apgsearch v2.2

[biggiemac]

only*

Had the code been written to support larger base tiles, generating symmetric soups would just be bit twiddling as it was in the python program. I am very interested in the possibility of increasing symmetric B3S23 soup volumes by a couple orders of magnitude, but I've proven to Adam before that I shouldn't be trusted modifying apgsearch. So here's another gentle request for him to tweak the initializion to support filling more than a 28x28 SIMD tile

[Billabob]

Is it a good idea to show who discovered an object on the object's page?

Also, this page has a strange name.

[isaacg]

The user info update is awesome - thanks for adding that!

Is it feasible to extend the current "This user discovered" to more/all objects? I'm particularly interested in newly discovered still-lives, since that's all I've discovered.

[biggiemac]

The user info update is awesome - thanks for adding that!

Is it feasible to extend the current "This user discovered" to more/all objects? I'm particularly interested in newly discovered still-lives, since that's all I've discovered.

I'm not too keen on that change, since with the addition of xp3 my page already bogs down anything I try to use to view it. I'm probably responsible for on the order of 10K still lives. If this is implemented I'd like to have the page display only the N most interesting objects, for some reasonable N < 200.

[calcyman]

If this is implemented I'd like to have the page display only the N most interesting objects, for some reasonable N < 200.

How does one objectively define interestingness?

[Scorbie]

Well, I have nothing, but I think a rough guideline would be
1) Still lifes < p2 objects < everything else -- not so sure...
2) The rarer, the more interesting.

[praosylen]

Well, I have nothing, but I think a rough guideline would be
1) Still lifes < p2 objects < everything else -- not so sure...
2) The rarer, the more interesting.

I'd say also 3) The larger, the more interesting. And I also think that prime-period oscillators are more interesting that composite-period oscillators (at least for p>3).

[Freywa]

Calcyman, please add my Twitter account to the user Parcly Taxel:
Here.

[calcyman]

Calcyman, please add my Twitter account to the user Parcly Taxel:
Here.

Codeholic owns the contributors.yml file, not me. The relevant line to append is:

Code: Select all

"Parcly Taxel": "@Parcly_Taxel"

[Freywa]

Oh, sorry, forgot.

[Apple Bottom]

If I may, here's a couple of suggestions for the Catagolue website, things that would be nice to have, in descending order of importance:

1) A lexicographically-sorted list of all rulestrings/symmetries that have been investigated, i.e. had hauls submitted.

2) Sample soup links broken down by symmetry. Using the tumbler as an example - instead of "There are 276 sample soups in the Catagolue", followed by "•"'s in various colors (which I can never remember myself!), have it say "There are 199 C1 sample soups", "There are 10 D4_+2 sample soups", and so on.

3) For a given rulestring, an overview of all objects found in that rule, regardless of symmetry.

[gameoflifeboy]

1) A lexicographically-sorted list of all rulestrings/symmetries that have been investigated, i.e. had hauls submitted.

I once liked this idea so much, I tried to make such a list myself. Although I'd actually prefer to have a complete list of rules ordered by the total number of soups across all symmetries, and the rule pages telling how many soups of each symmetry were searched.

2) Sample soup links broken down by symmetry. Using the tumbler as an example - instead of "There are 276 sample soups in the Catagolue", followed by "•"'s in various colors (which I can never remember myself!), have it say "There are 199 C1 sample soups", "There are 10 D4_+2 sample soups", and so on.

Hmmm... This would break the script find-object.py, which goes to catagolue and finds sample soups for a given object. I'm suggesting instead to have the symmetry of each soup display as alt text when the link is moused over.

3) For a given rulestring, an overview of all objects found in that rule, regardless of symmetry.

This would be good. We could have a page displaying the most common still life of each size, the most common, or smallest (or both) oscillator of each period, the most common spaceship of each period (and maybe speed too), and the most common linear growth pattern of each period.

[biggiemac]

have the symmetry of each soup display as alt text when the link is moused over.

It kind of already does that, if you count the URL preview as alt text. The URL holds the symmetry, and appears in the lower corner of the screen (for chrome at least) when you hover over a link.

[Apple Bottom]

I once liked this idea so much, I tried to make such a list myself. Although I'd actually prefer to have a complete list of rules ordered by the total number of soups across all symmetries, and the rule pages telling how many soups of each symmetry were searched.

Interesting idea. I'd be happy with any list of all available censuses, so long as it's one page.

An alternative idea: separate data and presentation; make the data available in e.g. JSON format (sorta like with the user contribution pie charts), and have a bit of Javascript render it the way the user likes it.

Hmmm... This would break the script find-object.py, which goes to catagolue and finds sample soups for a given object. I'm suggesting instead to have the symmetry of each soup display as alt text when the link is moused over.

It kind of already does that, if you count the URL preview as alt text. The URL holds the symmetry, and appears in the lower corner of the screen (for chrome at least) when you hover over a link.

There is that; but I'm more keen on knowing how many soups of each type there are for a given object, not what symmetry type a single specific soup has. In fact learning the various colors would be easier than hovering and checking the URL and/or alt text for each soup.

