Re: lineHeight

Thanks for asking, as this is a complex area of functionality when having
to wade through the more general models of both XSL-FO and CSS. See inline
below for clarifications and answers, and at the end some additional useful
information.

On Thu, Aug 22, 2013 at 4:55 PM, Michael Dolan <mdolan@newtbt.com> wrote:

> Glenn and all-****
>
> ** **
>
> A cursory reading of TTML suggests lineHeight=fontSize.  But digging
> deeper into XSL, one might reach different conclusions in red (also marked
> “CONCLUSION”).****
>
> ** **
>
> [TTML 8.2.12]:****
>
>  tts:lineHeight****
>
> If the value of this attribute is normal, then the initial value of the
> style property must be considered to be the same as the largest font sizethat applies to any descendant element.
> ****
>
> ….****
>
> The semantics of the style property represented by this attribute are
> based upon that defined by [XSL 1.1], § 7.16.4.
>

First, we should quote the current TTML1 ED text, which was updated on July
12th [1] and 16th [2] to read as follows:

"If the value of this attribute is normal, then the computed value of the
style property must be considered to be the same as the largest font size
that applies to any descendant element in the intermediate synchronic
document as determined by *9.3.2 Intermediate Synchronic Document
Construction*<https://dvcs.w3.org/hg/ttml/raw-file/default/ttml1/spec/ttml1.html#semantics-region-layout-step-1>
."

[1] https://dvcs.w3.org/hg/ttml/rev/00a9b799895d
[2] https://dvcs.w3.org/hg/ttml/rev/0355e9473d7e

This text, about the "largest font size", is intended to correspond to the
following language in:

XSL-FO [3]:

"When an element contains text that is rendered in more than one font, user
agents should determine the "line-height" value according to the largest
font size."

[3] http://www.w3.org/TR/2006/REC-xsl11-20061205/#line-height

and CSS2.1 10.8.1 [4]:

"When an element contains text that is rendered in more than one font, user
agents may determine the 'normal'
'line-height'<http://www.w3.org/TR/CSS2/visudet.html#propdef-line-height>
value
according to the largest font size."

[4] http://www.w3.org/TR/CSS2/visudet.html#propdef-line-height

Note that these two citations are identical except that CSS2.1 added
'normal', while XSL-FO, which quotes CSS2, didn't make it explicit that
this sentence was to be scoped by interpretation of what 'normal' means.

Note also the two changes [1][2] we made in TTML1 since PER:

   - correct "initial value" to read "computed value";
   - clarify that "descendant element" refers to descendants in the
   intermediate synchronic document, and not the source document;

In other words, this changed language is about determining what the
computed value of 'normal' means, and which descendants apply in making
that determination. For example, one may have read the prior language as
including the font size of descendant elements that are not temporally
active or are selected into a different region, both of which should not
apply when resolving the computed value for 'normal'.

> ****
>
> [XSL 7.16.4]: ****
>
> Normal: Tells user agents to set the computed value to a "reasonable"
> value based on the font size of the element. The value has the same meaning
> as <number>. We recommend a computed value for "normal" between 1.0 to 1.2.
> ****
>
> <length>: The box height is set to this length.****
>
> ** **
>
> CONCLUSION: Therefore as TTML states that a value of "normal" "must be
> considered to be the same as the largest font size", I assume that the XSL
> 7.16.4 definition that is in effect is "<length>" with the value of the
> largest font size.
>
Correct. For example, if one has:

<p id="p1" tts:fontSize="12pt" tts:lineHeight="normal">
<span tts:fontSize="10pt">FOO</span>
<span tts:fontSize="12pt">BAR</span>
<span tts:fontSize="14pt">BAZ</span>
</p>

Then the resolved computed value of tts:lineHeight on the paragraph p1 is
"14pt". In other words, we could have specified the equivalent with:

<p id="p1" tts:fontSize="12pt" tts:lineHeight="14pt">...</p>

On the other hand, if this content mapped were selected into two regions,
such as:

<p tts:fontSize="12pt" tts:lineHeight="normal">
<span region="r1" tts:fontSize="10pt">FOO</span>
<span region="r1" tts:fontSize="12pt">BAR</span>
<span region="r2" tts:fontSize="14pt">BAZ</span>
</p>

Then the computed line height that applies to p1 in region r1 would be
12pt, but in region r2 would be 14pt.

