ConTeXt+LMTX双行夹注的简单实现

欢迎加入本人建的"LuaTeX ConTeXt 学习互助"群：431714622，互助学习LuaTeX ConTeXt LaTeX相关技术，以及通过Lua、Python实现格式化文本、数据的自动排版。

这里的更新会比较慢，更新的详情请看项目的GitHub代码库

适用于ConTeXt和LuaMetaTeX(LMTX， LuaTeX的后继者，ConTeXt当前的实际引擎)的夹注模块，可以配合竖排功能和标点压缩功能使用。调整后应该可用于LuaTeX。

也是本人学习LuaMetaTeX、LuaTeX的练习项目，主要涉及：段落与行的结点列表的操控；使用tex.linebreak()函数等干预断行；使用回调函数。会保留较多学习性的注释和代码。

效果#

竖开明大学.jpg

横全角庄子.jpg

术语#

术语参考CTeX-org：

jiazhu(in Hanyu Pinyin): splitted annotation or inline cutting note, 夹注/双行夹注 in simplified Chinese, 割注/warichū in Japanese

大致思路#

在文档中段落插入夹注hbox，并给盒子加一个属性作为标记；
在"processors"类的"after"小类回调中处理：
1. 用一个一字宽rule结点代替段落中的夹注盒子，并在一个表中收集夹注盒子，循环：
  1. 试排段落，找到分行后的rule，获取从它到行末的宽度；
  2. 根据这个宽度和盒子的自然宽度预计、试排盒子的节点列表，循环：
    1. 预估小于或等于两行时，按盒子宽度的一半试排为两行（可略宽容），删除rule；
    2. 如果预估大于两行，则按空间宽度试排，取前两行；
    3. vpack插入rule之前，其后加罚点0，（当基本填满行末空间时罚点-10000，即强制断行，放置再次插入夹注）；
用新的节点列表代替原段落的节点列表（由系统断行）。

bug & TODO#

关于断行、分段的备用资料#

cjk排版原型#

http://wiki.luatex.org/index.php/Japanese_and_more_generally_CJK_typesetting

cjk断行相关的系统文件：#

语言脚本：D:\venvs\context-win64\tex\texmf-context\tex\context\base\mkiv\scrp-cjk.lua
行to段（tex.linebreak）：D:\venvs\context-win64\tex\texmf-context\tex\context\base\mkiv\node-ltp.lua
用于字间加空、加罚点的hanzi模块char-scr.lua scrp-cjk.lua

type-imp-mscore.mkiv

tex.linebreak的工作基于当前状态#

https://www.mail-archive.com/dev-luatex@ntg.nl/msg01805.html

It is basing itself on the current TeX state at that point in time (at least, that is what it should do). You can look up tex_run_linebreak() in ltexlib.c, it should be easy to follow.

也参见node-ltp.lua和 https://github.com/michal-h21/linebreaker

Hans论干预断行的思路#

https://www.mail-archive.com/luatex@tug.org/msg04329.html

By the way, Hans, I heard you already reimplemented the TeX linebreak algorithm in pure Lua, can your code be found somewhere?

on my machine ... i'm waiting for the new hz in the backend before putting it in the context distribution because i don't want to end up with several versions (and i then need to prune some test code) ... so some patience is needed ...

btw there was an article on it some time ago, also in: http://www.pragma-ade.com/general/manuals/hybrid.pdf, p 101

one thing you could play with is storing the paragraph before it gets broken, then let tex do the job, analyze afterwards, and when not ok, add some penalties and let tex do the job again

but it might be better to rethink this special case (which is unlikely to mix with all those things that can end up in paragraphs, i.e. it's a limited tex-only case), and cook up something dedicated, at least that's the approach i'd choose as soon as context users start asking for such features

Hans

另一种思路：在两个回调中处理#

https://www.mail-archive.com/luatex@tug.org/msg05587.html

I did set up pre_linebreak_filter to find out the cases where line breaking happens on the main vertical list (in contrast to being done in a box, say). Then inside pre_linebreak_filter I run

tex.linebreak

several times with different looseness (you cann see all my reports coming together :-) ) and record the different results

Then in post_linebreak_filter I replace the generated paragraph node list with something special (basically to drop it) and to be able to find out about the para on the contribution list (which I look at in buildpage_filter). Could probably be done differently but that was the simplest way to get this working for me.

关于parbuilder (and related hpacking)#

http://www.pragma-ade.com/general/manuals/hybrid.pdf, p95

in September 2008, when we were exploring solutions for Arabic par building, Taco converted the parbuilder into Lua code and stripped away all code related to hyphenation, protrusion, expansion, last line fitting, and some more.

末行长度与段落填充控制#

https://tex.stackexchange.com/questions/63762/minimum-length-of-last-line-of-a-paragraph

结点操控参考实例#

https://github.com/gucci-on-fleek/lua-widow-control