Opened 11 years ago
Last modified 9 years ago
#9774 closed enhancement
Use MathJax instead of jsMath — at Version 101
Reported by:  mpatel  Owned by:  jason, was 

Priority:  major  Milestone:  sage5.4 
Component:  notebook  Keywords:  sd41 
Cc:  rbeezer, robert.marik, jhpalmieri, rkirov, kcrisman, kini  Merged in:  
Authors:  Rob Beezer, Davide Cervone, John Palmieri, Jason Grout, Nathan Carter  Reviewers:  Jason Grout, John Palmieri, Punarbasu Purkayastha, Keshav Kini 
Report Upstream:  Fixed upstream, but not in a stable release.  Work issues:  
Branch:  Commit:  
Dependencies:  #13121  Stopgaps: 
Description (last modified by )
MathJax is the successor to jsMath.
apply:
 trac_9774scriptsmathjax.patch to the scripts repository
 trac_9774mathjaxtry7.patch to the sage library repository
Also install the new notebook from the sagenb git fork https://github.com/jasongrout/sagenb/tree/mathjax.
Change History (117)
comment:1 Changed 11 years ago by
 Cc rbeezer added
comment:2 followup: ↓ 4 Changed 11 years ago by
 Cc robert.marik added
comment:3 Changed 11 years ago by
comment:4 in reply to: ↑ 2 ; followup: ↓ 10 Changed 11 years ago by
Replying to rbeezer:
MathJax
is HUGE. Fonts for lots of Unicode points, I guess. SVN checkout is 53 MB, after unzipping fonts.zip it all occupies 171 MB.
I think MathJax? includes the equivalent of our jsmathimagefonts spkg. If we added MathJax? to Sage, it might be good to strip out the image fonts and distribute them separately as a mathjaximagefonts spkg (mathjax faq tells how to do this, I believe). Somewhere I have a halffinished prototype of this solution.
comment:5 Changed 11 years ago by
 Cc jhpalmieri added
comment:6 Changed 11 years ago by
 Description modified (diff)
comment:7 Changed 11 years ago by
 Description modified (diff)
comment:8 Changed 11 years ago by
 Status changed from new to needs_work
comment:9 Changed 11 years ago by
 Description modified (diff)
comment:10 in reply to: ↑ 4 ; followup: ↓ 11 Changed 11 years ago by
Replying to jason:
Replying to rbeezer:
MathJax
is HUGE. Fonts for lots of Unicode points, I guess. SVN checkout is 53 MB, after unzipping fonts.zip it all occupies 171 MB.I think MathJax? includes the equivalent of our jsmathimagefonts spkg. If we added MathJax? to Sage, it might be good to strip out the image fonts and distribute them separately as a mathjaximagefonts spkg (mathjax faq tells how to do this, I believe). Somewhere I have a halffinished prototype of this solution.
Just in case you haven't thought of this, what Gollum (the GitHub forum) folks did was host the MathJax? fonts on Amazon S3, and then not include them in their distribution. I don't know if that would work for you, but I wanted to point it out.
comment:11 in reply to: ↑ 10 Changed 11 years ago by
Replying to rminer:
Replying to jason:
Replying to rbeezer:
MathJax
is HUGE. Fonts for lots of Unicode points, I guess. SVN checkout is 53 MB, after unzipping fonts.zip it all occupies 171 MB.I think MathJax? includes the equivalent of our jsmathimagefonts spkg. If we added MathJax? to Sage, it might be good to strip out the image fonts and distribute them separately as a mathjaximagefonts spkg (mathjax faq tells how to do this, I believe). Somewhere I have a halffinished prototype of this solution.
Just in case you haven't thought of this, what Gollum (the GitHub forum) folks did was host the MathJax? fonts on Amazon S3, and then not include them in their distribution. I don't know if that would work for you, but I wanted to point it out.
That's a very interesting solution. I suppose we could host them on the main sage webserver, for example. Or maybe Google Code or something like that so we had redundant sources.
comment:12 Changed 11 years ago by
I just found the web page http://bitbucket.org/kevindunn/sphinxextensionmathjax/wiki/Home, which allows the use of Sphinx with MathJax. Assuming it works, we could upgrade all of Sage (not just the notebook but also the docs), to use MathJax instead of jsMath.
comment:13 Changed 11 years ago by
comment:14 Changed 11 years ago by
 Cc rkirov added
Steps that will install MathJax
with the new Flask notebook.
