summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
-rw-r--r--scripts/context/lua/mtxrun.lua6
-rw-r--r--scripts/context/stubs/mswin/mtxrun.lua6
-rwxr-xr-xscripts/context/stubs/unix/mtxrun6
-rw-r--r--tex/context/base/buff-ini.lua6
-rw-r--r--tex/context/base/char-ini.lua30
-rw-r--r--tex/context/base/cont-new.tex2
-rw-r--r--tex/context/base/context.tex2
-rw-r--r--tex/context/base/l-table.lua6
-rw-r--r--tex/context/base/pack-box.mkiv7
-rw-r--r--tex/context/base/page-sid.mkii2
-rw-r--r--tex/context/base/page-sid.mkiv2
-rw-r--r--tex/context/base/sort-ini.lua300
-rw-r--r--tex/context/base/sort-lan.lua3237
-rw-r--r--tex/context/base/strc-ref.lua173
-rw-r--r--tex/context/base/strc-ref.mkii18
-rw-r--r--tex/context/base/strc-reg.lua2
-rw-r--r--tex/context/base/strc-reg.mkiv5
-rw-r--r--tex/context/base/strc-sec.mkiv33
-rw-r--r--tex/context/base/strc-syn.mkiv14
-rw-r--r--tex/generic/context/luatex-fonts-merged.lua8
20 files changed, 1108 insertions, 2757 deletions
diff --git a/scripts/context/lua/mtxrun.lua b/scripts/context/lua/mtxrun.lua
index b175a35a4..622643f9c 100644
--- a/scripts/context/lua/mtxrun.lua
+++ b/scripts/context/lua/mtxrun.lua
@@ -615,7 +615,7 @@ local function sortedkeys(tab)
elseif tkey == "number" then
-- if kind == 1 then kind = 3 else kind = 2 end
kind = (kind == 1 and 3) or 2
- else
+ else -- if tkey then
kind = 3
end
end
@@ -631,7 +631,9 @@ end
local function sortedhashkeys(tab) -- fast one
local srt = { }
for key,_ in next, tab do
- srt[#srt+1] = key
+ if key then
+ srt[#srt+1] = key
+ end
end
sort(srt)
return srt
diff --git a/scripts/context/stubs/mswin/mtxrun.lua b/scripts/context/stubs/mswin/mtxrun.lua
index b175a35a4..622643f9c 100644
--- a/scripts/context/stubs/mswin/mtxrun.lua
+++ b/scripts/context/stubs/mswin/mtxrun.lua
@@ -615,7 +615,7 @@ local function sortedkeys(tab)
elseif tkey == "number" then
-- if kind == 1 then kind = 3 else kind = 2 end
kind = (kind == 1 and 3) or 2
- else
+ else -- if tkey then
kind = 3
end
end
@@ -631,7 +631,9 @@ end
local function sortedhashkeys(tab) -- fast one
local srt = { }
for key,_ in next, tab do
- srt[#srt+1] = key
+ if key then
+ srt[#srt+1] = key
+ end
end
sort(srt)
return srt
diff --git a/scripts/context/stubs/unix/mtxrun b/scripts/context/stubs/unix/mtxrun
index b175a35a4..622643f9c 100755
--- a/scripts/context/stubs/unix/mtxrun
+++ b/scripts/context/stubs/unix/mtxrun
@@ -615,7 +615,7 @@ local function sortedkeys(tab)
elseif tkey == "number" then
-- if kind == 1 then kind = 3 else kind = 2 end
kind = (kind == 1 and 3) or 2
- else
+ else -- if tkey then
kind = 3
end
end
@@ -631,7 +631,9 @@ end
local function sortedhashkeys(tab) -- fast one
local srt = { }
for key,_ in next, tab do
- srt[#srt+1] = key
+ if key then
+ srt[#srt+1] = key
+ end
end
sort(srt)
return srt
diff --git a/tex/context/base/buff-ini.lua b/tex/context/base/buff-ini.lua
index 4ae4df699..a590cf81a 100644
--- a/tex/context/base/buff-ini.lua
+++ b/tex/context/base/buff-ini.lua
@@ -153,6 +153,9 @@ end
function buffers.range(lines,first,last,range) -- 1,3 1,+3 fromhere,tothere
local first, last = first or 1, last or #lines
+ if last < 0 then
+ last = #lines + last
+ end
local what = settings_to_array(range)
local r_first, r_last = what[1], what[2]
local f, l = tonumber(r_first), tonumber(r_last)
@@ -172,6 +175,9 @@ function buffers.range(lines,first,last,range) -- 1,3 1,+3 fromhere,tothere
end
if r_last then
if l then
+ if l < 0 then
+ l = #lines + l
+ end
if find(r_last,"^[%+]") then -- 1,+3
l = first + l
end
diff --git a/tex/context/base/char-ini.lua b/tex/context/base/char-ini.lua
index 4893875c3..28a11a1e9 100644
