Talk:Main Page/Archive 13

From ProofWiki
Jump to navigation Jump to search
Filing cabinet.png
This is an article of past discussions, from 2-Aug-2017 to 8-May-2018.
Do not edit the contents of this page.
If you wish to start a new discussion or revive an old one, please do so on the current talk page.


Can we implement the CirrusSearch Extension? It makes searching more comfortable because it does spelling correction, among other improvements. --barto (talk) 06:35, 24 August 2017 (EDT)

It does seem to require an effort to install, but it is worth it. --barto (talk) 08:50, 31 August 2017 (EDT)
It has been a while, but this is currently being looked into. Thanks for the suggestion. — Lord_Farin (talk) 15:22, 15 November 2017 (EST)

Gadget Extension to see links to disambiguation pages

The Gadget Extension (which Wikipedia has) makes it possible for a user to display links to disambiguation pages in orange instead of blue. (Among other things.) This makes it possible to spot lazy links and fix them. --barto (talk) 15:44, 24 August 2017 (EDT)

Possibly, but there are rarely many. I go through them every so often and tidy them up. --prime mover (talk) 16:37, 24 August 2017 (EDT)

On whitespace before headings

I do support the increased whitespace before section headings, for the exact reason as explained in the FAQ. But the way in which it is added has some drawbacks, and in fact MediaWiki:common.css exists precisely for this purpose. I have tested the css-based solution by writing some code at user:barto/common.css so that, for me, the amount of whitespace is now always right. Here are some advantages and disadvantages of using css, after which I discuss alternative solutions and what css can and cannot do. I hope I'm not missing any important aspects. Any thoughts are welcome. (Please comment at the very end only.)

Tl;Dr: Replace all rules about blank lines around section headings by one simple rule, together with some other advantages including section editing, but at the cost of removing double blank lines before headings in the source text, possibly reducing its readability?


  • Easier to contribute. The less conventions (new) users have to keep in mind, the happier they will be to contribute.
  • Less discussions. If whitespace is fixed automatically, users are less likely to express their opinions about it, or to notice the increased whitespace in the first place.
  • Semantics. It is semantically wrong to use <p><br></p> (which is what happens now) to increase top margin.
  • Customizability. A coded solution is customizable, as opposed to relying on specific Wikitext usage on every page.
  • Section editing. Section editing could be allowed again. (See alternative solutions below for why it is not.) It has multiple advantages; some that come to mind are: Less need to scroll when editing Wikitext, more descriptive edit summaries, faster page load when editing/previewing (I've had problems with this when I was editing at places with slower connection than I have at home).
    One may argue that this is not such an advantage because pages at ProofWiki are, generally, small. Still, even relatively short proofs take quite some space in the Wikitext. Either way it's more elegant to only edit a section, especially if you want to correct a typo, add some extra explanation, add a link in the Also see section, add a source, ... There are other philosophical arguments in favor, such as the idea that unrelated edits should not be done at once.

Doubts you may have:

  • What with existing pages? Increasing the top margin using css, combined with the existing line break on those pages creates too much whitespace.
    Indeed. That's were jQuery comes to aid: it can remove those empty paragraphs, without affecting anything else.
    But then the page content jumps when the page is loaded.
    I don't know how to avoid that. But consider that it will always jump a bit due to MathJax.
  • There are currently about 50000 pages. They would all have to be brought up to the new standard.
    Unfortunately, yes. I would never punish anyone by having them fix that manually. Except a bot. As long as the job isn't finished, jQuery can take care of them. Remember that it won't have to be done ever again, because the styling of all headings will be automatized.
  • Wikitext readability. Blank lines in the Wikitext make it easier to spot section headings. Especially considering that there are many single blank lines in the source text already, because statements are separated by a blank line.
    I agree. On the other hand, with section editing there's less need for it. It will be mostly useful for refactoring and restructuring tasks, or when creating a page.
    To some extent, it's a matter of habit. Consider that the extra blank lines in source text are specific to ProofWiki, and are not used on e.g. Groupprops or Wikipedia, so not everyone will find the extra blank line needed for readability. I discussed this further at Alternative solutions below.
  • I don't think pages in [insert namespace] should be affected.
    No problem. Common.css can take care of that. See code details below.
  • Not all section headings need more whitespace. For example subsections that come immediately after a section heading, or transcluded definitions at equivalence proofs of multiple definitions.
    Again, this is no problem for css. Even better: css can be used here to make sure they all follow the same format, and any changes in style will be applied to them all at once.
  • Some templates create section headings, and should not be accessible through section editing.
    An administrator can lock them.

In summary, the only lasting disadvantage I see is readability of the source text. It has to be decided whether this outweighs the advantages above. Consider also that we could replace all rules about blank lines around headings by just one simple rule:

One blank line before a heading, one blank line after.

See also Alternative solutions below.

Alternative solutions

  1. Change the default number of blank lines at the end of a section when section editing.
    When submitting a section edit (or any edit), what the MediaWiki parser does is:
    1. Trimming
    2. Adding "\n\n" at the end
    It would be nice to make the parser add "\n\n\n" at the end instead. This would allow section editing while maintaining the right amount of whitespace. It does however not solve other things such as customizability, semantics and special cases like definitions at definition equivalences and perhaps others.
    Unfortunately, this is not configurable, or at least not without hacking into internal MediaWiki files.
  2. Change the way Wikitext is interpreted. Keep the double blank lines before headings, but don't let MediaWiki create empty paragraphs. Add the whitespace using css instead.
    This would allow multiple blank lines in the source text for readability, without affecting the result. A disadvantage is that section editing removes those blank lines again. Also, it still involves different source code rules on blank lines (though they won't affect the output anymore).
    I could not find anything about whether it's possible to customize wikitext interpretation (will update when I do) but it's very likely that it is not, again, at least not legally and free from danger.
  3. Keep it the way it is.

It must be noted that (unless we resort to more hacking) 1 and 2 would affect all pages, including user pages.

Code details

There are some things to be taken into account in common.css (and common.js for the temporary jQuery script):

  • Headings following headings. We probably don't want as much whitespace as usual between two headings if there's no text between them. No problem for css.
  • Definitions at definition equivalences. It appears that the current policy is to reduce whitespace between them. No problem for css, but requires some more code. Also does not depend on there being a blank line between them in the source code or not.
    (In general: 2n blank lines in source text = 2n+1 blank lines in source text = n empty paragraphs in output.)
  • Restricting to namespaces. There are some namespaces like Category:, User:, talk pages, that perhaps we don't want to affect. That's no problem, in fact MediaWiki has cues in the source code that are created specifically to distinguish between them when using common.css.
  • Browser compatibility. The css I wrote at user:barto/common.css uses some features of css3, but the most recent versions of all major browsers support them.

For the jQuery paragraph removing script: only empty paragraphs right before headings must be removed, but this is no problem. The following must be taken care of:

  • Empty paragraphs created by blank lines between list items; see e.g. Definition:Open.

The following cannot be affected, because they don't involve empty paragraphs:

  • Double blank lines before "then" in theorem statements or, similarly, in definitions. Or anywhere else between two paragraphs.

Thoughts? --barto (talk) 06:41, 4 September 2017 (EDT)

Whatever we need to do that:
a) gives me least work to do, I would rebel against a regime that made me do a lot of rework;
b) leaves it looking stylistically exactly the same as it is now.
Apart from that, I'm okay with whatever is decided. --prime mover (talk) 09:52, 4 September 2017 (EDT)
A lot of your points focus on enabling section editing. I consider section editing a flawed means of editing, and I know by experience that the MediaWiki parser code generating it is buggy and unpredictable, especially in combination with transclusion. That right, in combination with what is the single most used MW feature on ProofWiki.
Additionally there are very many cases where I would rather avoid that someone changes one part of the page (say, the proof), and accidentally uses different notation from the theorem statement. No thank you.
Those qualms regarding section editing (which makes up a good portion of the aims you seem to have with this change) having been mentioned, I think it can only add to the ease of contributing. Not a lot, admittedly, but a bit. And it helps to reduce frustration from header distance not behaving as wanted. So I would be in favour of experiments and a good style sheet.
Last point I want to mention here is the behaviour of transclusion. Because of the logical way this technique's tags are placed, we get extra line breaks for free. I would like to see a plan to address this. — Lord_Farin (talk) 15:11, 4 September 2017 (EDT)
On section editing: I'd personally like there to be an option to disallow editing transcluded sections (without having to use html tags for those headers). With the suggestion on transclusion below, I think it is possible.
I agree that section editing would slighty increase the possiblity that someone uses different notation than in the theorem statement. On the other hand, experienced users won't, and the bold unexperienced who are not informed about our policies on notations will probably do it anyway. Something we can already do now, is adding a line in ProofWanted as a reminder to use the same notation. --barto (talk) 12:28, 7 September 2017 (EDT)
On transclusion: It is infortunate that the placement of the onlyinclude tags matters so much. What would be ideal are tags that trim the text when transcluded.
Solution? There is no hope that the behaviour of onlyinclude will ever be changed because many wikis rely on it. However, what should be possible is to create a variant of the tag that does trim. (Alternatively, we could maybe combinine onlyinclude with a trimming template (I'm not sure, because of }'s in the source code), but the transcluding syntax will become inelegant.)
If it's possible, a key question (but not something to focus on now) will be: should the text be trimmed always, or only when transcluded?
Lastly, an advantage of our own onlyinclude tags is that (I think) we could disallow section editing of transcluded sections (by placing everything in a <span class=transcluded> or something, + common.css) --barto (talk) 12:28, 7 September 2017 (EDT)
Editing sections is easy, I'm editing one now. You just right click on the subheading.
I've said it before, I'll say it again, and I will continue to say it over and over again until the apocalypse: it ain't broke, don't fix it. If you feel the need for a programming exercise to keep you occupied, bend some thought to how we may be able to add some structure to Category:Specific Numbers.
All the dreaming aside, here is a realistic, practical suggestion: From now on, place the onlyinclude tags such that there are no additional linebreaks inside. This does not affect anything, again by mediawiki's whitespace handling. (It's how onlyinclude is supposed to be used.) --barto (talk) 05:40, 8 September 2017 (EDT)
"Improving" the MW transclusion mechanism is something that I have been looking into with regard to the ProofWiki extension (having options to increase heading levels and multiple transcludable sections, among other things). However, it turns out to be a pain because the parser code generating the table of contents is an absolute mess and by default will omit sections transcluded that way. While I would encourage such a mechanism to work (because as said, it enables more features than just whitespace control) I think it presupposes work on the table of content generation in MediaWiki. If you are interested I have some discussions on the MediaWiki issues board that should provide a good introduction.
Changing the placement of onlyinclude will introduce a major inconcistency in the source code formatting, and make it much harder to find the places where it starts and ends (due to the tags not being on their own line). I don't think this will improve things over time. Maybe we can rely on the jQuery to save us when it comes to transclusion.
In summary, I conclude that there is currently nothing in the source code conventions that can be suitably improved at this moment. However I still support the CSS/jQuery experiment to see if we can make steps towards achieving this layout with the proper technical tool. — Lord_Farin (talk) 05:14, 9 September 2017 (EDT)

css/jQuery proposal

There seems to be some support for testing out the css/jQuery approach. Shall we try it? It can always be limited to specific namespaces. --barto (talk) (contribs) 18:01, 30 October 2017 (EDT)

I would definitely prefer this to be thoroughly tested on the user pages before considering to bring a change of this magnitude live. — Lord_Farin (talk) 14:38, 4 November 2017 (EDT)
Sure. That would be namespace #2. I'm already using it via my user common.css, and I'm very happy with it; had no problems. --barto (talk) (contribs) 16:24, 7 November 2017 (EST)

Hosting and ads

Below is a discussion I moved from user:barto to here.

