Wikipedia talk:WikiProject Mathematics/Typography

Please discuss math typographical issues below. Ta!

—Nils von Barth (nbarth) (talk) 08:46, 15 April 2010 (UTC)[reply]

texhtml is too big

texhtml (TeX code rendered as HTML) is too big for me – do others have this problem?

Consider the following display (how it renders for me at left):

1+1

1 + 1

1+1

1+1\,

rendered by the following source:

1+1
1&nbsp;+&nbsp;1
<math>1+1</math>
<math>1+1\,</math>

For me, the 2nd and 4th rendering are acceptable (carefully and verbosely spaced HTML, and forced PNG rendering), while the 3rd rendering is far too big, but is how I would prefer to author (this is just for example – italics and subscripts are a better example of why I prefer to write in TeX if possible – compare <math>L^p</math> with ''L''''p'').

This behavior is due to the lines:

/* Adjust font-size for inline HTML generated TeX formulae */
.texhtml {
  font-size: 125%;
  line-height: 1.5em;
}

in monobook.css, and was implemented 25 May 2009 by Erwin Dokter, on a trial basis, following:

this discussion, and
this follow-up discussion.

I understand that rendering is significantly different in different browsers/operating systems/font combinations, but for me (using Firefox/Iceweasel 2.0 or 3.0 on Debian GNU/Linux, with default fonts; I can check other combinations), the result is that it makes inline TeX unusable – the output is ugly and likely to be painstakingly corrected by other editors to better-rendering but hard-to-edit HTML, which is suboptimal for all.

Do others have the same problem?

I’ve heard it with some other editors, and is noted at the second discussion thread, and seems clear, because it makes the text 125% the normal size, which only works if the default serif font is 80% (=1⁄125%) of the size of the sans serif font.

If so, I’d suggest that we:

talk to Erwin & other editors/admins about this and have them revert it, or
- figure out a different solution.
Place at note at the relevant CSS to “please discuss math typography changes at WP:WPMATHTYPE before making changes” so we aren’t blind-sided in future.

How does this sound?

—Nils von Barth (nbarth) (talk) 08:44, 15 April 2010 (UTC)[reply]

With Firefox 3.6.3 on Windows 7, both recently installed with no extra scripts or stylesheets or whatever, I find the second and third renderings to be about the same size, significantly smaller than the fourth. So for me, texhtml is pretty much the right size, not what you're seeing. Jheald (talk) 09:29, 15 April 2010 (UTC)[reply]

Thanks for the feedback Jheald – that’s very useful, and accords with some other editors (i.e., current sizing works for you, PNGs are v. big). I’ll take a look in different browsers/OSes/wiki configs/browser configs to see if I can pin down what the issue is.

I suspect that it’s because I use large font sizes (that’s the only unusual thing about my setup), but it may be for some other reason.

—Nils von Barth (nbarth) (talk) 20:13, 15 April 2010 (UTC)[reply]

I don't see any problem with the lines: they all look OK to me, the first three the same size, the fourth a larger PNG identical to yours. I tend to use HTML for inline formulae, and don't particularly find them harder to edit than math tags: they are usually things like a₁ which is quicker to type than than $\mathbf {a} _{1}$ , but it's a small issue: once I've typed one I use copy and paste to make more like it.

Browser font behaviour is very counter-intuitive. For example suppose I find another font I prefer for pages without fonts/styles specified but it is too big at its current size. If I make it bigger or smaller so it looks right it also annoyingly effects the sizes of fonts on pages which specify the font and size: I tried this recently and had to not use the font. This is using Safari on the Mac but I noticed something similar with Firefox on Windows a couple of years ago.--JohnBlackburne^words_deeds 20:56, 15 April 2010 (UTC)[reply]

My view (minority, I believe) is that LaTeX is best for everything mathematical. Yes, it comes out bigger. Even in in-line text and it comes out bigger it's better than rendering it as italic with html tags. The latter is difficult to write, difficult to read, more effort than it's worth to follow and utterly tedious to maintain. Subscripts and superscripts are terrible. LaTeX makes the maths render larger which is good, innit? If not, why not? --Matt Westwood 21:14, 15 April 2010 (UTC)[reply]

Please don't encourage the notion that all non-TeX math notation is italicized. Variables are italicized; digits, parentheses, and other punctuation are not. Obviously that is in accord with the style used by TeX. Everyone except newbies has this down; we shouldn't confuse the newbies. Some pretty good newbies seem to have been joining us lately. Michael Hardy (talk) 23:44, 15 April 2010 (UTC)[reply]

... I'd think comical mismatches in size are obviously bad. But you forgot to mention things like

2^{3}\,

where the "2" should be at the same level as the surrounding text and the "3" should be a superscript. Remember when Wikipedia first acquired TeX at the beginning of 2003? Instead of centering, it made the bottom of

2^{3}\,

line up with the bottoms of the letters, but then things like

\sum _{i=1}^{2}i

looked ridiculous. Michael Hardy (talk) 23:48, 15 April 2010 (UTC)[reply]

Thanks for your comments John – I’m coming to the conclusion that proper rendering, at least for now, may require some measure of calibration on the part of the user (as it currently does), though we’d like to minimize this.

