Difference between revisions of "LifeWiki:Tiki bar"

From LifeWiki
Jump to navigation Jump to search
(→‎Time for a consensus decision on pnames?: adding glider syntheses for auto-upload)
(13 intermediate revisions by 5 users not shown)
Line 368: Line 368:


:::::: (I should qualify this by stating that I'm not suggesting we delete existing manually-generated pattern files; merely that having Catagolue be able to do this would allow us to not have to worry about future ones, except for Patterns of Unusual Size™ etc).
:::::: (I should qualify this by stating that I'm not suggesting we delete existing manually-generated pattern files; merely that having Catagolue be able to do this would allow us to not have to worry about future ones, except for Patterns of Unusual Size™ etc).
::::::: I do like the idea of not having to upload pattern files for all the files that have apgcodes, theoretically.  On the other hand, the fastest way to get a picture of an object into an article about the object, these days, is to upload RLE to the RLE namespace and add the relevant infobox.  And the RLE-scraper script can then (with just a small amount of help from me) magically grab all the new pname-linked RLE files, put them in a ZIP archive, and send it to Nathaniel to put on the server -- once every month or three as needed.
::::::: I can add automatic conversion of everything to .cells format and add those files automatically to the ZIP file sent to Nathaniel.  So would we gain much by being able to link to a pattern file on Catagolue?  Not sure.
::::::: I'm starting to work on support in the script for {pname}_synth files as well, so that if people put something into RLE:{pname}_synth, it will get turned into {pname}_synth.rle and get thrown in the ZIP file along with everything else.  But here I'm really not sure if that's the best thing to do, at least for a lot of cases.  There are a couple of big sources of synthesis RLE:  chris_c's script (via Catagolue these days) and Mark Niemiec's database.  Maybe Catagolue could serve up RLE without the LifeViewer, or maybe the "Glider synthesis" section in the infobox could be tweaked so that it links to the LifeViewer page on Catagolue showing the relevant synthesis?
::::::: Then for anything that has a synthesis in Mark Niemiec's database, we just need to collect the identifiers and add them to the infobox template (somehow -- suggestions gratefully accepted), then set up the template to link directly to those files.  For example, the path to the beehive synthesis is [http://www.conwaylife.com/ref/mniemiec/0/6hv.rle "0/6hv.rle"], and the path to a beehive on cap synthesis is [http://www.conwaylife.com/ref/mniemiec/14/14-50.rle "14/14-50.rle"].
::::::: That way RLE:{pname}_synth pages would only have to be collected and uploaded for patterns where no synthesis is available on Catagolue or on Mark's site.
::::::: Something like this would prevent us from continuing to upload so many no-value-added syntheses that are just copies of something from elsewhere (and that won't get updated when Mark's database or chris_c's script gets updated).
::::::: I don't think we should necessarily start the big project of supplying all those Niemiec-database identifiers to the infoboxes ''quite'' yet:  Mark has said recently that he's close to rolling out a new version of his site, and it might make sense to wait and make sure nothing major has changed in the new version.  But we could try the experiment of getting everything set up correctly for, say, the [http://www.conwaylife.com/ref/mniemiec/p1-12.htm 12-bit still lifes] that were added to the LifeWiki recently.
::::::: TL;DR: Templates need tweaking. Anyone interested in trying some experimental additions to the Glider Synthesis section? [[User:Dvgrn|Dvgrn]] ([[User talk:Dvgrn|talk]]) 03:59, 10 January 2019 (UTC)
:::::::: It seems that MDN has helpfully highlighted objects in a different colour in his version of Extended RLE. I think this means that the .rle files can be automatically reverse-engineered to deduce the objects that are produced, enabling the generation of an apgcode-to-mniemiec-url mapping. [[User:Calcyman|Calcyman]] ([[User talk:Calcyman|talk]]) 16:10, 10 January 2019 (UTC)
::::::::: That might work, though it might also be more trouble than it's worth to get the automated reverse-engineering working.  It's not just the target objects that get the "x" color, but also the intermediate stable stages -- and there are a lot of those in some cases, when multiple construction paths are documented.  If you censused all the 'x' objects, the most common object is probably the target object -- or the one that's farthest to the right, I suppose.
::::::::: But that may not be necessary.  Mark sent an email to me over half a year ago saying that he had put together "...vastly improved search pages [which] should include everything necessary to perform searches... pattern lists, statistics for each pattern, and links to other sites (like Catagolue, Pentadecathlon, David Eppstein's Glider Repository, and LifeWiki)".  So it's possible that Niemiec Database Mark 2 (heh) might provide a list of the required apgcodes with no need for reverse engineering.
::::::::: We'll see when the time comes, I guess.  At worst, generating that lookup table manually or semi-manually would be a one-time effort -- painful, but finite. [[User:Dvgrn|Dvgrn]] ([[User talk:Dvgrn|talk]]) 21:44, 10 January 2019 (UTC)
== The list(s) of rules investigated on Catagolue ==
Short version: these have increasingly become a burden to maintain on-wiki, and with Catagolue now having [https://catagolue.appspot.com/rules its own endpoint] providing an overview over and an exhaustive list of all rules searched, they're largely irrelevant now. (The on-wiki list was only started because Catagolue didn't provide one at the time.)
So I'm giving up maintainership of these. If anyone wants to take over, please do! [[User:Apple Bottom|Apple Bottom]] ([[User talk:Apple Bottom|talk]]) 17:00, 7 January 2019 (UTC)
: If we are to retire the [[List of rules investigated on Catagolue]], how should we do this? Add a notice at the top saying:
: "This page is no longer actively maintained in favour of the [https://catagolue.appspot.com/rules equivalent Catagolue page]."
: or words to that effect?
: It still might be a good idea to actually keep the information in the LifeWiki, because Catagolue occasionally has outages when the daily quota has been exceeded, whereas conwaylife.com tends to be permanently accessible. [[User:Calcyman|Calcyman]] ([[User talk:Calcyman|talk]]) 18:26, 7 January 2019 (UTC)
:: The wiki page does have some advantages over the Catagolue one, such as listing the rule integers for outer-totalistic rules and being easier to edit (e.g. adding names for new rules). [[User:77topaz|77topaz]] ([[User talk:77topaz|talk]]) 23:51, 7 January 2019 (UTC)
== Time for a consensus decision on pnames? ==
I've run the RLE-scraper script to collect new RLE files for a bulk upload. The script found 321 new RLE files. Before I send them to Nathaniel, it looks like I'll be doing some more standardization, especially involving pnames.
The [[LifeWiki:Pattern_pages|guidelines for creating pnames]] say very clearly:
<pre>pname (required) The name of the pattern being described, but converted to lowercase and with all non-alphanumeric characters and spaces removed.</pre>
This has worked fine for us for the great majority of cases, but there are two related cases where blindly following that rule creates not-very-good pnames:
# '''apgcode-based names''', where removing the underscore can sometimes concatenate two strings of digits. For example, according to the rule, [[Xs15_3lkia4z32]] is theoretically supposed to have a pname of "xs153lkia4z32", which reads as if it's a 153-bit still life.  Underscores are confusing in article names because MediaWiki turns right around and renders the name without an underscore. But they do seem to work fine, and they're necessary in other article names, anyway -- raw RLE "pname_synth" synthesis files need them.
# '''patterns named after a Niemiec or pentadecathlon.com ID''', where removing a period causes similar problems with readability.  Examples:
* [[37P7.1]], created by Sokwe in 2009 with a pname of "37p7.1" -- including the period. Another similar case is [http://conwaylife.com/w/index.php?title=37P10.1&diff=13437&oldid=13432 37P10.1], where Sokwe changed the pname from Nathaniel's original "37p101" to "37p10.1", back in 2010.
* [[38P11.1]], with a pname of "38p111". Periods in filenames are definitely annoying because the part after the period can look like a file extension... but I think "38p11.1" would really be better here.
* Several patterns with pnames created by Entity Valkyrie recently: 14p2.1, 14p2.3, 14p2.4, 28p7.3, 28p7.3bumperbouncer, 28p7.3eatingss, 31.4, 33p3.1, 33p3.1bumper, 33p3.1eatingss, 33p3.1reactions, 34P6.1.