Some people tell me they don't want to use or contribute to $\mathsf{Pr} \infty \mathsf{fWiki}$ just because it has ads. That's sad. I think it is not hard to find a university or something that's willing to host the site for free. They tend to have huge servers anyway, so hosting $\mathsf{Pr} \infty \mathsf{fWiki}$ won't make a big difference for them. --barto (talk)

I use Adblock so I don't get ads. I'm wary about this site being hosted by a university because I don't go to university and you can bet bullets to bullshit that they'll disallow anyone not an alumnus from contributing. If that happens I'm going on another killing spree, but this time I won't stop when I've reached my age in weeks.
Many people don't use adblock, but that doesn't matter. The success of ProofWiki should not rely on Adblock. Whether it's a university or some mathematical society, we still decide what happens to the site. The idea is simply that someone would pay for ProofWiki as an act of generosity. --barto (talk) 17:11, 24 August 2017 (EDT)
Nobody ever does anything as an act of generosity. Universities are businesses out to make money, and they do that by charging people for the use of their facilities, and they do that by blocking everybody who isn't a fee-paying student.
But I'm just one person, and my opinion bears no more weight than that of everyone else, I just won't be around when it happens. No worries.

Another option is to become part of wikimedia. Reading their criteria it does seem that ProofWiki has a good chance of meeting them. Our goals and style do differ from Wikipedia:WikiProject Mathematics (we're not an encyclopedia but more a dictionary, quite the opposite and complementary to it), any Wikibooks on mathematics (we do not focus on one topic, dictionary style) and Wikiversity (not a collection of courses, nothing is done twice, dictionary style, nonlinear) --barto (talk) 17:05, 4 September 2017 (EDT)

If we're to preserve the current structure of requesting an account, the site would have to be independent of SUL. The problems that this site has faced in the past with spambots doesn't make opening up the site to anyone with a Wikipedia account an attractive proposition in my opinion. I'm unsure how open to compromise the sysadmins would be on this, as SUL is deemed to be a relatively important feature. There would be no harm in one of us creating a proposal on meta, however. I could always start drafting one in userspace. Caliburn (talk) 17:20, 4 September 2017 (EDT)
IF we were to merge with metawikipedia, then I am afraid that is where $\mathsf{Pr} \infty \mathsf{fWiki}$ and I would part ways. --prime mover (talk) 18:27, 4 September 2017 (EDT)
That would be very unfortunate, you are an exceptionally valuable contributor. Again, I do forsee problems involving SUL, and independence also. While we would not lose control of the project, we would most likely have to allow Wikimedia stewards and sysadmins reasonably escalated access. Thank god the WMF is nothing like Wikia, and do keep the interest of the community in mind in their actions. The migration would of course welcome a stream of new, inexperienced users, which is a monumental positive in some senses, and a significant negative in others. I quite like the status quo in terms of the editing environment, so I'm not sure how I'd feel about it, though I've only been here for a few weeks. Caliburn (talk) 18:47, 4 September 2017 (EDT)
Yeah well, I've not been here that long myself, I don't think I'd be missed. --prime mover (talk) 01:16, 5 September 2017 (EDT)
Good point. I don't know how spam-proof wikimedia is. There are surely other things I haven't thought about. Either way, everyone here would have to agree.
Quote from ProofWiki:About: The more people we have supporting the project, the bigger and better it will become. --barto (talk) 09:39, 5 September 2017 (EDT)
While there's no doubt that an influx of unexperienced editors requires more tidying and explaining policies (which can frustrate you if you let it), it is the only way in which a community can truly become great. It will be impossible for one user to patrol everything, and that's a good thing: there exists more than a lifetime of mathematics to read. --barto (talk) 10:07, 5 September 2017 (EDT)

Just to make clear why I think getting rid of ads is beneficial:

  • Makes the site more attractive
  • Less distraction for those who learn by reading ProofWiki

And for finding a generous hosting provider:

  • No more need for ads
  • The future of ProofWiki is secured: when one day the current administrators retire, the site will continue to exist
  • Currently people might refuse to contribute because they don't know who owns ProofWiki (a person or a (non-profit) organization?). (Or whether someone is able to shut the site down and take advantage of their contributions for personal gain. I know it sounds paranoid, but it wouldn't surprise me if some people think like that.)

in particular, wikimedia:

  • Likely to attract many contributors
  • (conjectured) Higher page rank in search engines
  • Technical support as much as we want

--barto (talk) 09:39, 5 September 2017 (EDT)

I'm currently preparing for Hurricane Irma. I'll try to address this stuff as well once I get some time. --Joe (talk) 10:41, 6 September 2017 (EDT)

Caching problem

Suppose you have a picture which you have uploaded, and included in a page, with a size, for example:


Suppose you then upload a new version of that file, using "Upload file".

However much you refresh the cache, purge, whatever you have to do, it will not update the page to show this new version.

If, instead, you change the size of the image on the page:


... the new version appears.

But then if you go back to the old size:


... you once more get the old version back.

Is there a way of purging these old views of the files, so you don't have to change it size to see the new version? --prime mover (talk) 07:40, 1 October 2017 (EDT)

Hm, strange. I checked today and all seemed to be fine when reverting to 400px. It must be caching somewhere, but I cannot determine now if it is server-side or client-side. Also couldn't find any similar cases online. We'll have to keep looking if it reoccurs. — Lord_Farin (talk) 17:48, 4 October 2017 (EDT)
Local caching problem. Sorted. I hate my new machine. --prime mover (talk) 18:27, 4 October 2017 (EDT)

Symbolic Evolution

I have come up against the difference in notational preferences for a sequence or a series: $\left\langle{a_n}\right\rangle$ or $\left({a_n}\right)$ or even $\left\{ {a_n}\right\}$?

For and against:

$\left\langle{a_n}\right\rangle$ is also used for the generator of a group, and so can be confused with it. The reason it has been settled on for this site is because it's what I learned in uni, and what I have in certain of my source works, and that is where it settled.
$\left\{ {a_n}\right\}$ can be conflated with the set notation, and so carries the subtext "order does not matter, uniqueness does" which are both fundamentally not the case for sequences. No takers for this, I hope.
$\left({a_n}\right)$ is what Barto, for one, uses by preference. I have been routinely changing them to $\left\langle{a_n}\right\rangle$ in my ongoing drive for consistency, but I have been minded that it may be "the most sensible" notation -- because it is the same sort of notation as the "ordered tuple", it brings its own subtext that "order matters, and terms may not be unique".

So, a big suggestion for a major refactoring task: should we settle on $\left({a_n}\right)$ for the standard notation on $\mathsf{Pr} \infty \mathsf{fWiki}$ for sequences and series? My vote is: yes, but we don't need to do the change all in one go please. --prime mover (talk) 08:29, 22 October 2017 (EDT)

Hmmm... I tend to use $(a_n)$ indeed, I guess because it's what I grew up with or what's easiest to write down (pen & paper, I mean). Regardless of that, I do see some objective reasons to prefer them: $\langle\cdot\rangle$ is sometimes used for generated substructures (e.g. linear span), and $\{\cdot\}$ makes it look like a set (without ordering). I'd like to hear other arguments I'm overlooking. --barto (talk) (contribs) 18:51, 25 October 2017 (EDT)
Well, $\langle\cdot\rangle$ can be confused with a generator just as much as $(\cdot)$ can be confused with an ideal... and with the general ubiquity of parentheses in mathematics (both in text, like here, and for order explication, and for function invocation, and ideals as mentioned, and probably for more things), it hardly seems that we need more things using parentheses.
Rather, it seems to me that sequences are one of the places where there is a reasonably common alternative with $\langle\cdot\rangle$. I see no harm using it and even less a need to refactor all of them away.
But I could live with the compromise that both are allowed (although consistency will suffer). — Lord_Farin (talk) 11:23, 29 October 2017 (EDT)

Proof count on main page

What about something like

{{ProofWiki}} currently has about {{PAGESINCATEGORY:Proven Results}} proofs.

at the main page? Or does this slow down page loading? --barto (talk) 10:02, 25 October 2017 (EDT)

We had this for a long time, I forget why we removed it (anyone remember?). Would be nice to overhaul the main page in general. I'm cool with adding it back if you have a proposal. --Joe (talk) 10:05, 25 October 2017 (EDT)
I think you removed it so as to de-clutter the front page a couple of years ago. We used to use it to keep an idea of the count of the proofs so we could add "landmark" proofs: "This is the "n"th proof on ProofWiki!" It would be good to have it back.
A count on the "Proven Results" category will in general be an overestimate, because parent pages which are the container for Proof 1, Proof 2, etc. are also counted, because they transclude the qed. But it would be a start. --prime mover (talk) 10:35, 25 October 2017 (EDT)
True. But there's no harm in exaggerating a bit, especially if we put "about" before the number. --barto (talk) (contribs) 19:09, 25 October 2017 (EDT)
Is there a better way we could estimate? I could add something to our extension. Eg. all pages that contain {{qed}} and don't contain any todos, etc. --Joe (talk) 11:20, 26 October 2017 (EDT)
If we could count pages with qed on them and not include the pages they are transcluded into, that would probably be the best technique, to my mind. I would say: keep it simple, and don't exclude pages with todos on them because they shouldn't have a qed on them in the first place. --prime mover (talk) 12:24, 26 October 2017 (EDT)
I have this implemented locally. Can anyone think of a good way to likewise count definitions? --Joe (talk) 14:56, 3 November 2017 (EDT)
Is it possible to count the contents of a category and all its subcategories? And exclude redirects in the process? --prime mover (talk) 16:05, 3 November 2017 (EDT)
The issue is that definition pages sometimes have several subpages including examples. How accurate do we want to be. Consider Definition:Rational_Number, it has several subpages, are these to be counted as more than 1. If so how do we ignore the "note" subpage. One idea I had was count the number of links to unique definitions from proofs. That total is around around 3,500 unique pages. All unique definition links from all namespaces is almost 10,000 (about the same as counting all pages in the definition namespace). Which of these feels more accurate? --Joe (talk) 16:15, 3 November 2017 (EDT)
If all subpages are themselves individual pages in the "Definition" category, then that should work surely? "note" subpages are in categories in a different namespace. Or am I missing something? --prime mover (talk) 20:07, 3 November 2017 (EDT)
Pages like Definition:Rational Number/Linguistic Note will also be counted. This is not a real definition. --Joe (talk) 23:10, 3 November 2017 (EDT)
But it's in the "Linguistic Notes" category, which is not a subcategory of Category:Definitions. This is what I'm trying to understand: is it possible to count all the pages in Category:Definitions and all its subcategories (excluding redirects)? Can something be set up that doesn't count the namespaces?
Here's a radical suggestion: like we have the qed template, select a symbol e.g. $\diamond$ or whatever the code is for a black one of them which can be used to signify the end of a definition, build a template with that and a category to put them all in like for qed, and then etc. etc. --prime mover (talk) 06:28, 4 November 2017 (EDT)
Okay, I see what you mean now. This should be pretty easy to scrape from the database. --Joe (talk) 10:56, 4 November 2017 (EDT)
So pages like Definition:(0,_1)-Matrix/Examples should not have a Definition category link? --Joe (talk) 18:55, 4 November 2017 (EDT)
Some of our structure may need to be tightened. There are some special cases like this which need to be thought about. Perhaps "examples" need to be removed from the Definitions category, or that we deliberately implement "Examples" categories which do not count, but then the "Rook Matrix" example should be a definition, and ... and... my brain hurts. --prime mover (talk) 19:12, 4 November 2017 (EDT)
Okay, well for now I won't worry about it and will just count definitions as you described above. --Joe (talk) 19:14, 4 November 2017 (EDT)

Subpages and the horror of refactoring

I've been working on an alternative for subpages: User:Barto/Sandbox/navigation. At least, for subpages except things like multiple (numbered) definitions, historical notes, multiple (numbered) proofs, ... Basically it's meant for the type of subpages that are commonly linked to (definitions, mostly).