Matt – I also prefer TeX, but many prefer HTML, and, further, the TeX should not render larger than running text. This can be done optionally (e.g., through user stylesheets or preferences), but standard convention and user expectations is that text and math should be the same size. A preliminary goal is to at least have consistent (and correct/acceptable) rendering across TeX and HTML, so that authors can use whichever they prefer – then we can worry about improving rendering.

"the TeX should not render larger than running text." Why?--Matt Westwood 20:59, 16 April 2010 (UTC)[reply]

Michael – good point about the baseline, thanks! – I’ve added a note about that.

—Nils von Barth (nbarth) (talk) 05:04, 16 April 2010 (UTC)[reply]

Hi Matt,

Typographical standards (in print) are that math is set at the same size as running text, so as a baseline we should use that on Wikipedia, following existing conventions and user expectations.

Some people (such as yourself) would rather the math be set bigger (say, for legibility), and there is certainly a case for this, but unless there is a consensus for it being bigger by default (which I do not believe there is), this should be user-configurable, and we should default to being the same size as running text.

—Nils von Barth (nbarth) (talk) 21:33, 17 April 2010 (UTC)[reply]

I'm afraid I can't agree with this consensus, so regretfully I won't be able to contribute much to mathematics pages because of lack of patience. As it is, I've found some pages impossible to read because of the author's slavish devotion to what is to me an arbitrary and unworkable convention. Never mind, I have other work to do - and I'll drop in the occasional comment / question here and there. Sorry and all that. --Matt Westwood 07:38, 18 April 2010 (UTC)[reply]

No worries Matt – hopefully we’ll someday have math typography sorted so that everyone’s preferences can be easily accommodated with but a simple tweak! Meanwhile, look forward to your (occasional) contributions, in whatever form they may be.

—Nils von Barth (nbarth) (talk) 18:40, 22 April 2010 (UTC)[reply]

mathJax, were it to be adopted universally, should render this entire discussion obsolete. --Matt Westwood 21:00, 21 August 2011 (UTC)[reply]

True, but that's not likely to happen soon. CRGreathouse (t | c) 20:34, 22 August 2011 (UTC)[reply]

relative sizes

the expression

X+Y

is too large, compared to

X+Y\,

and X + Y.

For me it's the other way around: the first expression fits (though bigger than plain text) and the second is clearly too big. —Tamfang (talk) 05:15, 21 April 2010 (UTC)[reply]

PS, shoulda mentioned, I'm using Firefox 3.6.3 on MacOS 10.5.8. —Tamfang (talk) 21:27, 22 April 2010 (UTC)[reply]

And it looks the same in Firefox 4.0.1 on MacOS 10.6.7. —Tamfang (talk) 20:33, 2 May 2011 (UTC)[reply]

Thanks Tamfang, that’s v. useful (particularly the level of detail)!

Jheald also reports PNG too big, and that accords with how some editors use \scriptstyle to shrink PNGs. So it looks like we’ve got 2 sizing issues; we’ll need to do some experiments to figure out how people should correctly calibrate (if anyone’s figured out how this works, please weigh in).

—Nils von Barth (nbarth) (talk) 18:45, 22 April 2010 (UTC)[reply]

Now (Firefox 115.0.3 on MacOS 13.5) the first looks maybe a hair bigger than the second. —Tamfang (talk) 23:27, 27 July 2023 (UTC)[reply]

To me the first two are equal in size and much bigger than the surrounding letters, whereas the third one fits. Michael Hardy (talk) 01:06, 3 June 2010 (UTC)[reply]

As you can see from above, the results are different on Firefox 4 on Win 7, which is what I assume the average user would use (default setting for everything). The 1st and 3rd are almost the same size, although the first uses serif fonts, which may be harder to read, while the 2nd one looks much bigger (about twice the size of text, and serif). The rendering is identical on IE 9, so I infer it's mostly an OS rather than browser font issue. Tijfo098 (talk) 17:06, 2 May 2011 (UTC)[reply]

Here's how the "ones test" from the beginning of this page looks like to me. Tijfo098 (talk) 17:25, 2 May 2011 (UTC)[reply]

Effect of \textstyle?