====Capitalization Bad====
The last pname in that list is also nonstandard due to capitalization, but that's a separate problem. The full list of capitalized pnames is 35P12, 53P13, 55P10, 113P18, BF20H, BFx59Hinjector, FMHEB, Gtolwss.rle, L112functions, L156reactions, L156variants, L200, Lightspeedcrawler, P5HWV, P58toadflipper, PT8P, PT9B, PT38P -- again all by Entity Valkyrie, I think.  I'll definitely have to go through and fix all of these, just because they're dangerous to cross-platform uses of the pattern collection:  "35P12.rle" will overwrite "35p12.rle" on a Windows operating system, but not on Linux.  And LifeViewer fails to find "RLE:35P12" when told given "pname = 35p12", because the LifeWiki's filesystem is case-sensitive.  So I think the no-capitalization part of the pname guidelines should continue to be very carefully enforced.
====Periods Not So Bad====
However, given the long precedent for pnames occasionally including periods, I'm not planning to change any of Entity Valkyrie's pnames if a period is the only non-standard part. Should probably do something about "31.4", but the rest seem okay.
-- Anyone know where the ".4" comes from in "31.4", by the way?  The problem with calling the thing just plain "Snark catalyst" is that there are several workable Snark catalysts. 31.4 is one of the two most common ones, but it's not exactly "the" Snark catalyst.  But no other common name has caught on. ('''Bellman Zero''', anyone? '''Catalyst B0'''?  '''31.4''' seems better than either of those.)
====Summary questions====
TL;DR:  Does anyone object if I adjust the pname guidelines to say that periods are okay, but "only where necessary", or something along those lines?  And also say that underscores are okay only in apgcode pnames and raw-RLE _synth articles?  Underscores are a minor nightmare, because MediaWiki automatically converts them into spaces, and pnames really aren't supposed to include spaces.  I'm reasonably sure that that underscore-to-space conversion is bound to cause coding difficulties somewhere sometime.  But unless someone wants to recommend consistently using periods in place of underscores in apgcode pnames, I just don't see any good alternative.
Comments, suggestions, disagreements? Please post 'em here! [[User:Dvgrn|Dvgrn]] ([[User talk:Dvgrn|talk]]) 17:56, 15 January 2019 (UTC)
: Also, not sure if anyone will find this note here, but gmc_nxtman's recent series of synthesis postings made for a good test case for reworking several pages recently updated by AwesoMan3000. It's been different changes for every article, but it tends to take a lot of fiddly adjustments to synchronize the pname, RLE, synthesis RLE, LifeViewer config, and any files already uploaded to the LifeWiki server.
: I've done half a dozen articles for starters:  [[very long snake]], [[trans-block on long hook]], [[integral with tub]], [[eater head siamese eater tail]], [[cis-block on long hook]], and [[aircraft carrier with feather]].  LifeViewer generally Just Works once there's a raw RLE article with the right pname, but the images come out too small by default, so I've been adding viewerconfig '''THUMBSIZE 2'''. This should probably be a default added to the template, with SUPPRESS, except I don't know if that will change the looks of a lot of existing articles).
: This leaves [[boat with long tail]], [[beehive with nine]], [[broken snake]], [[cis-boat with nine]], [[eater bridge eater]], [[long boat tie ship]], [[long shillelagh]], [[ortho-loaf on table]], [[snake siamese snake]], [[snake with feather]], [[snorkel loop]], [[trans-boat on table]], [[very long shillelagh]], and [[sesquihat]] that still need editing to add the latest syntheses. I could definitely use some help with updating these:
# decide whether the pname should change to match the article name
# new synthesis copied from '''Talk:{name}''' to '''RLE:{pname}_synth''', either updating or replacing any RLE that's currently there
# fix "synthesis = {n}" in infobox
# add to infobox: <pre>viewerconfig    = [[ THUMBSIZE 2 ]]</pre>
# remove Life1.05 and Life1.06 lines (optional)
# double-check that article name matches article text and Catagolue link -- there's often something wrong there
# add '''RLE:{pname}''' if it's not there already, to make LifeViewer show up instead of an old image file
: A couple other tasks are admin-only --
# If pname has been changed, delete {old pname}*.rle from LifeWiki server to keep things in synch
# If Life1.05 and Life1.06 lines have been removed, delete corresponding files from server. (I'm leaving the "plaintext" (.cells) links, because I'm hoping to generate those automatically for all sub-64x64 patterns currently on the LifeWiki.)
# When a good break point is reached, re-run the auto-upload script and collect all the new pattern and synthesis RLE text into a ZIP file for bulk upload.
: Here again, I'm leaving some broken links to pattern or synth files, which I'm planning to fix fairly soon (by putting the files back in place using the auto-upload script).
: This is the kind of project where I'm very unlikely to get everything exactly right. Independent reviews of all this stuff will be greatly appreciated, and just let me know what I've done wrong so far. [[User:Dvgrn|Dvgrn]] ([[User talk:Dvgrn|talk]]) 12:33, 21 January 2019 (UTC)
==Special pages broken?==
I have noticed several oddities in a few of the maintenance pages:
* [[Special:BrokenRedirects|Broken redirects]] claims there is an existing page named [[RLE:lobster]] that redirects to a non-existent page named [[RLE:83p7h1v1]], when the opposite is actually true.
* [[RLE:Loafer]] is listed in various places (such as [[Special:ShortPages|Short pages]]) despite being in the wrong namespace.
* [[ConwayLife.com:About]] is listed as in [[Special:DoubleRedirects|Double redirects]] despite not having any redirects.
* [[Special:FewestRevisions|Pages with the fewest revisions]] isn't even close to being correct.
* [[Special:UncategorizedPages|Uncategorized pages]] is quite incomplete; I noticed the articles [[Ruler]] and [[Alternating rule]] were not listed even before I added categories to them, and these are almost certainly not the only examples.
Anyone know what's up with these? [[User:Ian07|Ian07]] ([[User talk:Ian07|talk]]) 23:35, 15 January 2019 (UTC)
==Rulespace info for eaters, reflectors, conduits, etc.==
So lately I've been busy adding isorule parameters to the infoboxes for various patterns. So far I've stuck with oscillators, spaceships, still lifes, and infinite growth patterns, but I'd also like to expand this to other pattern such as conduits which are a bit more ambiguous. For example, the [[sidesnagger]] works in B/S23, but obviously there are no gliders for it to eat. I'm of the opinion that for these patterns we should show the rules they're actually ''useful'' in, since that's what makes them notable in the first place, (though with the possible exception of [[eater 1]] since it's such a small pattern) but I'd like to hear others' thoughts on this. [[User:Ian07|Ian07]] ([[User talk:Ian07|talk]]) 18:57, 19 January 2019 (UTC)
: My main thought is that all but the very simplest and lowest-step-size conduits are very close to rule-specific -- useful only in B3/S23. Adding a B8 or an S8 might be one of the few isotropic bits where some conduits would survive the rule switch (?). Even if a particular conduit works in another rule, it's only interesting if a large enough group of conduits works in the alternate rule to make it computationally universal.  (That would probably make it construction-universal, too, but only after someone re-did all the single-channel search work to produce new recipe libraries.)
: Anyway, from my point of view the reason why B38/S23 or B3/S238 doesn't get a lot of attention is that there's nothing really new and exciting about those rules to make up for the fact that the rule spec is just that little bit more complicated... pun maybe intended. [[User:Dvgrn|Dvgrn]] ([[User talk:Dvgrn|talk]]) 02:04, 20 January 2019 (UTC)
==[[apgsearch]] and [[Catagolue]]==
Would anyone be opposed to a reorganization of the information on these two articles? I noticed that a lot of the information in the Catagolue article really applies more to apgsearch rather than the site itself, and therefore might be worth moving. Such a change would be particularly easy to revert if need be, but I'd rather not go through the effort if that's the case, so I'd also like some feedback about that. [[User:Ian07|Ian07]] ([[User talk:Ian07|talk]]) 23:32, 19 January 2019 (UTC)
: No opposition here. I wouldn't think anyone would be likely to revert changes to those articles, just the usual quick review to see if the changes happen to serve as a reminder of anything else that should be added. [[User:Dvgrn|Dvgrn]] ([[User talk:Dvgrn|talk]]) 02:04, 20 January 2019 (UTC)