Pros: Prevents using direct links to subpages, Page titles become neater, Refactoring easier (hierarchy is easily changed, we can even go to much deeper levels just by adding links)

Cons: The absence of slashes in the title makes the hierarchy less clear, despite the "navigation bar". In some cases the structure of the title (e.g. "Continuous Real Function at Point") still reflects the hierarchy: "Continuous/Real Function/at Point"

Another con: having to learn yet another page structure template, and one which is horribly non-intuitive, for no benefit. --prime mover (talk) 13:34, 27 October 2017 (EDT)

Thoughts? Probably some stylistic improvements can be made. --barto (talk) (contribs) 18:41, 25 October 2017 (EDT)

How's that for a big suggestion for a major refactoring task? :) Might be the last of its kind though. --barto (talk) (contribs) 19:00, 25 October 2017 (EDT)
I'm going to need a lot of convincing. There are going to be some pages with a big ugly blue box on them which looks shit, and it's going to mean a colossal amount of work to get it into that shape in the first place. So, I'm out. --prime mover (talk) 01:42, 26 October 2017 (EDT)
Some examples: Definition:Big-O Notation/Real/Point could become Definition:Big-O Estimate of Real Function at Point and we could easily add another level "Asymptotic Notation" in the hierarchy, without making the page title longer. The theorems transcluded at Triangle Inequality (e.g. the reverse) could link back to that page, improving navigation.
We've already got Definition:Big-O Estimate for Real Function at Point so I really don't see what the point is. --prime mover (talk) 12:30, 26 October 2017 (EDT)
I don't mind doing the work. Style of the box can be changed to your liking.
If you want, I can test out an actual example, e.g. Triangle Inequality. --barto (talk) (contribs) 02:55, 26 October 2017 (EDT)

Made some stylistic changes, looks more intuitive now and fits within the color scheme: User:Barto/Sandbox/navigation. --barto (talk) (contribs) 05:54, 26 October 2017 (EDT)

Actually this is exactly how subpages work, and how the buildup of the backlink structure of subpages is set up. I don't see a need to improve upon this. Although the argument regarding the page titles is a fair one. (But who cares about the title anyway if there is a redirect? What we actually want is that the search functionality improves so that it becomes easier to find pages.)
As to refactoring, you are still going to transclude the subpages, and whether you do that with the redirect as the main page or not, the fact remains that they need to be changed. I am assuming here that a rename will also affect any and all subpages (which is ideally the case to ensure consistency). Therefore I struggle to see the benefit from this scheme.
It could be decided that we can improve upon the backlink structure by making it customisable, but then I would say we should invest into getting this a MediaWiki extension proper and not develop some hand-written workaround that sits beside the MediaWiki solution.
As to the labeled section transclusion extension mentioned, we had a look into it and it was deemed insufficient. Our own extension is supposed to fix this better but is not finished yet (the major problems with section headings and nested tags continue to exist; we might try to steal the syntax of LST, if it works them I'm all for it (although it is of course a bit against XML standards)).
But I'm open for hearing more arguments, and want to thank you for looking into it in any case! — Lord_Farin (talk) 06:54, 26 October 2017 (EDT)
Thanks for the feedback. Still, while there are redirects, it takes a few clicks to find them, which, apparently (and understandably), is too cumbersome for the average contributor.
As for refactoring, I don't think I understand "I am assuming here that a rename will also affect any and all subpages [...]" - the idea is to not use subpages anymore (except /Definition 1, /Proof 2, /Historical Note, /Lemma, /Corollary
I don't expect MediaWiki to create an extension for such navigation, as it is already possible with LST, I believe. I'd like to learn more about our own extension, although I don't think it is needed here. --barto (talk) (contribs) 07:24, 26 October 2017 (EDT)
Confirmed, LST suffices for the improved version: User:Barto/Sandbox/nav/2. --barto (talk) (contribs) 08:21, 26 October 2017 (EDT)
Well, regardless of whether it is a subpage or not, if we are transcluding onto Definition:Continuous Real Function the page Definition:Continuous Real Function at Point, then I assume that any rename of the former will naturally lead to a rename of the latter (by application of the same principle). Moreover, dropping this principle would open us up to losing consistency in naming (because the rename need not be consistently applied). But maybe you see this differently; if so I would like to hear an example.
As to LST, I was referring to its functionality as a whole, which makes the transclusion mechanism more flexible. We very much have a need for this with our deep transclusion practices (which can sometimes call for distinct subsections of pages to need transclusion, e.g. when transcluding to overview pages). This is where the need for our own, more versatile extension sprouted, see Help:ProofWiki Extension for how it works.
With regard to the extension I was mentioning, please be aware that it is perfectly possible to create a MediaWiki extension of your own. We don't have any dependency on MW here, I just imagined that it would help us if something like this would be included in the backli nk structure of subpages, instead of relying on a specific cumbersome application of LST, with the MW version sitting besides it to confuse the reader.
From a practical perspective, since the PW extension and LST both use the section tag, I fear there is a breaking incompatibility at work. — Lord_Farin (talk) 12:03, 26 October 2017 (EDT)
About the renaming: that's true, and the renaming job would be less efficient. But. :) It seems to me that most of the times such a page is renamed, it's either because (*) it has an ambiguous title or (*) something is wrong with the subpage hierarchy. What I want to say is, and I may be naive, I think a page with a title like Definition:Continuous Real Function at Point will never have to be renamed, with the hierarchy being independent of titles. (Okay, yes, there's always the possibility that someday a majority will prefer "Continuity of Real Function [...]". But hey, if we're talking about aesthetics, "Continuous Mapping/Real Function[/...]" isn't better :) In any case, "never" can safely be replaced by "much less often".) --barto (talk) (contribs) 13:52, 26 October 2017 (EDT)
It's only when someone decides to stride in and rename everything for reasons not completely understood that we have these renaming exercises in the first place.
As was established early on in the life of $\mathsf{Pr} \infty \mathsf{fWiki}$, there are multiple names for concepts, and different approaches that can be made. Having settled on one approach, and decided on a particular direction to come from, it was expected that this would be what it would be. It makes no sense at all just to arbitrarily rename and restructure everything. Yes, arbitrarily. --prime mover (talk) 16:12, 26 October 2017 (EDT)
I have worked out an example here to let you taste: Definition:N-Ary Operation Induced by Binary Operation.
It uses the ProofWiki transclude extension. You only have to give the title of the parent page.
Makes it very comfortable to restructure. No danger for links to subpages. --barto (talk) (contribs) 07:41, 27 October 2017 (EDT)

Okay so I've looked at Definition:Indexed Iterated Binary Operation, and I need to ask: do we really need to add all that stuff at the top of the page complete with all those "includeonly" and "noinclude" tabs and two "section" sections? Nope, no way.

Can you explain again why this is better than making Definition:Indexed Iterated Binary Operation a redirect to a structured hierarchical page in the manner fully supported by the MediaWiki software itself? --prime mover (talk) 09:46, 27 October 2017 (EDT)

I refer you to "Pros" in the beginning of the section, or the last line just above your comment. No need to repeat it. --barto (talk) (contribs) 10:01, 27 October 2017 (EDT)
Are those really pros?
"Prevents using direct links to subpages" -- that's not necessarily a pro, sometimes you might want to make a direct link to a subpage.
"Page titles become neater" -- what do you mean, "neater"? Clarity is always to be preferred over neatness -- and slash-separated page names make it completely clear what the hierarchy of page names is.
"hierarchy is easily changed" -- there's only one person here insistent on changing the hierarchy, and I for one do not want to make this easy, under any circumstances, for blindingly obvious reasons.
Now, can you find me some pros?
As for the heavy syntax, that's indeed unfortunate, and is probably the best reason not to implement it now. I'm planning to file a request at Phabricator to improve the functionality of LST, in a way that would allow to do this with a single template, as in {{nav|parent=...|abbr=...}} --barto (talk) (contribs) 10:06, 27 October 2017 (EDT)
Fun fact: the aspired version of LST would also allow to automatically populate, for each book, a page with links to all pages on ProofWiki where it is cited, in the same order as the results appear in the book. Without any changes to the wikitext (thanks to prev & next)). I.e. a detailed online version of it. --barto (talk) (contribs) 10:35, 27 October 2017 (EDT)

Nope, I can't live with it. Not at all. --prime mover (talk) 12:35, 27 October 2017 (EDT)

Some more cents: I agree with PM that the redirects take adequate care of the cry for "readable" page titles. I was thinking that the only thing that could sway me in favour of this mountain of work would be a solution to the issue that we might want the same page to be transcluded on more than one page (one of the things our extension addresses). But then if the hierarchy is still defined to have a single parent, we have no benefit over using the built-in subpage hierarchy.
If you manage to find a way to address that, then we might have something worthwhile that I would call a real "pro". Until then, I would call the suggested solution unpractical -- it would be more efficient to extend MW so that the displayed title can be one of the redirects to the page; such would solve the problem the same way, but without having to tamper with our existing pages in any way.
Again, thanks for your investigation, but it would just be too much work, and actually additional idiosyncrasy in the PW approach to MW as opposed to using the standard redirect functionality. I doubt this will increase usability for people... — Lord_Farin (talk) 12:58, 27 October 2017 (EDT)
"If you manage to find a way to address that" - I'm not sure here what you mean by "that", is it pages with multiple parents? The navigation template could allow that, yes (it could make recognize common ancestors and merge subtrees in one tree or whatever; design is not the problem).
It is already possible to change the displayed page title to whatever you want, see $wgRestrictDisplayTitle. Note also "[...] it breaks the wiki convention that a page's title is its name, and thus can be used for linking to it.", a convention that PW already violates by disrecommending direct links to subpages. --barto (talk) (contribs) 13:37, 27 October 2017 (EDT)
Allthough I do think $wgRestrictDisplayTitle=false is a good start. --barto (talk) (contribs) 13:41, 27 October 2017 (EDT)
I'm done arguing. --prime mover (talk) 13:40, 27 October 2017 (EDT)
arguing? brainstorming :) --barto (talk) (contribs) 13:48, 27 October 2017 (EDT)

The end of MathJax?

I just saw that the MathJax extension at MediaWiki has been archived: mediawikiwiki:extension:MathJax. Not sure what it implies. Are we using that extension or an independent one? --barto (talk) (contribs) 04:41, 27 October 2017 (EDT)

We're using that one with a few fixes. See --Joe (talk) 12:19, 30 October 2017 (EDT)

What is a "Proof"?

In the process of deriving some results regarding a particular logic, I wrote up Definition:Constructed Semantics/Instance 1. Now I want to refactor the validations on Hilbert Proof System Instance 2 is Consistent into separate pages.

