Page 16 of 23

Re: Catagolue Oddities

PostPosted: December 20th, 2018, 1:52 pm
by muzik
EDIT: both amended

Seems as though haul/ pages need updating, only the square symmetries are being listed here:
https://catagolue.appspot.com/haul/b2s3h/

Also, the new symmetries are being interpreted as unofficial:
https://catagolue.appspot.com/census/b2s3h/

Re: Catagolue Oddities

PostPosted: December 21st, 2018, 6:28 am
by muzik
Some Life patterns seem to have weirdly formatted names; for example compare
https://catagolue.appspot.com/object/xp2_31egge13/b3s23
to
https://catagolue.appspot.com/object/xp ... a221/b3s23
Shouldn't the latter be called "tub test-tube baby" instead?

Re: Catagolue Oddities

PostPosted: December 21st, 2018, 7:59 am
by danny
I always found that weird since 'tubebaby' isn't a word...

Re: Catagolue Oddities

PostPosted: December 21st, 2018, 9:58 am
by calcyman
danny wrote:I always found that weird since 'tubebaby' isn't a word...


Same here. The 'common names' mapping is a complete mess, and no-one has had time to go through all 3621 lines of it to ensure consistency. I think the 'tubebaby' slipped in from Achim's census:

http://wwwhomes.uni-bielefeld.de/achim/freq_top_life.html

Potentially the common names mapping could be automatically updated from the object pages on LifeWiki, if there's an API to download this data?

Re: Catagolue Oddities

PostPosted: December 21st, 2018, 4:41 pm
by calcyman
I previously wrote:Potentially the common names mapping could be automatically updated from the object pages on LifeWiki, if there's an API to download this data?


I've figured out how to do this from the command-line using curl:

$ curl http://conwaylife.com/wiki/Pond_on_dock?action=edit 2>/dev/null | grep '^|.*='
|name             = Pond on dock
|pname            = pondanddock
|c                = 18
|bx               = 8
|by               = 6
|rulemin          = B/S23
|rulemax          = B3678/S012345678
|rulespecial      = [[Conway's Game of Life|Conway Life]]
|rle              = true
|apgcode          = xs18_3lk453z3443
|niemiecid        = 18.2387
|pentadecathlonid = 18.2654
|viewerconfig     = #C [[ GPS 2 THUMBSIZE 2 ]]


Would anyone (@Nathaniel, @AppleBottom, @dvgrn) be vehemently opposed to me running over a thousand GET requests to LifeWiki to obtain a more comprehensive 'common names' list for use in Catagolue?

One appealing aspect is that the 'common names' list could also store auxiliary information, including links from the Catagolue object pages back to the LifeWiki.

Re: Catagolue Oddities

PostPosted: December 21st, 2018, 5:07 pm
by muzik
calcyman wrote:
I previously wrote:Potentially the common names mapping could be automatically updated from the object pages on LifeWiki, if there's an API to download this data?


I've figured out how to do this from the command-line using curl:

$ curl http://conwaylife.com/wiki/Pond_on_dock?action=edit 2>/dev/null | grep '^|.*='
|name             = Pond on dock
|pname            = pondanddock
|c                = 18
|bx               = 8
|by               = 6
|rulemin          = B/S23
|rulemax          = B3678/S012345678
|rulespecial      = [[Conway's Game of Life|Conway Life]]
|rle              = true
|apgcode          = xs18_3lk453z3443
|niemiecid        = 18.2387
|pentadecathlonid = 18.2654
|viewerconfig     = #C [[ GPS 2 THUMBSIZE 2 ]]


Would anyone (@Nathaniel, @AppleBottom, @dvgrn) be vehemently opposed to me running over a thousand GET requests to LifeWiki to obtain a more comprehensive 'common names' list for use in Catagolue?

One appealing aspect is that the 'common names' list could also store auxiliary information, including links from the Catagolue object pages back to the LifeWiki.

Currently we're trying to reach a consensus on the wiki for pattern naming, which might take a while, so that we can have a cleaner, more consistent methodology for systematically naming still lifes and oscillators, which should eventually carry on to Catagolue. (You're welcome to join in by the way.)

Re: Catagolue Oddities

PostPosted: December 21st, 2018, 6:38 pm
by calcyman
calcyman wrote:One appealing aspect is that the 'common names' list could also store auxiliary information, including links from the Catagolue object pages back to the LifeWiki.


I decided to do this anyway. Take a look at xs12, for instance:

https://catagolue.appspot.com/census/b3s23/C1/xs12

The H2 titles of the object pages are similarly updated:

https://catagolue.appspot.com/object/xq96_y8573w8aegvzys33zwg8gyvoozsvv0338gy2442cia049uo239e4z037e8x1z08464ogzy011zy3oozyf51guozyd2451032/b3s23

It makes Catagolue and LifeWiki feel more tightly coupled; their object pages link seamlessly (I hope!) in both directions.