 Install the Flask version of the notebook.
http://code.google.com/r/rkirovflask/
 Download v1.1 of
MathJax
as zip file (from bottom of page).http://www.mathjax.org/download/
 Unzip the MathJax? distribution into
$SAGE_ROOT/devel/sagenb/sagenb/data
and
rename the new directory:mathjaxMathJax5a7e4d7
(or whatever)
to be just the directory:MathJax
 jsmath compatibility. Edit
$SAGE_ROOT/devel/sagenb/sagenb/data/MathJax/config/default.js
by adding "jsMath2jax.js
" as the first entry of the "extensions" list, so it becomesextensions: ["jsMath2jax.js", "tex2jax.js"]
 Edit
$SAGE_ROOT/devel/sagenb/sagenb/data/sage/html/notebook/base.html
and change<script type="text/javascript" src="/javascript/sage/jsmath.js"></script>
to<script type="text/javascript" src="/static/MathJax/MathJax.js?config=default"></script>
This will install MathJax
and the notebook will use it. But the setup is buggy, so needs work. I'm sure there is some configuration on the Sage side and/or the MathJax
side that will need changes.
 A mix of black and green symbols.
 Only renders on a reload  adding new TeX via TinyMCE returns with an error about not finding fonts.
 Some stray tags are being rendered, or something.
Screenshots attached:
Double Integral:
http://wiki.math.toronto.edu/TorontoMathWiki/index.php/JsMath/MathJax_%28TeX_for_Web%29
$$ (2\pi h)^{d}\iint_{\{H(x,\xi) <\tau\}} dx d\xi $$
Matrix:
$\begin{bmatrix} x^2 & y^2\\ x^3 & \cos(z) \end{bmatrix}$
comment:15 Changed 11 years ago by
 Cc kcrisman added
comment:16 Changed 10 years ago by
I'm attaching a draft of a patch for the Sage library. The integration with Sphinx is completely untested, and is probably broken. See the top of the patch file for a list of things to do.
Changed 10 years ago by
comment:17 Changed 10 years ago by
Added patch to Flask notebook code to support MathJax
, built with Davide Cervone's help.
 This is a handedited patch to recover from an hg messup. I think it is OK, but be wary.
 Fonts are coming from
MathJax
CDN. Comment in code indicates change for localMathJax
installation. 
<script>
tags added by Palmieri's Sage library patch need to be handled with care on notebook side, so notebook does not evaluate them. Right now when "Typeset" box is checked, evaluated cells raise a Javascript error inevaluate_script_tags
(or a function with a similar name).
comment:18 followup: ↓ 19 Changed 10 years ago by
John's patch doesn't apply cleanly to stock 4.7; does it apply to the most recent alpha of 4.7.1?
applying mathjax.patch patching file sage/misc/latex.py Hunk #15 succeeded at 1750 with fuzz 2 (offset 2 lines). Hunk #23 FAILED at 2382 1 out of 25 hunks FAILED  saving rejects to file sage/misc/latex.py.rej patch failed, unable to continue (try v) patch failed, rejects left in working dir errors during apply, please fix and refresh mathjax.patch
comment:19 in reply to: ↑ 18 Changed 10 years ago by
Replying to jason:
John's patch doesn't apply cleanly to stock 4.7; does it apply to the most recent alpha of 4.7.1?
Yes.
rob@tiger:/sage/dev/devel/sage$ ../../sage version  Sage Version 4.7.1.alpha2, Release Date: 20110607  * Warning: this is a prerelease version, and it may be unstable. * rob@tiger:/sage/dev/devel/sage$ hg qimport P http://trac.sagemath.org/sage_trac/rawattachment/ticket/9774/mathjax.patch adding mathjax.patch to series file applying mathjax.patch now at: mathjax.patch
comment:20 followup: ↓ 22 Changed 10 years ago by
Don't we need the actual MathJax? spkg or files somewhere? Rob: do you have instructions for putting these in the directory, or do we use my halffinished patch mentioned in the description, or something else?
Changed 10 years ago by
comment:21 Changed 10 years ago by
Attached a patch to add sagenb/notebook/mathjax.py
New file is similar to, but not identical to, sagenb/notebook/jsmath.py
We did not delete the latter, but I think it can safely go away.
comment:22 in reply to: ↑ 20 Changed 10 years ago by
Replying to jason:
Don't we need the actual MathJax? spkg or files somewhere? Rob: do you have instructions for putting these in the directory, or do we use my halffinished patch mentioned in the description, or something else?
We need an spkg to be Internetindependent.
We did not experiment with your old one. I think it would be best to make a new one. This requires a small edit (once you are sure all works with CDN version).
Font suggestions coming up.
comment:23 Changed 10 years ago by
Davide Cervone says for an spkg we need only keep font subdirectories called
otf, eot, svg
In particular png
subdirectories are hugeest and can go away.
Rob
comment:24 Changed 10 years ago by
ah; I bet it's because the svg fonts replace the png fonts for all browsers we care about, or something. Okay, Davide is the final authority on what browsers support what...
comment:25 followup: ↓ 26 Changed 10 years ago by
I get an error from the line: response = make_response(render_template('js/mathjax.js', theme_mathjax_macros=mathjax_macros))
in flask_version/base.py. I see a jsmath.js template in sagenb/data/sage/js/jsmath.js
template. Is this another file that didn't get added to the patch?