The page says that inline equations should be set in \textstyle. Setting aside the silliness of having to do this by hand, I just tested this at Cauchy product: version with \textstyle vs version with default sizing. On my antique Firefox 2.0.0.14 the results are inconsistent: some formulae are smaller in textstyle, some are larger and some are unchanged. Can someone with a newer/different browser check this? Jowa fan (talk) 01:16, 3 May 2011 (UTC)[reply]

It seems to have no effect in Firefox 4 on Win 7. For instance "Let

\textstyle (a_{n})_{n\geq 0}

and

\textstyle (b_{n})_{n\geq 0}

be real sequences." displays the same way with and without \textstyle. Tijfo098 (talk) 14:13, 3 May 2011 (UTC)[reply]

mathJax

I signed up as a participant in this subproject because I am interested in how Wikipedia handles typography in mathematical notation. But it begins to seem as if those topics might be better discussed at the math wikiproject's talk page, because some of the best contributors to discussions of typography in math notation follow that page and regularly post there but don't go out of their way to follow this page.

Recently some interesting discussions of mathJax have taken place on the following three pages:

I would like to see mathJax be automatically used whenever someone reads a Wikipedia page with math tags. That would solve some of the problems we've had ever since February 2003. Those interested in typography should look at this, and I'd like to see supporters of this proposal express their support at the requests-for-comment page that is linked to above.

The mathJax software developers are now following these discussions and are interested in this proposal. Michael Hardy (talk) 22:09, 24 July 2011 (UTC)[reply]

I think that it would be great to have this as an option, and it's probably better than anything we have now. (I do have some concerns about the speed of rendering, based on my experience at math.se, but it's probably not as bad here since the time spent reading an article is presumably longer than the time spend reading a question and so a bit of loading is more acceptable here.)

CRGreathouse (t | c) 00:08, 25 July 2011 (UTC)[reply]

We already have it as an option; I've been using it for a few weeks now. Michael Hardy (talk) 04:27, 25 July 2011 (UTC)[reply]

Can you please explain how to set an "option"? Richard Gill (talk) 18:02, 21 August 2011 (UTC)[reply]

You can turn on MathJax by following the instructions at User:Nageh/mathJax. Cheers!

—Nils von Barth (nbarth) (talk)

In case anyone's interested in seeing how it looks on a heavy-duty project which is completely based around mathJax, feel free to check out ProofWiki (go google). It has considerable advantages over the MediaWiki LaTeX interpreter. --Matt Westwood 20:55, 21 August 2011 (UTC)[reply]

Hi all, may I point to effects of enabling mathjax as renderer by just simply logging out? (See below on my newbie question about "ddot"). I perceive this as a shame, what I had to read about the declining opinion of improving on this on the math project page. Purgy (talk) 09:12, 20 August 2014 (UTC)[reply]

The scourge that is \mbox

I have written this: Wikipedia:Manual of Style (mathematics)/Why you should never use \mbox within Wikipedia.

Opinions? Michael Hardy (talk) 23:38, 25 July 2011 (UTC)[reply]

Spot on. --Matt Westwood 20:57, 21 August 2011 (UTC)[reply]

xdvi

I've created a new article titled xdvi. It's still quite stubby. Have fun improving and extending it. Michael Hardy (talk) 16:54, 4 May 2012 (UTC)[reply]

Outdated

This page needs updating. For one, TeX rendered as HTML has been disabled for over a year. — Edokter (talk) — 13:03, 6 July 2013 (UTC)[reply]

A WikiProject for advanced wiki typesetting?

Ten days ago user:Edokter pointed out that the WikiProject Mathematics/Typography page is #Outdated. Why not to create something WP:WikiProject advanced wiki typesetting as a replacement? The goal should be not only to instruct how to do it, but also to contain various flamewars related to mathematical, physical, linguistic, and to lesser extent other scientific and technology articles. List of topic follows:

<math> vs "texhtml" vs bare wiki code
{{nobreak}} vs U+00A0 vs  
Full-width spaces vs thin spaces
Roman vs Italic
Arial vs serif
Accessibility of Unicode
- For example, ⟨⟩ vs $\langle \rangle$ vs <> vs ‹›
{{val}} vs whatever
{{chem}} vs bare wiki code
Math style (≠≤−×) vs ASCII style in articles about computing
(Uploaded) images vs whatever
vs table
letter-spacing vs deletionists
Some questions related to WP:navboxes

Incnis Mrsi (talk) 08:08, 16 July 2013 (UTC)[reply]

Sure, why not? M∧Ŝc²ħε И_τlk 08:25, 16 July 2013 (UTC)[reply]

