I'm moving this to the mathjax-dev@googlegroups.com list so it will be archived.
@Peter and/or Alfred, on the assumption that you will be able to get things working now with this additional info, would you be willing to write up what all it took to get MathJax running in an epub file on a page on the dev wiki:
You have to have a google account for me to add you with authoring priviledges I think, but that isn't a very big deal.
--Robert
From: Peter Krautzberger [mailto:pkrau...@umich.edu] Sent: Saturday, October 01, 2011 4:55 PM To: Davide Cervone Cc: Robert Miner; Alfred Wassermann Subject: Re: MathJax and epub3
Davide,
Thanks so much for the help!
Best, Peter. On Sat, Oct 1, 2011 at 5:36 PM, Davide P. Cervone <d...@union.edu> wrote: I have listed the files that can be deleted for several other people. Here is the upshot:
You can remove the MathJax/fonts/HTML-CSS/TeX/eot, svg, and png directories (since the WebKit engine will use the otf fonts). You can also remove the two oft fles that start with MathJax_Win (since they are for Windows). Then you can remove the MathJax/unpacked directory entirely, since you will be running the compressed versions of the files, and you can delete the MathJax/test and MathJax/docs directories completely. If you are only using TeX input (not MathML), then I would recommend using the TeX-AMS_HTML-full configuration file. In that case, you can remove the MathJax/jax/input/MathML and MathJax/jax/output/NativeMML directories, and the MathJax/extensions/mml2jax.js and MathJax/extensions/jsMath2jax.js extensions. You can remove the FontWarnings and v1.0-warnings extensions as well, and you can delete all the configuration files except the one you are using. Since you will be using the MathJax fonts, not the STIX fonts, you can remove the MathJax/jax/output/HTML-CSS/fonts/STIX directory and its subdirectories. If you want to remove every possible file, then you can open the configuration file that you are using and remove the files listed in the MathJax.Ajax.Preloading() call at the top of the file, as those will all be included in the configuration file itself. So you should be able to get things down to a smaller footprint if you need to. I was able to get it to about 1.2MB that way. To get more than that, you would have to limit the features you use.
Davide
On Sep 30, 2011, at 10:59 AM, Robert Miner wrote:
Peter,
Why not just ditch the image fonts, and set the config to use only system fonts. Then you could leave all the images of the individual characters out, and you would only be left with a few dozen (?) js files.
--Robert
From: Peter Krautzberger [mailto:pkrau...@umich.edu] Sent: Friday, September 30, 2011 9:49 AM To: Alfred Wassermann Cc: Robert Miner Subject: Re: MathJax and epub3
Dear Alfred,
I finally constructed an almost validating epub file -- including a full mathjax installation.
The biggest problem is the manifest. Since iBooks is very strict, all ~30.000 files have to be listed. I took me a while to find a reliable tool to create the manifest of 3.5MB (uncompressed). This, I think, was also the cause for most epub viewers to crash when trying to view the end product -- but surprisingly iBooks survived the test.
There are a lot of bugs in the displayed mathematics, but I'll have to test for the causes (on the one hand, the html content in my test file turned out not to be valid, on the other hand $\color{} ...$ seemed to consistently not render at all).
In any case, with the advent of epub3, a minimal mathjax focusing on mathml-rendering should make life much, much easier. I'll make both files (full mathjax, reduced mathjax) available after some more testing next week.
Best wishes, Peter.
On Fri, Sep 30, 2011 at 8:29 AM, Alfred Wassermann
Am 28.09.2011 05:16, schrieb Peter Krautzberger: Dear Alfred, Just a quick reply to your questions.
1) Converting to mathml ahead of time should be easy. I personally use pandoc to create epubs for my tests which creates very reliable mathml (out of pretty much anything). Of course, this brings us back to the lack of mathml support. I'm somewhat in favor of the idea to focus on successful conversion methods (back and forth).