Changed 10 years ago by
comment:26 in reply to: ↑ 25 Changed 10 years ago by
Replying to jason:
Is this another file that didn't get added to the patch?
Yes, same situation  "old" file could be removed.
"fileaddtwo" should create the missing file. Sorry about all the problems  my hg status
is useless with the hg messup.
This file will require an edit when we move from CDN to spkg.
Thankyou!!!!!!!!!!!!!!
comment:27 Changed 10 years ago by
I have a working version, but there still seems to be something slightly different about Typeset mode compared to before. I'm posting my patch, but there is still *lots* of debugging stuff in the patch that should be taken out before a final version is posted.
comment:28 Changed 10 years ago by
The problem with the Sage notebook code intercepting (and evaluating) the contents of <script>,</script>
tags can probably be rectified in one, or both, of the functions of the Flask notebook:
In sagenb/data/sage/js/notebook_lib.js: eval_script_tags separate_script_tags
comment:29 Changed 10 years ago by
rbeezer: yes, that's what my patch does. The current remaining problem is something different.
comment:30 Changed 10 years ago by
In the current version of mathjax.patch, look at the top of the file for some unfinished business. For example, the pathname at the end of doc/common/themes/sage/static/mathjax_sage.js_t might be wrong, which might break MathJax? in the Sphinx documentation.
comment:31 Changed 10 years ago by
In the current state, the mathjax switch seems to work with a build of the html version of the reference manual.
Mathematics in the documentation is being wrapped with \(, \)
pairs, which are the default MathJax? delimiters for inline mathematics. However these snippets are not being rendered when viewed in a browser.
In the page source of a page of the html documentation, I see a script in a file named mathjax_sage.js
which we are not generating. We do have jsmath_sage.js
lying about in the doc directories, which have the full jsmath configuration information. So this likely needs repair before an HTML documentation page with MathJax? support is built properly.
comment:32 Changed 10 years ago by
OK, I can force mathjax_sage.js
to be created by totally trashing some of the documentation output with
devel/sage/doc/output/html$ rm rf en
before rebuilding the HTML reference documentation. It comes from a template file mathjax_sage.js_t
that one of the patches creates properly.
Mathematics still does not render, but I think the MathJax? configuration is looking locally. Time for a local MathJax? install, I think.
comment:33 followup: ↓ 34 Changed 10 years ago by
In a page of the documentation, I inserted manually
<script type="text/javascript" src="http://cdn.mathjax.org/mathjax/latest/MathJax.js?config=TeXAMS_HTMLfull">
and the mathematics rendered properly with MathJax.
Then I installed the mathjax 1.1a distribution alongside jsmath in the Flask notebook tree and used
<script type="text/javascript" src="http://cdn.mathjax.org/mathjax/latest/MathJax.js?config=TeXAMS_HTMLfull">
which also caused the math to render properly. (/sage/notebook is my SAGE_ROOT
)
In both cases the fonts did not look too great to my eye.
Questions:
(a) How do we get these scripts commands to be placed properly in each page of the docs, with a proper relative path?
(b) Is there really a clean separation between Sage and the notebook if mathjax mode for documentation requires the MathJax code from the notebook?
comment:34 in reply to: ↑ 33 Changed 10 years ago by
Replying to rbeezer:
Then I installed the mathjax 1.1a distribution alongside jsmath in the Flask notebook tree and
which also caused the math to render properly. (/sage/notebook is my
SAGE_ROOT
)
Oops, the incantation for a local install of MathJax should have been
<script type="text/javascript" src="/sage/notebook/devel/rkirovflask/sagenb/data/mathjax/MathJax.js?config=TeXAMS_HTMLfull"></script>
comment:35 Changed 10 years ago by
Here's a new version of my patch. With a locally installed version of MathJax, the documentation seems to render properly. I don't know if this is the best solution, but at least it seems to work.
(I installed MathJax by downloading a zip file: https://github.com/mathjax/MathJax/zipball/v1.1a. I unzipped it and renamed the resulting directory to mathjax
, and moved it to SAGE_ROOT/local/lib/python/sitepackages/sagenb0.9.0py2.6.egg/sagenb/data/
, alongside of the old jsmath
directory. I also put a copy of the mathjax
directory in SAGE_ROOT/devel/sagenb/sagenb/data/
. I haven't tried to install local fonts.)