--- a/tex/context/base/char-ini.lua
+++ b/tex/context/base/char-ini.lua
@@ -771,10 +771,34 @@ characters.activeoffset = 0x10000 -- there will be remapped in that byte range
-- some day we will make a table
+characters.lccodes = allocate() local lccodes = characters.lccodes -- lazy table
+characters.uccodes = allocate() local uccodes = characters.uccodes -- lazy table
+characters.shcodes = allocate() local shcodes = characters.shcodes -- lazy table
+
+characters.lcchars = allocate() local lcchars = characters.lcchars -- lazy table
+characters.ucchars = allocate() local ucchars = characters.ucchars -- lazy table
+characters.shchars = allocate() local shchars = characters.shchars -- lazy table
+
+characters.lccharcodes = allocate() local lccharcodes = characters.lccharcodes -- lazy table
+characters.uccharcodes = allocate() local uccharcodes = characters.uccharcodes -- lazy table
+characters.shcharcodes = allocate() local shcharcodes = characters.shcharcodes -- lazy table
+
+setmetatable(lccodes, { __index = function(t,u) if u then local c = data[u] c = c and c.lccode or u t[u] = c return c end end } )
+setmetatable(uccodes, { __index = function(t,u) if u then local c = data[u] c = c and c.uccode or u t[u] = c return c end end } )
+setmetatable(shcodes, { __index = function(t,u) if u then local c = data[u] c = c and c.shcode or u t[u] = c return c end end } )
+
+setmetatable(lcchars, { __index = function(t,u) if u then local c = data[utfbyte(u)] c = c and c.lccode c = c and utfchar(c) or u t[u] = c return c end end } )
+setmetatable(ucchars, { __index = function(t,u) if u then local c = data[utfbyte(u)] c = c and c.uccode c = c and utfchar(c) or u t[u] = c return c end end } )
+setmetatable(shchars, { __index = function(t,u) if u then local c = data[utfbyte(u)] c = c and c.shcode c = c and utfchar(c) or u t[u] = c return c end end } )
+
+setmetatable(lccharcodes, { __index = function(t,u) if u then local c = data[utfbyte(u)] c = c and c.lccode or u t[u] = c return c end end } )
+setmetatable(uccharcodes, { __index = function(t,u) if u then local c = data[utfbyte(u)] c = c and c.uccode or u t[u] = c return c end end } )
+setmetatable(shcharcodes, { __index = function(t,u) if u then local c = data[utfbyte(u)] c = c and c.shcode or u t[u] = c return c end end } )
+
function characters.lower(str)
local new = { }
for u in utfvalues(str) do
- new[#new+1] = utfchar(data[u].lccode or u)
+ new[#new+1] = utfchar(lccodes[u])
end
return concat(new)
end
@@ -782,7 +806,7 @@ end
function characters.upper(str)
local new = { }
for u in utfvalues(str) do
- new[#new+1] = utfchar(data[u].uccode or u)
+ new[#new+1] = utfchar(uccodes[u])
end
return concat(new)
end
@@ -792,7 +816,7 @@ function characters.lettered(str)
for u in utfvalues(str) do
local d = data[u]
if is_letter[d.category] then
- new[#new+1] = utfchar(d.lccode or u)
+ new[#new+1] = utfchar(lccodes[u])
end
end
return concat(new)
diff --git a/tex/context/base/cont-new.tex b/tex/context/base/cont-new.tex
index 4c5e918ec..c9a207c34 100644
--- a/tex/context/base/cont-new.tex
+++ b/tex/context/base/cont-new.tex
@@ -11,7 +11,7 @@
%C therefore copyrighted by \PRAGMA. See mreadme.pdf for
%C details.
-\newcontextversion{2010.09.16 10:46}
+\newcontextversion{2010.09.22 09:35}
%D This file is loaded at runtime, thereby providing an
%D excellent place for hacks, patches, extensions and new
diff --git a/tex/context/base/context.tex b/tex/context/base/context.tex
index d67a066d0..681fd7312 100644
--- a/tex/context/base/context.tex
+++ b/tex/context/base/context.tex
@@ -20,7 +20,7 @@
%D your styles an modules.