2) Today, I created an epub with a reduced mathjax copy (of ~1.3MB) which drops all fonts and produces mathml only. I haven't tested it properly though. As before, the mathml problems remain.
3) Have you actually attempted to include all of mathjax? I'm curious if that would actually work on an iPad. From what you wrote, it should. I'll try that out.
Peter.
Dear Peter, method 3) has been tested by one of our power users. Unfortunately, without success. In the mathjax google groups some guys claim they could include mathjax in an iOS app. But I did nt see the code, yet. Best wishes, Alfred
The "almost validated" epub I mentioned is unreliable on the iPad //
in iBooks (which is, for now, the point as the only hardware device
supporting javascript in epub2).
Sometimes, mathjax is being rendered, sometimes it is not -- I'm
guessing the size/structure causes these problems. When rendered,
there were also rendering problems: characters like \varepsilon tended
to be rendered twice (i.e., like \varepsilon \varepsilon).
I have also followed Davide's advice from the above conversation to
reduce the mathjax installation. This has different problems. iBooks
runs mathjax (slowly but reliably) with severe rendering problems: all
rendered results are displayed in the wrong place (starting at the
bottom right of the correct position, but no doubling).
I should add: calibre renders the file just fine. I will create a file
from the mathjax test docs and share it later.
Thanks for the info. Can you clarify something for me? Is the positioning problem new to the stripped-down version, or is it also in the original unreliable one (with the double \varepslion)? I have not seen the misplacement that you describe before.
Can you send me the epub file so that I can take a look at it? I have not worked with epub creation before, so if you could also send me instructions for creating the file, I can see if I can figure out what is going on.
Davide
On Oct 3, 2011, at 11:38 AM, Peter Krautzberger wrote:
> The "almost validated" epub I mentioned is unreliable on the iPad // > in iBooks (which is, for now, the point as the only hardware device > supporting javascript in epub2).
> Sometimes, mathjax is being rendered, sometimes it is not -- I'm > guessing the size/structure causes these problems. When rendered, > there were also rendering problems: characters like \varepsilon tended > to be rendered twice (i.e., like \varepsilon \varepsilon).
> I have also followed Davide's advice from the above conversation to > reduce the mathjax installation. This has different problems. iBooks > runs mathjax (slowly but reliably) with severe rendering problems: all > rendered results are displayed in the wrong place (starting at the > bottom right of the correct position, but no doubling).
> I should add: calibre renders the file just fine. I will create a file > from the mathjax test docs and share it later.
If the "full" version didn't have the placement problems, then it looks like something important may have been lost in the removal of files. Not sure what, though. On the other hand, the fact that the full version placed thing properly (if unreliably and sometimes double), then it is hopeful that the problem is not too hard to fix.
Davide
On Oct 3, 2011, at 4:59 PM, Peter Krautzberger wrote:
I got the same result like Peter. All characters are in the wrong position. Also, the suggestion by Davide MathJax.Hub.Config({ "HTML-CSS": { availableFonts: ["TeX"], imageFont: null } });
> I got the same result like Peter. All characters are in the wrong > position. Also, the suggestion by Davide > MathJax.Hub.Config({ > "HTML-CSS": { > availableFonts: ["TeX"], > imageFont: null > } > });
And in particular, the section on Page Tests, does a lot of cross checking, logging of messages, etc. They aren’t merged to master at GitHub yet, but you can grab them from Sean’s fork:
From: mathjax-dev@googlegroups.com [mailto:mathjax-dev@googlegroups.com] On Behalf Of Alfred Wassermann
Sent: Wednesday, October 05, 2011 6:01 AM
To: mathjax-dev@googlegroups.com
Subject: Re: MathJax and epub3
Yes, I see the grey messages, but no error messages. But there are no error messages.
Alfred
Do you see any message about "loading TeX/Main/Italic" or some such things?
I know that you can't access the MathJax menu, or the other usual diagnostic tools, so it is hard to get the details of what is going on, but it appears to not be using the MathJax fonts, and that makes me think something is going wrong with the loading of the fonts or font data.