comment:36 Changed 10 years ago by
By the way, one comment about the latest version of my patch: it fixes the problem Rob mentioned about getting the appropriate invocation to MathJax in each piece of documentation. It does this by setting the variable mathjax_path
to
file://[path to MathJax.js]?config=TeXAMS_HTMLfull,file://[path to mathjax_sage.js]
The paths here are absolute ones, in SAGE_ROOT/devel/sage/doc/output/html/en/_static/
. This directory is created by sage docbuild website html
, so now if you build the docs with the j
flag (to enable MathJax), it first builds the website target, then builds whatever you requested. There might be a better way to do this, using relative paths, but I couldn't get it to work. For example, using
<script type="text/javascript" src="_static/MathJax.js?config=TeXAMS_HTMLfull"></script>
successfully turns on MathJax, but without the local configuration turning on the custom macros. Using
<script type="text/javascript" src="_static/MathJax.js?config=TeXAMS_HTMLfull,mathjax_sage.js"></script>
doesn't work, nor does any variation of it that I tried. (Note that when you change the path specifying the local configuration file, you also need to change the last line of that configuration file to match it, according to http://www.mathjax.org/docs/1.1/configuration.html.) If we could copy the file mathjax_sage.js
to _static/config/local/
, then we could use
<script type="text/javascript" src="_static/MathJax.js?config=TeXAMS_HTMLfull,local/mathjax_sage.js"></script>
for the invocation, and
MathJax.Ajax.loadComplete("[MathJax]/config/local/mathjax_sage.js")
for the last line in mathjax_sage.js. But I don't know how to put this file in the right place: it's autogenerated, so we can't put it in the MathJax local directory ahead of time, and I can't figure out how to get Sphinx to do it when it's building the documentation.
Another option would be to try to get Sphinx to write this into each html file:
<script type="text/xmathjaxconfig"> [contents of mathjax_sage.js] </script> <script type="text/javascript" src="_static/MathJax.js?config=TeXAMS_HTMLfull"></script>
We can get the last line by specifing mathjax_path in conf.py, but I don't know how to get the first part.
I'll keep working on it...
comment:37 Changed 10 years ago by
Hi John,
Are you testing with the Flask notebook? If not, it might be easier to build what look like absolute paths to the MathJax files in the data directory of the notebook, because of the way Flask sets up paths with the "route()" command? I can't tell if this will solve the problem you are wrestling with or not.
The <script>
tags were being messed with by the notebook on the server side just before being dispatched. But then the client sees two versions  wrapped and unwrapped, and totally trashes one of them to prevent running a script twice. I thought Jason had a patch for this too, but I haven't seen it. The symptom would be when you cycle through wrap, nowrap, and hide for some output, then one of them is missing. I guess we can press on and fix this eventually.
About to go offline shortly for the long weekend, but will come back to this when I can stick with it.
Rob
comment:38 Changed 10 years ago by
For building the documentation, the notebook isn't relevant: they get built from the command line. I think that when trying to make MathJax work with the notebook, we'll want to construct the URLs differently than I'm doing for the docs, but I don't know enough about the paths, etc., in the notebook to say exactly how to do it. For what it's worth, though, I have been using the flask notebook.
comment:39 Changed 10 years ago by
I've uploaded my current flask patch queue here: http://sage.math.washington.edu/home/jason/mathjaxflaskpatches.tar.gz My current (probably nonworking) patch queue for my sage library, including two mathjax patches, is here: http://sage.math.washington.edu/home/jason/mathjaxsagepatches.tar.gz
These are both worksinprogress, but currently stalled for at least a week or two.
comment:40 Changed 10 years ago by
I stalled working on it because of other time pressures with the singlecell, but also because I realized that the *next* version of Sphinx supports Mathjax out of the box, while I thought the current version would require patches.
comment:41 Changed 10 years ago by
I think that the implementation of MathJax in Sphinx comes from the file mathjax.py, which is available for download separately. So my patch just adds it.
I'm attaching another version ("v2") of my patch; this one writes the MathJax local configuration file to a file in SAGE_ROOT/local/lib/python/sitelibraries/sagenb.../
, which may not be a good idea. But it allows for a simpler invocation of MathJax. I'm really not sure which approach is better.
comment:42 Changed 10 years ago by
What is the current status on this now? Where should we start working on it again?
comment:43 Changed 10 years ago by
The last patch involving MathJax?, that I rebased from the ones here is on google code:
http://code.google.com/p/sagenb/issues/detail?id=46
Unless Jonathan Gutow did something else on top of it, this is latest work.
comment:44 Changed 10 years ago by
 Description modified (diff)
 Milestone set to sage5.0
comment:45 Changed 10 years ago by
 Description modified (diff)
comment:46 Changed 10 years ago by
 Description modified (diff)
 Reviewers set to Jason Grout
 Status changed from needs_work to needs_review
comment:47 Changed 10 years ago by
 Description modified (diff)
comment:48 Changed 10 years ago by
 Description modified (diff)
comment:49 Changed 10 years ago by
comment:50 Changed 10 years ago by
 Description modified (diff)
comment:51 Changed 10 years ago by
 Cc kini added
comment:52 Changed 10 years ago by
 Description modified (diff)
 Status changed from needs_review to needs_work
I assume the description was incorrect, but even with the correction I made, trac_9774mathjaxtry3.patch does not apply cleanly on top of 4.8.alpha6. On 4.8.alpha4 (at least by hg up 4.8.alpha4 && sage b
) it fails a couple of doctests. Rebasing and fixing doctests now.
