From c5fd2c4c0ee0e9f057c8d31ad529dca908b0427d Mon Sep 17 00:00:00 2001
From: Alex Maldonado <alex@scient.ing>
Date: Tue, 7 Jul 2026 01:26:09 -0400
Subject: [PATCH] feat: minus sign and number formatting

---
 _extensions/units/units.lua | 69 +++++++++++++++++++++++++++++++++++--
 1 file changed, 66 insertions(+), 3 deletions(-)

diff --git a/_extensions/units/units.lua b/_extensions/units/units.lua
index 0e4ccf7..1180506 100644
--- a/_extensions/units/units.lua
+++ b/_extensions/units/units.lua
@@ -1,6 +1,45 @@
 local NBSP = "\u{00A0}"
 local function nbsp() return pandoc.Str(NBSP) end
 
+-- The true minus sign, U+2212 -- NOT the ASCII hyphen-minus (U+002D) that
+-- string.format emits by default, and NOT an en dash (U+2013, reserved for
+-- ranges like "10-20 ns"). U+2212 is metrically designed to match the
+-- width and vertical centering of "+" (U+002B), so signed values align
+-- visually the way they would in typeset mathematics. This is applied
+-- only at the final display step -- internal parsing below still matches
+-- against the plain ASCII "-" that Lua's own %g/%e formatting produces.
+local MINUS = "\u{2212}"
+local function displayMinus(s)
+  return (s:gsub("^%-", MINUS))
+end
+
+-- Non-breaking DIGIT-GROUPING separator, per the SI Brochure / NIST SP811 /
+-- IUPAC Green Book convention: group digits in threes using a thin space
+-- (never a comma or period, since those mean different things -- decimal
+-- vs. thousands separator -- depending on locale). U+202F NARROW NO-BREAK
+-- SPACE is used rather than U+2009 THIN SPACE because the latter is an
+-- ordinary breakable space and could split a number across a line wrap.
+local THIN_NBSP = "\u{202F}"
+
+-- group a string of digits (no sign, no decimal point) into threes,
+-- e.g. "12345" -> "12" THIN_NBSP "345". Per NIST SP811, grouping is
+-- optional for exactly four digits and left ungrouped here; it is
+-- effectively mandatory from five digits up, since that's the point at
+-- which an ungrouped run of digits becomes hard to parse at a glance.
+local function groupDigits(digits)
+  local len = #digits
+  if len <= 4 then return digits end
+  local firstLen = len % 3
+  if firstLen == 0 then firstLen = 3 end
+  local groups = { digits:sub(1, firstLen) }
+  local i = firstLen + 1
+  while i <= len do
+    table.insert(groups, digits:sub(i, i + 2))
+    i = i + 3
+  end
+  return table.concat(groups, THIN_NBSP)
+end
+
 -- render a value already known to be in scientific-notation range
 -- returns a list of inlines: mantissa <NBSP> × <NBSP> 10^exp^
 local function formatSci(value)
@@ -14,13 +53,13 @@ local function formatSci(value)
   mantissa = mantissa:gsub("0+$", ""):gsub("%.$", "")
   local expNum = tonumber(sign .. exp) -- keeps sign, drops leading zeros (e.g. "+23" -> 23, "-06" -> -6)
 
-  local out = { pandoc.Str(mantissa) }
+  local out = { pandoc.Str(displayMinus(mantissa)) }
   if expNum ~= 0 then
     table.insert(out, nbsp())
     table.insert(out, pandoc.Str("×"))
     table.insert(out, nbsp())
     table.insert(out, pandoc.Str("10"))
-    table.insert(out, pandoc.Superscript({ pandoc.Str(tostring(expNum)) }))
+    table.insert(out, pandoc.Superscript({ pandoc.Str(displayMinus(tostring(expNum))) }))
   end
   return out
 end
@@ -38,7 +77,14 @@ local function formatPlain(value)
   if s:find("[eE]") then
     return formatSci(value)
   end
-  return { pandoc.Str(s) }
+  -- split into optional sign, integer part, and remainder (decimal point
+  -- + fractional digits, if any) so grouping applies only to the
+  -- integer part -- fractional digits are never grouped
+  local sign, intPart, rest = s:match("^(%-?)(%d+)(.*)$")
+  if intPart then
+    s = sign .. groupDigits(intPart) .. rest
+  end
+  return { pandoc.Str(displayMinus(s)) }
 end
 
 -- decide plain vs. scientific, unless overridden.
@@ -163,6 +209,22 @@ local function sci(args)
   return formatSci(value)
 end
 
+-- {{< number 12345 >}}          -> 12 345   (auto plain/sci, grouped)
+-- {{< number 12345 notation=plain >}} -> 12 345  (forced plain, still grouped)
+-- {{< number 12345 notation=sci >}}   -> 1.23 × 10⁴
+-- Bare-number equivalent of the unit shortcodes above, for cases where a
+-- value has no associated unit but should still receive the same digit
+-- grouping and notation handling.
+local function number(args, kwargs)
+  local value = tonumber(pandoc.utils.stringify(args[1]))
+  local mode  = kwargs and kwargs["notation"] and pandoc.utils.stringify(kwargs["notation"]) or nil
+  if not value then
+    io.stderr:write("[units] bad number() arg: '" .. pandoc.utils.stringify(args[1]) .. "'\n")
+    return pandoc.Str("[number error]")
+  end
+  return formatNumber(value, mode)
+end
+
 -- generic converter, parameterized by which table to use.
 -- optional kwargs.notation = "sci" | "plain" forces that format
 -- for both the source value and (if present) the converted value.
@@ -282,6 +344,7 @@ end
 return {
   ["unit"]          = unit,
   ["sci"]           = sci,
+  ["number"]        = number,
   ["energy"]        = makeQuantityShortcode(ENERGY_UNITS, "energy"),
   ["molarenergy"]   = makeQuantityShortcode(MOLAR_ENERGY_UNITS, "molarenergy"),
   ["distance"]      = makeQuantityShortcode(DISTANCE_UNITS, "distance"),