I would join. —Designate (talk) 20:32, 4 August 2013 (UTC)[reply]

Is all right with the proposed name? Particularly, is the word “advanced” optimal in this context? Incnis Mrsi (talk) 07:59, 9 August 2013 (UTC)[reply]

No, not really. Something perhaps to mean a style that may be adopted in an article, maybe "Wiki typesetting with care" or "Unified wiki typesetting". This is not an easy name to think up; if it sounds autocratic or all-encompassing, it may flounder on the adverse reaction. Also add:

{{sub}} mad {{sup}} vs  and

— Quondum 10:39, 9 August 2013 (UTC)[reply]

“Unified” is certainly not an option because our typesetting is objectively diverse and a drive to unification (that is implied by this name) would result in degradation. Typesetting with care is overly broad: it may refer, for example, to the use of “: ” and “ – ” but not “ - ”, and other similar trivial patterns. We do not need a WikiProject to teach the people that “blah - blah” is unprofessional and “blah- blah” is an egregious mistake. Incnis Mrsi (talk) 13:41, 9 August 2013 (UTC)[reply]

Newbie request for "dddot" in not-logged-in state

Hi all, the dddot renders fine when logged in and mathjax is enabled. But as soon as one logs out, a red error message appears in the same browser.

May I please ask, if there are any measures, work arounds, or secrets (to me), so I might use the Newtonian notation for time derivatives? Thanks in advance for any help. Purgy (talk) 10:40, 17 August 2014 (UTC)[reply]

It looks like \dddot{x} Failed to parse (unknown function "\dddot"): {\displaystyle \dddot{x}}

is not supported by the standard TexVC renderer, but is by MathJax. Help:Formula basically defines the syntax which could be use. The chance of TexVC being fixed to include \dddot is very small as it does not have active maintenance. One day we may have a better maths renderer as standard, there are moves in that direction. Using an overset \overset{...}{x} ${\overset {...}{x}}$ works and seems to render fine with both systems.--Salix alba (talk): 12:35, 17 August 2014 (UTC)[reply]

Thank you. Purgy (talk) 13:34, 17 August 2014 (UTC)[reply]

Consequences of a lack of consensus concerning inline text style mathematical formulae

CC: Template talk:Math#Consequences of a lack of consensus concerning inline text style mathematical formulae.

There appears to be no consensus concerning inline text style mathematical expressions, with the result that depending on one's whims, we have

LD (LaTeX display style aficionados): which means that ${\sqrt {x^{2}+1}}$ is whatever... (with <math>\sqrt{x^2 + 1}</math>)
LT (LaTeX text style aficionados): which means that $\textstyle {\sqrt {x^{2}+1}}$ is whatever... (with <math>\textstyle \sqrt{x^2 + 1}</math>)
LS (LaTeX script style aficionados): which means that $\scriptstyle {\sqrt {x^{2}+1}}$ is whatever... (with <math>\scriptstyle \sqrt{x^2 + 1}</math>)
HM (HTML with {{math}} aficionados): which means that $\sqrt x 2 + 1$ is whatever... (with {{math|{{sqrt|''x''2 + 1}}}})
HN (HTML without {{math}} aficionados): which means that √x² + 1 is whatever... (with {{sqrt|''x''2 + 1}})

In spite of all the disparaging quotes about consistency (http://www.goodreads.com/quotes/tag/consistency), I think Wikipedia would look more polished if all the inline mathematical expressions were done in a consistent fashion. — Tentacles^{Talk or ✉ mailto:Tentacles} 21:46, 24 March 2016 (UTC)[reply]

Now there's much worse consequences... Let's consider the following eventualities:

Wikipedia wants to use MathJax: oops, only the inline math expressed in LaTeX will be processed with MathJax...;
Wikipedia wants to change the appearance of HTML inline math via stylesheet or skin: oops, only the inline math expressed in HTML using a mathematical template (e.g. declaring a class="math") will be affected (missing out the inline math expressed in LaTeX; missing out the inline math not using the {{math}} template or any other mathematical template,
e.g. x² + 1 with ''x''2 + 1);

Now, Wikipedia has all those mathematical formatting templates for HTML inline math, without a consensus to use them inside the {{math}} template... without even a consensus to even use HTML instead of LaTeX for inline maths...

Disclosure: I am an "HTML with {{math}} aficionado" but, since there is no consensus (what if Wikipedia wants to use MathJax), I will stop converting LaTeX inline math expressions into HTML with {{math}} expressions. — Tentacles^{Talk or ✉ mailto:Tentacles} 22:02, 24 March 2016 (UTC)[reply]

There's even a {{bigmath}} template, so that standalone display style math can be written either in HTML or LaTeX, with the same consequences as above! — Tentacles^{Talk or ✉ mailto:Tentacles} 22:15, 24 March 2016 (UTC)[reply]

Calling it "LaTeX" is extraordinarily misleading. What if someone masters the language in which mathematical notation is coded in Wikipedia and mistakenly thinks he therefore knows LaTex, and then encounters actual LaTeX?? He's in for a major surprise, finding out that he knows virtually nothing about LaTeX.