However I struggle to determine where to put it. By analogy I would create e.g. a new proof page at Rule of Idempotence/Disjunction/Formulation 2/Reverse Implication but this seems off as the system does not represent "truth" in the ordinary sense. On the other hand, in the future this kind of result might be useful to be collected and described separately (especially for things like specific Kripke models whose theory is fascinating on its own).

It's also not really belonging in an examples category because it is used in a proof. An option I thought of was a subpage structure below the definition, but this puts proofs in the definition namespace.

Any thoughts where I can place these results? — Lord_Farin (talk) 14:16, 5 November 2017 (EST)

I now placed them on subpages for progress' sake, but marked them as candidates for renaming. — Lord_Farin (talk) 12:20, 7 November 2017 (EST)


Right that's it, I'm off. This site can sink or swim. --prime mover (talk) 15:05, 5 November 2017 (EST)

Sad to hear that. I hope you'll be back soon, your contributions are invaluable. — Lord_Farin (talk) 15:16, 5 November 2017 (EST)

Category definition template

I created Template:Category definition in order to standardize the way categories are defined. Comments (template name, parameter names) ? --barto (talk) (contribs) 02:36, 7 November 2017 (EST)

I don't like the template name so much because it is a departure from the established pattern (no template in use has spaces, or lowercase words). Maybe "DefineCategory"? The verb should keep it distinguished from the MW brand of categories, which at best need to be described.
I like the names of the arguments (matching the way they are denoted in TeX) and the proposal to have a template for this – such will make life easier. — Lord_Farin (talk) 12:20, 7 November 2017 (EST)
"DefineCategory" is fine for me. I don't mind so much about capitalization for non-content pages; I've always thought the reasons for capitalization are that (1) it makes links to theorems stand out and (2) it makes the reader focus on the end of page title (often containing important words) and not just the beginning. Finally, to some it just looks better.
What I want to say is that (1) and (2) don't apply to templates and short page names (help pages etc), so that title-casing of non-content pages isn't that important to me. For templates, I'd even say it's better not to capitalize, as it makes typing the template name slightly more complicated. But I really don't care so much that I will insist; it's all invisible anyway. --barto (talk) (contribs) 13:33, 7 November 2017 (EST)
For page titles in general, you are right. It helps people to focus on the more important words in the page name (also at the top of a definition page). I think the main argument for templates and help pages is consistency with the other namespaces. The difference in effort is not so big and it avoids people wondering about the discrepancy.
Alternatively dash-separated lowercase words could be considered like with Template:Begin-eqn c.s. But ultimately it is just a minor thing that is not worth the time to discuss, let alone to go and change throughout. I'm more worried about descriptiveness. — Lord_Farin (talk) 13:49, 7 November 2017 (EST)
Agreed. I don't understand the sentence The verb should keep it distinguished from the MW brand of categories, which at best need to be described. , but both "category definition" and "define category" are fine for me, whitespace or not. --barto (talk) (contribs) 13:59, 7 November 2017 (EST)
Well, MW categories don't need a definition, they just need to be created and have some content. Thus it makes no sense that "Category" in "DefineCategory" would refer to a MW category, and therefore there is less confusion. CategoryDefinition has this problem because of the obvious clash with DefinitionCategory. — Lord_Farin (talk) 14:06, 7 November 2017 (EST)

I decided to move ahead. Template:DefineCategory it is. — Lord_Farin (talk) 14:09, 7 November 2017 (EST)

Name of example categories

Some example categories are called "Examples of Xs", others "X examples", "X/examples" or "Xs/examples". I think the first one is best, because:

  • In a category X, the subcategory "Examples of X" will be sorted under E by default, which is where one would look for it.
  • In some cases (e.g. X=Group - and not "Group Theory"), "Category:Xs" does not exist, making it a bit strange to have "Category:Xs/Examples"
  • Alphabetical sorting in Category:Examples can always be fixed with key-value (can be done in the Template:ExampleCategory)
  • "Examples of Xs" is proper use of language; the other formats seem to be invented purely for alphabetic sorting in Category:Examples --barto (talk) (contribs) 15:03, 7 November 2017 (EST)
Good logic. Feel free to implement (although I have the impression you already started/finished without waiting for that message...). — Lord_Farin (talk) 09:56, 12 November 2017 (EST)
Thanks for the feedback. Indeed I started applying it to the case of categories shortly after the creation of {{DefineCategory}}. --barto (talk) (contribs) 10:05, 12 November 2017 (EST)

Adoption by the Wikimedia Foundation

Continuing the discussion on #Hosting and ads.

I'd like to bring up again the possibility of becoming part of the WMF. Summary of the comments so far:


  • No ads: more attractive, less distraction
  • Future is secured: when one day the current administrators retire, the site will continue to exist
  • Clarifies the status of ProofWiki. People may ask: Is it privately owned? Can someone, in theory, take advantage of my contributions for personal gain (cf ads)?
  • (Conjectured) higher page rank
  • Boosting growth speed, lots more contributors


  • Little to no control over who joins the community, due to WMF's Single User Login (SUL)
  • All those new users will have to learn our policies, in particular the presentation of proofs (detailed, links, linebreaks).
  • While we would not lose control of the project, we would most likely have to allow Wikimedia stewards and sysadmins reasonably escalated access. Thank god the WMF is nothing like Wikia, and do keep the interest of the community in mind in their actions. --Caliburn

Looking at site stats, there is clear evidence that this site is used. It agrees with WMF's mission to create educational content of different types. I have started drafting a proposal with more detailed thoughts here: user:barto/Sandbox/mw, emphasizing why it is different from existing projects, and why it is useful. --barto (talk) (contribs) 09:40, 9 November 2017 (EST)

To be clear, the idea is that ProofWiki will remain a project on its own. --barto (talk) (contribs) 10:32, 9 November 2017 (EST)
Mind if I work on the sandbox page a bit? May I also suggest alerting WikiProject Mathematics on the English Wikipedia of this discussion? Caliburn (talk) 10:05, 12 November 2017 (EST)
Feel free. (Perhaps a user sandbox would be a better place than the main sandbox.) Good idea, I think it is useful to hear opinions from Wikipedia contributors. --barto (talk) (contribs) 10:10, 12 November 2017 (EST)
I think there is a need for more internal discussion before setting any external wheels in motion. — Lord_Farin (talk) 10:13, 12 November 2017 (EST)
I agree. I'm not 100% sold on the idea. --Joe (talk) 10:11, 13 November 2017 (EST)


I wonder, how big of a problem is spamming at WMF projects these days? --barto (talk) (contribs) 09:49, 9 November 2017 (EST)

I've been reading a bit about it, and realized that one of the types of spam wikipedia has to deal with is (manual) self-promotion, which ProofWiki is unlikely to suffer from. --barto (talk) (contribs) 10:59, 9 November 2017 (EST)

Do you think we should take spamming and problems with unexperienced users into consideration, or is it a necessary evil if this site is to grow large? I think it we should not let it stop us. --barto (talk) (contribs) 09:49, 9 November 2017 (EST)

I'm still forming an opinion on this. I'll get back to it. — Lord_Farin (talk) 10:49, 12 November 2017 (EST)

Update on categorization

Hi all, this is to notify you that I made an attempt to organize the Categories, the philosophy being that (1) everything should be categorized (except perhaps redirects, that I didn't change), including templates etc. (2) smaller 'meta-categories' make browsing easier.

I got inspiration from Wikipedia, but didn't preserve the parts that seemed non-intuitive to me. The maintenance category still needs to be organized. --barto (talk) (contribs) 10:32, 12 November 2017 (EST)

I also cleaned up Category:Definitions by assigning more meaningful categories. --barto (talk) (contribs) 10:34, 12 November 2017 (EST)

Thanks for that. Next, it would be advised to settle on a standard for naming categories and templates and conform. The standard in place prefers usage of capitalisation. The pattern in your endeavours is a strict departure from that. Please elaborate. — Lord_Farin (talk) 10:49, 12 November 2017 (EST)
Categorisation of templates makes sense, especially if they are going to be more split up according to your suggestions (see also below).
Please also continue your efforts into describing templates and their usage in more detail. It is really helpful. — Lord_Farin (talk) 17:39, 15 November 2017 (EST)

New templates for specific unfinished tasks

This is to notify you that I have created/am in the process of creating templates for some specific types of tasks that often need to be done. Examples are {{ExtractTheorem}}, {{rewrite}}, {{ExpandList}}, {{IncomingLinks}}, {{LinkWanted}}, {{DefinitionWanted}}, {{image wanted}}, {{MissingDoc}}, {{AxiomReview}}. (Will open a discussion on template naming soon.)

The idea is that categories like Category:Refactoring In Progress, Category:Stubs, Category:Work To Do (previously WIP) do not distinguish e.g. between straightforward repetitive tasks or tasks requiring knowledge of a certain subject. Or simply, it can be pleasant if you can choose a specific task depending on your mood. Being able to focus on one type of tasks can also help you work more efficiently. --barto (talk) (contribs) 13:12, 12 November 2017 (EST)

They will be listed at Help:Templates; comments are welcome, including requests for deletion. --barto (talk) (contribs) 14:38, 12 November 2017 (EST)

I haven't focused much on design, because I personally would be happiest with a Scribunto-based solution (cf. the WMF mbox) that takes care of issues like when to add categories to the target page. --barto (talk) (contribs) 14:41, 12 November 2017 (EST)
Where applicable, I have used the same template names as those at WMF projects. The idea was to facilitate possible future collaboration, in the form of people who contribute to different projects, or importation of templates. --barto (talk) (contribs) 16:23, 12 November 2017 (EST)
No, we're going to stick with CamelCase. --prime mover (talk) 18:45, 13 January 2018 (EST)

Naming conventions for templates

Most template names are either CamelCase or all-caps abbreviations, exceptions being {{inuse}}, things like {{AoC}} some POTW templates, and recently created ones that led to this discussion. See here for all templates.

What (I think) may have been the reason to choose CamelCase, is the idea that templates play a role similar to functions in a programming language, whose names cannot (usually) contain spaces.

Personally, I find CamelCase harder to type and titles with spaces easier to read, especially when template names get longer. --barto (talk) (contribs) 13:20, 12 November 2017 (EST)

I also appreciate the convenience of typing lowercase template names. Especially for quickly added maintenance categories. However I would also strive for short template names, to make typing even easier. Maybe the maintenance task of "detailing maintenance template usages" should be acknowledged on its own.
I was pondering if there might be a value in distinguishing the maintenance templates from structure/content templates. Because somehow to my eye, a capitalised template name shows up more as a thing that does something else than writing text. But maybe that's just me. Or the desire for brevity makes it just more tempting to use acronyms, which are naturally capitalised.
Further ideas welcome. — Lord_Farin (talk) 17:39, 15 November 2017 (EST)

Capitalizing of non-content categories

When I recently organized the Categories, I did not capitalize the names of meta-categories, i.e. the categories that are not descendants of Category:Definitions by Topic or Category:Proofs, i.e. those that are not manually added when editing "content pages" and only when organizing ProofWiki.

This is a departure from the existing style, and thus understandably frowned upon.