Revision as of 12:33, 21 January 2019

Taka Tiki Break

Welcome, one and all, to the Tiki bar! This page is used to discuss the technical issues, policies, and operations of the LifeWiki. Or just sit down, relax, and enjoy a cocktail.

Welcome to the Tiki bar

Wikipedia has the Village pump, Wiktionary has the Beer parlour, but the LifeWiki's lacked a central page for discussion so far other than User talk:Nathaniel. So I took the liberty to create the Tiki bar to facilitate discussion in a friendly and relaxed atmosphere. Welcome! Apple Bottom (talk) 11:09, 13 June 2016 (UTC)

Archived discussions

Note: active discussions are never archived while still active.

Conduits and converters

I'm gradually gathering the necessary courage to tackle the new Life Lexicon items that start with "P". Looks like one of the big things I should do is to carefully figure out how to make proper use of Template:Reflector, but in this modern LifeViewer age I don't think I agree with the part about "The image in this infobox should NOT include the glider that is to be reflected...".

Seems to me these template recommendations should be updated to say something like "The image in this infobox should include the glider that is to be reflected -- optionally, two input gliders separated by the mechanism's minimum recovery time, and an output glider if that allows a smoother animation. However, the bounding box and population count should be calculated with these gliders removed."

It would actually be pretty annoying to provide RLE of a reflector and not at least show where the input is supposed to go. When copying and pasting one of these to use in a larger construction, it's usually pretty handy to have some kind of marker for where the the input goes and where the output comes from -- thus the ghost Herschels in recently added Herschel conduits.

[Ideally the marks are state-4 LifeHistory so you don't have to edit them out after pasting -- but we should probably stick with simple 2-state Life patterns on the LifeWiki and not open the LifeHistory can of worms.]

... And we can probably get rid of Template:Reflector/Doc while we're at it, no?

Before I start this I'll definitely undertake to review all the existing converters and reflectors and conduits -- there are a bunch with raw RLE and/or uploaded pattern files missing. That's relatively easy to fix, once we have an official decision about whether and how to show inputs and outputs. I'm currently puzzled by the mysterious Template:ConduitInput and Template:ConverterInputOutput. Not that that's surprising -- I'm easily confused by all this wiki template trickery. Dvgrn (talk) 15:26, 8 May 2018 (UTC)

I agree that reflector patterns should include the input glider. I'm the one who originally wrote that they shouldn't, and I'm not really sure why anymore. It can't have been very good reasoning, because I completely disagree with it now.
~Sokwe 07:08, 9 May 2018 (UTC)
Template:Reflector/Doc also asks users not to put animated images on pages, instead suggesting that one should "consider using a static image of the reflector with a caption that links to the animation". I think this does not match the general current LifeWiki practice regarding animated images, or generally animated content. Should we reconsider? Apple Bottom (talk) 05:25, 7 June 2018 (UTC)
It does seem to me that we have a developing consensus that LifeViewer-based illustrations are a good way to go. There are quite a few Help documents and templates that were written long before the advent of LifeViewer. I'd love to have the Help actually explain to a new user how exactly to add RLE to the RLE namespace, how to get that RLE to show up in an infobox or an embedded viewer, how to adjust the LifeViewer config so that animations (if any) look good, etc. It will take a while to get all the docs updated, no doubt. My edit yesterday was just a first attempt to start chipping away at the problem. Dvgrn (talk) 14:48, 7 June 2018 (UTC)
Definitely agree! Unfortunately writing documentation is one of things I'm hopelessly.. well, hopeless at. ;) Apple Bottom (talk) 10:33, 9 June 2018 (UTC)

Lexicon tags

Many of our articles (glossary, in particular) are based on, or at least synced with, Life Lexicon content. This creates a need to update these articles when the Lexicon changes.

Some of that has been handled in an ad-hoc manner on my userpage, but the process is fairly involved: look at the project page, find an article to work on, make sure it needs to be worked on, make the necessary edits, make the necessary changes to the project page to reflect the fact you edited the article.

It's also not easily found by newcomers who may want to help out. (OK, I'll admit, there likely aren't droves of eager newcomers to begin with, but that nonwithstanding, if you don't know said page exists you're not going to find it easily.)

So I was thinking, can't we improve on this? And I just had the idea of tagging articles themselves instead, indicating which version of the Lexicon they correspond to.

The Nethack wiki does something similar; for instance, take a look at their Foodless article, and you'll find that it has an indicator at the top right saying that the page reflects Nethack 3.4.3 (rather than the current 3.6.1), generated by this template.

We could use something similar. There wouldn't necessarily have to be a visible indicator (though there could be); at the very least, though, pages could be tracked in appropriate categories, and we'd know at a glance what needs to be updated (or at least reviewed) and what's current.

This way, all edits would be in one place: review an article and make edits as necessary, and also update the tag to indicate it now reflects a newer Lexicon version. And placing those tracking categories into an appropriate supercategory and placing that in the existing category tree in turn would allow editors interested in helping out find articles in need of review.

There would be two downsides. a) most of the Lexicon doesn't change in each Lexicon release, so we'd have a lot of articles tagged as (say) reflecting v28 when in fact they're also current, by virtue of not having changed since v28. And b) we wouldn't easily be able to see which articles are missing from the wiki entirely.

I still feel that this would be an improvement though, and there's no reason we couldn't combine these tags with a manually-curated project page to get the best of both worlds.

Also, re: downside a) specifically, I think this could be dealt with by also having an indicator on the wiki saying which Lexicon release is current; pages that haven't been tagged as reflecting the current version would then display a gentle, unobtrusive note, and anyone viewing such a page could quickly check that it does indeed match the current Lexicon release, and update the tag if so.

Thoughts? Apple Bottom (talk) 07:49, 18 July 2018 (UTC)

This seems like a fine idea to me. As far as downside a) goes, I think that the last year of Life Lexicon updates is highly unusual, since it involved catching up after over a decade of no maintenance at all.
The standard editing methodology for new Lexicon releases is to maintain a Changes section at the top of the raw Lexicon text file, carefully listing every "added" or "edited" entry since the last release, by name. Nobody is supposed to edit a Lexicon definition without updating the change log. This should make it trivial to find missing articles, and hopefully should also allow an easy update to the tags. Every Lexicon entry that's not listed in the change log can be automatically bumped to the latest lexicon release.
That's a lot of small changes to a lot of articles with every Lexicon release, though. Does it make sense to have the default Lexicon tag be just Template:LexiconLatest or some such, with a template to display on the page whatever the latest Lexicon release number actually is?
Then, for the next Lexicon release (30), we can just update the (relatively small) list of changed definitions to say "Release 29" -- and then after each definition is reviewed and patched, the tag is updated at the same time, back to Template:LexiconLatest? Dvgrn (talk) 22:15, 18 July 2018 (UTC)
I suppose that would work, but it's pretty much the opposite of what I was trying to accomplish. ;) I was thinking of this as a status checkbox of sorts where editors would check off that yes, this article has been reviewed for Life Lexicon release 30 or 50 or whatever, and any articles that lacked that virtual checkmark would automatically be herded and available for review and/or updating, as necessary.
Having a "LexiconLatest" tag instead would mean checkmarks that check themselves, by default, and that we'd then have to go and un-check. That's not so different from the current approach, with my TODO page.
But you raise a good point. We have a log of Life Lexicon changes, and once we're actually caught up with the Lexicon in general all we'd have to do is keep an eye on those. Hmmm.
Here's a thought, admittedly a rather complicated one. How about we do both? That is to say, how about a tag template that has both an explicit parameter and uses a default "low watermark", displaying the higher of both?
The explicit parameter would be used by editors to indicate that a page has been reviewed/updated to reflect a certain Lexicon release; the "low watermark" (kept in a template of its own) would be updated by us whenever we're sure that every Lexicon-related entry reflects a certain Lexicon version.
For instance, assume that all our articles conform to Lexicon release 30. Suppose that release 31 comes out now. We then go through the changelog, edit all articles that need updating, and after that's done, we conclude that no further changes are necessary, and bump the "low watermark" to 31, thus causing all articles (that reference the Lexicon) to declare that they match release 31.
One advantage of this would be that we'd still see when an article was last explicitly reviewed. For instance, an article might say it reflects Lexicon release 31, but the "version=" parameter might still say it was last reviewed for 28. If nothing else, this would make it easier to spot articles that haven't been reviewed for a long time and where discrepancies might've crept in.
Another idea: we've already got Template:LinkLexicon to link to Lexicon entries. We could repurpose this to also additionally display a tag, which would save us the need to edit 831 articles just to add a tagging template.
Apple Bottom (talk) 07:50, 19 July 2018 (UTC)
This all seems reasonable to me -- especially sneaking a displayed tag into Template:LinkLexicon. Now that Golly 3.2 and Release 29 are safely out the door, I'm sorta kinda planning to get back to work on the LifeWiki ToDo list for Lexicon updates, with the intention of getting everything up to date eventually -- hopefully well before Release 30 comes along to confuse things any more. We already have some kind of a tracking system set up for Release 28 and 29, so maybe it makes sense to keep using that, and design the new template/tag system to really come into use once everything has been updated to Release 29.
So in early 2019, if we end up with a list of say fifty articles that have changes for Lexicon Release 30, my thought would be to update just those fifty articles to specifically say "Release 29" (however we decide to do that exactly -- maybe with a template saying "this article is out of date, please help out by updating it"?). Then bump up the "low watermark" to 30 right away. As articles get updated, the "please help" template can get removed again with the same edit. Does that work? Dvgrn (talk) 18:49, 19 July 2018 (UTC)
OK, cool. :) Good to hear you'll have some time to devote to the Lexicon-to-LifeWiki TODO list. I'm not able to put in much effort there myself --- too much studying, too many exams. Ah well.
Re: marking e.g. fifty articles as needing updates and everything as conforming to e.g. release 30 by default --- that would be a lot easier if we had Lua scripting available! MediaWiki's templates only go so far and aren't really meant for pushing lots of structured data around.
Our options there would include at least the following:
  1. Manually edit each of those 50 articles (e.g. by setting an extra template parameter) to override the "low watermark". Not ideal --- we might as well just edit those 50 articles to update them if we're already editing them anyway.
  2. Provide a global "kill switch" for the low watermark that, when set, causes the low watermark to be ignored. Pages explicitely listing a conforming Lexicon release would then display that instead, so those 50 "release 29" articles would show up in the right category, etc. Also not ideal --- there might be many other articles that would also have the explicit "reviewed for release 29" tag, or older tags at that, which would NOT need to be updated.
  3. Keep a list of those 50 articles, and rig the template to display a notice if the title of the transcluding page happens to be on that list. Also not ideal --- we'd have to curate that list, and as I said, MediaWiki templates aren't really meant for this sort of thing.
Maybe there's another solution I'm not seeing, though.
That said I also have a feeling we're trying to overengineer the solution, though, or perhaps attacking the problem from the wrong angle. After all, what do we want to do? Keep the LifeWiki current as far as Lexicon content goes. How do we achieve that? By importing Lexicon as necessary, and (once done) keeping an eye on changes made to the Lexicon and mirroring them on the wiki (again, as necessary). And how do we do that? By rolling up our sleeves and working on it. Fancy templates and tagging nonwithstanding we won't get anywhere if we don't just jump in and do it.
(And by "we", I mean whoever's willing to do that job.)
Apple Bottom (talk) 18:16, 20 July 2018 (UTC)
Re: overengineering... yeah, offhand I don't see a better solution than the first one: manually edit 50 articles, copying and pasting the same "stub"-like template marker in as a header. This is a bit tedious, but that's what multiple browser tabs are for, and it can be done pretty easily in half an hour or so. The idea is that we can make a little bit of effort to spread the update work around. (Here "we" means the small group of people who have done the work so far -- a small group because it's kind of tricky to do everything right, so not many people have figured out all the fiddly details.)
I can add "needs Lexicon update" headers to 50 articles in half an hour, but I sure can't do a careful comparison and repair on 50 articles, especially if it will require adding new illustrations or modifying existing ones. But it seems to me that there's a larger (and growing) population of LifeWiki users who can perfectly well review a particular Lexicon definition when they trip over a "needs Lexicon update" template header begging for help. Often it isn't too hard to find what needs changing, make the required edits, and remove the "needs Lexicon update" tag at the same time.
Every one of these articles that someone picks up and fixes, is one that I don't have to do myself... and in the meantime, a half hour of work has already brought the LifeWiki more up to date, by specifically flagging the fact that there's newer information somewhere else that needs to be integrated into the article. Seems like this might be a good habit to get into, for as long as the Life Lexicon is kept more or less in synch with current reality.
Sound reasonable? And could you have a look at Template:NeedsLexiconUpdate and see if it has everything in it that this plan might need? Dvgrn (talk) 21:02, 20 July 2018 (UTC)
Redirect pages don't need any markers saying they're from a Lexicon entry -- do they? I've been trying to rebuild some momentum by getting the remaining redirects done... Dvgrn (talk) 21:57, 26 July 2018 (UTC)
Cool, good to see this is already progressing. Good job! :) I'm a little less swamped now, so I'll take a look at the Template'n all over the weekend.
No, I don't think redirect pages need markers. I don't consider these "content" in the strictest sense, in either the Lexicon or the LifeWiki --- they're just tools that help people find content.
Apple Bottom (talk) 08:31, 28 July 2018 (UTC)