At any rate, inline so-called "LaTeX" in Wikipedia often produced characters three times the size of the surrounding text and some really bad misalignments. But "displayed" "LaTeX" looks good. Michael Hardy (talk) 02:19, 26 March 2016 (UTC)[reply]

Concerning the inline LaTeX formulas, is there a way to configure Extension:Math so that we get a proper font size for inline mathematical formulas with the \textstyle LaTeX command (not the inappropriate kludge of using the \scriptstyle LaTeX command)? — Tentacles^{Talk or ✉ mailto:Tentacles} 19:50, 26 March 2016 (UTC)[reply]

I totally agree that displayed LaTeX looks good, but since we have a {{bigmath}} template we are now beginning to get an admixture of display style LaTeX formulas and HTML display style formulas throughout the wiki. So now the stand-alone display style math formulas can also have an inconsistent presentation... — Tentacles^{Talk or ✉ mailto:Tentacles} 19:50, 26 March 2016 (UTC)[reply]

One way round some of this might be to write a Wikipedia:Lua script in the Module: namespace. This could provide a limited subset of latex with good css based typography, similar to the quality of a MathJax install. 90% of inline maths only need a small subset of latex, superscripts, subscripts and a range of special character. Anything more difficult could be handled by the full latex. Lua allows more complex processing of the input than is possible by using normal templates and could easily do the parsing necessary. --Salix alba (talk): 20:01, 26 March 2016 (UTC)[reply]

We can handle quite a lot of cases with templates, and even more cases with Lua, but if there is no consensus for inline math formulas, we still have a hodge-podge of LaTeX (default \display style)/LaTeX (\textstyle)/LaTeX (\scriptstyle)/HTML (with {{math}})/HTML (without {{math}})... We would also need a consensus to avoid getting a hodge-podge of LaTeX (default \display style)/HTML (with {{bigmath}}) for stand-alone math formulas... — Tentacles^{Talk or ✉ mailto:Tentacles} 20:21, 26 March 2016 (UTC)[reply]

Unless/until the wikimedia developers give us a <math> mode that actually works, we are stuck with a hodgepodge. There is little we can do at a project level to fix that. At the article level, we can at least work to make the formatting consistent within each article. —David Eppstein (talk) 20:26, 26 March 2016 (UTC)[reply]

A Lua module is theoretically possible but it would be a significant undertaking, and it is unclear whether it would actually help. If it worked as you suggest, supporting a limited subset of LaTeX, then it would not be a replacement for exiting methods and so would fragment things further. It would be avoided by many editors who did not want to deal with its incompatibilities. It would only be really useful if it supported all LaTeX, but that would be a massive amount of work, so much so that it might never be completed satisfactorily.

We have all these variations as editors find them useful. The same editor can use more than one of them, even three or four of them, within the same article. Or they can use other things like {{var}} and dedicated math display templates like {{pi}}. There’s no correct way of doing it, not least as what readers see depends on their platform, browser and settings (if logged in).--JohnBlackburne^words_deeds 20:32, 26 March 2016 (UTC)[reply]

On https://www.mediawiki.org/wiki/Extension:Math/advancedSettings#Server-side_rendering_with_Mathoid it says "Mathoid is the rendering mode that is going to be used on future Wikipedia." Is that true? (A Mathoid server uses MathJax to convert texvc input on the server side to MathML+SVG rendering.) If that's true, then all our using {{math}} and {{bigmath}} to produce HTML+CSS math expressions will get in the way, since texvc input must be LaTeX. Oops... — Tentacles^{Talk or ✉ mailto:Tentacles} 23:13, 30 March 2016 (UTC)[reply]

Also, if Wikipedia ever wants to use the https://www.mediawiki.org/wiki/Extension:MathJax extension, then all our using {{math}} and {{bigmath}} to produce HTML+CSS math expressions will get in the way, since MathJax uses MathML, TeX and ASCIImath as input to produce HTML+CSS, SVG and MathML as output. Oops... — Tentacles^{Talk or ✉ mailto:Tentacles} 23:23, 30 March 2016 (UTC)[reply]