I saw that some maintenance categories are automatically added by MediaWiki, e.g.:

Category:Pages using duplicate arguments in template calls

These are hard-coded in the MediaWiki code (quote from mediawikiwiki:Manual:$wgTrackingCategories#Details); I don't know if they can be changed. The main argument to capitalize all categories seems to be consistency with content categories. It is indeed arguably strange to have both styles. This was briefly discussed at #Category definition template (blocks 3-4). --barto (talk) (contribs) 15:15, 12 November 2017 (EST)

Thanks for your explanations. I will sadly not have much time before Wednesday evening to go into everything, but at least it's now all made explicit and discussable. — Lord_Farin (talk) 16:09, 12 November 2017 (EST)
Having looked at it a bit the past days, capitalised categories have my definite preference. — Lord_Farin (talk) 17:39, 15 November 2017 (EST)
Ok. I'll take care of any inconsistencies. We can live with those few automatically created lowercase categories. --barto (talk) (contribs) 06:08, 16 November 2017 (EST)

Lowercase redirects to definitions?

What about redirects like group actionDefinition:Group Action?
For definitions only: not Parseval's theoremParseval's Theorem.
Specific links only: not continuousDefinition:Continuous Mapping (Topology). --barto (talk) (contribs) 16:13, 12 November 2017 (EST)

What about... "No". I think the efforts should now be in reducing clutter, not increasing it. That is, reducing the load of maintenance and ambiguity for prospective editors; not increasing it by having strange cross-namespace links that mostly are not there but sometimes are (which would be the net effect unless there were to be a project to "improve" current linking). It is a matter of time till someone starts polluting the main namespace with definitions "because that's how I (mistakenly) thought it was supposed to be done".
Actually I find it completely unproductive to consider this assimilation process any further. We are not reconsidering the namespace conventions, we are not reconsidering the capitalisation conventions for "main" namespaces. Too much impact for unproven benefit.
Instead please consider to invest time to facilitate an easy AutoHotkey script or whatever. This kind of redirect is a nightmare and an utter waste. — Lord_Farin (talk) 11:42, 13 November 2017 (EST)
I was about to add "Disclaimer: this is only meant to make linking easier and source code cleaner, the intention is not to change naming policies." :)
One nonnegative consequence would be that definitions would still show up in search results after CirrusSearch is installed. --barto (talk) (contribs) 12:15, 13 November 2017 (EST)
But the arguments about namespace mistakes and, to a lesser extent, proliferation of links, were also my main concerns. --barto (talk) (contribs) 12:58, 13 November 2017 (EST)

Allowed file extensions and InstantCommons

Can svg be allowed? --barto (talk) (contribs) 05:25, 16 November 2017 (EST)

Or we could allow InstantCommons, whose configuration seems to be the simplest I've ever seen :) That way we don't have to look for icons, download and upload them, and worry about copyright. --barto (talk) (contribs) 09:26, 16 November 2017 (EST)

How about both? I've enabled svg for now until others comment on InstantCommons. --Joe (talk) 10:02, 16 November 2017 (EST)
Why not? Can't see any reason not to enable both. Caliburn (talk) 10:44, 16 November 2017 (EST)
We need to be sure there are no licensing issues and that loading images from InstantCommons does not "overload" any of our own images (is there a prefix or anything, what is the search order).--Joe (talk) 11:46, 16 November 2017 (EST)
The former will be fine. The file description stored on the Commons will be stored as the local description until overwritten. Licensing and sourcing is done within this description. As to how it's used, you can call it like any local image. I am however unsure about file name conflicts, admittedly that did not cross my mind. Caliburn (talk) 13:18, 16 November 2017 (EST)
It says "Local filenames take precedence over Commons filenames." Avoiding name conflicts does seem preferable regardless. Don't know if there's a tool to find them. --barto (talk) (contribs) 15:58, 16 November 2017 (EST)

Categories for theorems with multiple proofs

It is a relatively recent practice to create a category for each theorem that has multiple proofs. To make sure everyone is up to date, a quick summary of how this is done:

  1. The category is given the same name as the theorem
  2. Theorem is placed in that category
  3. Proof subpages are placed in that category and no other
  4. The category is placed in the same categories as the theorem

As I understand it, the purpose of this is to (1) avoid overloading the various proof categories by having >=3 entries for one theorem and (2) avoid Uncategorized Pages.

I'm not very happy with #4, because giving theorems the status of subcategories makes the category tree disorderly: If you look at Category:Complex Numbers, you'll see that in fact only 3 of the subcategories are really subcategories, the others are theorems with their proofs and are repeated in the "Pages" section below.

Alternatives are:

  1. Keep doing #1-3, but place all those categories in Category:Theorems with Multiple Proofs and no other.
  2. Stop doing #1-4, and create a (customizable) {{proof subpage}} template that does something to avoid Uncategorized Pages. The template could also be used to automatize the relatively cumbersome procedure #1-3. --barto (talk) (contribs) 10:59, 18 November 2017 (EST)
Maybe add a sort key to the subcats separating them from theorems or vice versa? A sort key of an asterisks, say, would put the category ahead of those sorted alphanumerically. Did a quick example with fuzzy set theory and Category:Set Theory. Let me know what you think. This would create a major maintenance task though, best left to a bot if possible imo. Caliburn (talk) 12:28, 18 November 2017 (EST)
Edit: this is done either using {{DEFAULTSORT:Key}} (which will apply the key to all categories that the category/page is a member of) or [[Category:Category|Key]], (which will apply the key to only that category) forgot to add. Caliburn (talk) 12:29, 18 November 2017 (EST)
Rather not to subcats, as it breaks alphabetic sorting. Changing sorting only makes the cat tree look a bit less messy, so I'm still in favor of hiding those thm-cats away. After all, the thms will still appear in their categories, under pages (where there supposed to be), so nothing will be lost. --barto (talk) (contribs) 12:43, 18 November 2017 (EST)
Alright, sounds good. At least that'll be far less of a monumental maintenance task. Caliburn (talk) 13:23, 18 November 2017 (EST)

Difference between Help and Proofwiki namespaces

Does it make sense to say:

Proofwiki: is for information about ProofWiki and other administration (e.g. POTW);
Help: is for information about how to use ProofWiki (edit, search, ...)? --barto (talk) (contribs) 11:28, 20 November 2017 (EST)
The need to formalise the distinction eludes me. Aren't there more relevant projects in need of attention? — Lord_Farin (talk) 12:05, 20 November 2017 (EST)
I wanted to give Help:Editing/House Style a separate page, chose ProofWiki: having a certain namespace distinction in mind; then realized that Help: may have been a better choice and that the situation can use some clarification. --barto (talk) (contribs) 12:16, 20 November 2017 (EST)
It's not a big thing, just want to clarify it. --barto (talk) (contribs) 12:19, 20 November 2017 (EST)
Was there a discussion on moving this page? I'm not sure I'm sold on it being the PW namespace. --Joe (talk) 12:22, 20 November 2017 (EST)
@barto Changes like this should be first discussed on the relevant talk page. I understand that you are enthusiastic but please take other contributor's opinions and schedules into account before making structural changes. --Joe (talk) 12:24, 20 November 2017 (EST)

I just want that you get your hands off restructuring some of the main pages on this site without discussion. This is not your personal project and the iterated violation of this polite request has just earned you a one-week block. You don't just "want to give Help:Editing/House Style a separate page" and especially don't proceed when you are not sure about how these different namespaces work.

I'm sad to take this decision (as you are bringing quite some quality contributions) but the friendly way did not work repeatedly. There is no choice remaining. — Lord_Farin (talk) 12:25, 20 November 2017 (EST)

Loss of a contributor

I have just received the sad news that Grzegorz Bancerek, who has made great contributions towards the exercise of bringing aspects of $\mathsf{Pr} \infty \mathsf{fWiki}$ and the "Mizar" project into alignment has recently passed away.


--prime mover (talk) 04:28, 23 November 2017 (EST)

Tragic news. Deepest condolences to his friends and family. No doubt he'll always be remembered in this community for his superb contributions. Caliburn (talk) 10:28, 23 November 2017 (EST)

Subject Classification

I want to suggest a subject classification page similar to the Mathematics Subject Classification 2010 or the tags system used on . The classification page would help users to find definitions, theorems and exercises of interest as readily as possible. It would also help to guide users to other potential interesting proofs and results they may not be aware of. I also think that a classification based on topics which are usually covered in an undergraduate and graduate program would help the site become more user friendly for students. I don't know if such a feature already exists because it doesn't seem to be visible.

--Ddanndt (talk) 09:53, 28 November 2017 (EST)

We did make a start at MSC, a long while ago, but the mountain was too steep to climb. As for stackoverflow, we don't want to get too tied up with another website, as ultimately all websites are ephemeral, and we'll be left supporting a structure of a non-existent hollow. We've been burned like that in the past. --prime mover (talk) 14:47, 28 November 2017 (EST)
For the current scope of the website, it seems to me that the notion of categories should do. See Category:Proofs. — Lord_Farin (talk) 15:35, 28 November 2017 (EST)

Template reorganisation

I grew tired of the mountains of work remaining regarding Template:CategDef and Template:Subcategory. They have been given implementations calling the respective superseding templates. This breaks some pages. It fixes and improves many others. There are about 800 calls remaining. They will be processed as and when I find enough empty brain cycles. — Lord_Farin (talk) 18:00, 11 December 2017 (EST)

Yes, that was a seriously substandard design. Better is expected of our contributors, who are supposed to be smarter than that. --prime mover (talk) 18:19, 11 December 2017 (EST)
The only constant in software development is change. Even when software development is crafting MW templates. Therefore it is not surprising that over time, templates change. But I somehow suspect you know all that and just want to make a point that the old was not so bad. No, it wasn't. It was good, but not perfect, and something new came along. The new won't prove perfect either. Such is life. — Lord_Farin (talk) 18:33, 11 December 2017 (EST)
Template:CategDef has been successfully eliminated. It took some time but we improved the structuring of categories by doing so. — Lord_Farin (talk) 14:52, 26 February 2018 (EST)
It took a further enhancement of the superseding templates (they are now built in Lua/Scribunto), but now also Template:Subcategory is replaced everywhere by Template:DefinitionCategory and its variants. — Lord_Farin (talk) 06:02, 10 May 2018 (EDT)

Hidden category category

Despite MediaWiki:Hidden-category-category hidden categories are still placed in Category:Hidden categories. --barto (talk) (contribs) 08:21, 24 December 2017 (EST)

Do we need it? --prime mover (talk) 08:43, 24 December 2017 (EST)

Deleting redirects to moved pages?

When I move a page, especially one with subpages, and delete the redirect (which in the case of a subpage should no longer be used) I often think there may be external links to it (forums etc) which will be broken. Should be keep those redirects for a certain period? Thoughts: 1) if we delete them someday, they will be broken someday anyway. 2) nonbroken links increase page rank. 3) on some (not all, cf. StackEx) forums, old posts (with broken links) are hardly ever read again. --barto (talk) (contribs) 17:03, 8 January 2018 (EST)

When moving high-profile pages, it might be good to keep this in mind. To be explicit one can investigate the amount of links to such a page using a search engine. On the other hand, life moves on and the history of the page is preserved, so it's not a huge concern. — Lord_Farin (talk) 13:17, 12 January 2018 (EST)
It's a higher-priority matter to make sure you don't break links to pages on $\mathsf{Pr} \infty \mathsf{fWiki}$ itself. I have found rather more places where this has happened than is optimal recently. Another priority is to make sure the source work links are not compromised either -- which is, I admit, a more challenging task. Again, it is too easy to break the flow by injudicious renaming and shuffling things around. --prime mover (talk) 15:17, 12 January 2018 (EST)

Focus welcome message more on writing style?

We've seen it many times: There's a new user who doesn't know about our writing style and adds a proof with blocks of text, words like "evidently", parenthetical remarks and too few wikilinks. My first contribution looked the same, and I only became aware of the conventions when someone tidied it.

