How to force inline equation not to break without baseline shift?

[Pakorn]

Let’s see the example below:

#set page(height: 3cm, width: 4cm)
#show math.equation: math.display 
#set box(stroke: red)
#set text(top-edge: "bounds", bottom-edge: "bounds")

if $w x = 100$ so $x = 100/w$, in conclusion.

if $w x = 100$ so #box($x = 100/w$), in conclusion. // baseline shift! 

if $w x = 100$ so #box(baseline: 35%, $x = 100/w$), in conclusion. // manually set the baseline. 

let $2^y^y^y = 1$,  $y^y^y = log_2 (1)$ as well.

let $2^y^y^y = 1$,  #box(baseline: 30%, $y^y^y = log_2 (1)$) as well. // different value of baseline needed. 

It will be rendered like this:

What I want is a way to create display-style inline equations that does not have linebreak in between. What I have tried is using box but it causes baseline of the equation body to be different from the text. This can be alleviated by manually set the baseline shift, but it is painful because different body of the equation cause different size of the box, thus different baseline shift. How do I do?

[Y.D.X]

Unfortunately, we meet Boxing an inline equation shifts the baseline · Issue #4796 · typst/typst · GitHub here. As shown in A snippet to debug font by visualize baseline, cap-height, etc - #4 by Y.D.X, heights in math equations are tricky.

This is because a multi-line math expression, like a fraction, needs to set its baseline at the math axis height (a constant provided by the font, then scaled by the font size), but a boxed frame probably set its baseline with another value or never set it.

[Pakorn]

I tried the following snippet and it seems to work:

#let pick = [#none#label("x")]
#let bind(x) = [$#x^#pick$]
#let group(body, baseline: label("x")) = {
  let puppet = [#box($a^#[#none#label("current-line")]$ + box(body + $bind(x)$))<puppet>]
  show <puppet>: it => box(place(hide(it)))
  puppet 
  context {
    let locs = query(selector.or(<current-line>, baseline).before(here()))
    let loc-base = locs.filter(it => it.label == baseline).last()
    let loc-floor = locs.filter(it => it.label == <current-line>).last()
    let base-y = loc-base.location().position().y
    let floor-y = loc-floor.location().position().y
    box(body, baseline: 0% - base-y + floor-y,)
  }
}

So basically I stuff the boxed equation with non-boxed equation tagged with <current-line> label, and attach the normal text tagged with <x> label to the original equation body.

After that I place the equation and hide it, so that I can measure the distance between the shifted baseline, and use that information to set the box’s baseline.

Here are some tests:

#set page(height: 3cm, width: 4cm)
#show math.equation: math.display 
#set box(stroke: red)
#set text(top-edge: "bounds", bottom-edge: "bounds")

#let pick = [#none#label("x")]
#let bind(x) = [$#x^#pick$]
#let group(body, baseline: label("x")) = {
  let puppet = [#box($a^#[#none#label("current-line")]$ + box(body + $bind(x)$))<puppet>]
  show <puppet>: it => box(place(hide(it)))
  puppet 
  context {
    let locs = query(selector.or(<current-line>, baseline).before(here()))
    let loc-base = locs.filter(it => it.label == baseline).last()
    let loc-floor = locs.filter(it => it.label == <current-line>).last()
    let base-y = loc-base.location().position().y
    let floor-y = loc-floor.location().position().y
    box(body, baseline: 0% - base-y + floor-y,)
  }
}

if $w x = 100$ so $x = 100/w$, in conclusion.

if $w x = 100$ so  #group($x = 100/w$), in conclusion. // fixed


if $w x = 100$ so #box($x = 100/w$), in conclusion. // shifted

let $2^y^y^y = 1$,  #group($y^y^y^(1/y/y/y/y) = log_2 (1)$) as. // fixed

let $2^y^y^y = 1$,  #box($y^y^y = log_2 (1)$) as. // shifted

You can discuss about the limitations, potential problems from this snippet, and also simplification of the code, if you have any. Thank you

[bluss]

Very nicely done!

I did some similar but not the same hackery for equation baselines: baselinefix/main.pdf · main · bluss / typst-recipes · GitLab (code exists in same directory) Since I worked with display equations and you inline, our solutions are not interchangeable, at least I don’t think so.

The biggest thing I learned was the sensitivity to floating point accuracy, and I added rounding to avoid the issue.

You have put both the measurement points in supscripts but I think that if you place the probe point as just an appendix to the equation that its location will be on the actual baseline itself? Since you do a relative measurement it doesn’t matter.

[Pakorn]

I tried putting the probes on the subscript and with normal text. However, I found that putting on the superscript works best. I don’t know why…