The wikimedia developers seem to have their hearts set on transmitting mathematical formulas from the server to the browser in mathml format, even though the big browsers have moved away from that and now the mathjax project itself is moving away from that. As a result, we are stuck with a <math> implementation whose server-side-rendered mathematics still continues to clash badly in font, text size, and baseline with the inline text of articles, that cannot be properly colored to show the presence of links, and that (on browsers such as Chrome that use the svg fallback) cannot be copied and pasted. As a result, it is significantly worse than {{math}}. It is likely that as long as the developers continue down this path, <math> rendering will remain bad. And as a result, it is also likely that many mathematics editors on Wikipedia (myself included) will prefer using {{math}}, which despite its own disadvantages (like much less flexibility in what it can handle) produces formatting that is much better integrated into the surrounding text. I.e. mathematics formatting on Wikipedia is bad, there is no evidence that it will get better any time soon, and your hope that Wikipedia might want to use the MathJax extension is contradicted by past actions of the developers to rip out earlier implementations of MathJax support. I could go on at much greater length than this, and already have, at http://11011110.livejournal.com/314841.html. —David Eppstein (talk) 00:09, 31 March 2016 (UTC)[reply]

Then I wished there was the following consensus:

use the {{math}} template for all the inline text style HTML math expressions (there is currently no consensus to avoid e.g. ''x''2 + 1 yielding x² + 1 without the {{math}} template);
use <math>...</math> for all the stand-alone display style LaTeX math expressions (maybe in the future we could use the {{bigmath}} template if it was sophisticated enough to handle e.g. 99.9% of the cases with typographic quality comparable to LaTeX, with the advantage of a crisper display of fonts than PNGs, but the disadvantage that we won't have copyable LateX code via the contextual menu).

Since the inline text style math expressions are typically much simpler than stand-alone display style math expressions, I am quite confident that the {{math}} template could get sophisticated enough to handle e.g. 99.99% of the cases with typographic quality comparable to LaTeX, without the drawbacks (wrong font, wrong vertical alignment, no wikilinks in math expressions...). — Tentacles^{Talk or ✉ mailto:Tentacles} 01:28, 31 March 2016 (UTC)[reply]

By the way, I've been working on improving {{intmath}} (check {{intmath/sandbox}}, {{intmath/int/sandbox}} and {{intmath/testcases}}) to get nicer HTML+CSS integrals. — Tentacles^{Talk or ✉ mailto:Tentacles} 01:33, 31 March 2016 (UTC)[reply]

On https://www.mathjax.org/ it says that MathJax uses "MathML, TeX and ASCIImath as input and produce HTML+CSS, SVG and MathML as output." Will Wikipedia ever consider using MathJax to produce HTML+CSS, or to produce SVG? If Wikipedia ever wants to use those options, then all the math expressions should be left in LaTeX, since all those HTML+CSS math expressions produced by {{math}} and/or {{bigmath}} will get in the way. Oops... again. — Tentacles^{Talk or ✉ mailto:Tentacles} 19:27, 31 March 2016 (UTC)[reply]

Wikipedia already does use MathJax on the server side to generate the SVG that you get if you set your rendering preference to MathML and then use a non-MathML-supporting browser such as Chrome, as I understand it. What the Wikimedia (not Wikipedia) developers refuse to do is provide an option for MathJax-generated HTML, whether the generation is done server-side or browser-side. This is because they think that MathML provides clean semantics (even though the clean-semantic variant of MathML and the variant of MathML that you can use for displaying math in your browser are two different things), that semantics is more important than usability and readability, and that refusing to support non-MathML solutions will put pressure on browser makers to properly support MathML. Obviously, I disagree. —David Eppstein (talk) 19:38, 31 March 2016 (UTC)[reply]

On Wikipedia:Manual_of_Style/Mathematics#Typesetting of mathematical formulae it says "One should not change formatting boldly from LaTeX to HTML, nor from non-LaTeX to LaTeX without a clear improvement." I'm glad I stopped changing formatting boldly from LaTeX to HTML (using the {{math}} template): there is no consensus on anything, it's time consuming grunt work, someone else might revert/change it back boldly from HTML to LaTeX, what a monumental waste of time... On the {{radic/doc}} template doc page, I added a recommendation (with examples) to use the the {{radic}} template inside the {{math}} template and it was reverted because there is no consensus to even use the {{math}} template for HTML math expressions! — Tentacles^{Talk or ✉ mailto:Tentacles} 21:34, 1 April 2016 (UTC)[reply]

Also, people who use the "MathML with SVG or PNG fallback" preference are probably unhappy that the LaTeX is changed into HTML. In a few months or years, when Wikipedia wants to convert the LaTeX (using MathJax) into MathML or HTML or SVG, lots of people will be unhappy that a lot of LaTeX formulas have been manually converted to HTML with math templates. The whole thing is very paralysing... — Tentacles^{Talk or ✉ mailto:Tentacles} 21:43, 1 April 2016 (UTC)[reply]