comment:53 Changed 10 years ago by
Whoops, looks like "a couple of doctests" was an understatement. I guess it should be done in a separate patch. Here's the new patch, rebased only.
comment:54 Changed 10 years ago by
The try3 patches were against alpha4. Late last night I patched 5.0.beta1, and realized there were some patch failures, which I fixed. I'll compare your rebase with mine.
I also realized last night during dinner that I hadn't run doctests, so those should be run on sage and sagenb. Thanks for checking this.
comment:55 followup: ↓ 58 Changed 10 years ago by
You don't happen to have your changes as a separate patch on top of mine, do you? I'd love to review your changes, but it's a bit hard to see where your additions are.
If you don't have it, I'll just diff the two patches, but that's a little harder to read than a separate patch.
comment:56 Changed 10 years ago by
I wonder if the reported problems with "<" and ">" are due to line 1685 in latex.py:
x = x.replace('<', '<').replace('>', '>')
This was put in for use with jsMath, but maybe MathJax doesn't have the same issues.
comment:57 Changed 10 years ago by
It was more complicated than that, but I've fixed it by getting rid of the math_parse function and conditional typesetting we were doing. We've configured MathJax? to make the math_parse function obsolete, and it's inconsistent to typeset only conditionally. See https://github.com/jasongrout/sagenb/commit/6a4eed4d6baf78cd19fe634144c727843d1af08d and https://github.com/jasongrout/sagenb/commit/8d6b0a071c13850cea983faaa6b910be498cd0ae
I've updated test.sagenb.org with the newest mathjax changes.
comment:58 in reply to: ↑ 55 Changed 10 years ago by
Replying to jason:
You don't happen to have your changes as a separate patch on top of mine, do you? I'd love to review your changes, but it's a bit hard to see where your additions are.
If you don't have it, I'll just diff the two patches, but that's a little harder to read than a separate patch.
Er, no, I don't, unfortunately. You can get such a diff by doing hg qfinish
on my patch on top of 4.8.alpha6, then hg up 4.8.alpha4
, then hg qimport P [your patch]
, then hg diff r tip r [revid of my patch]
(you would probably want to hg strip [revid of my patch]
later).
But I don't see how that's very useful to you since it will be some gigantic patch subsuming everything that was merged in alpha5 and alpha6. In your position I would just diff the two patches, but maybe I don't understand what you're asking for. If you just want diffs for the files touched by the patch, that would also be possible, but maybe a bit messier. Do as above, except also qfinish your patch, and then merge the two qfinished patches' commits into a new commit, with your patch as the primary parent. Then do hg log r [the merge commit] p [the file names]
. (I think that should work.) Or I guess you could just make the gigantic patch and ignore the parts that are on irrelevant files.
comment:59 Changed 10 years ago by
 Description modified (diff)
 Status changed from needs_work to needs_review
kini: your try4 patch looks fine to me. It applies cleanly on 5.0.some_version_of_prealpha1_or_beta1
comment:60 Changed 10 years ago by
I seem to have a broken install here (2nd try). I followed this sequence of steps to install mathjax + jmol (output of history
in zsh).
663 local/bin 664 ../../sage hg qimport http://trac.sagemath.org/sage_trac/rawattachment/ticket/9774/trac_9774scriptsmathjax.patch *** ../../sage hg qpush # This line is missing because of my zsh setup 666 ../../ 667 l 668 cd devel/sage 669 ../../sage hg qimport http://trac.sagemath.org/sage_trac/rawattachment/ticket/9774/trac_9774mathjaxtry4.patch 670 ../../sage hg qpush 671 ../../sage b 672 ../.. 674 ./sage f http://sage.math.washington.edu/home/jason/mathjax/sagenb0.9.0.spkg 675 ./sage f http://sage.math.washington.edu/home/jason/jmol/jmol12.0.45.p4.spkg 676 ./sage b 678 ./sage n 679 BROWSER="firefox" ./sage n
Mathjax is broken and so is jmol. Do I need to install something extra for mathjax + jmol to work? These steps were followed on a freshly compiled sage4.8. Screenshot: http://i.imgur.com/A6wrQ.png
comment:61 Changed 10 years ago by
 Dependencies set to #11080
comment:62 Changed 10 years ago by
And thanks for looking at this! This ticket is also live at test.sagenb.org.
