[css-values] Fallback value of ex height

[fantasai]

In #814 (comment) @faceless2 wrote:

Another quick addendum as this doesn't warrant a post on it's own - but as I was just banging on about font metrics for 100 lines, it's relevant to note that the other significant metric extracted from the font is x-height, used for the "ex" unit and middle alignment. This is extracted from "OS2.sxHeight. If it is zero, Firefox, Chrome and Safari all set it to 0.8em - not the 0.5em currently recommended in css-values-4.

which does in fact deserve a post on its own. :)

[jfkthame]

If it is zero, Firefox, Chrome and Safari all set it to 0.8em

I'm a bit surprised by the 0.8em figure here; from what I see in the Firefox code, there are a few fallback options used on various platforms, including 0.5em (on Freetype-based platforms, Linux and Android) or 0.56*ascent (on macOS), or examining the actual ink bounds of an 'x' glyph if present. It may also use a value retrieved from OS font APIs (directwrite, core text), in which case exactly where the value comes from is opaque to us.

(A lot of this is, I think, historical baggage that could be cleaned up without too much harm.)

used for the "ex" unit and middle alignment

Perhaps worth noting that the x-height measurement is also important as an input to font-size-adjust computations. So changing how it is derived from the font (or the fallback used) could affect the used font size in cases where adjustment is in effect.

(FWIW, of the various fallbacks present in different parts of the Firefox code, I suspect the 0.56*ascent value was probably the one based most directly on observation of what worked well with actual fonts.)

[faceless2]

I'd derived that 0.8em figure a couple of years back, when I was still delving into font-metrics - I'd be a more specific about the platforms and fonts I tested with these days, not least because I now understand the variation across them! It's certainly based on the metrics tested on my Mac, and possibly not verified on other platforms.

[litherum]

WebKit does:

            Glyph xGlyph = glyphForCharacter('x');
            if (xGlyph)
                xHeight = -CGRectGetMinY(platformBoundsForGlyph(xGlyph));
            else
                xHeight = CTFontGetXHeight(m_platformData.font());

CTFontGetXHeight() uses a few different signals: OS2.sxHeight, the bounding boxes of representative characters, and additional heuristics. I don't think I can be more specific than that.

The spec should be flexible enough to:

1. Not require browsers to parse font files themselves
2. Not dictate the implementation of system-wide text libraries

[fantasai]

Fwiw, current spec text is

Equal to the used x-height of the first available font [CSS3-FONTS]. The x-height is so called because it is often equal to the height of the lowercase "x". However, an ex is defined even for fonts that do not contain an "x". The x-height of a font can be found in different ways. Some fonts contain reliable metrics for the x-height. If reliable font metrics are not available, UAs may determine the x-height from the height of a lowercase glyph. One possible heuristic is to look at how far the glyph for the lowercase "o" extends below the baseline, and subtract that value from the top of its bounding box. In the cases where it is impossible or impractical to determine the x-height, a value of 0.5em must be assumed.

[faceless2]

I've made some versions of Ahem with the "OS2.sxHeight" set to 0, and put a testcase up at https://jsbin.com/lexuren/edit?html,output

[dbaron]

Some pointers to Gecko code: the default of 0.56 and places that use it and the freetype code that doesn't use it and uses 0.5 instead.

[faceless2]

My apologies, I've sent this off in the wrong direction. On my Mac, when the OS2.sxHeight is set to 0 it appears to be falling back to some other metric in the font, which happens to be 0.8em. I haven't yet figured out which metric it is.

[Myndex]

Hi @faceless2

My apologies, I've sent this off in the wrong direction. On my Mac, when the OS2.sxHeight is set to 0 it appears to be falling back to some other metric in the font, which happens to be 0.8em. I haven't yet figured out which metric it is.

It depends entirely on the specific font. 0.8em is approximately the cap-height for some.

AND: what is needed is an actual font-x-size: property that sets the font size BASED ON the font's x-height, not the other way around. !!!

[fantasai]

@Myndex You're looking for font-size-adjust. Go look it up.

Anyway, I'm closing out this issue since afaict, the spec doesn't need any changes.

[Myndex]

@Myndex You're looking for font-size-adjust. Go look it up.

Anyway, I'm closing out this issue since afaict, the spec doesn't need any changes.

I am quite familiar with font-size-adjust, it does not do nor behave in the way needed for directly setting font size.