[PATCH v2] Add new entity \-- serving as markup separator/escape symbol

[Ihor Radchenko]

Max Nikulin <manikulin@gmail.com> writes:

>>> The good point in your patch is that \- is still work as shy hyphen
>>> (that, by the way, may be used in some cases instead of zero width
>>> space: *intra*\-word). On the other hand I have managed to find a case
>>> when your approach is not ideal:
>>>
>>> *\--scratch\--*
>>>
>>> <p>
>>> <b>&#x00ad;-scratch</b></p>
>> 
>> Well. I think that it is impossible to use the same escape construct to
>> both force emphasis and escape it.
>
> Let's articulate the problem as follows: when some characters ("*". "/". 
> etc.) besides used literally are overloaded with 2 additional roles that 
> are start emphasis group and terminate emphasis group, in addition to 
> lightweight markup heuristics, it is necessary to provide a way to 
> disambiguate which of 3 roles is associated with particular character.
>
> "Activate" and "deactivate" characters or entities for emphasis markers 
> are alternative and perhaps not so clear terms have used before.
>
> The advantage of zero width space is that "[:space:]" is part of 
> PREMATCH and POSTMATCH (outer) regexps in 
> `org-emphasis-regexp-components' and "[:space:]" is forbidden at the 
> inner borders of emphasized span of text. The latter is mostly 
> meaningful, however I am unsure if bold space has the same width as 
> regular one, and space in fixed width font is certainly distinct.
>
> The problem with the "\--" entity is that it is not handled properly at 
> the start of emphasis region. It neither disables emphasis nor parsed as 
> complete entity, instead it becomes combination of "\-" shy hyphen and 
> literal "-".
>
> Unsure if it can be solved consistently. Possible ways:
> - It addition to space-like (in respect to current regexp) entity add 
> another one that acts as a part of word, but like "\--" stripped from 
> output. Likely it should be accompanied by more changes in the parser 
> and regexps.
> - Provide some new explicit syntax for literal character, start of 
> emphasis group, end of emphasis group.

The fact that \-- was not parsed in your example is because entities
cannot be directly followed by a letter (see 12.4 Special Symbols).

You need

*\--{}scratch\--*

Concerning the 3 listed roles of the *_/+ markup, I propose to simplify
the problem a bit and not try to make \-- serve as a proper escape symbol.
Instead, we can document the already existing quoting entities:

 ("slash" "/" nil "/" "/" "/" "/")
 ("plus" "+" nil "+" "+" "+" "+")
 ("under" "\\_" nil "_" "_" "_" "_")
 ("equal" "=" nil "=" "=" "=" "=")
 ("star" "\\star" t "*" "*" "*" "⋆")

Then, your example should better be written as

\star{}scratch\star

\-- may better work between markup, not inside.

> Concerning zero width space workaround, I may be wrong, but Nicolas 
> might consider using U+200B zero width space as the escape character for 
> itself: single one is filtered out during export, double zero width 
> space becomes single character. (I do not like this kind of "white 
> space" programming language".)

This is too complex, IMHO.
If desired, we can again go the entity road and introduce
\zws entity.

Note that we already have

 ("nbsp" "~" nil "&nbsp;" " " " " " ")
 ("ensp" "\\hspace*{.5em}" nil "&ensp;" " " " " " ")
 ("emsp" "\\hspace*{1em}" nil "&emsp;" " " " " " ")
 ("thinsp" "\\hspace*{.2em}" nil "&thinsp;" " " " " " ")

Generally, it is a good idea to advertise entities in the manual.
Zero-width space is not only limited, it is impossible to use, e.g. in
tables when you want to quote "|". The only solution is using \vert or
\vbar entity.

> Another question is whether U+2060 word 
> joiner (or some other character) should be added either as alternative 
> to zero width space or to allow =    verbatim    = fixed width text 
> surrounded by fixed width spaces.

This particular example is tricky.
If we put escape symbol _inside_ the verbatim, it is never possible to
know if the user intents to use that symbol literally or not.
But non-space before/after opening/closing markup char is hard-coded and
changing it is fragile.

Instead of using some kind of "escape" symbol here, I suggest turning to
the idea about inline special blocks. We can introduce a more verbose
markup that will allow spaces inside at the beginning/end of the
contents.

https://orgmode.org/list/87a6b8pbhg.fsf@posteo.net
Manuel Macías [ML:Org mode] (2022) About 'inline special blocks'

Instead of using the tricky *bold text*, we may allow _*{bold text}*_ or
something similar, with _name{...}name_ being inline special block.

Best,
Ihor