Object frequency classes

I do apologize for my somewhat extended absence. That said, I had an idea (long ago actually) about adding information on the commonness of objects to pattern infoboxes, using data from Catagolue (specifically, B3/S23/C1).

I don't think saying "this still life is the 1,691th most common object on Catagolue" is useful, of course. What I'm proposing instead is the frequency class, defined as follows: a pattern is in frequency class X if the most commonly-occurring object (the block, in this case) is 2X times more common. X need not be an integer; to strike a balance I'd suggest using one decimal digit.

Let me give an example. The twin hat has appeared 240,372,408 times on Catagolue (as of this morning), whereas the block has appeared 71,146,901,659,666 times. So the block is approximately 295,986 ≈ 218.17517 times more common, and the twin hat's frequency class is 18.2, rounded to one decimal digit.

I think this is a fairly intuitive way of capturing commonness. An additional nice property is that if an object has occurred sufficiently often, its frequency class is unlikely to change much, if at all; this is true even for objects whose commonness is very similar and who might switch ranks regularly, with one or the other having occurred more often at any given moment. So once this information's added, we wouldn't need to edit it much, if at all ever.

Like I said, only sufficiently common objects should have this information added; there's too much uncertainty about the frequency class of an object that has only appeared once, say. I unfortunately lack the statistical background to suggest a good cut-off value ("objects should only have this information in their infoboxes if they have occurred at least n times"), but unless there are objections I'll add this, or at least do the necessary template work.

...heck, I'll just go ahead and do it, it's been a while since I've edited anything here. If anyone thinks that this is a load of bull, please just speak up and say so. :) Apple Bottom (talk) 17:56, 1 September 2018 (UTC)

(P.S. --- although the block is the most common in B3/S23/C1, it isn't necessarily for other B3/S23 censuses; in some symmetries, the blinker is more common.)

Replying to myself, I've started doing this; there is a new template parameter, fc=, currently only for Template:Stilllife, Template:Oscillator, Template:Spaceship and Template:Puffer (no other types of object have appeared on Catagolue anyway). I've also added a short glossary entry at Frequency class, and added frequency calss data to a couple of object infoboxes, including all with FC ≤ 10.0. The script used to generate the necessary data from Catagolue's textcensus is this:
#!/usr/bin/perl

# usage eg.:
# perl ../frequencyclasses.pl b3s23.C1.txt >frequencyclasses.txt

use Modern::Perl '2016';

# only patterns with more than $cutoff occurrences should be considered.
# mark all other patterns with an asterisk.
our $cutoff = 10;

# throw away header line
<>;

my %objects = ();
my $mostcommon = -1;
my $mostcommoncode = "";
while(<>) {
    chomp;
    next unless m/^"([^"]*)","(\d+)"$/;

    my ($apgcode, $count) = ($1, $2);
    $objects{$apgcode} = $count;

    if($count > $mostcommon) {
        $mostcommon = $count;
        $mostcommoncode = $apgcode;
    }
}

my %frequencies = ();
foreach my $apgcode (keys %objects) {
    my $frequencyclass = sprintf("%.1f", (log($mostcommon / $objects{$apgcode})) / log(2));
    $frequencies{$frequencyclass}->{$apgcode} = $objects{$apgcode};
}

foreach my $frequency (sort { $a <=> $b } keys %frequencies) {
    foreach my $apgcode (sort { $frequencies{$frequency}->{$a} <=> $frequencies{$frequency}->{$b} } keys %{ $frequencies{$frequency} }) {
        print "*" if($frequencies{$frequency}->{$apgcode} <= $cutoff);
        say "$frequency\t $apgcode\t $frequencies{$frequency}->{$apgcode}";
    }
}
(I'm sure there's better ways of doing this, but this worked for me.) Apple Bottom (talk) 18:52, 1 September 2018 (UTC)
Another reply to myself --- Goldtiger997 suggested a cut-off value of 10 (non-inclusive). This strikes me as sensible. So unless there's objections, how about we run with this, and only add frequency class information to objects having appeared more than 10 times?
Also --- right now the information is Catagolue-specific, which is sensible but still somewhat arbitrary; if we want to include more information later (e.g. from Achim's, Andrzej's and Nathaniel's censuses, or from whatever future censuses people may come up with), we can easily adjust the infoboxes to include a new "Commonness" section, and re-interpret fc= as "frequency in [c]atagolue". Apple Bottom (talk) 09:08, 2 September 2018 (UTC)
Sounds good. However, I already broke my "suggestion" of the 10 occurence cut-off twice; for the Coe ship and Achim's p8. Is it worth removing the fc parameter for those two articles, or should they just be left as they are? Goldtiger997 (talk) 09:26, 2 September 2018 (UTC)
I think we can grandfather those in --- would be good if you could keep an eye on them in case the information changes, of course, but it's just two articles, so that should be fine. I've also added the cutoff of >10 to the script above; patterns not reaching that cutoff are marked with an asterisk. Apple Bottom (talk) 09:35, 2 September 2018 (UTC)

Infobox vs. EmbedViewer

All this interminable Life Lexicon import work has been leading me to believe that there are two classes of articles that can use LifeViewer animations. There are the named patterns, where if you say "Pattern X" there's really only one likely Pattern X that you could be referring to. These get put into an infobox category, with appropriate statistics collected and so forth. The most recent example of this kind of imported Lexicon article is line crosser.