comment:63 Changed 10 years ago by
There seems to be some problem with mathjax rendering with LatexExpr
. Compare the jsmath rendering here: http://sagenb.org/home/pub/4337/ to the mathjax here: http://test.sagenb.org/home/pub/19/
comment:64 Changed 10 years ago by
Some doctest failures with this patch:
...sage5.0.beta2/devel/sage> ../../sage t sage/misc/latex.py sage t "devel/sagemainbackup/sage/misc/latex.py" ********************************************************************** File "/home/punarbasu/Installations/sage5.0.beta2/devel/sagemainbackup/sage/misc/latex.py", line 1765: sage: MathJax().eval(type(3), mode='inline') Expected: <html><script type="math/tex">\newcommand{\Bold}[1]{\mathbf{#1}}\hbox{ < type 'sage.rings.integer.Integer' > }</script></html> Got: <html><script type="math/tex">\newcommand{\Bold}[1]{\mathbf{#1}}\verb<type\phantom{\verb!x!}\verb'sage.rings.integer.Integer'></script></html> ********************************************************************** File "/home/punarbasu/Installations/sage5.0.beta2/devel/sagemainbackup/sage/misc/latex.py", line 2199: sage: sys.displayhook Expected: <html>...\verb<function\phantom{x}\verbpretty_print...</html> Got: <html><script type="math/tex">\newcommand{\Bold}[1]{\mathbf{#1}}\verb<function\phantom{\verb!x!}\verbpretty_print\phantom{\verb!x!}\verbat\phantom{\verb!x!}\verb0x150b050></script></html> ********************************************************************** 2 items had failures: 1 of 7 in __main__.example_52 1 of 7 in __main__.example_59 ***Test Failed*** 2 failures. For whitespace errors, see the file /home/punarbasu/.sage//tmp/latex_7938.py [3.1 s]  The following tests failed: sage t "devel/sagemainbackup/sage/misc/latex.py" Total time for all tests: 3.1 seconds
comment:65 Changed 9 years ago by
 Status changed from needs_review to needs_work
 Work issues set to does not work in DOT_SAGE
rebased patch to sage5.1beta2.
There is however a very weird problem with mathjax, both 1.0 and 2.0 branches. The math is not rendered as long as the sagenb directory is under DOT_SAGE
. To reproduce, try the steps below. Use any branch (mathjax1 or mathjax2) for instance jasongrout/mathjax (mathjax1) in github.
mkdir /tmp/a
DOT_SAGE=/tmp/a sage n
. First, note that the above command launches the nb in port 8000 instead of 8080. But if we use some other command such as the one below, then it does open in port 8080.
sage n directory=/tmp/a.sagenb
 Open a new worksheet, and try to print latex, for instance run the command
view('?')
. You will get a js error popup saying (in Opera 12.00RC, and Opera11.64)SyntaxError: at index 0 in "\newcommand{\Bold}[1]{\mathbf{#1}}?": invalid character escape sequence
In firefox10.0.4 I simply get the error: SyntaxError: illegal character
. This error is not present if I run sage instead as sage n directory=/tmp/a.sagenb
, and the mathjax output is all good.
comment:66 Changed 9 years ago by
 Description modified (diff)
comment:67 followup: ↓ 68 Changed 9 years ago by
It looks like Samuel Ainsworth has independently got MathJax working in his "newui" branch (I haven't tested it). At a glance it seems he's using the MathJax CDN.
comment:68 in reply to: ↑ 67 Changed 9 years ago by
Replying to kini:
It looks like Samuel Ainsworth has independently got MathJax working in his "newui" branch (I haven't tested it). At a glance it seems he's using the MathJax CDN.
Has he applied the mathjax patches to the sage repo? Without those patches it won't work. And these patches were not applying for quite sometime. I just updated one today.
There are more changes to make actually. SAGE_ROOT/devel/sage/doc
has lots of jsmath references, which all need to be converted to mathjax.
If you can independently confirm the behavior in comment:65 then it will be good.
comment:69 Changed 9 years ago by
 Description modified (diff)
 Reviewers changed from Jason Grout to Jason Grout, John Palmieri
 Status changed from needs_work to positive_review
Regarding your comment: the spkg in the ticket description is quite old, and that was probably the problem. The notebook has now been upgraded to use MathJax, and it does not have the problem you described.
comment:70 Changed 9 years ago by
 Work issues does not work in DOT_SAGE deleted
comment:71 Changed 9 years ago by
 Milestone changed from sage5.1 to sage5.2
Fantastic news!
This was merged in the notebook at this github pull request, just for completeness.
