\noindent The code in this project was written to illustrate a blog post and is not a full implementation of a paragraph ``parser''. It ignores a number of Lua\TeX's node types that might be encountered during parsing of the paragraph because it is designed purely to illustrate concepts discussed in the blog post which accompanies this project. Note too that the parser code is not recursive: for example, if you detect an {\tt hlist} or {\tt vlist} during parsing then you would need to use recursive behavior to descend into those boxes. This is not difficult to implement but, again, for simplicity it is not done here.
\vskip10mm
\let\lua\directlua % Don't change this.
\catcode`\~=12
{\noindent \bf Here is a typeset paragraph\vskip3mm }
\noindent For example, when \TeX{} typesets a paragraph of text and breaks it into a series of lines, it considers the paragraph's text as a sequence of boxes and uses the dimensions of those character boxes to find the best linebreaks. Each line of the paragraph is itself a box (containing other boxes---e.g., characters) and the typeset paragraph lines (boxes) are stacked (vertically) to produce the paragraph. Eventually, the largest box of all is produced: the typeset page. Clearly, this is a very simplified picture because you also need the ability to arbitrarily position those boxes and \TeX{} does this using so-called glue. Knuth commented (page 70 of The \TeX book) that ``glue'' probably should have been referred to as ``spring'' but the term glue was adopted early on and, to use Knuth's pun, it stuck.
\vskip10mm
{\noindent \bf Here is the same paragraph processed by a Lua\TeX{} callback \vskip3mm}
\lua{%
local HLIST = node.id("hlist")
local GLUE = node.id("glue")
local GLYPH = node.id("glyph")
local WHAT = node.id("whatsit")
function lineparse(line)
local box_order=line.glue_order
local box_ratio=line.glue_set
local box_sign=line.glue_sign %0 = normal, 1 = stretching, 2 = shrinking
% This loop uses a function to look for all lines of the typeset paragraph
% This function scans all the items in a typeset line of the paragraph
for n in node.traverse(line.head) do
% We have seen some glue: calculate its size, draw a box
if n.id == GLUE then
% glue parameters
local gwidth=n.width %horizontal or vertical displacement
local gstr=n.stretch %extra (+ve) displacement or stretch amount
local gstrord=n.stretch_order %factor applied to stretch amount
local gshr=n.shrink %extra (-ve) displacement or shrink amount
local gshrord=n.shrink_order %factor applied to shrink amount
local width=0
% This section is how TeX calculates glue settings
if box_sign==0 then %(normal)
width = gwidth
elseif box_sign==1 then %(stretch)
if gstrord == box_order then
width = gwidth + (box_ratio * gstr)
else
width = gwidth
end
elseif box_sign == 2 then %(shrink)
if gshrord == box_order then
width = gwidth - (box_ratio * gshr)
else
width = gwidth
end
end %if box_sign
% now calculate the glue's width in bp + draw a box.
if width > 0 then % don't bother with zero width glues
local pdfn=node.new(WHAT, "pdf_literal")
local wp = (width/65536)*(72/72.27)
pdfn.data=" 0.2 w 0 0 "..wp.." 5 re q 0.9 0.9 0.0 rg f Q "
pdfn.mode=0
% insert into the typeset paragraph
local np = n.prev
np.next=pdfn
pdfn.prev=np
pdfn.next=n
n.prev=pdfn
end
end % if n.id == GLUE
% Within the typeset line we've seen another hbox
if n.id == HLIST then
local nw = n.width
local nd= n.depth
local nh = n.height
local ns = n.shift
local hw = ((nw/65536)*(72/72.27))
local hd = ((nd/65536)*(72/72.27))
local hh = ((nh/65536)*(72/72.27))
local hs= ((ns/65536)*(72/72.27))
local pdfn=node.new(WHAT, "pdf_literal")
local tot=hd+hh %total height + depth
local ts = hd + hs %total height + any shift
pdfn.data=" 0.2 w 0 -"..ts.." "..hw.. " "..tot.." re q 0.9 0.0 0.0 rg f Q"
pdfn.mode=0
% insert into the typeset paragraph
local np = n.prev
np.next=pdfn
pdfn.prev=np
pdfn.next=n
n.prev=pdfn
end % if n.id == HLIST then
if n.id == GLYPH then
local nw = n.width
local nd= n.depth
local nh = n.height
local hw = ((nw/65536)*(72/72.27))
local hd = ((nd/65536)*(72/72.27))
local hh = ((nh/65536)*(72/72.27))
local tot=hd+hh %total height + depth
local pdfn=node.new(WHAT, "pdf_literal")
pdfn.data=" 0.2 w 0 -"..hd.." "..hw.. " "..tot.." re q 0.0 0.0 0.0 RG S Q"
pdfn.mode=0
% insert into the typeset paragraph
local np = n.prev
if np ~= nil then
np.next=pdfn
pdfn.prev=np
pdfn.next=n
n.prev=pdfn
elseif np == nil then
%First glyph at start of the list: special steps needed... May not always work...?
n.prev=pdfno
pdfn.next=n
line.head=pdfn
end
end % if n.id == GLYPH then
% Insert the line bounding box
local lw = line.width
local ld = line.depth
local lh = line.height
local lhw = ((lw/65536)*(72/72.27))
local lhd = ((ld/65536)*(72/72.27))
local lhh = ((lh/65536)*(72/72.27))
local ltot=lhd+lhh
local pdfn=node.new(WHAT, "pdf_literal")
pdfn.data=" 0.1 w 0 -"..lhd.." "..lhw.. " "..ltot.." re q 0.9 g f Q"
pdfn.mode=0
%local ln=line.head.next
pdfn.next=line.head
line.head=pdfn
end % for n in node.traverse
end % end of function
parsepara = function (head)
for line in node.traverse_id(HLIST, head) do
lineparse(line)
end
return head
end
}
\lua{callback.register("post_linebreak_filter", parsepara)}
\noindent For example, when \TeX{} typesets a paragraph of text and breaks it into a series of lines, it considers the paragraph's text as a sequence of boxes and uses the dimensions of those character boxes to find the best linebreaks. Each line of the paragraph is itself a box (containing other boxes---e.g., characters) and the typeset paragraph lines (boxes) are stacked (vertically) to produce the paragraph. Eventually, the largest box of all is produced: the typeset page. Clearly, this is a very simplified picture because you also need the ability to arbitrarily position those boxes and \TeX{} does this using so-called glue. Knuth commented (page 70 of The \TeX book) that ``glue'' probably should have been referred to as ``spring'' but the term glue was adopted early on and, to use Knuth's pun, it stuck.
\bye