Obviously breaking the script should be avoided, but couldn't there be both a list of soups, like now, and an overview over what types of soups there are? Alternatively, how easy would fixing the script be?

On a more general note this is also something that could benefit from a clean separation of data and presentation. If the script's currently screen-scraping census pages, then having a well-defined API providing machine-readable data would be a boon, and make it much easier to adjust the census pages without breaking nothing.

And on the subject of symmetry: is there a non-technical explanation of what the symmetry types mean? I found this post David made, and I read up on rotational and dihedral symmetry on Wikipedia, but I don't quite understand the difference between, say, D4_+1 and D4_+2 yet. If my web searching skills aren't failing me, that notation's not common outside Conway Life soup-searching.

This would be good. We could have a page displaying the most common still life of each size, the most common, or smallest (or both) oscillator of each period, the most common spaceship of each period (and maybe speed too), and the most common linear growth pattern of each period.

And also the usual lists of all objects of a given type, just for all symmetries combined. (Perhaps updated once per day if keeping it current all the time would be too resource-intensive/expensive.)

Speaking of spaceships, it would be awesome if spaceships' pages listed their velocities as well. And I also notice the Life wiki lists, among other things, heat, temperature and volatility. Would it be possible to eventually have those calculated automatically and displayed on census pages where appropriate?

[Apple Bottom]

Apologies for double-posting.

I noticed that some objects, e.g. the HWSS, have "25%" and "75%" sample soups. However, the links for these are broken, likely because the percent sign is special in URLs. (And manually changing it to its encoded form doesn't give you the right soup.)

Also, a few objects still have hexadecimal nicknames:

xs19_354c8716853 (066304a102ae0118)
xs20_i5lmzdlk9 (01b202a502950136)
xs25_259arhe0ehr (063605590142055c0630)
xs26_660uhe0ehu066 (031802a81aab1aab0110)

[biggiemac]

The % fiasco was me.. I tried to add 1/4 and 3/4 average densities into the code as symmetries way back when but the string I sent had a percent sign. They should work if you change the % char into "pct".

(This is one of a few reasons for my statement earlier that Adam shouldn't trust me modifying his program)

[flipper77]

...how easy would fixing the script be?

It would be very easy for me to fix the script if something like that does end up happening, so I wouldn't mind if catagolue adopted the idea of specifying number of soups with specific symmetry, not to mention that I've been thinking of how great it would be if the soups were organized that way.

[Billabob]

Silly "milestones" we can expect to see soon:

(sqrt(20)^20)/20 (Roughly 19 trillion)
20/sqrt(20^-20) (Roughly 20 trillion)
13^13 (Roughly 30 trillion)
17! (Roughly 35 trillion)

[gameoflifeboy]

Silly "milestones" we can expect to see soon:

(sqrt(20)^20)/20 (Roughly 19 trillion)
20/sqrt(20^-20) (Roughly 20 trillion)
13^13 (Roughly 30 trillion)
17! (Roughly 35 trillion)

Don't forget 3^3^3 * 3, or 22,876,792,454,961. Or better yet, 3^3^3 * 3 + 3, which is 22,876,792,454,964!

Also, why is http://catagolue.appspot.com/help/ not linked to anymore? I think it's quite a useful page.

[gmc_nxtman]

Unfortunately, I won't be able to commit any more hauls until I get a new computer that can run overnight without overheating. No more griddle variants, I guess

[Billabob]

Don't forget 3^3^3 * 3, or 22,876,792,454,961. Or better yet, 3^3^3 * 3 + 3, which is 22,876,792,454,964!

I always miss the semi-obvious ones!

[gmc_nxtman]

I can't resist pointing out our perfect number of hauls right now:

hauls.png (25.72 KiB) Viewed 15509 times

[calcyman]

Silly "milestones" we can expect to see soon:

(sqrt(20)^20)/20 (Roughly 19 trillion)
20/sqrt(20^-20) (Roughly 20 trillion)
13^13 (Roughly 30 trillion)
17! (Roughly 35 trillion)

We've just passed 2 ^ 44 = 17592186044416 objects.

By the way, there are 39 different two-spaceship flotillae (see http://codercontest.com/mniemiec/flotilla.htm) once you exclude the three pseudo-spaceships. According to Catagolue, we've found 25 of them (see https://catagolue.appspot.com/census/b3s23/C1/xq4). Do you think that the other 14 will be found before any other natural spaceship, or that (say) a loafer may emerge first?

[codeholic]

In order to add some incentive, I put €50 bounty on the first new natural spaceship not being an interaction of two standard spaceships.

Please retweet and spread the word other ways.

[Freywa]

I was looking at the soup counts generated by apgsearch and I had some idea.

Note the irregular soup counts: 11720000, 11750000, 11780000…

A.png (49.8 KiB) Viewed 15436 times

Apparently the way OpenMP is used to parellelise the soup searches is set up such that each core gets an equal number of soups – I can see short gaps of 30000 soups here, since I'm running three cores (on a new laptop). This leads to delays when one core finishes first. Maybe when that happens, assign some of the remaining soups to it so that the workload remains equal?