Since #11080 is for Sage 5.2, this should be too.
comment:72 Changed 9 years ago by
 Reviewers changed from Jason Grout, John Palmieri to Jason Grout, John Palmieri, Punarbasu Purkayastha, Keshav Kini
comment:73 Changed 9 years ago by
comment:74 followup: ↓ 75 Changed 9 years ago by
 Status changed from positive_review to needs_work
We have doctest failures:
sage t long force_lib devel/sagenbmain/sagenb/notebook/jsmath.py # 5 doctests failed sage t long force_lib devel/sagenbmain/sagenb/misc/sphinxify.py # 1 doctests failed sage t long force_lib devel/sage/doc/de/tutorial/latex.rst # 27 doctests failed sage t long force_lib devel/sage/doc/en/tutorial/latex.rst # 27 doctests failed sage t long force_lib devel/sage/sage/matrix/matrix0.pyx # 1 doctests failed
The problem with sphinxify is easy to fix. Should we delete jsmath.py, or is it still needed for something?
The problems with the tutorial are a bit more involved, since (a) it explicitly discusses using JSMath and (b) I don't speak German.
I don't understand the matrix0.pyx problem at all, at least not yet.
comment:75 in reply to: ↑ 74 Changed 9 years ago by
The problems with the tutorial are a bit more involved, since (a) it explicitly discusses using JSMath and (b) I don't speak German.
There are a number of Sage folks who do, though, so hopefully the second one won't be a problem. Once you fix the English one let us know.
I wonder why the other tutorial versions don't cause problems? In particular, there is a tutorial in Russian and one in French in 5.1.beta3 that you don't mention as having caused problems  maybe they're based on different versions of the English tutorial.
Changed 9 years ago by
comment:76 followup: ↓ 80 Changed 9 years ago by
 Description modified (diff)
 Work issues set to fix sagenb
Here is a patch fixing the doctests in the Sage library. For the English tutorial, I did a little bit more than a searchandreplace, changing "JSMath" to "MathJax". For the German tutorial, that's essentially all I did.
sagenb also needs fixing, and I don't have the energy to do this right now. If someone else wants to, that would be great. Here's what I think should be done: delete notebook/jsmath.py, and patch misc/sphinxify.py:

sagenb/misc/sphinxify.py
diff git a/sagenb/misc/sphinxify.py b/sagenb/misc/sphinxify.py index 837b40a..85c9f4d 100644
a b def sphinxify(docstring, format='html'): 71 71 sage: sphinxify('**Testing**\n`monospace`') 72 72 '\n<div class="docstring"...<strong>Testing</strong>\n<span class="math"...</p>\n\n\n</div 73 73 sage: sphinxify('`x=y`') 74 '\n<div class="docstring">\n \n <p><span class="math"> x=y</span></p>\n\n\n</div>'74 '\n<div class="docstring">\n \n <p><span class="math">\\(x=y\\)</span></p>\n\n\n</div> 75 75 sage: sphinxify('`x=y`', format='text') 76 76 'x=y\n' 77 77 sage: sphinxify(':math:`x=y`', format='text')
comment:77 Changed 9 years ago by
Regarding the change in matrix0.py: the main patch already modified that file, and this doctest was missed. (jsMath can't handle the command \hline
, so matrices with subdivisions had to be handled specially in the notebook. MathJax handles this command just fine, so the specialcasing went away, so the extra doctest for the notebook needed to go away, too.)
comment:78 Changed 9 years ago by
test.sagenb.org and alpha.sagenb.org are running these patches up through try5 (but not the doctest patch just posted).
comment:79 Changed 9 years ago by
Regarding my comment about sagenb: apparently jsmath.py is already gone in the new sagenb; I don't know why I thought otherwise. sphinxify.py still needs to be patched.
comment:80 in reply to: ↑ 76 ; followup: ↓ 84 Changed 9 years ago by
Here is a patch fixing the doctests in the Sage library. For the English tutorial, I did a little bit more than a searchandreplace, changing "JSMath" to "MathJax". For the German tutorial, that's essentially all I did.
That looks to have been enough. I have no idea if all the stuff in English is actually true any more, of course. As long as
In the case of Sage, the notebook is always connected to a server used to execute the Sage commands, and this server also provides the necessary jsMath fonts. So there is nothing extra to set up to have typeset mathematics in your web browser when you use the Sage notebook.
is still true for MathJax you should be fine.
comment:81 Changed 9 years ago by
 Status changed from needs_work to needs_review
I just submitted a pull request for the sphinxify change in sagenb. So I think this is ready for review. The only thing that needs review is the doctest patch, which just fixes doctests. I'm running doctests now.
comment:82 Changed 9 years ago by
 Work issues fix sagenb deleted
comment:83 Changed 9 years ago by
All tests pass with the new patches.
comment:84 in reply to: ↑ 80 Changed 9 years ago by
Replying to kcrisman:
As long as
In the case of Sage, the notebook is always connected to a server used to execute the Sage commands, and this server also provides the necessary jsMath fonts. So there is nothing extra to set up to have typeset mathematics in your web browser when you use the Sage notebook.is still true for MathJax you should be fine.