The link to Help:House Style in the welcome message seems little effective. Suprisingly, new users don't read that entire page before they start writing. [irony] I believe most users don't even read the welcome message, or at least have the misconception that «yeah all right that help page will just have some typical remarks I could guess».

What I suggest is to focus the welcome message more on the fundamental guidelines:

  1. you cannot write in any style
  2. do use wikilinks

with some bold words and a link to a shorter help page, where these are repeated and which refers you to the general house style guide for more. Some filler words to keep it friendly. Putting the word important in bold can work like magic.

The idea is that, while everything a new user needs to know is in the house style guide, we can introduce them to it in a more efficient way. --barto (talk) (contribs) 06:44, 14 January 2018 (EST)

Worth a try, I suppose. Could you draft something in your user space? — Lord_Farin (talk) 07:32, 14 January 2018 (EST)
Wouldn't bother, myself. The first posting by a new user is *always* substandard, unless they are extremely careful and take their time to read all the house rules before they start. People don't want to read this stuff, no matter how loudly you shout at them. The only thing to do is to correct their style and hope they learn. Sometimes they do (in which case good), and sometimes they don't (in which case we just routinely slap a Tidy on every page they produce and go back to it when time and inclination coincide. (Some people even justify their non-conforming attitude by insisting either that their way is better, or that what they are doing is too important to attend to such details, in which case we just shake our heads, let them get on with it, and hope they don't stay around long.) --prime mover (talk) 07:43, 14 January 2018 (EST)
I mostly share this point of view, but decided to park my reservations for a moment. But I'm also not sure if there will be any positive effect. Maybe because it's easier for people to find the house rules after they find out that they are different from "whatever I think best". — Lord_Farin (talk) 07:57, 14 January 2018 (EST)

Big Proofs

The question was asked the other day: "what is the recommended technique for structuring big proofs (of the order of 100 pages)?"

While we haven't got to the stage of posting up proofs that long, it is probably worth codifying our technique for proofs which benefit from being broken up into sections.

As an example of this, I offer up the Completion Theorem (Metric Space), which has been broken down into 4 lemmata, each of which has its own page. Please feel free to study, complete the details, and comment on the workability of this approach, as it can be suggested that it be used as a paradigm. --prime mover (talk) 07:08, 27 January 2018 (EST)

I think this approach makes sense. Furthermore I think it can be applied to more pages than just big ones, if amended with the habit of introducing a "proof summary" or "outline".
I would be interested in ideas how to make that fit in consistently with the rest of the site, with multiple proofs, etc.
A point that needs to be addressed is how the "big proof" strategy is handling multiple proofs (with current subpages being called /Lemma X, which doesn't suitably generalise).
I don't see a big problem with that.
a) Multiple proofs of a lemma: Big Proof/Lemma 1/Proof 1, /Proof 2, etc. etc. We only show the result in the main page, so the multiple proofs are hidden under the engine housing.
b) Multiple proofs of the big proof itself may well both use Lemma x, Lemma y and Lemma z, but in between one uses Lemma p and the other uses Lemma q, for example. The modular approach still functions.
We have in fact done that sometimes, when we have an iff proof, where the necessary condition has n proofs and the sufficient condition has m. We just separate out those unidirectional proofs as separate pages, and multiproof them in this way. I haven't come up against a proof that can't be handled like that.

That having been said, if the lemma approach is handled suitably, it can be a good addition to the site (as opposed to the "verbose chapter with intermittent lemmata and a nebulous proof claim to conclude" seen in some textbooks). — Lord_Farin (talk) 12:11, 28 January 2018 (EST)

Hence $\mathsf{Pr} \infty \mathsf{fWiki}$ which tries not to do that. --prime mover (talk) 18:18, 28 January 2018 (EST)

Topology Refactoring

As a result of recent discussions and suggestions, the Topology work has been significantly refactored:

  • Definitions now mainly define nouns, rather than adjectives. This has unfortunately resulted in more fragmentation than might have been anticipated, where we have, for example, "Connected Space" and "Connected Set" rather than the rather more all-embracing "Connected" which would do for both concepts.
  • Further to the above, I have experimented with two approaches regarding "space" and "set", where the former concept refers to a specific topological space in its entirety, and the latter as a subset of such a space -- hence, following on from the above point, "Connected Space" refers to a connected space and "Connected Set" refers to just a set within that space. This has led to some rather more unwieldy distinctions in the body of the text to ensure the correct concept is linked to -- and there also exists, with this added further fragmentation, that any specific given result may be more difficult to find, as it may be in the "Space" category instead of the "Set" category.
  • The big question: "Set" or "Subset"? The original work on $\mathsf{Pr} \infty \mathsf{fWiki}$ defined the concept "Open Set" rather than "Open Subset", because that was the term that could be found in the works immediately at hand to the author of those pages. While it is appreciated that an open set is of course indeed a subset of the underlying set of a topological space, the term "open set" (and similar such terms) embraces that fact within its definition, and so "open subset" adds unnecessary length to a basic concept. In fact, the difference is more philosophically subtle than that, as the definition of a "subset" has no other structure than the fact that "each element of one set is in another", and the structure of the topological space into which that subset is embedded is not taken into consideration. It is only when you consider whether such a set is "open" or "closed" (or neither or both) that the topology on the space is considered. Hence an "open set" cannot be discussed without the topology also being discussed, while a "subset" is independent of the topology in which it is embedded.
  • Notation: I have tried to be consistent (work is still in progress here) and refer to a general topological space as "$T = \left({S, \tau}\right)$" when first introduced, rather than just call it, say, "$X$". This is for two reasons: a) consistency, and b) to allow the subtle distinction between (see the above point) a "subset of $S$" and a "set in $T$". In my mind, this is the greatest source of confusion for students of topology: confusing the topological space itself (that is: $T$) with its underlying set (that is, $S$). We have had discussions in the past about how the extra symbols in "$T = \left({S, \tau}\right)$" cluttering up the page when in fact only $T$ is referred to within the body of the rest of the page, but I believe that the extra detail in the specification improves clarity, as does a rigorous separation of statements one per line on the page.
  • The work may still be incomplete -- please feel free to refine it, or discuss the results with a view to improving it.
  • There are other points to consider, in particular the distinction between "Locally (property)" and "Weakly Locally (property)" which is an important avenue to investigate, as it appears in the literature to have been glossed over. While I requested that work on this area be placed on hold while I did that refactoring job, please feel free to continue. --prime mover (talk) 05:43, 28 January 2018 (EST)
Your work is much appreciated and definitely adds to the clarity and maturity of Topology as a subject covered on $\mathsf{Pr} \infty \mathsf{fWiki}$. Thanks! — Lord_Farin (talk) 12:11, 28 January 2018 (EST)

To coin a new notation

How convenient would it be to have a symbol understood in set theory to be interpreted as the binary relation "is disjoint from"?

Thus, say:

$A \mathrel{\overline \varnothing} B$

would mean:

$A \cap B = \varnothing$

Does anyone know of a usage of such a notational convention anywhere in the literature? I haven't been able to find one. Are we up for innovation? --prime mover (talk) 02:05, 13 February 2018 (EST)

Don't see the need for it. Could be that there is a symbol already. This is also discussed here on MSE and here, where the OP similarly combined symbols into a new one, and one of the commentators suggested $\perp$ instead. I haven't found this to be used anywhere (searched MSE and MO). --barto (talk) (contribs) 05:42, 13 February 2018 (EST)
Reading those appalling posts from, quite frankly, typical mathematicians, I despair of the universe.
Okay, forget I spoke. I have been schooled. When can I die? --prime mover (talk) 07:43, 13 February 2018 (EST)

What next?

I'm tired of topology. I don't know what to do. Anyone got anything they want me to do? --prime mover (talk) 02:12, 22 February 2018 (EST)

Maybe the direction of Fourier analysis could be something? We're definitely lacking there... and it might provide fertile soil for future developments in the direction of function spaces ($L^2$ and the like).
It could be that our analysis section is lacking rigour and cohesion to support this endeavour, but this is then something to be investigated. — Lord_Farin (talk) 12:40, 25 February 2018 (EST)
I believe I have one or two texts on fourier analyis, I'll take it on. --prime mover (talk) 02:23, 26 February 2018 (EST)
I've made a start, and the groundwork is there, but it needs restructuring so that the "most general" is at the top, and the various simpler forms are implemented as special cases. --prime mover (talk) 04:13, 14 March 2018 (EDT)
As has been pointed out, I'm not able to do this after all, it's too difficult for me. Not sure what I am capable of, it's sure not maths. --prime mover (talk) 17:27, 22 March 2018 (EDT)

Examples -- how do we categorise them?

Before I go and do a lot of rework concerning how we categorise examples ... how are we going to manage "examples" categories?

a) as we have for metric spaces: "Examples of Metric Spaces"
b) as we have done for groups: "Groups/Examples" (and all their subgroups)
c) as I have just done for topology "Topology/Examples" (as opposed to "Topologies/Examples")
d) as we have done for norms "Norm Examples"

That is:

a) Examples of (Plural Object Type)
b) (Plural Object Type)/Examples
c) (Mathematical Branch)/Examples
d) (Singular Object Type) Examples

All have merit, and there will be plenty of rework consolidating. And yes I know we have probably discussed this before, but clearly nothing was decided.

Once we have decided, we can go and rework the category templates and move forward. (And yes I understand that in this post I have deliberately conflated the grammatical "we", the royal "we" and the kindergarten "we".) --prime mover (talk) 02:17, 23 February 2018 (EST)

I think a) generalizes the best. Although we could do with the sort key being automatically defined as "Plural Object Type". Another alternative in a similar vein would be:
e) Examples/(Plural Object Type)
in similarity to "Definitions/...". However I don't have a strong opinion either way. What is however important is that we straight away start using appropriate templates to link to and populate these categories. This will prevent future frustration. — Lord_Farin (talk) 12:40, 25 February 2018 (EST)
I have amended DefinitionExamplesCategory accordingly, and I am rationalising everything now. --prime mover (talk) 15:46, 25 February 2018 (EST)
Cool. Hold on a sec on applying Template:ExampleCategory everywhere, then I hook it up just like Template:DefinitionCategory (using def parameter for definition, and unnamed params for parent transclusion). — Lord_Farin (talk) 15:57, 25 February 2018 (EST)
I think the aforementioned template is now more or less up to shape to be used.
A slight downside of the current setup is that it is not so easy to link with just the MediaWiki tools to or from an examples category, because it is difficult to link the names together. However I don't think that this can easily be avoided. But sadly it means a bit more manual work in tying everything together... — Lord_Farin (talk) 16:25, 25 February 2018 (EST)
I have renamed all the categories using existing invocation of the DefinitionExamplesCategory and ExampleCategory template.
There is a small problem with ExampleCategory. If you use the unnamed parameter technique to link to other categories, then it does not include the automatic link to the "Examples" category, which is a bad thing (the latter is sorted on entity type, so that everything is not put into "E").

