Supporting `calc(100% + 0.333em)` in the `width` attribute of `<mpadded>`

[tmke8]

<mpadded> is only really useful for horizontal spacing when it's possible to add something to the width. For example, if one sets lspace on <mpadded> but does not change the width, then the lspace eats into the width of the content.

MathML 3 had the width="+0.333em" notation for this (with + as the prefix) but this was apparently replaced by a new notation in MathML 4, see this example:

<mrow>
  <mi>x</mi>
  <mpadded lspace="0.3em" width="calc(100% +0.6em)">
    <mi>y</mi>
  </mpadded>
  <mi>z</mi>
</mrow>

The problem for MathML Core here is that the 100% here refers to the width of the child, and not to the width of the parent, as is the case in CSS.

But there is actually a solution in CSS, with the newly proposed calc-size function. The following actually works correctly in Chrome (which is the only browser currently implementing calc-size):

<math display="block">
  <mi>x</mi>
  <mpadded lspace="0.16667em" style="width: calc-size(min-content, size + 0.33333em)">
    <mi>sin</mi>
  </mpadded>
  <mi>θ</mi>
</math>

So, I think the spec should be either changed so that:

1. width on <mpadded> accepts any CSS value (currently only allows length-percentage), so that the new calc-size() can be used there
2. support for calc(100% + 0.333em) is explicitly hard-coded in the spec and is internally translated to calc-size()
3. (as David says below) width accepts calc-size() in addition to length-percentage

[davidcarlisle]

Thanks for this, It means that the example in the MathM4 draft is wrong (and I just confirmed it is as you say) so we should change that to use calc-size (probably with a warning that it's not generally implemented) I'm a bit confused as it's been in the draft for quite a while I'm sure we tested that at the time, but perhaps we were confused by the implementation status.

so the question would be whether we should allow length-percentage or calc-size everywhere that we currently allow length-percentage or just on mpadded (doing it for all length attributes seems more consistent)

support for calc(100% + 0.333em) is explicitly hard-coded in the spec and is internally translated to calc-size()

I'm not sure we need to do that, it seems just to be a spec error (in mathML4 draft) not something that needs to be supported.

[dginev]

During the MathML Core meeting yesterday this issue was discussed and I encountered the percentage widths being ignored in Chromium. I have now reread the related discussion in #183 that motivates that behavior.

Here is a codepen example of what I witnessed:
https://codepen.io/dginev/pen/yyJPmBQ

Writing that in here, since I assume an mpadded width of calc() based on a percentage value will not be well-founded until #183 is resolved in some way.

[tmke8]

Writing that in here, since I assume an mpadded width of calc() based on a percentage value will not be well-founded until #183 is resolved in some way.

Basing it on calc will indeed not work, as I show above in the first example. But, as I wrote after that, it does work when it's based on calc-size(). Your example works then:

<math>
  <mpadded style="width: calc-size(min-content, 3 * size); height: calc-size(min-content, 3 * size)">
    <mi>x</mi>
  </mpadded>
</math>

[davidcarlisle]

@tmke8 yes calc-size works but there was feeling it was too early to suggest that as a general replacement at the current time as its implementation status is uncertain. We can always come back to that,

[davidcarlisle]

@dginev I missed this during the meeting, but percentages not working for width is actually specified behaviour in core.

https://www.w3.org/TR/mathml-core/#inner-box-and-requested-parameters

... If the width attribute is present, valid and not a percentage then that attribute is used as a presentational hint setting the element's width property to the corresponding value.