As far as I can tell, it still is. If it's not, it's close enough.
comment:85 Changed 9 years ago by
I see that after this mathjax1.1 is merged into sagenb. Shouldn't we move to mathjax2 directly?
comment:86 Changed 9 years ago by
MathJax? 2.0 should be a separate patch. Let's not delay this yearinthemaking positivelyreviewed patch any longer.
comment:87 Changed 9 years ago by
 Summary changed from Upgrade the notebook to use MathJax instead of jsMath to Use MathJax instead of jsMath
comment:88 Changed 9 years ago by
Once my make ptestlong
finishes, I'm just going to give this positive review. The pull request has already been merged into sagenb (and so everyone will need to apply these patches anyway if they want to test the latest revision of the notebook), the patch looks good to me (though I don't know much about the code it touches), and from some poking around it seems to work as advertised.
comment:89 Changed 9 years ago by
 Status changed from needs_review to positive_review
All systems go!
comment:90 Changed 9 years ago by
 Keywords sd41 added
comment:91 Changed 9 years ago by
 Status changed from positive_review to needs_work
This obviously needs a proper sagenb spkg before it can be merged.
comment:92 Changed 9 years ago by
Important note: if the sagenb people can write me a script which automatically creates a new sagenb spkg from a git repository, I'll be happy to use it.
comment:93 Changed 9 years ago by
 Dependencies changed from #11080 to #13121
 Report Upstream changed from N/A to Fixed upstream, but not in a stable release.
 Status changed from needs_work to positive_review
No, we'll (I'll) do it manually. SPKGs being autogenerated is at odds with having repositories for SPKGs.
I have written scripts which at least simplify the process, though  they create a directory which can be used as the new SPKG's src/
directory without modification of the spkginstall
script. Those are at this pull request, which I'd appreciate it if someone reviewed :)
I'm setting this ticket to positive review because there is nothing left to be done on the Sage end other than wait for #13121 to get in (i.e. Sage to upgrade to sagenb 0.9.1).
comment:94 Changed 9 years ago by
 Description modified (diff)
comment:95 Changed 9 years ago by
 Milestone changed from sage5.2 to sagepending
comment:96 Changed 9 years ago by
 Description modified (diff)
Rebased to Sage 5.1.beta5 (because of #11775).
comment:97 Changed 9 years ago by
I rebased trac_9774mathjaxtry6.patch to sage 5.1beta6
comment:98 Changed 9 years ago by
Dumb question. If the changes to the latex of some vectors is to work around a jsmath bug, maybe we shouldn't remove that comment; after all, maybe that workaround is no longer needed, but now no one will no that it was a workaround in the first place... Maybe it works now completely, and we should change it or remove it!
This was apparently introduced in the try3 patch.
comment:99 followup: ↓ 100 Changed 9 years ago by
Can you tell which part is "weird"? Is it the braces around \mapsto
? If I remove those braces, it works with both jsMath and MathJax.
comment:100 in reply to: ↑ 99 Changed 9 years ago by
Can you tell which part is "weird"? Is it the braces around
\mapsto
? If I remove those braces, it works with both jsMath and MathJax.
I wish I could tell you! I didn't write it. I was just trying to point this out, in case it was clear to someone else. If it's not clear what the problem is, maybe we leave the comment in? Maybe Jason can explain.
Also, both representations have the braces, and braces are fine for demarcating things in TeX, so I don't think that's the issue. hg blame says that this dates from before the symbolics switch to Pynac.
Anyway, not a huge deal, but pointing it out.
comment:101 Changed 9 years ago by
 Description modified (diff)
Updated patch to the changes introduced in #11775. Otherwise, typeset was broken.
I've been experimenting a bit with
MathJax
, outside of the notebook. Mostly thinking about how it will handle tex4ht output in jsMath mode, but here are some some observations that might be useful:MathJax
. Just insert something likeNo need for a "process()" call at the end, etc.
\(..\)
. This would be a good thing, since if a user adds text (via TinyMCE) right now jsMath tries to parse the following as math.$$..$$
and\[..\]
both work for display math. We would break lots of old worksheets if we stopped recognizing$..$
.config/MathJax.js
which is just one huge wellcommented Javascript object. It can be overridden in a web page by adding a new version into the script block mentioned above. Maybe we want to make this easy for users to access, or maybe it is easy already, or maybe we don't want to bother.http://www.mathjax.org/resources/docs/?configuration.html
MathJax
is HUGE. Fonts for lots of Unicode points, I guess. SVN checkout is 53 MB, after unzipping fonts.zip it all occupies 171 MB.html.table()
usesclass="math"
which is a jsMath way to tag span's or div's for processing. I haven't found how to do something similiar inMathJax
, though this will work in jsMath compatibility mode. A very small test would indicate that the two modes can be used at the same time.