The other class of article is for a term that might refer to a variety of different patterns, so that there are various examples but no specific example should really be considered to be the one canonical one. In these cases I've been using an embedded viewer but haven't been bothering with an infobox. The most recent examples along these lines are line-cutting reaction and line-mending reaction. I like the way these are turning out, but am curious to hear if anyone thinks that these should also be infoboxed somehow, or if anything else should be added as standard practice. Dvgrn (talk) 09:45, 12 September 2018 (UTC)

I think this is eminently sensible. Off the top of my head, aren't there a few articles that have infoboxes despite being about a family of patterns rather than a specific individual one? (Or patterns with variants, anyway --- the bee shuttles come to mind there.) I've never been quite sure how to handle those, though that's not limited to LifeViewer and embedded patterns: the same goes for other infobox'ed information, such as bounding box, population etc. Apple Bottom (talk) 07:54, 13 September 2018 (UTC)
Yup, those are the ones where I find the infoboxes to be not-helpful. Would suggest in those cases maybe just using an embedded viewer to show one of the family, or maybe a small stamp collection would be better. The most recent example I dealt with was HFx58B from the Life Lexicon. Rather than pick a variant, and/or leave out perfectly good information that the Lexicon had, I just threw caution to the winds and put both patterns in the same infobox, but picked the older variant to do the infobox stats about. Probably this will puzzle somebody sometime, but sometimes Life can be confusing... Dvgrn (talk) 11:20, 13 September 2018 (UTC)

Semi-automated collection of raw RLE

I now have a completed Python script -- working on my system, at least! -- that goes through all articles in the main namespace looking for pname definitions in infoboxes and embedded viewers. If it finds any defined pnames, it checks www.conwaylife.com/patterns/{pname}.rle; if a Not Found error comes back, it creates an appropriate file using RLE from conwaylife.com/w/index.php?title=RLE:{pname}&action=edit, in a reasonably standard format including pname, discoverer and discovery year if available, and links to the relevant article and the RLE file itself.

On the last pass the script found 190 missing RLE files in the main namespace. These have now been uploaded to the server and added to all.zip. You can sort the contents of all.zip by date to see the new additions. Since this was a mostly automated process, the script may have picked up a few patterns that shouldn't really be part of the collection. If anyone wants to do a quick independent review, I'd appreciate it!

I think this will make the process of getting RLE uploaded to the server a lot easier for non-admins. If raw RLE is created in the RLE: namespace, and is used in a pattern infobox or an embedded viewer, then it will make it to the all.zip collection eventually. To give time for new raw-RLE additions to be peer-reviewed, I would think this script would only be run quarterly or so, with the resulting new RLE files sent as a ZIP file to Nathaniel to do a bulk upload to the server. There shouldn't be any problem with good files getting overwritten with bad ones, since the script only generates an RLE file if no existing file is found.

It occurs to me that another semi-automated survey might be looking for articles with nofile=true, that do in fact now have raw RLE and/or uploaded RLE files. I'll try adjusting the script to include any such files it finds in its final report. It's a little trickier to automatically update all such "nofile = true" to say "rle = true" instead -- it's doable, but it needs a different kind of automation.

Thoughts, suggestions, worries, bug reports? I'll add a link here to the RLE-scraper Python script when I've made it available on GitHub. Dvgrn (talk) 14:21, 23 October 2018 (UTC)

Link! Dvgrn (talk) 00:58, 26 October 2018 (UTC)
The next item on the RLE-scraper script TODO list will be to check for raw-RLE {pname}_synth files, and upload them if they aren't already there. Longer term, the script can make a report of any differences it finds between files already on the server and the current contents of the RLE: namespace. Probably best not to upload changed files automatically -- it seems worth having a human review any changes, and take the time to revert any changes that aren't approved for upload to the pattern collection. Dvgrn (talk) 10:46, 25 October 2018 (UTC)

Oscillator mods

I noticed that all the recently created oscillator pages from the latest Lexicon update (example: p29 pentadecathlon hassler) have their mods listed along with their periods even if they're equal. Is it agreed upon that this should be the case? Because if so I can go through the unknown mod list later and add in all the mods if no one objects to it. Ian07 (talk) 14:53, 28 October 2018 (UTC)

I can't claim to have made a really deliberate decision to include the mod when it's the same as the period -- I was just blindly filling in values in the oscillator template I was using. I don't have any objection to listing both period and mod, but let's see if anyone else has a different opinion. Many thanks for all the cleanup work you've been doing recently, by the way! Dvgrn (talk) 17:04, 1 November 2018 (UTC)
Just to chime in --- I think listing both the period and the mod is valuable even if they match. Otherwise, if an infobox doesn't have mod information, a user won't know if that's because we haven't filled in the info or because it's the same as the period.
And I agree, thanks are due to Ian07 for all the clean-ups and other work. MediaWiki has barn (heh) stars; do we have something similar? Maybe we should. Apple Bottom (talk) 21:01, 3 November 2018 (UTC)

Conduit orientations and ghost Herschels

Quick question, y'all: is "T" a standard designation for a "turned" conduit output orientation? I'm asking because of this edit to Template:Conduit --- I lack the expertise whether this is standard terminology or not.

If it is, it should be documented in the template. Apple Bottom (talk) 21:27, 3 November 2018 (UTC)

Similarly, this edit to Template:Conduit/Doc doesn't seen to be adding anything useful to the page. As I've just started here I'm a bit hesitant to go around rolling back changes to Template pages though. Wildmyron (talk) 03:31, 12 December 2018 (UTC)
Thanks for pointing these out. I rolled back the "gray eaters" edit. The "T" option for symmetrical signals is a little more complicated. It is definitely something that I tried as a classifier a few years ago, but it never really caught on. Now just in the last few days Freywa has done a new update of the Elementary Conduits Collection with a better idea than "T", so I'll have a go at documenting that instead.
Another conduit-related topic, for @Sokwe and anyone else interested: it looks like there isn't universal agreement about whether ghost Herschels are a good idea or not, in conduit patterns. I recklessly ported them in from the Life Lexicon, and I'm certainly going to keep them there because they're so darn useful. You can copy conduits out of the Lexicon and string them together immediately. My theory is that the same is true of patterns on the LifeWiki, and therefore that there should be a ghost eater in Syringe 2, to match the one in syringe and the dozens of other instances in various recently added conduits.
I can see how this looks a little bit like pollution, though, especially if it's extended to other types of outputs (which isn't very clean or easy to do in general, so let's not do that). I've been careful to link to ghost Herschel every time I use one (I think), so they shouldn't be mysterious for long -- and I'm seeing them getting a fair amount of use as markers in constructed patterns lately. Anyone want to contribute other opinions on this? Dvgrn (talk) 13:33, 13 December 2018 (UTC)

Categories and User Pages

Entity Valkyrie has been using pattern templates on user pages, causing those user pages to show up in category:patterns and other categories. It is my opinion that patterns on user pages should not be included in the categories. One way to fix this would be to detect the namespace of the page that the pattern template is being used on. Something like the following:

{{#ifeq:{{NAMESPACE}}|User||[[category:patterns]]}}

This would categorize a non-user page as a pattern, but would do nothing on a user page.

Are there any objections to this proposal? Comments?
~Sokwe 08:11, 2 December 2018 (UTC)

I like that plan, especially if someone else implements it who is less likely than me to break templates. I've been trying to keep user namespace stuff out of the main namespace in general, with fairly good success so far I think -- but I don't usually go in and edit things like categories when a page moves from the main namespace to the User namespace. Dvgrn (talk) 11:53, 2 December 2018 (UTC)

Documenting 12-Bit Still Lifes

AwesoMan3000 has been working on a fairly ambitious project to add articles for all of the 12-cell still lifes. Most of them have systemic names, and it looks like negotiations have been at least partially successful about not just making names up when long-standing existing names are available. As of this writing, swimming cap is still an unnecessary neologism for "integral with tub and tail", I believe, and there may be others -- it's hard to keep up, but this is a work in progress with lots of ongoing adjustments. The plan is for it to be complete by the end of the year.