This has now been changed. Category:Examples is now always included. — Lord_Farin (talk) 12:51, 26 February 2018 (EST)

Bear in mind the original reason for using the "unnamed parameter" technique in the original design of Subcategory template was only to make sorting easy in the parent category (note it has to use the {{#titleparts: {{SUBPAGENAME}} | 1 | 2 }} to sort it properly) and so linking as unnamed parameters was expedient. It seems like overkill to do the same thing for normal categories where the sorting is by default. --prime mover (talk) 18:28, 25 February 2018 (EST)

Point regarding this is that, as you can see, we now use the def parameter to guide sorting of categories. This to avoid problems with all categories being sorted as "Examples of ..." under E. Maybe we can suffice with only including Category:Examples automatically (with the mentioned sorting key, or possibly even adding a sortkey parameter) and leaving all the others to be added explicitly using standard syntax (and hence sorted under E). What do you say? — Lord_Farin (talk) 12:51, 26 February 2018 (EST)

If we can make all "unnamed parameters" sort on the def parameter like we do with Examples, and yes, okay, override def with sortkey if you can think of any instances where this is needed, then yes, that would work. --prime mover (talk) 13:54, 26 February 2018 (EST)
On the other hand, do we want this? Take "Examples of Semigroups", for example. In "Examples", it makes sense to stick it under "S" for Semigroups. But in the Semigroups category, it may make better sense to stick it under "E" as it is the "examples"-ness of it that makes it interesting in the category that is already about Semigroups. --prime mover (talk) 13:59, 26 February 2018 (EST)
This was exactly what I was thinking. Should we make things more transparent and remove support for it? — Lord_Farin (talk) 14:48, 26 February 2018 (EST)

It keeps it simple. I'm all for simplicity. In this context a template to add categories in a vanilla style is probably over-complication. At least, in my view. --prime mover (talk) 17:09, 26 February 2018 (EST)

The template has been amended accordingly. — Lord_Farin (talk) 15:07, 28 February 2018 (EST)
Oh yes, and there is also a problem with SubjectCategoryNodef which fails to transclude [[Definition:pagetitle]] when there is no parameter. Don't know why. --prime mover (talk) 18:30, 25 February 2018 (EST)

SubjectCategoryNodef was designed to not transclude the definition, that's why the template exists. But maybe I miss something here. — Lord_Farin (talk) 12:51, 26 February 2018 (EST)

No, it's designed not to provide a link to the definition category, like when the definition does not have its own category to sit in, but instead sits in a larger category. --prime mover (talk) 13:54, 26 February 2018 (EST)
I've found the mistake. I caused it myself, my apologies. I have included an explanatory comment to prevent it from reoccurring. — Lord_Farin (talk) 14:48, 26 February 2018 (EST)

Okay, the categories have had an overhaul.


CategExampNodef (which is the same in effect as a SubjectCategory template but links to a result page, not a definition page)
Some instances of Subcategory (in particular the Category:Definitions/Groups of Order 4 etc. categories, which may need the same sort of treatment as the GroupExamples categories)
Some other instances of Subcategory where the category linked to is an Axiom category

It's all a bit better than it was, but there may be some construction debris lying around. Not much, I hope. --prime mover (talk) 17:09, 26 February 2018 (EST)

Too many redirects

Does anyone fancy going through and deleting the ridiculous number of pointless redirects that have exploded like a ruptured boil over the website? --prime mover (talk) 19:31, 9 March 2018 (EST)

There are some arguments for keeping them, like better search results both here and in search engines. Which kind of redirects are you thinking of, specifically? --barto (talk) (contribs) 04:45, 10 March 2018 (EST)
There's so much stuff to wade through I can't find the examples that stood out for me now. I'll get back to you on this. --prime mover (talk) 05:24, 10 March 2018 (EST)
Yes, here goes: Definition:Intersection of Set. --prime mover (talk) 05:34, 10 March 2018 (EST)


I've changed the way we load MathJax and updated to the latest version. It will now read its configuration from MediaWiki:MathJax.js so it's easy for admins to modify. I haven't yet found any downsides to the new implementation compared to the old one, but if something is broken please let me know. --Joe (talk) 20:08, 13 March 2018 (EDT)

What immediately comes to mind are the shortcuts like $\rd$ and $\HH$ and so on which I did not know about. Good job.
We might also want to consider relaunching the project to create further shortcuts, which fell over some years back through lack of coordination.
Not me, not now, I have an ongoing family crisis that will take a few days to resolve at the moment. --prime mover (talk) 04:12, 14 March 2018 (EDT)
I'm not sure if this is related to the update but the cases environment now seems to be a bit broken when indented. Example 1, Example 2. Might be worth taking a look to see if other environments have problems. Using Chrome 65 if that matters. Caliburn (talk) 12:01, 14 March 2018 (EDT)
The issue seems to be a switched order between MathJax and MediaWiki. Formerly, the dollar sign environments were grouped before MediaWiki rendering. This meant that:

The code following the above sentence has been commented out because it broke the page. This needs to be resolved. --prime mover (talk) 18:16, 25 March 2018 (EDT)

This should now be resolved. —Joe (talk) 10:14, 27 March 2018 (EDT)
Not sure if it has. This is on Basel Problem/Proof 9:
:$f \left({x}\right) = \begin{cases}
\left({x - \pi}\right)^2 & : 0 < x \le \pi \\
\pi^2 & : \pi < x < 2 \pi \end{cases}$
This currently gets rendered like this:
$f \left({x}\right) = \begin{cases}

\left({x - \pi}\right)^2 & : 0 < x \le \pi \\ \pi^2 & : \pi < x < 2 \pi \end{cases}$

So we're back to where we were. The fix you did undid the fix that fixed this. --prime mover (talk) 18:13, 27 March 2018 (EDT)

\complicated \TeX$
worked, because the $ were connected. However now MediaWiki takes precedence and accordingly indents the first line more than the second. Obviously the cases environment is an obvious candidate for this situation. Joe, can you look into this? — Lord_Farin (talk) 14:20, 14 March 2018 (EDT)
Ok, doing some more complicated logic and fixed the ordering. Let me know what I broke. --Joe (talk) 15:24, 14 March 2018 (EDT)

Take a page, any page. Click on the "Page" button top left. The MathJax rendering is no longer performed, and you just see the raw LaTeX. Then try this: navigate to Basel Problem. It renders the MathJax, but not the rich and pretty style we know and love, but in a thin and manky style that looks inferior. If you can get it to render at all. It's not rendering Basel Problem at all at the moment. --prime mover (talk) 15:34, 14 March 2018 (EDT)
Yeah, just noticed. It's a caching issue I didn't see as dev env doesn't use caching. Working on a fix. --Joe (talk) 15:36, 14 March 2018 (EDT)
The caching issue should be resolved now. --Joe (talk) 22:49, 14 March 2018 (EDT)

Additionally, the XyJax package seems not to be included. See Help:LaTeX Editing/Commutative Diagrams. I know it has been deprecated for a long time, but still... I don't know of any other package enabling the same functionality. — Lord_Farin (talk) 15:37, 14 March 2018 (EDT)

XyJax is now working again! --Joe (talk) 19:51, 14 March 2018 (EDT)
Aweseome, good work on the upgrade! — Lord_Farin (talk) 16:28, 15 March 2018 (EDT)

Another issue just surfaced [here] and I confirmed it myself: on mobile, the size of the MathJax is not proportional to the rest of the text. Also, not everything seems to be properly rendered to begin with (I saw some white squares). I wonder what the problem could be, no obvious ideas. — Lord_Farin (talk) 06:35, 17 March 2018 (EDT)

Just seeing this on mobile, right? Chrome and Safari on my iPhone seem fine (they use the same engine), what browser/OS are you using. Maybe it's an issue with the new MathJax version or the default config we're loading? --Joe (talk) 12:34, 17 March 2018 (EDT)
We should also take a closer look at the Combined Configurations page. We're currently using TeX-AMS-MML_HTMLorMML, maybe it's not the best default for us. Any thoughts? --Joe (talk) 12:52, 17 March 2018 (EDT)
Not sure what could be the problem here. I'm using Chrome on Android (5.1, if it matters (yes I know that's old; blame Motorola)) and I am guessing that the unfortunate user who surfaced on maths.SE does the same. But on my desktop everything looks fine. — Lord_Farin (talk) 16:17, 19 March 2018 (EDT)

I'm sad to say that there is another problem popping up, one could say it is the converse of the problem with the MathJax/MW order. See Definition:Constructed Semantics/Instance 1. The eqn template should be expanded before the MathJax rendering is attempted. Hopefully this is possible while keeping the other problem solved as well... — Lord_Farin (talk) 14:49, 21 March 2018 (EDT)

Annoying, but the workround is not to nest latex into a \text. It's neater not to use an "if" anyway, as the structure of a case statement implies the if, making it redundant. If you do feel you need it, put just the "if" in a \text on its own and close. --prime mover (talk) 17:39, 21 March 2018 (EDT)
There was an issue with recursion, should be fixed now! --Joe (talk) 19:42, 21 March 2018 (EDT)
Awesome, thanks! — Lord_Farin (talk) 03:51, 24 March 2018 (EDT)

Two comments on brackets in MathJax

There are two longstanding points on brackets I want to bring up:

  1. The use of curly braces in addition to brackets. Examples:
    \left({\frac 1 2}\right) instead of \left( \frac 1 2 \right)
    \left\{{x : x \in S}\right\} instead of \left\{ x : x \in S \right\}
  2. The use of \left-\right delimiters for brackets with few enclosed tokens. Examples:
    f\left(x\right) instead of f(x)
    2 \cdot \left(3 + 4\right) instead of 2 \cdot (3 + 4)

I understand that:

  • 1. is intended for users of a specific Firefox plug-in, as explained in the FAQ, which does not otherwise highlight brackets.
  • 1. and 2. avoid forgetting to close every bracket (because the math processing breaks if a closing \right or } is omitted).
  • 2. because of the wish to have one rule ("use left-right") apply to all cases.
  • 2. because some people find the extra spacing created with left-right more visually appealing.

The motivation is questionable. First, there are definitely other, entirely customizable and free text-editors out there. Put differently, if the editor does not understand LaTeX syntax, I'd say change the editor, not the syntax.

People use whatever editor they prefer. Editors with syntax highlighting may not be common, but I can't see why this should not be the case in the future.
I understand that some people may use something like texNicCentre or whatever it's called, which does have extensive syntax highlighting and editing helpers and macros and the like, which, etc. etc.
Actually you are misreading the primary motivation for 1.; it clearly is "every \left is paired with the intended \right", as you conveniently omitted. Or more generally that braces match up. More below. — Lord_Farin (talk) 13:15, 19 March 2018 (EDT)

Regarding left-right, I'd be happy to move from "because I think it looks better", "because you might be dull and forget to close a bracket [once every 500 times] if you don't follow this rule", "because it's tradition" to case-by-case decisions based on good judgment and the question: Will left-right make the math look better [structurally], or, the source code more readable? (i.e. as soon as they contain some 5-10+ish tokens, or if there is vertical material). After all, forgetting to close a bracket is only like a typo. See one, click 3 times, it's gone. (Here's a fun question: is a user more likely to forget to close a bracket, or to use left-right + curly braces when writing f(x)?) Personal preferences will always cause slight frustration depending on how perfectionist one is; the idea is to let good judgment and tolerance take precedence. --barto (talk) (contribs) 06:33, 18 March 2018 (EDT)