\edef\contextformat {\jobname}
-\edef\contextversion{2010.09.16 10:46}
+\edef\contextversion{2010.09.22 09:35}
%D For those who want to use this:
diff --git a/tex/context/base/l-table.lua b/tex/context/base/l-table.lua
index 996e36a83..1bd9921cb 100644
--- a/tex/context/base/l-table.lua
+++ b/tex/context/base/l-table.lua
@@ -99,7 +99,7 @@ local function sortedkeys(tab)
elseif tkey == "number" then
-- if kind == 1 then kind = 3 else kind = 2 end
kind = (kind == 1 and 3) or 2
- else
+ else -- if tkey then
kind = 3
end
end
@@ -115,7 +115,9 @@ end
local function sortedhashkeys(tab) -- fast one
local srt = { }
for key,_ in next, tab do
- srt[#srt+1] = key
+ if key then
+ srt[#srt+1] = key
+ end
end
sort(srt)
return srt
diff --git a/tex/context/base/pack-box.mkiv b/tex/context/base/pack-box.mkiv
index 09af1d633..c0d3741ff 100644
--- a/tex/context/base/pack-box.mkiv
+++ b/tex/context/base/pack-box.mkiv
@@ -497,9 +497,14 @@
\fi
\advance\scratchdimen\@@bgvoffset
\xdef\bleedheight{\the\scratchdimen}%
+ %
+ \bgroup
\hsize\bleedwidth
\vsize\bleedheight
- \setbox\scratchbox\hbox{#2}%
+ \global\setbox\globalscratchbox\hbox{#2}%
+ \egroup
+ \setbox\scratchbox\box\globalscratchbox
+ %
\doif\@@bgpage\v!yes
{\setbox\scratchbox\topskippedbox{\box\scratchbox}}%
\setbox\scratchbox\hbox to \@@bgwidth
diff --git a/tex/context/base/page-sid.mkii b/tex/context/base/page-sid.mkii
index 0848c9d8c..806acc6f1 100644
--- a/tex/context/base/page-sid.mkii
+++ b/tex/context/base/page-sid.mkii
@@ -906,7 +906,7 @@
\divide\!!counta \baselineskip
\advance\!!counta \plusone
\parskip\zeropoint
- \dorecurse\!!counta{\line{}}
+ \dorecurse\!!counta{\line{}}%
\kern-\!!counta\baselineskip
\penalty\zerocount }
diff --git a/tex/context/base/page-sid.mkiv b/tex/context/base/page-sid.mkiv
index e54b4ecd5..9365f7508 100644
--- a/tex/context/base/page-sid.mkiv
+++ b/tex/context/base/page-sid.mkiv
@@ -770,7 +770,7 @@
\divide\!!counta \baselineskip
\advance\!!counta \plusone
\parskip\zeropoint
- \dorecurse\!!counta{\line{}}
+ \dorecurse\!!counta{\line{}}%
\kern-\!!counta\baselineskip
\penalty\zerocount }
diff --git a/tex/context/base/sort-ini.lua b/tex/context/base/sort-ini.lua
index eebbb9aff..386f1a45f 100644
--- a/tex/context/base/sort-ini.lua
+++ b/tex/context/base/sort-ini.lua
@@ -15,8 +15,7 @@ local utf = unicode.utf8
local gsub, rep, sort, concat = string.gsub, string.rep, table.sort, table.concat
local utfbyte, utfchar = utf.byte, utf.char
local utfcharacters, utfvalues, strcharacters = string.utfcharacters, string.utfvalues, string.characters
-local chardata = characters.data
-local next, type, tonumber = next, type, tonumber
+local next, type, tonumber, rawget, rawset = next, type, tonumber, rawget, rawset
local allocate = utilities.storage.allocate
@@ -24,48 +23,157 @@ local trace_tests = false trackers.register("sorters.tests", function(v) trace_
local report_sorters = logs.new("sorters")
-local comparers = { }
-local splitters = { }
-local entries = allocate()
-local mappings = allocate()
-local replacements = allocate()
-local ignoredoffset = 0x10000
-local replacementoffset = 0x10000
-local digitsoffset = 0x20000
-local digitsmaximum = 0xFFFFF
+local comparers = { }
+local splitters = { }
+local definitions = allocate()
+local tracers = allocate()
+local ignoredoffset = 0x10000 -- frozen
+local replacementoffset = 0x10000 -- frozen
+local digitsoffset = 0x20000 -- frozen
+local digitsmaximum = 0xFFFFF -- frozen
+
+local lccodes = characters.lccodes
+local shcodes = characters.shcodes
+local lcchars = characters.lcchars
+local shchars = characters.shchars
+
+local variables = interfaces.variables
sorters = {
- comparers = comparers,
- splitters = splitters,
- entries = entries,
- mappings = mappings,
- replacements = replacements,
- constants = {
+ comparers = comparers,
+ splitters = splitters,
+ definitions = definitions,
+ tracers = tracers,
+ constants = {
ignoredoffset = ignoredoffset,
replacementoffset = replacementoffset,
digitsoffset = digitsoffset,
digitsmaximum = digitsmaximum,
+ defaultlanguage = variables.default,
+ defaultmethod = variables.before,
}
}
-local ssorters = sorters
+local sorters = sorters
+local constants = sorters.constants
-local language, defaultlanguage = 'en', 'en'
+local data, language, method
+local replacements, mappings, entries, orders, lower, upper
-local currentreplacements, currentmappings, currententries = { }, { }, { }
+local mte = {
+ __index = function(t,k)
+ local el
+ if k then
+ local l = lower[k] or lcchars[k]
+ el = rawget(t,l)
+ end
+ if not el then
+ local l = shchars[k]
+ if l and l ~= k then
+ el = rawget(t,l)
+ if not el then
+ l = lower[k] or lcchars[l]
+ if l then
+ el = rawget(t,l)
+ end
+ end
+ end
+ el = el or k
+ end
+ -- rawset(t,k,el) also make a copy?