> ****
>
> ** **
>
> [TTML 8.2.12] continued:****
>
> Furthermore, it is the intention of this specification that the allocation
> rectangle of a line be consistent with the per-inline-height-rectangle as
> defined by [XSL 1.1], § 4.5, i.e., that a CSS-style line box stacking
> strategy be used.
>
Here's where things get complex to interpret (particularly when trying to
decode the language in XSL-FO). However, before we do that, let's take note
of some simplifying factors that hold in TTML than apply to XSL-FO and CSS:

In particular, in TTML1 (though perhaps not TTML2 when we introduce images):

   - tts:lineHeight only applies to the paragraph element p, and does not
   apply to span or br;
   - all inline children of p are non-replaceable (text) elements; i.e.,
   there are no replaceable elements like <img/> [if some profile introduces
   replaceable elements in a superset extension, then this simplifying
   assumption may not hold];

In general, in both XSL-FO and CSS, these assumptions don't hold, so the
descriptive text in XSL-FO is more complex to interpret.

Before I attempt to explain the XSL-FO text, I will summarize the three
line stacking strategies defined there:

*font-height*

Produces constant distance between the baselines of adjacent line areas.
The height of a line area's allocation rectangle is the height of the *
nominal-requested-line-rectangle*, which is the sum of the altitude and the
depth, i.e., font size, of the font that applies to the containing block
element (ignoring altitude and depth of the fonts that apply to the line
area's child areas).

Line areas of a block are stacked (in the block progression dimension) so
that the distance between baselines are constant and equal to the computed
value of the line height property of the containing block element. This is
accomplished by (1) computing the half-leading trait of a block area as
half the difference between the computed value of the line-height property
and the sum of the computed values of the text-altitude and text-depth
properties, then (2) assigning this half-leading to the space-before and
space-after traits of each line area generated by the block.

Note that half-leading is negative if line-height is less than altitude
plus depth (font size), in which case line area rectangles will overlap.

*max-height*

Produces constant distance between the adjacent edges of consecutive line
areas, i.e., between the after edge of line N and the before edge of line
N+1. The height of a line area's allocation rectangle is the height of the *
maximum-line-rectangle*, which is the greater of (a) the height of the
nominal-requested-line-rectangle described above and (b) the sum of the
maximum altitude and maximum depth of the *allocation rectangles* of the
line area's child inline areas.

If no child inline area has an altitude (distance above baseline) greater
than the altitude of the containing block's font and no child inline area
has a depth (distance below baseline) greater than the depth of the
containing block's font, then the result will be identical to the *
font-height* strategy.

Note that the *allocation rectangle* of an inline area generated by an
element comes in two flavors [5]:

   - normal-allocation-rectangle
   - large-allocation-rectangle

In general, non-replaceable elements, e.g., text, generate normal
allocation rectangles, while replaceable elements, e.g., images, generate
large-allocation-rectangles. The former "normal" variety *does not* include
padding and border when computing its height, while the latter "large"
variety *does* include padding and border.

[5] http://www.w3.org/TR/2006/REC-xsl11-20061205/#area-geo

Note further, and importantly (for what follows), that this strategy does
not take into account (a) a different line height property of any
non-replaceable child element and (b) the space (margin) before (top) and
after (bottom) properties of any replaceable child element.

Half-leading computation and usage follows the same rules described above
for the font-height strategy, meaning that half-leading may be negative,
and thus cause line area rectangle overlap.

*line-height*

Effectively the same as the max-height strategy, except that (a)
half-leading is computed on a per-line basis, (b) the half-leading of
non-replaceable child inline elements is applied (rather than ignored), and
(c) the margin (space) in the block progression dimension of replaceable
child inline elements is applied (rather than ignored).

More specifically, the height of a line area's allocation rectangle is the
height of the *per-inline-height-rectangle*, which is the greater of (a)
the height of the expanded-nominal-requested-line-rectangle (see below) and
(b) the sum of the maximum altitude and maximum depth of
the expanded-rectangle (see below) of each of the line area's child inline
areas.

The *expanded-nominal-requested-line-rectangle* is the same as the
nominal-requested-line-rectangle described above except that it is *expanded
* in the before and after directions by an amount equal to the half leading
of the containing block area. [If this half leading is negative, it will be
contracted rather than expanded.]

The *expanded-rectangle* of an inline area child of a line area is the same
as its allocation-rectangle described (under max-height strategy) above
except that it is *expanded* in the before and after directions by either
(1) an amount equal to the half leading of the inline area, if generated by
a non-replaceable element, or (2) amounts equal to the margin-{top,bottom}
(or space-{before,after}), respectively, of the inline area, if generated
by a replaceable element.

Because the half-leading of the containing block is already taken into
account by the expanded-nominal-requested-line-rectangle, the space-before
and space-after traits of each line area are set to zero.

This strategy corresponds to the line stacking strategy adopted by CSS, and
is adopted by TTML, Section 9.3.3, step (5), when mapping intermediate
synchronic documents to XSL-FO [6].

