Vanilla 1.1.9 is a product of Lussumo. More Information: Documentation, Community Support.
@algori: I've never seen that. What browser/OS are you using?
Also, does this behavior happen on the MathJax preview page (I ask because we'll probably switch from jsMath to MathJax once MathJax is out).
I see this behaviour (as described by algori in Room 101, above) from time to time. At first, I would try to edit the post as I ascribed it to a fault in the syntax, but then when I was unable to find the fault I would "return to question/answer" only to find it rendering correctly again. I have no ideas as to why it might be happening, though. I'm using Firefox 3.0.17 on Linux.
@Charles: I can't reproduce the problem. What browser/OS combination are you using? My guess is that there's nothing we can do about it now, but that the problem won't come up again once we switch to MathJax, which seems to be much more stable than jsMath. Unfortunately, right now MathJax is still very slow compared to jsMath, so we can't switch yet.
I have a similar problem with jsMath rendering, and much more often in the last days (often enough to convince me to subscribe at meta). An example is this page
I don't get the rendered math with either Firefox 3.5.7, Chrome 5.0 or Opera 10.10 on Ubuntu 9.10, even after many reloads. The dollars disappear and the content is shown as is.
@Andrea: I don't understand. The post you mention (the question, and the answers) don't seem to be using jsMath at all. Not a dollar sign in sight when I look at the source.
Uhm... so we have one question and two answer using LaTeX notation and everyone managed not to put it in dollar signs?
By the way, is it possible for everyone to check the source, so that if I run into this problem again (as I told it is far from the first time) I can check if this is a jsMath bug or a poorly formatted question/answer?
If you want to see the source, but don't have 2000+ reputation (so you can't edit), you can still do it by looking at the revision history ... see tip number 5.
The post you mention (the question, and the answers) don't seem to be using jsMath at all.
It looks pretty silly, but that question (and those two answers) were posted October 17 and 18, and we didn't have any LaTeX support back then ... MO had just launched. Since the question was bumped up to the home page, I'll go edit it now to use dollar signs.
A couple of other people have mentioned jsMath requiring a page refresh to work sometimes, which shouldn't be happening. If we don't solve that problem (it's hard to since I can't reproduce it), hopefully it will go away when we switch to MathJax.
That's pretty weird, because we haven't changed anything about jsMath for a long time. Maybe they've changed something on the servers and they're having trouble getting the files to you. I've added a line of text for debugging purposes. Whenever jsMath is loaded on a page, the very bottom of the footer should say "loading jsMath files". Next time jsMath fails to process the page, can you please check to see if these words appear?
@Georges: I've tried adding an additional bit of debugging. The bottom of the page should now say "loading jsMath files. (Re)process math" where clicking on the words "(Re)process math" instructs jsMath to do it's thing. Does clicking this link (without reloading the page) get math to render properly?
@Sonia,
adding backticks around that latex string fixes the symptoms, but I don't know the cause.
@Georges: I still don't understand what's going on. I hope that the problem goes away when we switch to MathJax. For now, I've moved the "(Re)process math" link to the sidebar (which I assume you prefer). Sorry I couldn't find a better solution.
@Akhil: the < characters are being interpreted as html markup for some reason. You can fix the problem with the usual trick: add some backquotes, so it becomes `$0<f'(0)<1$`
In my browser, writing $0<f'(0)<1$ introduces another interesting bug. Since the < are interpreted as html markup, it messes up the rest of the page. In particular, I can't click the submit button. I have this webinar soon, but I'll post a bug report about this later today (or somebody else can do it).
for whatever reason, my browser [Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9.2.3) Gecko/20100401 Firefox/3.6.3] together with mathoverflow's jsMath, etc., makes some really weird character choices inside formulae; commas become semicolons, full stops become colons, greek letters named in TeX become various decorated roman characters. The last at least does not happen with the MathJax preview page, nor with greek letters named in html-escaped forms.
Running in debian [2.6.31-trunk-686 #1 SMP Tue Oct 13 22:34:42 UTC 2009 i686 GNU/Linux] with packages jsmath-fonts and ttf-jsmath installed.
@qnoodles: Do you have the same problem on other sites that use jsMath? For example, does the math come out funny on planetmath? What happens if you click the jsMath button is the bottom right-hand corner of your browser, click "options", and tell it to use image fonts?
The last at least does not happen with the MathJax preview page
Are you suggesting that the other problems (e.g. commas become semicolons) do occur on the MathJax preview?
@Anton: Yes, commas become semicolons on planetmath.
Selecting image fonts resolves the odd substitutions. They don't scale well, but I can deal with that...
Are you suggesting that the other problems (e.g. commas become semicolons) do occur on the MathJax preview?
No, it's just that there are no visible test cases there.
I've just updated the way that jsMath is loaded in the footer. Please let me know if you notice a difference in speed, or if you find that you need to reprocess the page (see this thread) more or less frequently.
@qnoodles: I'm able to reproduce your problem on my home computer. It looks like using image fonts "fixes" the problem. It looks like jsMath actually does express greek letters as decorated roman letters, but in spans with special classes. These classes are translated into font families via some css that jsMath injects into the header. I think firefox is failing to apply those css rules for some reason.
Here is a strange bug I don't know how to reproduce: Sometimes, when I load this page, several of the equations are replaced by the error message "Unknown symbol \binom". Reloading fixes this. Does anyone else see this?
As discovered on the above mentioned page, $\begin{array}{c}x\\y\end{array}$ will produce "Unknown symbol \y". On a true LaTeX installation, the second backslash will bind to the first, not to the y.
I wrapped all the displayed mathematics in '<p>' tags (which is better style than backticks as '<p>' tags says "Markdown keep out" whereas backticks says "Markdown treat this as code").
You are both right. My bad, I thought I had tried backquoting those but, based on a test case, I had not.
@rwbarton: when I updated the way jsMath is loaded, I accidentally introduced an error in the definition of \operatorname (I left something as \\\\ instead of \\). It should be fixed now.
I'm experiencing the problem right now that the live preview isn't working. Is anybody else getting this? I'm looking into it. (Edit: this problem happens in chrome, but not firefox)
@Charles: it seems to be having some trouble unwrapping the command \varprojlim. Sometimes it works correctly for me, sometimes it complains that \varprojlim is undefined, and sometimes it complains that \underleftarrow is undefined (this presumably comes up in the definition of \varprojlim). Perhaps the extra macro packages are sporadically failing to load. I'll see if I can do anything about it.
@Peter: the underscores are causing the problem because markdown tries to turn underscores into italics sometimes. For example, if you type something like _this_, it gets interpreted as <em>this</em>. So when you type the top line in the block below, it get converted to the second line before jsMath gets a chance to look at it (and jsMath doesn't want to touch anything with html markup inside of it), so it gets rendered as the third line:
$E[x]_\sigma$ is equal to $E[x]_\sigma$
$E[x]<em>\sigma$ is equal to $E[x]</em>\sigma$
$E[x]\sigma$ is equal to $E[x]\sigma$
The accepted solution is to put backquotes around your math whenever this problem shows up, like this:
`$E[x]_\sigma$` is equal to `$E[x]_\sigma$`
jsMath is not rendering for me on question pages. Reloading the page or reprocessing does not seem to so anything. :/ Did anything change lately?