+ return el
+ end
+}
-function sorters.setlanguage(lang)
- language = lang or language or defaultlanguage
- currentreplacements = replacements[language] or replacements[defaultlanguage] or { }
- currentmappings = mappings [language] or mappings [defaultlanguage] or { }
- currententries = entries [language] or entries [defaultlanguage] or { }
+local function preparetables(data)
+ local orders, lower, method, mappings = data.orders, data.lower, data.method, { }
+ for i=1,#orders do
+ local oi = orders[i]
+ mappings[oi] = 2*i
+ end
+ local delta = (method == variables.before or method == variables.first or method == variables.last) and -1 or 1
+ local mtm = {
+ __index = function(t,k)
+ local n
+ if k then
+ local l = lower[k] or lcchars[k]
+ if l then
+ local ml = rawget(t,l)
+ if ml then
+ n = ml + delta -- first
+ end
+ end
+ if not n then
+ l = shchars[k]
+ if l and l ~= k then
+ local ml = rawget(t,l)
+ if ml then
+ n = ml -- first or last
+ else
+ l = lower[l] or lcchars[l]
+ if l then
+ local ml = rawget(t,l)
+ if ml then
+ n = ml + delta
+ end
+ end
+ end
+ end
+ end
+ if not n then
+ n = 0
+ end
+ else
+ n = 0
+ end
+ rawset(t,k,n)
+ return n
+ end
+ }
+ data.mappings = mappings
+ setmetatable(data.entries,mte)
+ setmetatable(data.mappings,mtm)
+ return mappings
+end
+
+local function update() -- prepare parent chains, needed when new languages are added
+ for language, data in next, definitions do
+ local parent = data.parent or "default"
+ if language ~= "default" then
+ setmetatable(data,{ __index = definitions[parent] or definitions.default })
+ end
+ data.language = language
+ data.parent = parent
+ data.mappings = { } -- free temp data
+ end
+end
+
+local function setlanguage(l,m)
+ language = (l ~= "" and l) or constants.defaultlanguage
+ data = definitions[language or constants.defaultlanguage] or definitions[constants.defaultlanguage]
+ method = (m ~= "" and m) or data.method or constants.defaultmethod
if trace_tests then
- report_sorters("setting language '%s'",language)
+ report_sorters("setting language '%s', method '%s'",language,method)
end
- return currentreplacements, currentmappings, currententries
+ data.method = method
+ replacements = data.replacements
+ entries = data.entries
+ orders = data.orders
+ lower = data.lower
+ upper = data.upper
+ mappings = preparetables(data)
+ return data
+end
+
+function sorters.update()
+ update()
+ setlanguage(language,method) -- resync current language and method
end
--- maybe inline code if it's too slow
+function sorters.setlanguage(language,method)
+ update()
+ setlanguage(language,method) -- new language and method
+end
local function basicsort(sort_a,sort_b)
if not sort_a or not sort_b then
@@ -118,16 +226,27 @@ function comparers.basic(a,b) -- trace ea and eb
local na, nb = #ea, #eb
if na == 0 and nb == 0 then
-- simple variant (single word)
- local result = basicsort(ea.e,eb.e)
- return (result == 0 and result) or basicsort(ea.m,eb.m)
+ local result = basicsort(ea.m,eb.m)
+ if result == 0 then
+ result = basicsort(ea.c,eb.c)
+ end
+ if result == 0 then
+ result = basicsort(ea.u,eb.u)
+ end
+ return result
else
-- complex variant, used in register (multiple words)
local result = 0
for i=1,nb < na and nb or na do
local eai, ebi = ea[i], eb[i]
- result = basicsort(eai.e,ebi.e)
if result == 0 then
- result = basicsort(eai.m,ebi.m) -- only needed it there are m's
+ result = basicsort(eai.m,ebi.m)
+ end
+ if result == 0 then
+ result = basicsort(eai.c,ebi.c)
+ end
+ if result == 0 then
+ result = basicsort(eai.u,ebi.u)
end
if result ~= 0 then
break
@@ -177,59 +296,50 @@ local function firstofsplit(entry)
split = split.s
end
local entry = split and split[1] or ""
- return entry, currententries[entry] or "\000"
+ return entry, entries[entry] or "\000"
end
sorters.firstofsplit = firstofsplit
--- beware, numbers get spaces in front
-
-function splitters.utf(str)
- if #currentreplacements > 0 then
- for k=1,#currentreplacements do
- local v = currentreplacements[k]