A number of issues have appeared that I'd be interested in trying to get some kind of consensus about:

  1. when a still life is renamed for consistency -- e.g., "X and Y" names have been rapidly changing to "X on Y", as in block on cap -- changing the pname to match the name tends to break lots of links, especially for older 12-bit objects where Life 1.05, Life 1.06, and .cells formats are available. I'd like to suggest that we can take the radical step of dropping support for Life 1.05, 1.06 and .cells formats, and removing the links for those formats whenever we have an opportunity. Is there any disagreement on this? After that, it should be fairly workable to add just RLE:pname.rle and RLE:pname_synth.rle pages wherever necessary. That will eventually fix the remaining broken links (see below).
  2. if the pname is changed, or if the still life in question didn't have an existing article on LifeWiki, AwesoMan3000 has been promising that there is an RLE file, using rle = true in the infobox parameters, and then uploading raw RLE under that pname. If I remember right, the rle = true is currently needed to get the infobox template to notice that LifeViewer can be used to display the pattern, instead of looking for an image file. There isn't actually an uploaded RLE file in the LifeWiki pattern collection, so .rle and _synth.rle links will give Not Found errors for the time being.
  3. Several still life names have been changed due to an alphabetization rule, e.g., barge siamese loaf instead of loaf siamese barge. This poses the same dangers of link breakage as above, fixable in the same way. Or... it also seems workable to change the name of the page but keep the pname the same, at least until replacement raw RLE is uploaded. See Talk:Hat_siamese_vase. This is being done for the moment in several of the "X on Y" pages. Am I missing other possible problems with this brave but potentially foolhardy renaming-for-consistency project?
  4. When no systemic or traditional name is available for a still life, either on Catagolue or in Mark Niemiec's database, it's hard to avoid the temptation to invent new names that no one has ever used before, and most likely no one will ever use and they'll just cause confusion. One way around this would be to make it standard practice to write the article using the apgcode as the systemic name. I don't like this idea all that much, just because the pname would have to have an underscore in it for readability -- and MediaWiki likes to change underscores to spaces in some contexts, and I'm relatively sure that Murphy's Law will produce some unintended consequences somewhere. Still, it's been tried, and it appears to work okay -- see xs15_3lkia4z32. Does that seem like a reasonable stopgap solution for unnamed objects? We can always move apgcode-named articles later if a better systemic naming convention shows up.

Next Steps

When all raw RLE files have been added in the RLE: namespace for this project, I can re-run the auto-uploader script and make a set of new RLE files for a bulk upload to the LifeWiki server. The script will have to be updated to check for RLE:pname_synth pages as well as RLE:pname files. If the RLE namespace has been populated correctly, this will fix all the remaining broken links. I'll plan to do a round of auto-updating early in 2019.

As before, if a pname.rle or pname_synth.rle pattern has already been uploaded to the LifeWiki server, it won't be overwritten by anything added to the RLE namespace. Eventually the script might check whether the uploaded pattern is the same as the raw RLE, and produce a report of any discrepancies so they can be resolved. Not sure I'll get around to adding that feature in this next update, though.

Comments, concerns, suggestions? Dvgrn (talk) 17:23, 21 December 2018 (UTC)

