csswg-drafts
csswg-drafts copied to clipboard
w3c
[css-text] [css-values] Circular dependency between letter-spacing and x-height
https://drafts.csswg.org/css-text-3/#letter-spacing-property
When the effective spacing between two characters is not zero (due to either justification or a non-zero value of letter-spacing), user agents should not apply optional ligatures
letter-spacing takes a <length>, which means that it accepts font-relative units, like 2ex for example. Some (all?) browsers can measure the ex length by actually rendering a literal "x" character and inspecting its height (at least in the case where the metrics inside the font appear to be bogus). However, the liga font feature can change which glyph the "x" character renders with, so if we're supposed to not apply optional ligatures, the height of the "x" character can change depending on whether we're disabling ligatures or not. This means that, if someone applies letter-spacing on an element with ex units, we may have to measure a glyph, but in order to know which glyph to measure, we have to know whether that glyph has a height of 0 or not.
It's a weird fiddly corner case that probably wouldn't ever make a difference in real content, but it's probably worth at least considering how to fix it.
Note: Allowing letter-spacing to accept font-relative units is useful. Also, disabling ligatures when letter-spacing is nonzero is also required for good typography. Whatever solution we pick shouldn't break either of these things.
The best solution is probably "don't run shaping when calculating the value of font-relative units."
... and the ch and ic units should be calculated without letter-spacing applied (which I think is already the case).
The best solution is probably "don't run shaping when calculating the value of font-relative units."
Definitely. I would think "liga" is the least of your problems - salt, swsh, smcp and others would have a much bigger impact on units if they were applied, and some of the substitutions could be language dependent too. The metrics really have to be taken from the original glyph as specified by the glyf or CFF tables.
The spec doesn't define exactly how to find the value of an 'ex' from the font, leaves that up to the UA with some pointers. So I don't think there's anything to fix here wrt the specs? https://www.w3.org/TR/css-values-3/#ex
Right, we could arguably be more precise, but given the current lack of precision, no circular dependency seem mandated.
@litherum Shall I close this no change, or is there some change to the spec you want to see here?
@litherum @kojiishi Added
The [=font-relative lengths=] are calculated in the absence of shaping.
Let me know if you would like some different phrasing.