[6]
https://dvcs.w3.org/hg/ttml/raw-file/default/ttml1/spec/ttml1.html#semantics-region-layout-step-2

As mentioned above, however, TTML1 does not presently apply line-height to
inline elements (span or br), and thus the half-leading of each inline
element is zero, and does not support a replaceable element type. Consequently,
the line-height strategy, when applied to TTML1, produces equivalent
results as produced by the max-height strategy. That is, each
per-inline-height-rectangle is the same height as the equivalent
maximum-line-rectangle to which half-leading (of the block) has been added
(instead of applying this half-leading to the space-{before,after} of the
line area).

So, with TTML1, this produces line areas that abut one another in the block
progression dimension, but whose height includes half leading both before
and after, as contrasted with max-height, where line areas don't abut but
rather use space before and after on line areas to represent half leading.

> ****
>
> ** **
>
> [XSL 4.5]****
>
> The per-inline-height-rectangle for a line-area is the rectangle whose
> start-edge and end-edge are parallel to and coincident with the start-edge
> and end-edge of the nominal-requested-line-rectangle, and whose extent in
> the block-progression-dimension is determined as follows.****
>
> ….****
>
> The extent of the per-inline-height-rectangle in the
> block-progression-direction is then defined to be the minimum required to
> enclose both the expanded-nominal-requested-line-rectangle and the
> expanded-rectangles of all the inline-areas stacked within the line-area;
> this may vary depending on the descendants of the line-area.****
>
> ….****
>
> The expanded-rectangle of an inline-area is the rectangle with start-edge
> and end-edge coincident with those of its allocation-rectangle, and whose
> before-edge and after-edge are outside those of its allocation-rectangle
> by a distance equal to either (a.) the half-leading, when the area's
> allocation-rectangle is specified to be the normal-allocation-rectangle by
> the description of the generating formatting object , or (b.) the
> space-before and space after (respectively), when the area's
> allocation-rectangle is specified to be the large-allocation-rectangle. **
> **
>
> …..****
>
> The expanded-nominal-requested-line-rectangle is the rectangle with
> start-edge and end-edge coincident with those of the
> nominal-requested-line-rectangle, and whose before-edge and after-edge are
> outside those of the nominal-requested-line-rectangle by a distance equal
> to the half-leading.****
>
> ……****
>
> The nominal-requested-line-rectangle for a line-area is the rectangle
> whose start-edge is parallel to the start-edge of the content-rectangle of
> the nearest ancestor reference-area and offset from it by the sum of the
> start-indent and the start-intrusion-adjustment of the line area, whose
> end-edge is parallel to the end-edge of the content-rectangle of the
> nearest ancestor reference-area and offset from it by the sum of the
> end-indent and the end-intrusion-adjustment of the line area, whose
> before-edge is separated from the baseline-start-point by the
> text-altitude of the parent block-area, and whose after-edge is separated
> from the baseline-start-point by the text-depth of the parent block-area.
> It has the same block-progression-dimension for each line-area child of a
> block-area.****
>
> …****
>
> [XSL 6.5.2]****
>
> The .minimum, .optimum, and .maximum components of the half-leading trait
> are set to 1/2 the difference of the computed value of the line-height property
> and the computed value of the sum of the text-altitude and text-depth
> properties. The .precedence and .conditionality components are copied from
> the line-height property.****
>
> ** **
>
> CONCLUSION: Therefore lineHeight for each line of text in a TTML document
> is given by: largest fontSize on the line + half leading == 1.5 * largest
> fontSize
>
First, we must distinguish between the height of a line area and the
computed value of the lineHeight property on an element. Here, you use the
term "lineHeight", so I'm not sure to which of these you are referring.

As I indicated above, the computed value of the 'normal' lineHeight
property value is derived from the maximum font size over the entire set of
descendants of a paragraph element (in a particular intermediate synchronic
document). Or, if the lineHeight property is specified as a length, then
that is resolved as usual (without reference to any descendant element's
font size).

Once the computed lineHeight value is resolved, now you can compute half
leading and then derive individual line area rectangle heights according to
the line stacking strategy rules.

So, if I were to state the correct conclusion, it would read:

*CONCLUSION: In TTML1, the height of a line area rectangle generated by a
paragraph (<p/>) element is the greater of (a) the computed value of the
lineHeight property that applies to the paragraph and (b) the sum of the
maximum altitude and maximum depth of each font that applies to the line
area's child inline areas.*

For those that wish to delve even deeper, I'd suggest reading:

[7] https://lists.w3.org/Archives/Member/w3c-xsl-fo-sg/2000Nov/0027.html(Member
Only)
[8] http://meyerweb.com/eric/css/inline-format.html

Regarding [7], I'll check with the author to see if he minds me resending
his message to this public list.

G.