Also, AwesoMan3000 put the following in checkin comments for beacon on dock:
can we get a tiki bar discussion on whether all disambiguation pages should have (disambiguation) on the end, by the way? i'd rather cut down on unnecessary redirects where possible
So, okay, here's a Tiki Bar discussion. I don't see why the page shouldn't be moved to beacon on dock (disambiguation) to match the original beacon and dock (disambiguation) page. I'm not a big fan of disambiguation pages in general, especially where they can be avoided by making a single page for the most common definition, and linking from that page to other possible meanings under different names. But that doesn't work for cases like this where there isn't a most common definition. Dvgrn (talk) 20:44, 21 December 2018 (UTC)
Re: Made-up names
I agree that we should stick to established names when they're available and not make up new ones. OTOH, when an object does not have any established name (and I really do mean does not, not just "whoever wrote the article couldn't remember it"), I think it's in the best tradition of Life (and life) to give it one.
The question then is whether the LifeWiki is the right place for this. My general feeling is that we're a relatively conservative part of the Life universe: we aim to document, not invent. We're not as anal about it as Wikipedia with its "no-original-research" and "verifiability-not-truth" criteria for inclusion, and neither do we have to be; but just like we're asking people to not create wiki pages for new discoveries of their own but instead share them on e.g. the forums, we can also ask people to not name previously-unnamed on the wiki but instead resort to (again) the forums, etc.
OTOH this may very well lead to a situation where a user wanting to name an unnamed object will simply suggest the name on the forum in an appropriate thread (is there one yet?), wait for a few days/weeks/months, and conclude from the resulting thundering silence that there are no objections -- all in favor --, and go ahead and name the object on the wiki. The end result would be the same, modulo the extra pain of that extra waiting time.
And would this gain us anything?
FWIW, what are we looking to gain from a policy that forbids new names from being put on the wiki first anyway? If we don our documentationist's hats (documentationist --- I hope that's a word!) and take no position on whether a named object is somehow preferable to an unnamed ones, if we merely want to document the Life's community works, passively and from the outside, then yes, this is preferable. If we see ourselves as being an active part of the community, we might find named objects preferable to unnamed ones (and "proper" names preferable to systematic ones, etc).
I think I'm in the latter camp myself. I like named objects.
What I am worried about is that, if we allow new names to be "born" on the LifeWiki, … shall we say, easily excited users might get carried away and go on an editing spree, adding hundreds of new names to previously unnamed objects without any discussion or consensus.
Going off on a tangent for one paragraph, I also think the creative naming in Life stem in no small part from the need to be able to talk about objects. These days we have apgcodes, so we can indeed refer to xs15_3lkia4z32 without it having a proper names.
The best solution I have is what one might call a four-eyes approach, where
  1. New names are valued in principle;
  2. But the person proposing them can't be the one naming the object on the Wiki;
  3. If someone wants to add page for a (notable) object that doesn't yet have a name, they should use the object's apgcode.
Whether names are proposed on the forum or elsewhere is largely irrelevant than, though I'd suggest a dedicated thread on the forum. I'd also suggest a certain cool-down period between the proposal and the wiki edit, so that others have a chance to speak up. (What I'm worried about there is the possibility that two easily-excited users, might join forces after one of them proposed a new name elsewhere, say on Discord: one would propose it, and right after the other would accept it).
Re: pname changes, Life 1.05/... support
No objects to dropping Life 1.05, 1.06, and cells; RLE has clearly emerged as the standard at this point. I don't know if anyone's still using a CA simulator that's not capable of using RLE, but the lack of complaints about new patterns only having RLE files available and nothing else leads me to believe there isn't. I'd say let's remove support for these and see if anyone speaks up. If there's no complaints, we can flip the switch for good.
As for pname changes in general, they're still a pain. I'd suggest that
  1. Whoever changes a pname is responsible for cleaning up the resulting mess; and
  2. pname changes shouldn't be mandatory without a good reason.
What is a "good reason"? A typo, say --- "pname = 2enginecrodership" would obviously require correction. OTOH, if Beluchenko's p51 has "pname = 112p51", I see no problem with that at all. (If anyone else does, they're welcome to change it, provided they clean up afterwards, as per 1. above).
Re: infobox parameters
These should always correctly reflect the status quo. If a pattern doesn't have an RLE file uploaded, the infobox template call shouldn't have "rle=true". Don't lie to the infobox templates! It's the responsibility of those who create articles to make sure all parameters are correct to the best of their knowledge.
Re: RLE files and raw RLE snippets
I'll leave this in your capable hands. :)
Final note
I think having articles on all 12-bit still lifes is a worthwhile endeavor, but -- in light of who's doing this, and admittedly without actually having looked at anything that was created recently -- I'd like to remind everyone who's creating articles that they have a duty to exercise care when doing so. In particular, this means not just starting a whole lot of articles and leaving them in half-broken states.
I'd also like to remind people that they should learn from their mistakes. Nobody's perfect, especially newcomers; the LifeWiki can be difficult to get used to, I imagine. We have the right to make mistakes, but we have the duty to learn from them. He who keeps making the same mistakes time and again is either ignorant, or careless, or both, and those are not qualities a LifeWiki editor should have.
(And that's about all I can think of, off the top of my head.) Apple Bottom (talk) 11:20, 22 December 2018 (UTC)
Thanks for the thoughtful review. A couple of template issues have come up where some expert advice would be helpful:
1) A change made yesterday to Template:Oscillator seems to be not working to populate Category:Periodic_objects_with_minimum_population_3 and all other, um, categories in that category. Is there a misplaced character somewhere in those hideous piles of parentheses, or is the problem actually somewhere else?
2) Is it possible to do Template Magic (TM) to make glider syntheses behave the same way as regular pattern files do? That is, if synthesisRLE = true, then the standard link to an uploaded pattern file should appear in the Glider Synthesis section of the infobox -- but otherwise, if there's a page at RLE:{pname}_synth, then a "raw RLE" link should appear instead.
Here's an example of an update currently in process: we used to have an uploaded eateronboat.rle --
#N Eater on boat
#C A 12-cell still life consisting of an eater 1 and a boat.
#C http://www.conwaylife.com/wiki/index.php?title=Eater_on_boat
-- and an uploaded eateronboat_synth.rle --
#N Eater on boat_synth
#O Mark D. Niemiec
#C Glider synthesis of eater on boat.
#C www.conwaylife.com/wiki/index.php?title=Eater_on_boat
Then along comes AwesoMan3000, who probably thinks that that's a lousy name for this object because it can mean two different things, and isn't really a common use of "on" anyway. So the name gets upgraded to the more specific "boat tie eater tail", and all the appropriate changes get made to the text of the article. And there's a redirect from eater on boat, so damage is limited to other pages that link to this page.
However, AwesoMan3000 can't do anything directly about those uploaded pattern files, or the comments in those files. The simple solution is to just leave the pname the same, still pointing to the old pattern and synth files. That doesn't break any links, but it's a bit confusing because the name doesn't match the article.
If we want to get boattieeatertail.rle and boattieeatertail_synth.rle files uploaded to the LifeWiki server, currently what we can do is add those as raw RLE, and change the pname in the article to boattieeatertail. I've done that experimentally for this example. As a nice side effect, as soon as the pname changes, LifeViewer shows up in the infobox instead of a static image.
That's not too exciting when we're dealing with still lifes... but it does start to get people used to an easier way of getting a copy of the RLE to paste into Golly or wherever -- click to launch LifeViewer, then Ctrl+C.
(If anyone is following along, Beehive_at_loaf is currently an example of the old style of article, with a separately uploaded static image -- no LifeViewer, just because no raw RLE has been uploaded to RLE:beehiveatloaf.)
Okay, so here is where the difficulty shows up!
PROBLEM: as soon as the pname is changed from "eateronboat" to "boattieeatertail", the rle = true and synthesisRLE = true lines in the infobox suddenly turn into lies. They're only temporary lies, because there's a plan to run the auto-uploader script and get the new RLE uploaded. But when RLE is being added for dozens or hundreds of still lifes, there's no way that an admin is going to keep up with running the auto-uploader after every change.
In fact, it just plain doesn't work that way -- the auto-uploader does a scan of the entire main LifeWiki namespace and creates an archive ZIP file of a big pile of changes, intended to be sent to Nathaniel to do a bulk upload to the server. That should really only happen a few times a year, not every time a change happens. So we wait until a batch of changes have been made, then collect and send them.
Theoretically we could remove the rle = true and synthesisRLE = true lines from each article, then add them right back again after the auto-upload is done. But when the timeline of a project is short enough, that looks like a highly irritating waste of time -- basically, Life is too short.
I hope everyone is okay with the idea of saying that these particular deliberate temporary inaccuracies in infobox parameters are actually not "lies", but rather something along the lines of "promises". Dvgrn (talk) 13:40, 22 December 2018 (UTC)
As for the first issue, I checked and double-checked the template and didn't find any syntax errors. What's weird is if you go to the Blinker article you can see that it has the Category:Periodic objects with minimum population 3 category, but the category page says it's empty. This makes me think that there's some sort of glitch with MediaWiki rather than a syntax error, especially considering that, as you pointed out on Discord, LifeWiki is running a pretty outdated version of it. Ian07 (talk) 14:00, 22 December 2018 (UTC)
Can the links to pattern files be directed to a dynamic endpoint that creates the pattern file? For all patterns with apgcodes, for instance, Catagolue could theoretically include endpoints of the form /objfile/<rule>/<apgcode>/<format> (with support for RLE, Life 1.05, and Life 1.06). That would mean that humans only need to provide static RLE files for large and/or aperiodic patterns. Calcyman (talk) 00:36, 23 December 2018 (UTC)
Late reply, but yes, that would definitely be possible. Pattern file links are handled by Template:PatternDownload, which already gets passed the apgcode (if we have it, for a given object), so it'd only be a matter of tweaking that template. And I think doing away with manually-generated and -uploaded RLE files where possible would save us a lot of work.
(I should qualify this by stating that I'm not suggesting we delete existing manually-generated pattern files; merely that having Catagolue be able to do this would allow us to not have to worry about future ones, except for Patterns of Unusual Size™ etc).
I do like the idea of not having to upload pattern files for all the files that have apgcodes, theoretically. On the other hand, the fastest way to get a picture of an object into an article about the object, these days, is to upload RLE to the RLE namespace and add the relevant infobox. And the RLE-scraper script can then (with just a small amount of help from me) magically grab all the new pname-linked RLE files, put them in a ZIP archive, and send it to Nathaniel to put on the server -- once every month or three as needed.
I can add automatic conversion of everything to .cells format and add those files automatically to the ZIP file sent to Nathaniel. So would we gain much by being able to link to a pattern file on Catagolue? Not sure.
I'm starting to work on support in the script for {pname}_synth files as well, so that if people put something into RLE:{pname}_synth, it will get turned into {pname}_synth.rle and get thrown in the ZIP file along with everything else. But here I'm really not sure if that's the best thing to do, at least for a lot of cases. There are a couple of big sources of synthesis RLE: chris_c's script (via Catagolue these days) and Mark Niemiec's database. Maybe Catagolue could serve up RLE without the LifeViewer, or maybe the "Glider synthesis" section in the infobox could be tweaked so that it links to the LifeViewer page on Catagolue showing the relevant synthesis?
Then for anything that has a synthesis in Mark Niemiec's database, we just need to collect the identifiers and add them to the infobox template (somehow -- suggestions gratefully accepted), then set up the template to link directly to those files. For example, the path to the beehive synthesis is "0/6hv.rle", and the path to a beehive on cap synthesis is "14/14-50.rle".
That way RLE:{pname}_synth pages would only have to be collected and uploaded for patterns where no synthesis is available on Catagolue or on Mark's site.
Something like this would prevent us from continuing to upload so many no-value-added syntheses that are just copies of something from elsewhere (and that won't get updated when Mark's database or chris_c's script gets updated).
I don't think we should necessarily start the big project of supplying all those Niemiec-database identifiers to the infoboxes quite yet: Mark has said recently that he's close to rolling out a new version of his site, and it might make sense to wait and make sure nothing major has changed in the new version. But we could try the experiment of getting everything set up correctly for, say, the 12-bit still lifes that were added to the LifeWiki recently.
TL;DR: Templates need tweaking. Anyone interested in trying some experimental additions to the Glider Synthesis section? Dvgrn (talk) 03:59, 10 January 2019 (UTC)
It seems that MDN has helpfully highlighted objects in a different colour in his version of Extended RLE. I think this means that the .rle files can be automatically reverse-engineered to deduce the objects that are produced, enabling the generation of an apgcode-to-mniemiec-url mapping. Calcyman (talk) 16:10, 10 January 2019 (UTC)
That might work, though it might also be more trouble than it's worth to get the automated reverse-engineering working. It's not just the target objects that get the "x" color, but also the intermediate stable stages -- and there are a lot of those in some cases, when multiple construction paths are documented. If you censused all the 'x' objects, the most common object is probably the target object -- or the one that's farthest to the right, I suppose.
But that may not be necessary. Mark sent an email to me over half a year ago saying that he had put together "...vastly improved search pages [which] should include everything necessary to perform searches... pattern lists, statistics for each pattern, and links to other sites (like Catagolue, Pentadecathlon, David Eppstein's Glider Repository, and LifeWiki)". So it's possible that Niemiec Database Mark 2 (heh) might provide a list of the required apgcodes with no need for reverse engineering.
We'll see when the time comes, I guess. At worst, generating that lookup table manually or semi-manually would be a one-time effort -- painful, but finite. Dvgrn (talk) 21:44, 10 January 2019 (UTC)

The list(s) of rules investigated on Catagolue

Short version: these have increasingly become a burden to maintain on-wiki, and with Catagolue now having its own endpoint providing an overview over and an exhaustive list of all rules searched, they're largely irrelevant now. (The on-wiki list was only started because Catagolue didn't provide one at the time.)

So I'm giving up maintainership of these. If anyone wants to take over, please do! Apple Bottom (talk) 17:00, 7 January 2019 (UTC)

If we are to retire the List of rules investigated on Catagolue, how should we do this? Add a notice at the top saying:
"This page is no longer actively maintained in favour of the equivalent Catagolue page."
or words to that effect?
It still might be a good idea to actually keep the information in the LifeWiki, because Catagolue occasionally has outages when the daily quota has been exceeded, whereas conwaylife.com tends to be permanently accessible. Calcyman (talk) 18:26, 7 January 2019 (UTC)
The wiki page does have some advantages over the Catagolue one, such as listing the rule integers for outer-totalistic rules and being easier to edit (e.g. adding names for new rules). 77topaz (talk) 23:51, 7 January 2019 (UTC)

Time for a consensus decision on pnames?