Davide
On Oct 5, 2011, at 7:01 AM, Alfred Wassermann wrote:
It is hard to debug but I see the following sequence of messages (approximately): * Processing Math: 0% * Loading Web-font: TeX/Math/Italic * Loading Web-font: TeX/... * Processing Math: % * Loading ???.js * Processing Math: 100%
By the way in epub3 all the text files have to be XHTML files.
OK, thanks. That does look like it is trying to load the MathJax fonts, and it does seem to think they are available (otherwise there would be "Failed to load TeX/math/Italic" and other similar messages). Can you send a higher resolution image of the messed up math (can you zoom in and take the picture)? I want to see the characters more carefully to see if I can tell what font it is. It appears NOT to be the actual MathJax fonts.
I will probably have to look into this myself, since it seems that it isn't going to be just a configuration change.
Davide
On Oct 5, 2011, at 12:07 PM, Alfred Wassermann wrote:
> It is hard to debug but I see the following sequence of messages (approximately): > * Processing Math: 0% > * Loading Web-font: TeX/Math/Italic > * Loading Web-font: TeX/... > * Processing Math: % > * Loading ???.js > * Processing Math: 100%
> By the way in epub3 all the text files have to be XHTML files.
> -----Original Message----- > From: mathjax-dev@googlegroups.com [mailto:mathjax- > dev@googlegroups.com] On Behalf Of Davide P. Cervone > Sent: Wednesday, October 05, 2011 12:35 PM > To: mathjax-dev@googlegroups.com > Subject: Re: MathJax and epub3
> OK, thanks. That does look like it is trying to load the MathJax > fonts, and it does seem to think they are available (otherwise there > would be "Failed to load TeX/math/Italic" and other similar messages). > Can you send a higher resolution image of the messed up math (can you > zoom in and take the picture)? I want to see the characters more > carefully to see if I can tell what font it is. It appears NOT to be > the actual MathJax fonts.
> I will probably have to look into this myself, since it seems that it > isn't going to be just a configuration change.
> Davide
> On Oct 5, 2011, at 12:07 PM, Alfred Wassermann wrote:
> > It is hard to debug but I see the following sequence of messages > (approximately): > > * Processing Math: 0% > > * Loading Web-font: TeX/Math/Italic > > * Loading Web-font: TeX/... > > * Processing Math: % > > * Loading ???.js > > * Processing Math: 100%
> > By the way in epub3 all the text files have to be XHTML files.
No, I have one; it is lack of time while teaching and grading. (I had originally left the iPad at home in WV when I started teaching, but I picked it up this past weekend, so will have it to work with now whenever I get the time to do it.)
I suppose I should have said "yes", and you might have offered one. On the other hand, this is an original iPad, not an iPad2. :-)
I finally got the chance to view you book in iBooks, and I certainly see the alignment issues that you have been having. I also can confirm that the fonts in use for your book ARE the MathJax web fonts, so that is good. But it leaves me in the dark about the problem with the alignment. I will have to do some experimenting myself so that I can get some debugging information out of MathJax in the epub file. Make sure you didn't delete the mathjax/jax/output/HTML-CSS/fonts/TeX directory (only the STIX one). Can you send me a directory listing of the MathJax files that are actually included in the epub file (I haven't yet determined how to unpackage the epub into its component files).
Davide
On Oct 3, 2011, at 5:18 PM, Peter Krautzberger wrote:
OK, I was able to unpack the file, and see that the right files appear to be there. I will have to look deeper. That means working out how to create epub files. My initial attempts to unpack your epub file and then repack it as an epub have failed. I guess I'll have to learn how to do it properly. Sigh.
Davide
On Oct 3, 2011, at 5:18 PM, Peter Krautzberger wrote:
It looks like Alfred’s script is doing the hard part of making the EPUB. In general, it isn’t very hard, and I’d be happy to try to help. I’ve done it by hand with small examples. The main thing is the manifest has to correspond to the files. Otherwise, you should just be able to stick with the file structure and zip it.