+function splitters.utf(str) -- we could append m and u but this is cleaner, s is for tracing
+ if #replacements > 0 then
+ -- todo make an lpeg for this
+ for k=1,#replacements do
+ local v = replacements[k]
str = gsub(str,v[1],v[2])
end
end
- local s, e, m, n = { }, { }, { }, 0
- for sc in utfcharacters(str) do -- maybe an lpeg
- local ec, mc = currententries[sc], currentmappings[sc] or utfbyte(sc)
- n = n + 1
- s[n] = sc
- e[n] = currentmappings[ec] or mc
- m[n] = mc
+ local s, u, m, c, n = { }, { }, { }, { }, 0
+ if method == variables.last then
+ for sc in utfcharacters(str) do
+ local b = utfbyte(sc)
+ local l = lower[sc]
+ l = l and utfbyte(l) or lccodes[b]
+ if l ~= b then l = l - 1 end -- brrrr, can clash
+ n = n + 1
+ s[n], u[n], m[n], c[n] = sc, b, l, mappings[sc]
+ end
+ elseif method == variables.first then
+ for sc in utfcharacters(str) do
+ local b = utfbyte(sc)
+ local l = lower[sc]
+ l = l and utfbyte(l) or lccodes[b]
+ if l ~= b then l = l + 1 end -- brrrr, can clash
+ n = n + 1
+ s[n], u[n], m[n], c[n] = sc, b, l, mappings[sc]
+ end
+ else
+ for sc in utfcharacters(str) do
+ local b = utfbyte(sc)
+ n = n + 1
+ s[n], u[n], m[n], c[n] = sc, b, mappings[sc], b
+ end
end
- local t = { s = s, e = e, m = m }
---~ table.print(t)
+ local t = { s = s, m = m, u = u, c = c }
+ -- table.print(t)
return t
end
--- we can use one array instead (sort of like in mkii)
--- but for the moment we do it this way as it is more
--- handy for tracing
-
--- function splitters.utf(str)
--- if #currentreplacements > 0 then
--- for k=1,#currentreplacements do
--- local v = currentreplacements[k]
--- str = gsub(str,v[1],v[2])
--- end
--- end
--- local s, e, m, n = { }, { }, { }, 0
--- for sc in utfcharacters(str) do -- maybe an lpeg
--- local ec, mc = currententries[sc], currentmappings[sc] or utfbyte(sc)
--- n = n + 1
--- ec = currentmappings[ec] or mc
--- s[n] = sc
--- e[n] = ec
--- if ec ~= mc then
--- n = n + 1
--- e[n] = mc
--- end
--- end
--- return { s = s, e = e }
--- end
-
function table.remap(t)
local tt = { }
for k,v in next, t do
@@ -286,45 +396,3 @@ function sorters.sort(entries,cmp)
end)
end
end
-
--- some day we can have a characters.upper and characters.lower
-
-function sorters.adduppercasereplacements(what)
- local rep, new = replacements[what], { }
- for i=1,#rep do
- local r = rep[i]
- local u = chardata[utfbyte(r[1])].uccode
- if u then
- new[utfchar(u)] = r[2]
- end
- end
- for k, v in next, new do
- rep[k] = v
- end
-end
-
-function sorters.adduppercaseentries(what)
- local ent, new = entries[what], { }
- for k, v in next, ent do
- local u = chardata[utfbyte(k)].uccode
- if u then
- new[utfchar(u)] = v
- end
- end
- for k, v in next, new do
- ent[k] = v
- end
-end
-
-function sorters.adduppercasemappings(what,offset)
- local map, new, offset = mappings[what], { }, offset or 0
- for k, v in next, map do
- local u = chardata[utfbyte(k)].uccode
- if u then
- new[utfchar(u)] = v + offset
- end
- end
- for k, v in next, new do
- map[k] = v
- end
-end
diff --git a/tex/context/base/sort-lan.lua b/tex/context/base/sort-lan.lua
index 322af2598..338ba747b 100644
--- a/tex/context/base/sort-lan.lua
+++ b/tex/context/base/sort-lan.lua
@@ -7,758 +7,319 @@ if not modules then modules = { } end modules ['sort-lan'] = {
}
-- Many vectors were supplied by Wolfgang Schuster and Philipp
--- Gesang.
---
--- Replacements are indexed as they need to be applied in sequence
---
--- Maybe we should load these tables runtime, just like patterns.
-
-local utf = unicode.utf8
-local uc = utf.char
-local ub = utf.byte
-
-local sorters = sorters
-
-local mappings = sorters.mappings
-local entries = sorters.entries
-local replacements = sorters.replacements
-
-local adduppercasereplacements = sorters.adduppercasereplacements
-local adduppercaseentries = sorters.adduppercaseentries
-local adduppercasemappings = sorters.adduppercasemappings
-
-local replacementoffset = sorters.constants.replacementoffset
+-- Gesang. However this is a quite adapted and reformatted variant
+-- so it needs some checking.