On all those sites that use MathJax, I can choose the Math renderer to be from: MathML, HTML+CSS, SVG and more options... If Wikipedia had all those options, we could leave everything as LaTeX and all the zillion hours spent on math typography (and writing all those templates for HTML maths) could be spent on writing actual mathematical content. Offering the MathJax HTML+CSS rendering option does not get in the way of their project of supporting the MathML rendering option; NOT offering the MathJax HTML+CSS rendering option does get in the way of their project of supporting the MathML rendering option because people are then inclined to write the math in HTML with templates instead of the LaTeX that MathJax needs as input. Oops... — Tentacles^{Talk or ✉ mailto:Tentacles} 22:04, 1 April 2016 (UTC)[reply]

Tell it to the Wikimedia developers. I don't think they're paying much attention to Wikipedia talk pages. —David Eppstein (talk) 22:43, 1 April 2016 (UTC)[reply]

https://meta.wikimedia.org/wiki/Developers (moved to MediaWiki.org)

https://www.mediawiki.org/wiki/Talk:Developers (I told them...) — Tentacles^{Talk or ✉ mailto:Tentacles} 01:31, 2 April 2016 (UTC)[reply]

Thanks. The current mix of bad options is really frustrating; I don't have much hope for change any more but it can't hurt to try. —David Eppstein (talk) 01:39, 2 April 2016 (UTC)[reply]

Under topic: "NOT offering the MathJax HTML+CSS rendering option does get in the way of their project of supporting the MathML rendering option" — Tentacles^{Talk or ✉ mailto:Tentacles} 01:43, 2 April 2016 (UTC)[reply]

You know, I think it's not so much the lack of consistency that is the problem but the fact that all of these options are ugly. It's not much of a choice if you have variation, but all variants are bad. It kind of shows that Wikipedia grew up in the early 2000s, at which time, of course, this stuff was top-of-the-line.

I suppose the long term solution will be to use latter-day web technology to produce nice looking math. Look at this example here: I didn't look too closely at the source, but somehow this guy uses scripts and (I guess) html5 layout options to typeset math. It would be really, really nice to aim at something of this kind for Wikipedia in the mid term. --dab (𒁳) 17:41, 6 April 2016 (UTC)[reply]

The displayed math notation on the page to which you link looks good, just as on Wikipedia, but the inline notaiton looks terrible. Variables are not italicized and there's actually a hyphen where a minus sign should be, in several places. And he's actually got

\lambda _{max}

where he should have

\lambda _{\max }

--- how clumsy can you get? Michael Hardy (talk) 02:41, 28 October 2016 (UTC)[reply]

That site, and as far as I know all modern websites that support math including MathOverflow and MathSciNet, does its math formatting with MathJax, which is easy to add to a site, works well, and produces very good looking Math. Wikipedia used to support MathJax. But the developers ripped it out in order to pursue their MathML chimera instead. —David Eppstein (talk) 18:02, 6 April 2016 (UTC)[reply]

So it seems to be a good thing that I work on templates to render nice HTML+CSS mathematical expressions, especially for inline text style formulas. For example, I've been working on improving {{intmath}} (check {{intmath/sandbox}}, {{intmath/int/sandbox}} and {{intmath/testcases}}) to get nicer HTML+CSS integrals. — Tentacles^{Talk or ✉ mailto:Tentacles} 19:00, 6 April 2016 (UTC)[reply]

Apparent bug in rendering \operatorname*

Tracked in Phabricator
Task T185552

\operatorname* does not seem to render as I believe it should. Take a look, for example, at this revision, in which

<math>\widehat B = \operatorname*{median}_i \operatorname*{median}_{j\,\ne\,i} \operatorname{slope}(i, j)</math>

is rendered like this. -- The Anome (talk) 14:28, 21 February 2018 (UTC)[reply]

It was _not_ rendered like that when I viewed the article after editing it and it is not rendered like that when I look at the old version just after I edited it. Maybe it's something in your preferences and mine that causes the difference.