Here you can see which objects are on the LifeWiki (common names have hyperlinks), which ones have common names from elsewhere (no hyperlinks), and which ones have no common name (apgcode only). I imagine this could be useful for LifeWiki maintainers.

muzik wrote:Currently we're trying to reach a consensus on the wiki for pattern naming, which might take a while, so that we can have a cleaner, more consistent methodology for systematically naming still lifes and oscillators, which should eventually carry on to Catagolue. (You're welcome to join in by the way.)


Great! When that's been accomplished, I can just rerun the scripts to scrape LW and generate an updated common names list.

Re: Catagolue Oddities

PostPosted: December 21st, 2018, 8:38 pm
by mniemiec
calcyman wrote:I decided to do this anyway. Take a look at xs12, for instance: ...

This looks very nice! One thing that concerns me, however, is that Catagolue seems to sort object by APG, irrespective of rule; e.g. comments appear to be shared by patterns in different rules that share the same APG code, even if the patterns don't behave the same way. This isn't an issue with still-lifes (which DO behave the same way), but oscillators and spaceships often have different behaviors in different phases, and spaceships may even have different velocities, but they're lumped in together as long as they share the same period.

Re: Catagolue Oddities

PostPosted: December 22nd, 2018, 6:40 am
by Apple Bottom
calcyman wrote:Would anyone (@Nathaniel, @AppleBottom, @dvgrn) be vehemently opposed to me running over a thousand GET requests to LifeWiki to obtain a more comprehensive 'common names' list for use in Catagolue?


Certainly not me --- only Nathaniel can say for sure, since he's the one ultimately running the server, but MediaWiki is made for handling lots of page views; it's got rather sophisticated caching, so reads are cheap.

Since you're a Python aficionado, BTW, there's a an actively-maintained Python wrapper around the MediaWiki API, pymediawiki. Chances are you've already discovered it, but I thought I'd toss the link out here anyway. There's also a tool for generating backups of MediaWiki installations that you could use to all current page versions, if you like dealing with XML. :)

EDIT: ah, I see you've already gone ahead. Good stuff, it's nice to see the various sites in the Life universe being woven together more tightly.

Re: Catagolue Oddities

PostPosted: December 26th, 2018, 4:12 pm
by PushDecimal
Two distant spaceships being recognised as one:
http://catagolue.appspot.com/census/b3-es23/C1/xq14

Re: Catagolue Oddities

PostPosted: December 26th, 2018, 6:32 pm
by Ian07
The × character used on the wiki shows up incorrectly on Catagolue:
https://catagolue.appspot.com/object/xs ... 6221/b3s23

Re: Catagolue Oddities

PostPosted: December 28th, 2018, 6:21 pm
by 77topaz
I just noticed an enormous pile of submitted but unverified/uncommitted hauls sitting in the b3s23/C1 census backlog. The weird thing is that, despite this backlog, apgsearch's /verify function claims there are no more hauls to verify. What's up with that?

Re: Catagolue Oddities

PostPosted: December 28th, 2018, 6:33 pm
by Ian07
77topaz wrote:I just noticed an enormous pile of submitted but unverified/uncommitted hauls sitting in the b3s23/C1 census backlog. The weird thing is that, despite this backlog, apgsearch's /verify function claims there are no more hauls to verify. What's up with that?


Initially, I thought this had something to do with peer review automatically being enabled for censuses with over 10^12 objects, but Adam told me it was actually an error caused by someone searching Day & Night. The error's now been fixed, and I'm currently working on verifying the backlog of hauls this caused.

Re: Catagolue Oddities

PostPosted: December 28th, 2018, 6:49 pm
by 77topaz
The strange thing is that my apgsearch keeps saying "No more hauls to verify." even when there are clearly still many hauls in the "?committed=0" log. This didn't happen, for example, when rliston was submitting many b3s23/C1 hauls quickly; those just verified normally and that backlog was eventually cleared when I edited my apgsearch to essentially work as a full-time verification script. But this approach isn't working for me now, because apgsearch isn't even recognising that this backlog of hauls needs to be verified.

Re: Catagolue Oddities

PostPosted: December 28th, 2018, 7:03 pm
by calcyman
77topaz wrote:The strange thing is that my apgsearch keeps saying "No more hauls to verify." even when there are clearly still many hauls in the "?committed=0" log. This didn't happen, for example, when rliston was submitting many b3s23/C1 hauls quickly; those just verified normally and that backlog was eventually cleared when I edited my apgsearch to essentially work as a full-time verification script. But this approach isn't working for me now, because apgsearch isn't even recognising that this backlog of hauls needs to be verified.


The ones awaiting verification are in 'committed=1'. The system is:

  • Hauls received from apgsearch immediately have committed=0;
  • The cron process moves them to either committed=1 (if verification is enabled) or commits them immediately (if verification is disabled);
  • Instances of apgsearch verify hauls in the committed=1 group, moving them to either committed=3 (successful) or committed=4 (unsuccessful);
  • If verification is enabled, the cron process commits anything with committed=3 (moving them to committed=2).

