TeX input, amsmath: change \sideset implementation to generate multiscripts

[pkra]

As per F2F and additional feedback from AMS math folks, multiscripts seem to be the desired output.

[pkra]

(Maybe try to push it into 2.6...)

[dpvc]

Does the self-assignment mean you are planning to implement it yourself? If so, I could give some suggestions about one approach that should work fairly easily.

[pkra]

Does the self-assignment mean you are planning to implement it yourself?

Yes. Since I realized my other plan is basically done. But I don't mind being beaten to it.

If so, I could give some suggestions about one approach that should work fairly easily.

That'd be great!

[dpvc]

My suggestion is to use ParseArg() to get the arguments and parse them as MathML. This should give you a MathML element for each argument. Since the prescript and postscript arguments should look like (_a^b), you should get back an msubsup (or potentially an msup or msub, or possibly just an empty mrow if the argument is empty). You can check this using the type property of the returned MathML, or via instanceof. The children of the mml element are in its data property, and its sub and sup properties give the indices into the array for the proper element (since they differ for msubsup, msub, and msup). E.g.,

SideSet: function (name) {
  var sub, sup, presub, presup;
  var mml = this.ParseArg(name);
  if (mml instanceof MML.msubsup) {sub = mml.data[mml.sub], sup = mml.data[mml.sup]}
    else this.Error(["SideSetArg","Argument should include only super- and subscripts"]);
  ...
}

(Here MML = MathJax.ElementJax.mml that can be set globally.)

I think that should give you a good start.

As for the multi scripts macro that you cite above, it is part of the mhchem package, and only handles a single pre sup, pre sub, sup, and subscript, so is not a completely general solution. The tensor package and \prescript from math tools both allow multiple scripts on either side. Doing that is a bit more complicated, and my technique from above can't really be used for that. So there is still work to be done for a complete mmultiscripts implementation.