I've run the RLE-scraper script to collect new RLE files for a bulk upload. The script found 321 new RLE files. Before I send them to Nathaniel, it looks like I'll be doing some more standardization, especially involving pnames.

The guidelines for creating pnames say very clearly:

pname	(required) The name of the pattern being described, but converted to lowercase and with all non-alphanumeric characters and spaces removed.

This has worked fine for us for the great majority of cases, but there are two related cases where blindly following that rule creates not-very-good pnames:

  1. apgcode-based names, where removing the underscore can sometimes concatenate two strings of digits. For example, according to the rule, Xs15_3lkia4z32 is theoretically supposed to have a pname of "xs153lkia4z32", which reads as if it's a 153-bit still life. Underscores are confusing in article names because MediaWiki turns right around and renders the name without an underscore. But they do seem to work fine, and they're necessary in other article names, anyway -- raw RLE "pname_synth" synthesis files need them.
  2. patterns named after a Niemiec or pentadecathlon.com ID, where removing a period causes similar problems with readability. Examples:
  • 37P7.1, created by Sokwe in 2009 with a pname of "37p7.1" -- including the period. Another similar case is 37P10.1, where Sokwe changed the pname from Nathaniel's original "37p101" to "37p10.1", back in 2010.
  • 38P11.1, with a pname of "38p111". Periods in filenames are definitely annoying because the part after the period can look like a file extension... but I think "38p11.1" would really be better here.
  • Several patterns with pnames created by Entity Valkyrie recently: 14p2.1, 14p2.3, 14p2.4, 28p7.3, 28p7.3bumperbouncer, 28p7.3eatingss, 31.4, 33p3.1, 33p3.1bumper, 33p3.1eatingss, 33p3.1reactions, 34P6.1.

Capitalization Bad

The last pname in that list is also nonstandard due to capitalization, but that's a separate problem. The full list of capitalized pnames is 35P12, 53P13, 55P10, 113P18, BF20H, BFx59Hinjector, FMHEB, Gtolwss.rle, L112functions, L156reactions, L156variants, L200, Lightspeedcrawler, P5HWV, P58toadflipper, PT8P, PT9B, PT38P -- again all by Entity Valkyrie, I think. I'll definitely have to go through and fix all of these, just because they're dangerous to cross-platform uses of the pattern collection: "35P12.rle" will overwrite "35p12.rle" on a Windows operating system, but not on Linux. And LifeViewer fails to find "RLE:35P12" when told given "pname = 35p12", because the LifeWiki's filesystem is case-sensitive. So I think the no-capitalization part of the pname guidelines should continue to be very carefully enforced.

Periods Not So Bad

However, given the long precedent for pnames occasionally including periods, I'm not planning to change any of Entity Valkyrie's pnames if a period is the only non-standard part. Should probably do something about "31.4", but the rest seem okay.

-- Anyone know where the ".4" comes from in "31.4", by the way? The problem with calling the thing just plain "Snark catalyst" is that there are several workable Snark catalysts. 31.4 is one of the two most common ones, but it's not exactly "the" Snark catalyst. But no other common name has caught on. (Bellman Zero, anyone? Catalyst B0? 31.4 seems better than either of those.)

Summary questions

TL;DR: Does anyone object if I adjust the pname guidelines to say that periods are okay, but "only where necessary", or something along those lines? And also say that underscores are okay only in apgcode pnames and raw-RLE _synth articles? Underscores are a minor nightmare, because MediaWiki automatically converts them into spaces, and pnames really aren't supposed to include spaces. I'm reasonably sure that that underscore-to-space conversion is bound to cause coding difficulties somewhere sometime. But unless someone wants to recommend consistently using periods in place of underscores in apgcode pnames, I just don't see any good alternative.

Comments, suggestions, disagreements? Please post 'em here! Dvgrn (talk) 17:56, 15 January 2019 (UTC)

Also, not sure if anyone will find this note here, but gmc_nxtman's recent series of synthesis postings made for a good test case for reworking several pages recently updated by AwesoMan3000. It's been different changes for every article, but it tends to take a lot of fiddly adjustments to synchronize the pname, RLE, synthesis RLE, LifeViewer config, and any files already uploaded to the LifeWiki server.
I've done half a dozen articles for starters: very long snake, trans-block on long hook, integral with tub, eater head siamese eater tail, cis-block on long hook, and aircraft carrier with feather. LifeViewer generally Just Works once there's a raw RLE article with the right pname, but the images come out too small by default, so I've been adding viewerconfig THUMBSIZE 2. This should probably be a default added to the template, with SUPPRESS, except I don't know if that will change the looks of a lot of existing articles).
This leaves boat with long tail, beehive with nine, broken snake, cis-boat with nine, eater bridge eater, long boat tie ship, long shillelagh, ortho-loaf on table, snake siamese snake, snake with feather, snorkel loop, trans-boat on table, very long shillelagh, and sesquihat that still need editing to add the latest syntheses. I could definitely use some help with updating these:
  1. decide whether the pname should change to match the article name
  2. new synthesis copied from Talk:{name} to RLE:{pname}_synth, either updating or replacing any RLE that's currently there
  3. fix "synthesis = {n}" in infobox
  4. add to infobox:
    viewerconfig     = [[ THUMBSIZE 2 ]]
  5. remove Life1.05 and Life1.06 lines (optional)
  6. double-check that article name matches article text and Catagolue link -- there's often something wrong there
  7. add RLE:{pname} if it's not there already, to make LifeViewer show up instead of an old image file
A couple other tasks are admin-only --
  1. If pname has been changed, delete {old pname}*.rle from LifeWiki server to keep things in synch
  2. If Life1.05 and Life1.06 lines have been removed, delete corresponding files from server. (I'm leaving the "plaintext" (.cells) links, because I'm hoping to generate those automatically for all sub-64x64 patterns currently on the LifeWiki.)
  3. When a good break point is reached, re-run the auto-upload script and collect all the new pattern and synthesis RLE text into a ZIP file for bulk upload.
Here again, I'm leaving some broken links to pattern or synth files, which I'm planning to fix fairly soon (by putting the files back in place using the auto-upload script).
This is the kind of project where I'm very unlikely to get everything exactly right. Independent reviews of all this stuff will be greatly appreciated, and just let me know what I've done wrong so far. Dvgrn (talk) 12:33, 21 January 2019 (UTC)

Special pages broken?

I have noticed several oddities in a few of the maintenance pages:

Anyone know what's up with these? Ian07 (talk) 23:35, 15 January 2019 (UTC)

Rulespace info for eaters, reflectors, conduits, etc.

So lately I've been busy adding isorule parameters to the infoboxes for various patterns. So far I've stuck with oscillators, spaceships, still lifes, and infinite growth patterns, but I'd also like to expand this to other pattern such as conduits which are a bit more ambiguous. For example, the sidesnagger works in B/S23, but obviously there are no gliders for it to eat. I'm of the opinion that for these patterns we should show the rules they're actually useful in, since that's what makes them notable in the first place, (though with the possible exception of eater 1 since it's such a small pattern) but I'd like to hear others' thoughts on this. Ian07 (talk) 18:57, 19 January 2019 (UTC)

My main thought is that all but the very simplest and lowest-step-size conduits are very close to rule-specific -- useful only in B3/S23. Adding a B8 or an S8 might be one of the few isotropic bits where some conduits would survive the rule switch (?). Even if a particular conduit works in another rule, it's only interesting if a large enough group of conduits works in the alternate rule to make it computationally universal. (That would probably make it construction-universal, too, but only after someone re-did all the single-channel search work to produce new recipe libraries.)
Anyway, from my point of view the reason why B38/S23 or B3/S238 doesn't get a lot of attention is that there's nothing really new and exciting about those rules to make up for the fact that the rule spec is just that little bit more complicated... pun maybe intended. Dvgrn (talk) 02:04, 20 January 2019 (UTC)

apgsearch and Catagolue

Would anyone be opposed to a reorganization of the information on these two articles? I noticed that a lot of the information in the Catagolue article really applies more to apgsearch rather than the site itself, and therefore might be worth moving. Such a change would be particularly easy to revert if need be, but I'd rather not go through the effort if that's the case, so I'd also like some feedback about that. Ian07 (talk) 23:32, 19 January 2019 (UTC)

No opposition here. I wouldn't think anyone would be likely to revert changes to those articles, just the usual quick review to see if the changes happen to serve as a reminder of anything else that should be added. Dvgrn (talk) 02:04, 20 January 2019 (UTC)