"case-by-case decisions based on good judgment" does not translate into automatic proof-writing tools, which is an ongoing project being undertaken by a number of research teams worldwide, some of which (e.g. Mizar) are particularly interested in constructing an interface between us and them.
Besides, I don't trust "good judgment" -- we have a plethora of pages written by a particularly disagreeably arrogant contributor a while back who refused point-blank to make any attempt to tidy up the pages he wrote, because he has more important things to do. As a result, not only the source code but also the actual appearance of the finished pages look appallingly unfinished and careless.
I, too, am skeptical of the rigour with which one will ask oneself the question about making source code more readable. As any developer can confirm, readability is not something usually taken into account (by other developers, obviously).
Moreover I am annoyed by the proposal to replace an existing rule which is perceived to be poorly motivated, by a suggestion that uses hyperbole like "once every 500 times". If you have good arguments, why the exaggeration? It just distracts from the point you're trying to make.
What would have been wrong with a simple suggestion that the strictness be brought back to "at least when there are brackets or long compound expressions inside a pair of brackets"? This could have been argued to be reasonable and proportional. Regardless the approach chosen I consider the braces to be an integral part of the guideline as they safeguard against confusing pairing problems. — Lord_Farin (talk) 13:15, 19 March 2018 (EDT)
Thanks for the replies! I'll answer with some further thoughts shortly.
PS: I could definitely have made the proposal simpler, as you say. I meant to skip the part where someone explains the pros and cons to me after which I get to reply...
PPS: Woops, didn't mean to exaggerate. You can replace the parenthesis by [which doesn't occur that often] if you wish. I actually did made a quick calculation when I put that number:) Anyway, so much for the presentational style:) --barto (talk) (contribs) 15:03, 19 March 2018 (EDT)
Bottom line is that if you consistently refuse to adhere to our house rules, you will end up being barred, at least until such time as we clean up the mess you make.
I'm not sure we'll lose so much, you not being around, just hundreds of unsourced and unverified definitions, with practically no proofs to use them. --prime mover (talk) 17:58, 24 March 2018 (EDT)
I must have misread that sentence at the FAQ, thinking it was about \left-\right instead of {}... Which makes I don't understand that part now: LaTeX by default requires a \right for each \left, so I'm thinking it's about avoiding situations like \left( \left[ \right) \right], do I understand correctly? --barto (talk) (contribs) 10:04, 25 March 2018 (EDT)
$\left({\left[{...}\right)}\right]$ is completely valid $\LaTeX$ should you care to write it.
What we are guarding against is things like $(f(s, g(t, h), j(t(a, b))))$. --prime mover (talk) 18:14, 25 March 2018 (EDT)

Piecewise Continuity

Anyone who has an interest in this: please feel free to comment on the direction I'm taking the definition of Definition:Piecewise Continuous Function.

Incomplete yet, still working on it. --prime mover (talk) 04:17, 19 March 2018 (EDT)

Take care before deleting

A quick reminder to anybody who does refactoring work:

Remember, remember: before you delete the automatically-generated redirect, always do a "What links here" to make sure that all pages that use that redirect have their links updated. There appear at the moment to be many cases I'm falling over at the moment where links have been broken because this has not been done.

It is easy and fun to just rename pages because you want to impose your own personal stamp on the website. Remember: with great power comes great responsibility. Please do not neglect to amend all pages pointing to your renamed page, just because it is dull and mechanical, and because you want to rename another.

Thank you for your cooperation. --prime mover (talk) 10:35, 19 March 2018 (EDT)

A note about definitions and redlinks

I would like to make a plea to anyone who is working on a project to add definitions to the database.

Two points:

a) It is often the case that the definitions being added depend on other more fundamental definitions. If those more fundamental definitions do not exist, then the contributor is strongly encouraged to add those definitions. By "strongly encouraged", I mean very strongly encouraged. The philosophy of $\mathsf{Pr} \infty \mathsf{fWiki}$ is that everything is defined in terms of simpler concepts, and that everything is therefore completely understandable by referring back to those simpler concepts. When a large number of those concepts do not exist, the value of $\mathsf{Pr} \infty \mathsf{fWiki}$ is compromised.

b) There is always the temptation to define a large number of "composite definitions". There is a precedent for this: we have for example "finite subdivision", which is a "subdivision" which is "finite"; we have "commutative ring with unity", which is a "commutative ring" which is also a "ring with unity". There is a specific reason for constructing such a composite definition: it is because this concept crops up over and over again in the course of analysis. However, I would suggest that until such definitions are needed to be made, in the process of using these definitions in the course of a proof, then it may be better to delay constructing such a composite definition until such time as it is needed. Otherwise we will be in danger of cluttering up the database with ever more specialised and detailed definitions which are not actually doing anything but causing confusion (especially if there are 3 or 4 redlinks on those pages).

At base, remember that the site is called $\mathsf{Pr} \infty \mathsf{fWiki}$, not "DefinitionWiki" or "RedlinkWiki". --prime mover (talk) 05:59, 22 March 2018 (EDT)

Probably another MathJax problem -- but something seriously wrong

When I access Fourier Series/Square of x minus pi, Square of pi, nothing appears on the page.

I know something is there, because the history shows 9644 bytes.

Last version that can be viewed is Revision as of 10:15, 4 March 2018.

Diffing shows the big edit between that and Revision as of 11:31, 4 March 2018.

In case this helps, performing "Inspect" on that latter revision shows this in comments in the dev window:

NewPP limit report
Cached time: 20180328065935
Cache expiry: 86400
Dynamic content: false
CPU time usage: 0.080 seconds
Real time usage: 0.089 seconds
Preprocessor visited node count: 2059/1000000
Preprocessor generated node count: 3223/1000000
Post‐expand include size: 39340/2097152 bytes
Template argument size: 6841/2097152 bytes
Highest expansion depth: 4/40
Expensive parser function count: 0/100

Transclusion expansion time report (%,ms,calls,template)
100.00%   80.040      1 -total
 72.07%   57.681     50 Template:Eqn
  5.85%    4.681      1 Template:BookReference
  4.11%    3.286      1 Template:Proofread
  4.06%    3.248      4 Template:Qed
  1.91%    1.526     13 Template:Begin-eqn
  1.83%    1.465      1 Template:AuthorRef
  1.72%    1.375     13 Template:End-eqn
  1.36%    1.089      1 Template:=

Has the page broken a page limit size or something?

Many thanks if anybody is able to help. --prime mover (talk) 03:05, 28 March 2018 (EDT)

EDIT: You can of course edit the source code, and this works as well as ever, but you can't actually see the result of your edits. --prime mover (talk) 03:31, 28 March 2018 (EDT)
Okay, that's a weird one. I'm investigating. --Joe (talk) 11:12, 28 March 2018 (EDT)
Try now! Let me know if I've inadvertently broke something else. --Joe (talk) 13:04, 28 March 2018 (EDT)
Thank you -- I see it again in all its wondrous glory. --prime mover (talk) 14:56, 28 March 2018 (EDT)

MathJax on page preview

I've added MathJax rendering to page previews when editing ("Show preview"). Let me know if you have any issues. --Joe (talk) 11:52, 19 April 2018 (EDT)

Contours, paths, directed smooth paths -- sorry about the mess

I am in the process of attempting a general refactoring of the definition of "contour integral", currently defined only in the context of complex analysis. The intention is to define it in the context of a general real cartesian space. There is a lot of work to extract the definitions out into their more general context, and this has turned into a bit of a mess. It is going to take a lot of time and brainwork to do it, and both of these are unfortunately limited at the moment.

Please be tolerant of the unbearable mess in this area at the moment with redlinks everywhere. This is an ongoing task which I thought I could get done in a day or two, and it has turned out to be rather more heavy. --prime mover (talk) 12:12, 28 April 2018 (EDT)

Nightmare. I've made a mess of it. The results upon which some of the underlying definitions rest are themselves proved only in the complex plane, so attempting to move all this into the general vector space (even the general real cartesian space) is going to take more patience than I can muster today. Unfortunately, moving everything back to where it was may also be a little tricky. Bad news all round, I'm afraid. --prime mover (talk) 12:11, 29 April 2018 (EDT)

Namedfor with pseudonyms

Unsure how to use the namedfor template in the case of a pseudonym. (the only example I can think of is Gosset/Student) Do I credit it to Gosset or credit it to Student (as in the name) and redirect it to Gosset? Caliburn (talk) 04:21, 30 April 2018 (EDT)

Good question. Off the top of my head: credit it to Gosset, and then we would do well to add a Historical Note (the form is more-or-less well-developed) explaining the origin of the result and its name. It's interesting enough to merit a section in its own right. --prime mover (talk) 04:35, 30 April 2018 (EDT)

Tidying up of Exponential Function

As suggested, the definition page for Exponential has been renamed Definition:Exponential Function. During the course of this, a lot of cleaning up of some multiple redirects was done, and a hopeless tangle has been sort of resolved. However, there may be some subdefinition pages which are misdirected, because somewhere in the course of development, the names were changed of some of the definitions, and it is not completely clear whether they have all been redirected appropriately. I believe we can redraft this by changing the numbered definitions to names, which allows us to be consistent between Real and Complex definitions at last. Work in progress. --prime mover (talk) 15:10, 4 May 2018 (EDT)

Template Categories

As you may notice, I have flattened the Maintenance category, which now has (just about) everything in it on one level. This makes it easy to find the category you want, because they are all presented before you in one go.

I am about to go through the same operation on the Templates category, and remove all the categorisation of all of them (or at least flatten them).

Can anyone think of a good reason why we would want lots of categories (e.g. Axiom Templates, Category Namespace Templates, Character Replacement Templates) rather than putting them all just in the "Templates" category?

The advantage to having them all in one place is simply to make the easier to find. If they are all in one place, you don't have to hunt through a whole load of categories to find the one you want. --prime mover (talk) 05:12, 5 May 2018 (EDT)

I'm all for flattening so that stuff can be found easily. However I would like to preserve a distinction between, say, Template:Eqn, Template:Tidy, Template:Namedfor and Template:DefinitionCategory. More so as having these listed between one another leads to a kind of information overflow. Anything that does not fit in a given set of subcategories can stay at the highest level, so that we are free to invent new categories.
In general I support pragmatic categorisation, i.e. keeping the average size of a category within reasonable bounds. (In that light we might want to split up the numbers category into 1-1000, 1000-1000000, >1000000, for example). — Lord_Farin (talk) 06:24, 6 May 2018 (EDT)
Any progress towards tidying up the numbers category will be good progress. But there's a lot of work there, and I wonder if there are ways of setting up some automation in the area. Not my highest priority at the moment, but it's on my list of things to think about. --prime mover (talk) 06:46, 6 May 2018 (EDT)

Further tidying up

I have finally given some attention to the "source review" pages that were on my to-do list. All the simple stuff that I have the source works for has now been done. I am left with a few pages which still need to be written. This will probably not be immediate.

I note there is a considerable pile of source review tasks outstanding for works I do not have. Some of these are in the inbox of people who are no longer contributing. Hence this work will need to wait till somebody else obtains the works in question.

I will now continue to process another source work. --prime mover (talk) 18:17, 7 May 2018 (EDT)

Thanks, keep up the good work! I'll try to see if I have some of the works in question. — Lord_Farin (talk) 01:48, 8 May 2018 (EDT)

Lua/Scribunto conversion of Template:DefinitionCategory

This is to inform all of you that I and Joe have worked to enable Lua/Scribunto support.

As a first proof of concept, I have converted the in some ways problematic and complicated Template:DefinitionCategory to this new structure. It has gained support for other namespaces than definition (notably Axiom, see Category:Axiom of Foundation) as well as for an arbitrary number of categories to be added.

I have tried to test and see how it works, but if you have time, please take some time to see if the template behaves according to expectations (i.e. no observable change).

Users of the template are not affected in any way. — Lord_Farin (talk) 17:41, 8 May 2018 (EDT)

I'll keep my eyes open. It all looks fine to me so far. --prime mover (talk) 18:14, 8 May 2018 (EDT)