At the moment Catagolue seems to be committing the maximum allowed rate of 120 hauls every cron process (i.e. every 12 minutes), so it should soon catch up with the backlog.

Re: Catagolue Oddities

PostPosted: January 4th, 2019, 7:29 pm
by 77topaz
For range-1 Generations rules, a LifeViewer is rendered if the pattern has no state-2+ cells in its canonical phase (example), but only a static image is shown otherwise (example). It's a bit weird, because both of these examples are in the same rule and both have a functional apgcode, yet one has a LifeViewer and one does not.

Also, I think hexagonal-neighbourhood rules should perhaps have a separate rule folder from the Moore-neighbourhood non-totalistic rules. (EDIT: This was done within a few minutes of this post being written.)

EDIT: The transcluded-information-from-Fano boxes for outer-totalistic rules seem to have disappeared.

Re: Catagolue Oddities

PostPosted: January 5th, 2019, 12:13 am
by muzik
77topaz wrote:EDIT: The transcluded-information-from-Fano boxes for outer-totalistic rules seem to have disappeared.

Is it just me, or is fano also timing out every time I try to connect to it?



Another oddity: the site says that you should select a symmetry type for a rule even when said rule hasn't yet been searched. https://catagolue.appspot.com/census/bs1234567

Re: Catagolue Oddities

PostPosted: January 8th, 2019, 1:00 pm
by muzik
Rather problematic oddity: this list page

https://catagolue.appspot.com/rules/isotropic

houses a particularly long rulestring which results in the table being unnaturally stretched. The end result is that all named rules, as well as numbers of censuses, all of which are written in black text, end up being ousted from the light background onto the dark background, making them particularly hard to read.


There are a few ways to fix this problem;

- All rules with non-canonical rulestrings could be removed from this table (and tabulated elsewhere; there are likely many rules of interest which have not yet been searched under a canonical rulestring), which would help keep the table at a reasonable size; the maximum permittable length of a rulestring would probably not cause such dramatic stretching.

- A script could be run that automatically renames/merges all censuses from their noncanonical forms into their canonical forms, also eliminating this problem and cleaning up the site as well as a result.

Re: Catagolue Oddities

PostPosted: January 8th, 2019, 2:07 pm
by danny
or just delete the rule

Re: Catagolue Oddities

PostPosted: January 8th, 2019, 7:36 pm
by A for awesome
Or abbreviate (/ellipticize [is that a word]) the offending rules, or just wrap them at a certain character length. Either of those could be useful for apgcodes as well — I'd personally lean toward wrapping the rules at about 60 characters and apgcodes at some other value appropriate to the formatting (but I'm not @calcyman, so it's really just a thought).

Re: Catagolue Oddities

PostPosted: January 9th, 2019, 5:54 am
by muzik
Those would also work, but then that results in a lot of duplicate rules remaining in the list, taking up space, which isn't ideal.

-

It seems that for outer-totalistic and isotropic non-totalistic rules, catagolue only checks if p99 and below patterns make sense:

https://catagolue.appspot.com/object/xq ... 6952/b3s23
https://catagolue.appspot.com/object/xq ... 6952/b3s23

Other rulespaces (including B0) don't seem to care:

https://catagolue.appspot.com/object/xq ... b0235s1234
https://catagolue.appspot.com/object/xq ... b0235s1234

Re: Catagolue Oddities

PostPosted: January 9th, 2019, 7:07 am
by 77topaz
That limit is because if the server had to double-check a large amount of generations each time someone wanted to load a pattern page (the xp apgcodes can go up to 1048576), it would slow the site down and possibly also strain the server quota.

Re: Catagolue Oddities

PostPosted: January 10th, 2019, 8:03 pm
by muzik
Isn't it because only period 100 and up patterns are actually simulated in image form?

Re: Catagolue Oddities

PostPosted: January 10th, 2019, 10:19 pm
by 77topaz
Those are both for the same reason: preventing server lag/strain. And having an invalid apgcode does not prevent apgsearch from simulating a GIF for it in a census, even if the period is below 100.

Re: Catagolue Oddities

PostPosted: January 11th, 2019, 9:50 am
by muzik
77topaz wrote:For range-1 Generations rules, a LifeViewer is rendered if the pattern has no state-2+ cells in its canonical phase (example), but only a static image is shown otherwise (example). It's a bit weird, because both of these examples are in the same rule and both have a functional apgcode, yet one has a LifeViewer and one does not.


In addition, LifeViewer doesn't follow the same colour scheme as the previewed pattern image files. There are no green cells to be seen in the evolution of this:

https://catagolue.appspot.com/object/xp13_318c/g13b3s23

-

The isotropic non totalistic rules page seems to have its stretching fixed, although the non-canonical rules are still there...