39411 - Default style rules for <mtable> need tuning

Status: RESOLVED FIXED

(Core :: MathML, defect, P3)

Description

[rbs]

While full and proper support of <mtable> is not yet there, mathml.css
needs appropriate style rules for <mtable> and its related elements
so that <mtable> can yield interesting layout in common situations.
[http://lxr.mozilla.org/seamonkey/source/layout/mathml/tests/mathml.css]

One particular problem with the current rules is that they leave
a spacing around the table. Hence if for example someone does
v = ( <mtable>column_vector_here</mtable> ), there is an ugly
gap between the stretchy parenthesis and the mtable content... Of interest
are also rules to map things like <mtable rowlines="solid">, <mtable
columnlines="solid"> (as per the description given the URL).

With MathML-enabled binaries now available, hope you now have a setup
to try things out :-)

Comment 1

[Hixie (not reading bugmail)]

Accepting for now, but I am currently overwhelmed by Layout bugs so moving 
to M20 for now... :-/

Comment 2

[rbs]

Not sure if the collapsing table code has changed as of late.
I have seeing some ugly layout of <mtable> due to this bug...
Any chance you can have a look at it these days?

Comment 3

[rbs]

Here is what is needed by default (all horizontal units are em, all
vertical unitis are ex)

   0.4em                                                0.4em
   /--/-------------------------------------------------/--/
   |                 0.8em            0.8em                | 0.5ex
   +  +---------------/-/--------------/-/--------------+  /
   |  |               | |              | |              |  |
   |  |               | |              | |              |  |
   |  |  content-box  | |              | |              |  |
   |  |               | |              | |              |  |
   |  |               | |              | |              |  |
   +  +---------------+-+--------------+-/--------------+  +
   |  |               | |              | | 1ex          |  |
   +  +---------------+-+--------------+-/--------------+  +
   |  |               | |              | |              |  |
   |  |               | |              | |              |  |
   |  |               | |              | |              |  |
   |  |               | |              | |              |  |
   |  |               | |              | |              |  |
   +  +---------------+-+--------------+-/--------------+  +
   |  |               | |              | | 1ex          |  |
   +  +---------------+-+--------------+-/--------------+  +
   |  |               | |              | |              |  |
   |  |               | |              | |              |  |
   |  |               | |              | |              |  |
   |  |               | |              | |              |  |
   |  |               | |              | |              |  |
   +  +---------------+-+--------------+-+--------------+  /
   |                                                       | 0.5ex
   +-------------------------------------------------------/

Comment 4

[Hixie (not reading bugmail)]

The collapsing borders code is in a rather ugly state right now.

I assume in your diagram that the gaps are gaps, not borders, right?

I'll try to look at this soon, but no promises...

Comment 5

[rbs]

Yep, the outlines are for illustration only. They show the edges on the
content boxes. There are no borders.

Comment 6

[Hixie (not reading bugmail)]

I actually had a look at this, finally. However, MathML tables don't seem to
be displaying right at all. See, for example, tests 23 and 24 of the torture
test on mozilla.org: 
   http://www.mozilla.org/projects/mathml/demo/texvsmml.xml

A small version is available here:
   http://www.hixie.ch/tests/adhoc/mathml/001.xml

I have to fiddle with the font size to get them to either appear at all or to
appear as a grid, respectively.

So there are two ways I can see to do this: either fiddle with :first-child
and :last-child rules and use the separate border model and margins and padding
on the cells, which will look bad if anyone uses borders, or use transparent
collapsed borders, which will also look weird if people use borders, although
less so. The second option is blocked on bug 41262.

Which would you prefer?


Reading further through the rules in the MathML spec, I'm having some difficulty
trying to come up with CSS rules that fully implement what they ask for. As I 
understand it, they say each element inside an <mtr> is automatically imagined
to be inside an <mtd>, but this is "hard" in CSS (i.e., impossible). Also, some
of the attributes (e.g. "framespacing") seem to me to require some C++ love.

Also, "rowlines" and "columnlines" (and friends), while possible to do in CSS,
kind of clash with your request for spacing... I'm not sure how to handle those.
Just use the same width as your spacing, but for the lines?

Comment 7

[rbs]

<mtable> is simply mapping to <table>, but at each milestone something seems to
go wrong somehere when used in the mathml context. Because it is a direct
mapping, if you can customize the rules for a <my:table> to display as required,
then that will still be fine in the long run when they are passed to <mtable>.

Actually, I don't have any preferences regarding how your do it. Any that
works in an efficient way around the existing table bugs will do.

The MathML 2.0 spec has now removed these optional 'inferred' <mtr> and <mtd>.
So you can safely assume that the table is well-formed in the strict XML sense.

Comment 8

[rbs]

Also don't worry about attributes that need some C++ support. Just setting rules
that emulate their default values is fine. Then, these could be overriden later
in the C++ side.

Comment 9

[Hixie (not reading bugmail)]

Excellent news about the inferred stuff.

Comment 10

[Hixie (not reading bugmail)]

I had a go at implementing this, but <mtable> elements are not matching the
mtable rule in mathml.css. They _are_ matching the :-moz-math-table table.

Is this expected? Should I simply style the pseudo-element?

Comment 11

[rbs]

Are things okay when you combine the two:

:-moz-math-table, mtable {
  ...
}


[I might have to revisit the frame constrution code to eliminate one or the
other]

Comment 12

[Hixie (not reading bugmail)]

That should work (merely because the <table> rule is ignored).

Comment 13

[rbs]

Taking the bug & targeting m0.9.9. I have iterated on the code to cover most of
what is needed in common situations.

Comment 14

[rbs]

Improvements checked in.