Another thing that doesn't work is \mathbin{} and \mathrel. Michael Hardy (talk) 17:07, 21 February 2018 (UTC)[reply]

@The Anome: It doesn't seem clear that anything at all on this page ever gets attention. Michael Hardy (talk) 17:08, 21 February 2018 (UTC)[reply]

@Michael Hardy: I think you're right. I've currently got the "Math" setting within the "Appearances" section of my preferences set to "MathML with SVG or PNG fallback (recommended for modern browsers and accessibility tools)". I'm not sure where we should report bugs in wikitext rendering now: perhaps there is some sort of WMF bug tracke for these? -- The Anome (talk) 18:53, 21 February 2018 (UTC)[reply]

There was up until fairly recently a workaround for this, namely to use \limits after \operatorname. Unfortunately, something was done that broke this from working too. I reported this at phabricator, but I have a sinking feeling that no one really cares. –Deacon Vorbis (carbon • videos) 20:27, 21 February 2018 (UTC)[reply]

How about using underset

<math>\widehat B = \operatorname{\underset{\mathit{i}}{median}} \operatorname{\underset{\mathit{j\,\ne\,i}}{median}} \operatorname{slope}(i, j)</math>

gives

{\widehat {B}}=\operatorname {\underset {\mathit {i}}{median}} \operatorname {\underset {\mathit {j\,\neq \,i}}{median}} \operatorname {slope} (i,j)

--Salix alba (talk): 00:23, 22 February 2018 (UTC)[reply]

test

abcdefg $\sum \limits _{i=1}^{n}a_{i}$ hijklmnopqr stuvwxyz. Etc. abcdefg $\sum _{i=1}^{n}a_{i}$ hijklmnopqr stuvwxyz. Etc.

abcdefg ${\textstyle \sum \limits _{i=1}^{n}a_{i}}$ hijklmnopqr stuvwxyz. Etc. abcdefg ${\textstyle \sum _{i=1}^{n}a_{i}}$ hijklmnopqr stuvwxyz. Etc.

On the line above this one, I used <math display="inline"> and in the first expression I used \limits and in the second I did not. On my browser window I see the first expression with i = 1 and n directly below and above the Sigma. In the second they are to the right of Sigma. Thus \limits works. @Deacon Vorbis: @The Anome: Michael Hardy (talk) 01:22, 23 February 2018 (UTC)[reply]

Yeah, it works for sums, integrals, "big" operators, etc., but it also used to work with \operatorname{} in order to force subscripts directly below the operator, a la \lim. –Deacon Vorbis (carbon • videos) 01:49, 23 February 2018 (UTC)[reply]

A bit of investigation seems to suggest this is due to MathJax changes and not a local mediawiki thing. If I run an old MathJax version 2.7.0. It looks like the old \operatorname{A}\limits_a^b syntax was a bug in MathJax which didn't match LaTex rendering. This is now fixed.[1] So the real problem is that

our local installation is not recognising \operatorname*.--Salix alba (talk): 06:48, 23 February 2018 (UTC)[reply]

@The Anome: : Have you reported this to whoever is in charge of fixing software bugs here? I don't think reporting it on this page has the effect of bringing it to their attention. Michael Hardy (talk) 21:44, 14 March 2018 (UTC)[reply]

@Michael Hardy: Yes, a Phabricator bug has been filed: https://phabricator.wikimedia.org/T185552 -- The Anome (talk) 21:47, 14 March 2018 (UTC)[reply]

Looking at the phab task, it looks like getting \operatorname* to work would be really hard. It involves completely removing the old texvc parser, which is a big job. One possibility would be defining a convenience short-cut \operatorname[A]{B} where the part in square brackets as an underset, but this would add non latex syntax.

It might be better just to use the underset workaround above, which seems to produce identical output. As its not clear which the best thing to do the task is marked as stalled. --Salix alba (talk): 22:13, 14 March 2018 (UTC)[reply]

MathJax, please!

Can't believe we still haven't implemented a modern and unified way of coding maths.

What will it take for WikiMedia to make the change to fully supporting MathJax?

yoyo (talk) 20:20, 8 October 2018 (UTC)[reply]

Prejudice against {{sqrt}}/{{radic}}

Can anybody suggest what [2] (by LLarson) is about? Incnis Mrsi (talk) 07:59, 18 May 2019 (UTC)[reply]

Hi Incnis Mrsi, I can answer. From both templates, there were gaps between the √ and ‾ symbols when viewing the page on mobile using Safari 12.1.2 on an iPhone running iOS 12.4. It wasn’t prejudice.

I felt like it was an especially low-stakes edit because my change, from

{{sqrt|''xy''}} = {{sqrt|''x''}}{{sqrt|''y''}}

to

<math>\sqrt{xy} = \sqrt{x}\sqrt{y}</math>

better matched the example it is being compared to, which also uses <math>. —LLarson (said & done) 13:32, 18 May 2019 (UTC)[reply]

LaTeX help

My LaTeX isn't brilliant and I am having trouble getting the equals signs to align for the first formula in the caption at Matrix multiplication#Definition. Could someone give me a hand? Thanks. Yaris678 (talk) 13:12, 18 December 2019 (UTC)[reply]