+
+local utfchar = utf.char
+local sorters = sorters
+local definitions = sorters.definitions
+local replacementoffset = sorters.constants.replacementoffset
+
+definitions["default"] = {
+ method = "before",
+ replacements = {
+ -- no replacements
+ },
+ entries = {
+ ["a"] = "a", ["b"] = "b", ["c"] = "c", ["d"] = "d", ["e"] = "e",
+ ["f"] = "f", ["g"] = "g", ["h"] = "h", ["i"] = "i", ["j"] = "j",
+ ["k"] = "k", ["l"] = "l", ["m"] = "m", ["n"] = "n", ["o"] = "o",
+ ["p"] = "p", ["q"] = "q", ["r"] = "r", ["s"] = "s", ["t"] = "t",
+ ["u"] = "u", ["v"] = "v", ["w"] = "w", ["x"] = "x", ["y"] = "y",
+ ["z"] = "z",
+ },
+ orders = {
+ "a", "b", "c", "d", "e", "f", "g", "h", "i", "j",
+ "k", "l", "m", "n", "o", "p", "q", "r", "s", "t",
+ "u", "v", "w", "x", "y", "z",
+ },
+ lower = {
+ -- no replacements
+ },
+ upper = {
+ -- no replacements
+ }
+}
+
+sorters.setlanguage("default")
-- english
-entries['en'] = {
- ["a"] = "a", ["b"] = "b", ["c"] = "c", ["d"] = "d", ["e"] = "e",
- ["f"] = "f", ["g"] = "g", ["h"] = "h", ["i"] = "i", ["j"] = "j",
- ["k"] = "k", ["l"] = "l", ["m"] = "m", ["n"] = "n", ["o"] = "o",
- ["p"] = "p", ["q"] = "q", ["r"] = "r", ["s"] = "s", ["t"] = "t",
- ["u"] = "u", ["v"] = "v", ["w"] = "w", ["x"] = "x", ["y"] = "y",
- ["z"] = "z",
- ["A"] = "a", ["B"] = "b", ["C"] = "c", ["D"] = "d", ["E"] = "e",
- ["F"] = "f", ["G"] = "g", ["H"] = "h", ["I"] = "i", ["J"] = "j",
- ["K"] = "k", ["L"] = "l", ["M"] = "m", ["N"] = "n", ["O"] = "o",
- ["P"] = "p", ["Q"] = "q", ["R"] = "r", ["S"] = "s", ["T"] = "t",
- ["U"] = "u", ["V"] = "v", ["W"] = "w", ["X"] = "x", ["Y"] = "y",
- ["Z"] = "z",
-}
-
-mappings['en'] = {
- ["a"] = 1, ["b"] = 3, ["c"] = 5, ["d"] = 7, ["e"] = 9,
- ["f"] = 11, ["g"] = 13, ["h"] = 15, ["i"] = 17, ["j"] = 19,
- ["k"] = 21, ["l"] = 23, ["m"] = 25, ["n"] = 27, ["o"] = 29,
- ["p"] = 31, ["q"] = 33, ["r"] = 35, ["s"] = 37, ["t"] = 39,
- ["u"] = 41, ["v"] = 43, ["w"] = 45, ["x"] = 47, ["y"] = 49,
- ["z"] = 51,
- ["A"] = 1, ["B"] = 3, ["C"] = 5, ["D"] = 7, ["E"] = 9,
- ["F"] = 11, ["G"] = 13, ["H"] = 15, ["I"] = 17, ["J"] = 19,
- ["K"] = 21, ["L"] = 23, ["M"] = 25, ["N"] = 27, ["O"] = 29,
- ["P"] = 31, ["Q"] = 33, ["R"] = 35, ["S"] = 37, ["T"] = 39,
- ["U"] = 41, ["V"] = 43, ["W"] = 45, ["X"] = 47, ["Y"] = 49,
- ["Z"] = 51,
- --
- -- uppercase after lowercase
- --
- -- ["A"] = 2, ["B"] = 4, ["C"] = 6, ["D"] = 8, ["E"] = 10,
- -- ["F"] = 12, ["G"] = 14, ["H"] = 16, ["I"] = 18, ["J"] = 20,
- -- ["K"] = 22, ["L"] = 24, ["M"] = 26, ["N"] = 28, ["O"] = 30,
- -- ["P"] = 32, ["Q"] = 34, ["R"] = 36, ["S"] = 38, ["T"] = 40,
- -- ["U"] = 42, ["V"] = 44, ["W"] = 46, ["X"] = 48, ["Y"] = 50,
- -- ["Z"] = 52,
-}
+definitions["en"] = { parent = "default" }
-- dutch
-replacements['nl'] = { { "ij", 'y' }, { "IJ", 'Y' } }
-entries ['nl'] = entries ['en']
-mappings ['nl'] = mappings['en']
-
---~ -- czech (defined later)
---~
---~ local cz_ch = uc(replacementoffset + 1)
---~ local cz_CH = uc(replacementoffset + 2)
---~
---~ replacements['cz'] = {
---~ [1] = { "ch", cz_ch }
---~ }
---~
---~ entries['cz'] = {
---~ ['a'] = "a", -- a
---~ [uc(0x00E1)] = "a", -- aacute
---~ ['b'] = "b", -- b
---~ ['c'] = "c", -- c
---~ [uc(0x010D)] = uc(0x010D), -- ccaron
---~ ['d'] = "d", -- d
---~ [uc(0x010F)] = "d", -- dcaron
---~ ['e'] = "e", -- e
---~ [uc(0x00E9)] = "e", -- eacute
---~ [uc(0x011B)] = "e", -- ecaron
---~ ['f'] = "f", -- f
