Expose the content of <annotation-xml>/<annotation> to the "Show Math As" menu

[fred-wang]

The element is often used by generators to provide the source code of a formula in an alternative syntax (StarOffice Math, content MathML, LaTeX, etc). It could be useful to make the source code available via the "Show Math As" menu so that it can be transfered to other applications.

[PaulTopping]

It would be really nice if there were more robust standards for this sort of thing. An annotation may also be used for data that the user should not see or wouldn’t want to see. There’s also no way for the data itself to fill in ‘X’ in “Show Math As X”. Of course, this also introduces localization issues, though I guess “LaTeX” is the same in all languages. There are also metadata standards being developed as part of the HTML5 family of standards that might be better for this sort of thing.

[fred-wang]

It seems to me that is the standard method recommended by the MathML REC in chapter 6 and I know that people use it. I'm not aware of any use of HTML5 metadata formats for that purpose. It probably does not make sense to make all the annotations available, though. I was thinking to make visible only text annotations (like LaTeX) and XML annotations from a predefined list of "known" flavors (e.g. content MathML or OpenMath).

[pkra]

I believe Paul is referring to #330 and other discussions we've had about improving MathJax output to enable better interoperability, accessibility etc -- in particular for the HTML output but I think it would also apply to MathML output in the sense of enhancing existing MathML in a page (mathjax as an interface to math etc).

For an easy example, the TeX, MathML and Asciimath source could be added as data-* tags instead of the current script-tags. This might improve copy&paste if people adopt our hypothetical standards for doing so -- script tags will never get copied to the clipboard as far as I can see.

Then there are things like webintents and web activities that have great potential. I once made a small demo adding webintents to the MJ menu -- it's extremely easy to write code for that. But these are very early (e.g., Chrome has taken the webintents code out of v24 again, I think).

Anyway, I think there are two different topics here:

• how to add (what) data
• how to expose it to the user.

I think it's too early to discuss this on the issue tracker though. Maybe the dev-list or the dev-wiki is more suitable to collect some data on it.

Regarding the original question, I think it's a good idea in general. I'm wondering how reliable that information is though. For example, how usable is the TeX that you find (does it use macros etc). @fred-wang Could you create a page on the wiki with some use cases that provide such annotation tags?

[dpvc]

I don't think the script tags were ever intended as something that was going to get copy-or-pasted.

As for TeX annotations, there is some commented out code in the TeX input jax that included the original TeX as an annotation in a <semantics> element in the MathML that it generates. I commented it out because I didn't want the output jax to have to bother with it (it would add more nesting to the output unnecessarily). But the Show As MathML could put that in again, I suppose.

But you are right, the issue tracker is probably not the place for this discussion.

[fred-wang]

OK, I see. For this issue ticket, I was really thinking about the annotation-xml/annotation data imported from the MathML source code that third-party tools generate (this is a standard method defined in the MathML REC) and exposing them to the "Show Math As". Note that we already do that, except that the user has to show the original MathML source and extract the annotation himself.

As I understand it, the discussion about "Show Math As" is the other direction: is there a better way to export MathJax's own LaTeX/AsciiMath source? A natural method is to use the standard element but that will only work in the native MathML output and the rendering agent has to handle the info itself (as far as I know, only MathPlayer is able to extract data from the at the moment). Other metadata formats would be more appropriate in HTML-CSS/SVG output mode. Anyway, that's a separate issue: I didn't mean to change the way the data is exposed here, just to make more info available.

[fred-wang]

Could you create a page on the wiki with some use cases that provide such annotation tags?

Here is a first draft: https://github.com/mathjax/MathJax/wiki/Usage-of-the-semantics-element

[pkra]

Awesome -- thanks!

[fred-wang]

I set this "Address Later". We will need to refactor a bit the menu code do add localization support. Peter also mentions elsewhere the idea of a Menu API that will make this easier.

[fred-wang]

OK, I've created a branch that fixes this issue. I added a new configuration option to the MathMenu, that lists some known math format and the corresponding encoding used:

semanticsAnnotations: {
  "TeX": ["TeX", "LaTeX", "application/x-tex"],
  "StarMath": ["StarMath 5.0"],
  "Maple": ["Maple"],
  "Content MathML": ["MathML-Content", "application/mathml-content+xml"],
  "OpenMath": ["OpenMath"]
},

=> Ready For Review

[pkra]

Very cool, Fred!

[dpvc]

This is very nice. There is one addition that should be made: the semanticsAnnotations object should also appear in the MathMenu.js extension, so that it is defined even when default.js isn't the configuration used.

I also wonder if the annotation menu should be disabled rather than hidden when there are no annotations. I found the appearance and disappearance of it a bit disconcerting (but could get used to it). To disable rather than hide, change hidden to disabled in MathEvents.js#L169 and L175 and MathMenu.js#L1032.

Other than that, it is ready for release.

[dpvc]

I did some more testing of this branch and found three more issues.

First, the <math> and <mrow> methods for getAnnotation() can call getAnnotation() on elements that don't implement it. For example

<math>
  <mi>x</mi>
</math>

In this case, the menu code will crash and you won't get MathJax's contextual menu when you click it.

The easiest solution is to add

getAnnotation: function () {},

to the mbase class so that getAnnotation is defined for all elements (and returns null), with math, mrow and semantics overriding that method with their own.

Second, in the semantics implementation of getAnnotation (line 1187), you use this.data[i].attr.encoding, but the attr property is only available on math that was processed by the MathML input jax, so it is possible that this will fail in the future. Currently, neither the TeX nor the AsciiMath input jax produce <semantics> nodes, but future jax might (and future extensions might).

The easiest solution here is to use this.data[i].Get("encoding"), which will look up the encoding from the proper location.

(This code actually points out a problem with how attr is handled in the MathML input jax, as the encoding value should have been stored as a property of the node itself, not its attr object. I will fix that separately, but it's another reason why the Get() method should be used.)

Finally, the indexOf method for Arrays is not available in IE8 and below. One approach would be to use a hash instead of an array, like {TeX: 1, LaTeX:1, "application/x-tex": 1} and then (if this is stored as encodingList), use

if (encodingList[encoding]) {return this.data[i]}

Of course, this would mean converting the array to the hash (either in the configuration data itself, which is ugly, or when the jax is initialized, which means it is harder to reconfigure afterward, or in the semantics getAnnotation call, which is inefficient).

Alternatively, you could loop through the array by hand.

[fred-wang]

to the mbase class so that getAnnotation is defined for all elements (and returns null), with math, mrow and semantics overriding that method with their own.

oops, I thought I had done that... I'll check that and the other issues. Thanks.