--Robert
From: mathjax-dev@googlegroups.com [mailto:mathjax-dev@googlegroups.com] On Behalf Of Alfred Wassermann
Sent: Thursday, October 06, 2011 2:41 AM
To: mathjax-dev@googlegroups.com
Subject: Re: MathJax and epub3
Yes, I did figure that out eventually. (But you do need to make sure the mimetype file is the first one in the epub file, apparently). In any case, I was able to repack an epub file and view it, so I should be able to modify it and do some debugging. It looks like it might be the mechanism for measuring the widths and heights isn't working in the ebook, so that could be a problem. But that is just a guess at the moment.
One thing I'd like to try is the SVG output jax, since that should work more reliably. If I could just get an hour free to try it!
> It looks like Alfred’s script is doing the hard part of making the EPUB. In general, it isn’t very hard, and I’d be happy to try to help. I’ve done it by hand with small examples. The main thing is the manifest has to correspond to the files. Otherwise, you should just be able to stick with the file structure and zip it.
> --Robert
> From: mathjax-dev@googlegroups.com [mailto:mathjax-dev@googlegroups.com] On Behalf Of Alfred Wassermann > Sent: Thursday, October 06, 2011 2:41 AM > To: mathjax-dev@googlegroups.com > Subject: Re: MathJax and epub3
I have made some progress. I can now create epub files that include MathJax, and I wrote a little debugging environment to help me be able to see what is going on. Crude, but enough to get by. It does look like there is a problem with the measuring of sizes of things, but I haven't tracked it down exactly yet. I tried out the v2.0-candidate branch, and that does a better job, at least, but is still getting things to be too wide (though the vertical positioning seems better). But the scaling is wrong (everything is twice as big as it should be), so there are several things still to work out. In any case, it looks like it should be able to be made to work, with a little more testing.
I also tried the SVG output, and that worked out better, but is also twice as large as it should be, so that needs work, too. But I'm confident that one or the other can be made to do the job. I'm out of time at the moment, but will look into it further at my next opportunity. I'm sorry it is not a quick fix.
Davide
On Oct 6, 2011, at 3:41 AM, Alfred Wassermann wrote:
Sorry for being so quiet around here. I haven't had time to extend my
how-to on epub publically yet.
I also tried to use other alternatives to otf fonts and nothing
worked. Could you share an example of your SVG solution (just the
configuration or whatever else it needs)?
Your message seems to imply that SVG fonts have the wrong size, but
that's the only problem (i.e., they are placed correctly). That would
be a big step forward, I think.
The SVG code is not yet in the repository, but it is the next thing that I hope to do when I have enough time to get the final pieces in place.
Yes, the placement is correct for the SVG code, but there are some scaling issues (it is not just that the whole expression is too big, which it is, but the super and subscripts are not being reduced in size properly, and I haven't had the chance to check out what is causing that). So it is not quite ready for prime time yet, but does look good for the future.
On the other hand, the v2.0-candidate code does a better job than v1.1a, and doesn't have the "piling up" problem that you are seeing. It is also scaled too large, and it looks like the widths of the expressions aren't being calculated correctly. It think that the HTML- CSS output can also be made to work, but it will take some time to check into it, and I ran out of it. I'm hoping to get a few hours next week to work on it again.
Davide
On Oct 7, 2011, at 1:37 PM, Peter Krautzberger wrote:
> Sorry for being so quiet around here. I haven't had time to extend my > how-to on epub publically yet.
> I also tried to use other alternatives to otf fonts and nothing > worked. Could you share an example of your SVG solution (just the > configuration or whatever else it needs)?
> Your message seems to imply that SVG fonts have the wrong size, but > that's the only problem (i.e., they are placed correctly). That would > be a big step forward, I think.
I just had the chance to upload a mathml-epub onto an ipad running
iOS5 . It didn't look too shabby. Will test more on the weekend (but
this means my silly how-to draft will be delayed).
