others joining the Synthesis Wizard team as time went on.

2025 seems like it's been pretty good all in all. I guess there isn't one single over-the-top thing from 2025 yet that's the obvious winner of Pattern of the Year. But the year isn't quite over... and anyway, we could pick any of dozens of 2025 discoveries, send them back a quarter century, and completely wow every Lifenthusiast of the previous millennium.

[hth3]

Are there self-supporting wicks, waves, and/or agars?
Like, an infinite version of a self-supporting spaceship?

[dvgrn]

Are there self-supporting wicks, waves, and/or agars?
Like, an infinite version of a self-supporting spaceship?

That's certainly a thing. The Silverfish from the Life textbook is a seriously antiquated construction nowadays, but you could just stack any number of rephasers onto a Silverfish spine. On an appropriate-sized torus the rephasers would count as self-supporting. Could do something similar for spines for a waterbear or Caterpillar or any other self-supporting spaceship, I think.

There could be self-constructing agars, too, easily enough. It would simplify the engineering quite a lot if each unit of an infinite agar provided the target block during each cycle, for a single-channel salvo coming from the next unit.

One reason why nobody has designed self-constructing infinite stuff, is that Golly only barely tolerates toroidal universes at that kind of size. It would be a terribly painful experience to watch a self-constructing agar get through a full cycle on a torus, unless we cheated with a custom script.

[hth3]

Can you make a metacell that self-replicates in Life, and then emulate a Seeds pre-block using it, making an explosive Life pattern?

[PK22]

Can you make a metacell that self-replicates in Life, and then emulate a Seeds pre-block using it, making an explosive Life pattern?

Yes - the 0E0P metacell allows for these sorts of constructions. Unfortunately, it is so large that simulating one cell takes about half a core-year. Emulating an explosive pattern with the metacell to the point where it is clearly explosive could take core-centuries or even core-millenia, and so until a faster metacell (performance-wise) is made or personal supercomputers become common, such patterns will be highly impractical.

[WhiteHawk]

what is the fastest HWSS eater?

[102564]

The following pattern consists of two non-interacting oscillators, one p200 and one p300. (Both are simply capped versions of guns in the collection on Catagolue.) However, there are cells that oscillate at period 600 = lcm(200, 300), where the two glider streams intersect. So, would the pattern be considered an LCM oscillator?

Code: Select all

x = 95, y = 76, rule = B3/S23
61bo9bo10b2ob2o2b2o$61b3o3b2o4bo7bobobo2bob3o$64bo2bo5bo4bo2bobobob2o
4bo$63b2o3b2obo2bo2bobob5o3b3obo$64b4o5b4o2bobo2bo2b2o3bo$55b2o8b3o6b
o4bo3b3o5bo$55bo2b2o4b2o2bo6bo2b4o2b2obo$56b2obo4bo14b2o3bo4bo2bo$60b
o3b3o8bob2o3bob4ob4o$60bo3b2obo7bo3bobo6bo$56b2obo5bob2o7b2obobob3o2b
obo$55bo2b2o5b2o11bo3bobo4b2o$3b2o2b2ob2o20bo22b2o9bob2o5bobo8bo$b3ob
o2bobobo8bo3b2o3b3o32b2ob2o5b2o8b2o$o4b2obo3bo2bo3b3o2bobo2bo35b2ob2o
$ob3o3bo3bobob2o3bo3bo3b2o36bo$bo2b2obo4bobo2bo2bo59b3o$5b2o3b3o4bo2b
o16b2o41b3obo$7bo4bo2bo11bo6b2o2bo20bo21bobobo$bo2bobo9b2o8b3o4b2ob2o
21bo22bo2bo$b4ob5obo2bobo7bo2b2o3b2o30b3o15b2o$5bo6bobo3bo14b2o24bo29b
2o$3bobo2b3obobob2o6b2o7b2ob2o20bobo3bo3bo20bobo$3b2o4bobo3bo8b2o8b2o
2bo26b3o23bo$7bo8bobo4bo2bo10b2o19b3o30bob2o$7b2o8b2o6b3o60b2obobo$31b
2o50b2o3b2obobo$12b3o17bo2bo22b3o22b2o6bob2o$10bob3o17bobobo28b3o4b2o
14b4o2bo$9bobobo17b2obobo21bobo3bo3bo2bobo14bo3b2o$9bo2bo17bo3b2o23bo
11bo18b2o$10b2o9bo10b2o31b3o2b2o19bo$4b2o15bobo9bo25bo29bo$3bobo15b2o
7b3o26bo29b2o$3bo22b3o2bo$2obo23bobo3bob2o$bobob2o5b2o12b4obob2obo$bo
bob2o5b2o11bo4bobo$2obo21b4obobo$o2b4o14b2o5bob2o$b2o3bo14bobo3bo$3b2o
18bo2bo$3bo19bobo$4bo17b2ob2o$3b2o27$37b2o$36bobo7b2o$36bo9bobo$35b2o
11bo$48b2o!

[hotdogPi]

The following pattern consists of two non-interacting oscillators, one p200 and one p300. (Both are simply capped versions of guns in the collection on Catagolue.) However, there are cells that oscillate at period 600 = lcm(200, 300), where the two glider streams intersect. So, would the pattern be considered an LCM oscillator?

Code: Select all

snip

No, since it's two oscillators, not one.

[b-engine]

No, since it's two oscillators, not one.

But what about the small region of overlapping rotors that has period of 600? This makes the entire oscillator non-trivial.

[hotdogPi]

No, since it's two oscillators, not one.

But what about the small region of overlapping rotors that has period of 600? This makes the entire oscillator non-trivial.

For an oscillator to be an LCM, it must be one oscillator, not two.

I know apgsearch is properly able to separate this into two oscillators (it couldn't in older versions).

[islptng]

Is T-and-Tub notable enough to have its own "Rule:"-namespace page on LifeWiki?

And should I add the emulator to the page APGsembly?

EDIT; Why isn't the result of OCA DOTY 2024 on LifeWiki?

[hotcrystal0]

What is the oldest version of LifeViewer that can still be used online?