Discretionaries in new liners

[Omikhleia]

We went merging #1977 a bit too soon.

Edge case:

Discretionaries still need more refactoring so as to avoid such cases (and ugly guesses everywhere they are used)

[alerque]

It's okay, this isn't a regression since multi-line effects like this were not even possible before. I'm okay with iterating on the develop branch as long as we're sure we're moving the bar forward, and we're not locked into supporting any given API shape until we release.

Do you have sources for this test case to add (even without proper expectations)?

[Omikhleia]

Do you have sources for this test case to add (even without proper expectations)?

Currently using a minimized "parbox" from parbox.sile after I had a sudden doubt, so I still have to craft a test that doesn't need such an overkill.

[Omikhleia]

A possible MWE

\begin[papersize=a7, class=resilient.book]{document}
\lua{
local class = SILE.documentState.documentClass
class:registerCommand("xredacted", function (options, content)
  local bs = SILE.measurement("0.9bs"):tonumber()
  local bsratio = 0.3 -- hard-coded here, should use font metrics
  SILE.typesetter:liner("xredacted", content,
    function (box, typesetter, line)
      local outputWidth = SU.rationWidth(box.width, box.width, line.ratio)
      local H = SU.max(box.height:tonumber(), (1 - bsratio) * bs)
      local D = SU.max(box.depth:tonumber(), bsratio * bs)
      local X = typesetter.frame.state.cursorX
      SILE.outputter:pushColor(SILE.color("gray"))
      SILE.outputter:drawRule(X, typesetter.frame.state.cursorY - H, outputWidth, H + D)
      SILE.outputter:popColor()
      -- MVE: Use the box width to advance position, rather
      -- than rely on internal content to move it.
      typesetter.frame:advanceWritingDirection(outputWidth)
    end
  )
end)
}
\language[main=pl]
% In liner
\xredacted{logi-zoty logi-zoty logi-zoty logi-zoty logi-zoty logi-zoty
logi-zoty logi-zoty logi-zoty}

% Vs. normal
logi-zoty logi-zoty logi-zoty logi-zoty logi-zoty logi-zoty
logi-zoty logi-zoty logi-zoty

\end{document}

Current = wrong, see blue lines

Fixed as per #1985 = correct

Analysis

With underline/striketrough etc., we don't use the box width, but rely on the inner content being output to move the frame cursor.

Here we used the target outputWidth to move the cursor (ignoring content): it proves the box's width is incorrect, due to not taking into account (on line 2) the discretionary. EDIT: on line 1, it's not a discretionary, it's the dash; but one line 2, the computation misses both the initial postbreak (repeated hyphen in Polish) and the final prebreak.

[Omikhleia]

(Sorry if that MWE is non-trivial + uses resilient, that was one quick way to get sile-x shim and test with 0.14.16)

[alerque]

Thanks for the MWE. I simplified it a bit more, stripped out some of the extra noise, and made it work (er, fail) with the plain class, then verified the PR made it work.