---~ ['g'] = "g", -- g
---~ ['h'] = "h", -- h
---~ [cz_ch] = "ch", -- ch
---~ ['i'] = "i", -- i
---~ [uc(0x00ED)] = "i", -- iacute
---~ ['j'] = "j", -- j
---~ ['k'] = "k", -- k
---~ ['l'] = "l", -- l
---~ ['m'] = "m", -- m
---~ ['n'] = "n", -- n
---~ ['ň'] = "n", -- ncaron
---~ ['o'] = "o", -- o
---~ ['p'] = "p", -- p
---~ ['q'] = "q", -- q
---~ ['r'] = "r", -- r
---~ ['ř'] = "ř", -- rcaron
---~ ['s'] = "s", -- s
---~ [uc(0x0161)] = uc(0x0161), -- scaron
---~ ['t'] = "t", -- t
---~ [uc(0x0165)] = "t", -- tcaron
---~ ['u'] = "u", -- u
---~ [uc(0x00FA)] = "u", -- uacute
---~ [uc(0x016F)] = "u", -- uring
---~ ['v'] = "v", -- v
---~ ['w'] = "w", -- w
---~ ['x'] = "x", -- x
---~ ['y'] = "y", -- y
---~ [uc(0x00FD)] = "y", -- yacute
---~ ['z'] = "z", -- z
---~ [uc(0x017E)] = uc(0x017E), -- zcaron
---~ }
---~
---~ mappings['cz'] = {
---~ ['a'] = 1, -- a
---~ [uc(0x00E1)] = 3, -- aacute
---~ ['b'] = 5, -- b
---~ ['c'] = 7, -- c
---~ [uc(0x010D)] = 9, -- ccaron
---~ ['d'] = 11, -- d
---~ [uc(0x010F)] = 13, -- dcaron
---~ ['e'] = 15, -- e
---~ [uc(0x00E9)] = 17, -- eacute
---~ [uc(0x011B)] = 19, -- ecaron
---~ ['f'] = 21, -- f
---~ ['g'] = 23, -- g
---~ ['h'] = 25, -- h
---~ [cz_ch] = 27, -- ch
---~ ['i'] = 29, -- i
---~ [uc(0x00ED)] = 31, -- iacute
---~ ['j'] = 33, -- j
---~ ['k'] = 35, -- k
---~ ['l'] = 37, -- l
---~ ['m'] = 39, -- m
---~ ['n'] = 41, -- n
---~ ['ň'] = 43, -- ncaron
---~ ['o'] = 45, -- o
---~ ['p'] = 47, -- p
---~ ['q'] = 49, -- q
---~ ['r'] = 51, -- r
---~ ['ř'] = 53, -- rcaron
---~ ['s'] = 55, -- s
---~ [uc(0x0161)] = 57, -- scaron
---~ ['t'] = 59, -- t
---~ [uc(0x0165)] = 61, -- tcaron
---~ ['u'] = 63, -- u
---~ [uc(0x00FA)] = 65, -- uacute
---~ [uc(0x016F)] = 67, -- uring
---~ ['v'] = 69, -- v
---~ ['w'] = 71, -- w
---~ ['x'] = 73, -- x
---~ ['y'] = 75, -- y
---~ [uc(0x00FD)] = 77, -- yacute
---~ ['z'] = 79, -- z
---~ [uc(0x017E)] = 81, -- zcaron
---~ }
---~
---~ adduppercaseentries ("cz")
---~ adduppercasemappings("cz") -- 1 can be option (but then we need a runtime variant)
---~
---~ entries ['cz'][cz_CH] = entries ['cz'][cz_ch]
---~ mappings['cz'][cz_CH] = mappings['cz'][cz_ch]
---~
---~ replacements['cs'] = replacements['cz']
---~ entries ['cs'] = entries ['cz']
---~ mappings ['cs'] = mappings ['cz']
-
---~ print(table.serialize(mappings.cs))
+definitions['nl'] = {
+ parent = 'default',
+ replacements = {
+ { "ij", 'y' }, { "IJ", 'Y' },
+ },
+}
-- French
-entries ['fr'] = entries ['en']
-mappings['fr'] = mappings['en']
+definitions['fr'] = { parent = 'default' }
-- German (by Wolfgang Schuster)
-- DIN 5007-1
-entries ['DIN 5007-1'] = entries ['en']
-mappings ['DIN 5007-1'] = mappings['en']
+definitions['DIN 5007-1'] = { parent = 'default' }
-- DIN 5007-2
-replacements['DIN 5007-2'] = { -- todo: adduppercasereplacements
- { "ä", 'ae' },
- { "ö", 'oe' },
- { "ü", 'ue' },
- { "Ä", 'Ae' },
- { "Ö", 'Oe' },
- { "Ü", 'Ue' },
+definitions['DIN 5007-2'] = {
+ parent = 'default',
+ replacements = {
+ { "ä", 'ae' }, { "Ä", 'Ae' },
+ { "ö", 'oe' }, { "Ö", 'Oe' },
+ { "ü", 'ue' }, { "Ü", 'Ue' },
+ },
}
---~ adduppercasereplacements('DIN 5007-2')
-
-entries ['DIN 5007-2'] = entries ['en']
-mappings ['DIN 5007-2'] = mappings['en']
-
-- Duden
-replacements['Duden'] = { { "ß", 's' } }
-entries ['Duden'] = entries ['en']
-mappings ['Duden'] = mappings['en']
-
--- new german
-
-entries ['de'] = entries ['en']
