mrow-like elements and space-like / embellished operator / core operator

[fred-wang]

The following elements are laid out as an mrow in MathML Core:

Unknown MathML Elements (e.g. mfenced)

maction
mphantom
mrow
mstyle
semantics
merror

math

mprescripts
none

After #139 #160 and #180, most of these elements use the same definition for space-like and embellished operators. @rwlbuis pointed out that it is not the case of merror, but that sounds a clear mistake to me.

The math element does not have any parent so the fact of being being embellished operator or space-like is not really important in general. The only difference is the case of a math with a single operator and other space-like elements, where the mrow-layout will or will not add space around the operator depending on whether math is treated as an embellished operator. If that's really important, we can add a special case in the https://mathml-refresh.github.io/mathml-core/#layout-of-mrow

The two last ones are generally empty, without padding/border/margin and in any case only used as "markers" inside mmultiscripts. So in general being embellished operator or space-like or laid out as mrow is not important at all ; this change should not be a big deal.

So my proposal is to define "mrow-like" elements to be one of the elements above i.e. the ones using the mrow layout (with perhaps additional CSS rules). Then change the definitions of space-like and embellished operators:

• https://mathml-refresh.github.io/mathml-core/#definition-of-space-like-elements
  A MathML Core element is a space-like element if is is:
  An mtext or mspace,
  Or an mrow-like element all of whose in-flow children are space-like.

• https://mathml-refresh.github.io/mathml-core/#dfn-embellished

  A MathML Core element is an embellished operator if is is:
  Or one scripted MathML element or an mfrac, whose first in-flow child exists and is an embellished operator;
  Or an mrow-like element whose in-flow children consist (in any order) of one embellished operator and zero or more space-like elements.

• https://mathml-refresh.github.io/mathml-core/#dfn-core-operator

  The core operator of an mo element; is the element itself.
  The core operator of an embellished scripted element or is the core operator of its first in-flow child.
  The core operator of an embellished mrow-like element is the core operator of its unique embellished operator in-flow child.

From the spec point of view, introducing a new concept simplifies definitions, improve readability and facilitate maintenance. Maybe the most obvious way to show the latter point is that the definition of Dictionary-based attributes is already out-of-sync: https://mathml-refresh.github.io/mathml-core/#dictionary-based-attributes (I'll open a separate issue for it)

Similarly from the implementer point of view, it's also easier to handle similar elements consistently rather than adding particular cases.

Note that I already followed that approach by introducing https://mathml-refresh.github.io/mathml-core/#dfn-scripted-mathml-elements

---

So to summarize, the proposal is:

• Introduce mrow-like element concept corresponding to all the elements currently laid out as mrow.
• Just use this in the definitions of space-like / embellished operator / core operator.
• This is changing (fixing) the case of merror.
• It does not matter for mprescripts and none.
• We can add an exception for the math root if we really want to preserve the layout of a math with a single mo and space-like children.

[NSoiffer]

The following elements are laid out as an mrow in MathML Core:
Unknown MathML Elements (e.g. mfenced)

The spec seems to be silent on unknown elements other than a list of unknown MathML elements. Does HTML dictate what should happen for other elements? E.g.,span or my-element or my-custom-element (if there is a difference in how a custom element is handled from an unknown one)? If HTML doesn't say what should happen, then the core spec should.

[fred-wang]

I think the spec does say it e.g. "Unknown MathML elements with display: math behave the same as the mrow element" https://mathml-refresh.github.io/mathml-core/#new-display-math-value and in several other places.

[NSoiffer]

In a call, @davidcarlisle pointed out that the behavior for non-MathML elements not in mtext are specified by the HTML parser to break out of MathML. However, @fred-wang also pointed out that one can manipulate the DOM via Javascript and put them in MathML, so we do need to say what should happen.

[fred-wang]

@NSoiffer I'm not sure I follow the connection with this issue?

[fred-wang]

consensus from feb 3: approved with special case for the math root

[fred-wang]

w3c/mathml-core@b29bdb2

[NSoiffer]

@fred-wang: can this be closed? If not, what's left to do?

[fred-wang]

@fred-wang: can this be closed? If not, what's left to do?

tests for space-like definition have been updated but we need to do that for embellished op too. Doing it in web-platform-tests/wpt#22381