un-styling of nested text

[davidg-sil]

If the default stylesheet marks a style as, say, italic but the user does not want to specify if the the text is italic or not (e.g. they are applying a colour or before/after marks), there is no way to tell the style sheet to negate the default without specifying /never/ italic.
E.g:

\Marker add
\Italic -

Applies the non-italic style to this marker. This is different to making it as though \Italic had not been specified by an earlier style sheet.

E.g. \qt And the Lord God said \add to me\add* <<Sit at my right hand......\qt
The \qt (by default) applies italic and the nested \add removes it, leaving to me regular.

Other than knowing enough about the internals of how styles function, of course, and adding something like this into ptxprint-mods.tex:
\expandafter\let\csname add:italic\endcsname\relax

[davidg-sil]

It is trivial to alter the definition of \defbinp@ram to accept another flag, e.g. = or 0 indicating No value here. Altering line 209 of ptx-stylesheet to

\def\defbinp@ram#1#2{\edef\t@mp{#2}\ifx\t@mp\h@phen\defp@ram{#1}{false}\else\ifx\t@mp\equ@ls\clearp@ram{#1}\else\defp@ram{#1}{true}\fi\fi}

Would match on =, for example, and delete any previously read value.

However, there are two remaining issues:

1. How to make this accessible to users. (Hand-crafting entries in a custom style sheet is probably less familiar than edtiting ptxprint-mods to most people.)
2. What is the best flag, which won't confuse people?

Note that we need to not build an expectation that people can cancel an effect from a less-specific marker definition (add+s2 overriding a value set by add), because that would not only need the 'cancel' value to be recorded and tracked, but also for the style-propagation mechanics to change so that instead of simply identifying the first value it found as it went down the style stack, it would somehow need to distinguish styling levels and do something very clever on finding a 'cancel', skipping some list elements and then restarting its search later.

[mhosken]

Thinking about this is very helpful in explaining a number of problems in the styleeditor. I think we don't need to extend the stylesheet language for this since undefined means no property is output. But I do have a question. For these overlay character styles do we need full tristate as in: undefined, true, false or can we keep it bistate: undefined, true. It would be much easier on the user if they don't have to deal with tristate controls. My thinking is that people don't want to unset a style property in a character style, while they do want, as you ask, for an undefined.

[davidg-sil]

Hmm. if someone had a large chunk of italic text, e.g. because they used that for pm/pmo, I can imagine them wanting a hard-off for it+pm.

At present we don't have a mechanism like LaTeX's \em, which toggles italic on and off, so you get the other one for embedded text, (unless there is something in the font-feature maths, I can't remember....) But if we did have it, then I would expect that most people would actually be happy with 'undefined' and 'toggle'.

[mhosken]

No toggle is not going to happen since it is both really hard to do and very confusing for users. I think we are coming around go undefined and on only. What will happen is if a style property is not defined in the base stylesheets then 'off' becomes 'undefined' (i.e. don't output the property) otherwise make it 'off'. 'on' is always 'on'.

[davidg-sil]

The problem this issue was created for was the very fact that the base style sheet defines it as on. Based on that suggested algorithm, it then becomes hard-off, which is problematic...
How about this:

• UI response to boolean:

  • Paragraph styles: on /off (+ or - in stylesheet)
  • Char styles, milestones, etc: on / no effect (+ or = in stylesheet)

• Stylesheet (TeX) accepts (for all style types):

  • + = true, - = false, = = set any values previously read to \relax

[davidg-sil]

Some examples:
Character styles that include italic in the basic usfm_sb.sty stylesheet and are most likely to be embedded in other chunks of text (or each other, even) are probably: \sls, \add, \qt. With other contenders including \litl, \tl, \sig.

E.g. some \add text included for clarity in the middle of an OT quote (marked up with \qt) (as above), or in a Byz reading (marked up with \sls), included for consistency with the preferred majority scripture.

[mhosken]

There are a number of ways of addressing this and given how often it has come up, I would prefer we not try to swing the whole UI around this situation. First you can create qt+add and turn on italic for that one context. The other alternative is that we say for markers with a value not set we allow unset to be undefined and you can create a new marker and switch out \add to the new marker using a change. But you say that is not acceptable, so I suggest we just revert to the first solution.

Turning on full tristate values in the style editor is going to confuse people so much that I don't want to go there if I can possibly avoid it.

[davidg-sil]

Yes, I agree about changing the UI to full tristate being a pain and causing too much user-pain for the benefit.
Is what you suggested above the algorithm below? (getting applied at stylesheet-reading time), or am I misinterpreting it:

1. If bold/italic/etc. is undefined, then '' / '-' -> true/false set as at present. This makes setting hard-on and hard-off in a new / complex marker possible.
2. If it is already defined as true/false, then the opposite setting makes it undefined.

[mhosken]

Close: 2. If already defined then if the setting is the same as the underlying setting, then make it undefined, otherwise set it to the value specified.