-mappings ['de'] = mappings['en']
-
--- old german
-
-entries ['deo'] = entries ['de']
-mappings ['deo'] = mappings['de']
-
--- german - Germany
-
-entries ['de-DE'] = entries ['de']
-mappings ['de-DE'] = mappings['de']
-
--- german - Swiss
+definitions['Duden'] = {
+ parent = 'default',
+ replacements = { { "ß", 's' } },
+}
-entries ['de-CH'] = entries ['de']
-mappings ['de-CH'] = mappings['de']
+definitions['de'] = { parent = 'default' } -- new german
+definitions['deo'] = { parent = 'de' } -- old german
+definitions['de-DE'] = { parent = 'de' } -- german - Germany
+definitions['de-CH'] = { parent = 'de' } -- german - Swiss
-- german - Austria
-entries['de-AT'] = {
- ["a"] = "a", ["ä"] = "ä", ["b"] = "b", ["c"] = "c", ["d"] = "d",
- ["e"] = "e", ["f"] = "f", ["g"] = "g", ["h"] = "h", ["i"] = "i",
- ["j"] = "j", ["k"] = "k", ["l"] = "l", ["m"] = "m", ["n"] = "n",
- ["o"] = "o", ["ö"] = "ö", ["p"] = "p", ["q"] = "q", ["r"] = "r",
- ["s"] = "s", ["t"] = "t", ["u"] = "u", ["ü"] = "ü", ["v"] = "v",
- ["w"] = "w", ["x"] = "x", ["y"] = "y", ["z"] = "z",
--- ["A"] = "a", ["Ä"] = "ä", ["B"] = "b", ["C"] = "c", ["D"] = "d",
--- ["E"] = "e", ["F"] = "f", ["G"] = "g", ["H"] = "h", ["I"] = "i",
--- ["J"] = "j", ["K"] = "k", ["L"] = "l", ["M"] = "m", ["N"] = "n",
--- ["O"] = "o", ["Ö"] = "ö", ["P"] = "p", ["Q"] = "q", ["R"] = "r",
--- ["S"] = "s", ["T"] = "t", ["U"] = "u", ["Ü"] = "ü", ["V"] = "v",
--- ["W"] = "w", ["X"] = "x", ["Y"] = "y", ["Z"] = "z",
+definitions['de-AT'] = {
+ entries = {
+ ["a"] = "a", ["ä"] = "ä", ["b"] = "b", ["c"] = "c", ["d"] = "d",
+ ["e"] = "e", ["f"] = "f", ["g"] = "g", ["h"] = "h", ["i"] = "i",
+ ["j"] = "j", ["k"] = "k", ["l"] = "l", ["m"] = "m", ["n"] = "n",
+ ["o"] = "o", ["ö"] = "ö", ["p"] = "p", ["q"] = "q", ["r"] = "r",
+ ["s"] = "s", ["t"] = "t", ["u"] = "u", ["ü"] = "ü", ["v"] = "v",
+ ["w"] = "w", ["x"] = "x", ["y"] = "y", ["z"] = "z",
+ },
+ orders = {
+ "a", "ä", "b", "c", "d", "e", "f", "g", "h", "i",
+ "j", "k", "l", "m", "n", "o", "ö", "p", "q", "r",
+ "s", "t", "u", "ü", "v", "w", "x", "y", "z",
+ },
}
-mappings['de-AT'] = {
- ["a"] = 1, ["ä"] = 3, ["b"] = 5, ["c"] = 7, ["d"] = 9,
- ["e"] = 11, ["f"] = 13, ["g"] = 15, ["h"] = 17, ["i"] = 19,
- ["j"] = 21, ["k"] = 23, ["l"] = 25, ["m"] = 27, ["n"] = 29,
- ["o"] = 31, ["ö"] = 33, ["p"] = 35, ["q"] = 37, ["r"] = 39,
- ["s"] = 41, ["t"] = 43, ["u"] = 45, ["ü"] = 47, ["v"] = 49,
- ["w"] = 51, ["x"] = 53, ["y"] = 55, ["z"] = 57,
--- ["A"] = 2, ["Ä"] = 4, ["B"] = 6, ["C"] = 8, ["D"] = 10,
--- ["E"] = 12, ["F"] = 14, ["G"] = 16, ["H"] = 18, ["I"] = 20,
--- ["J"] = 22, ["K"] = 24, ["L"] = 26, ["M"] = 28, ["N"] = 30,
--- ["O"] = 32, ["Ö"] = 34, ["P"] = 36, ["Q"] = 38, ["R"] = 40,
--- ["S"] = 42, ["T"] = 44, ["U"] = 46, ["Ü"] = 48, ["V"] = 50,
--- ["W"] = 52, ["X"] = 54, ["Y"] = 56, ["Z"] = 58,
-}
-
-adduppercaseentries ('de-AT')
-adduppercasemappings('de-AT',1)
-
-- finish (by Wolfgang Schuster)
-entries['fi'] = {
- ["a"] = "a", ["b"] = "b", ["c"] = "c", ["d"] = "d", ["e"] = "e",
- ["f"] = "f", ["g"] = "g", ["h"] = "h", ["i"] = "i", ["j"] = "j",
- ["k"] = "k", ["l"] = "l", ["m"] = "m", ["n"] = "n", ["o"] = "o",
- ["p"] = "p", ["q"] = "q", ["r"] = "r", ["s"] = "s", ["t"] = "t",
- ["u"] = "u", ["v"] = "v", ["w"] = "w", ["x"] = "x", ["y"] = "y",
- ["z"] = "z", ["å"] = "å", ["ä"] = "ä", ["ö"] = "ö",
+definitions['fi'] = {
+ entries = {
+ ["a"] = "a", ["b"] = "b", ["c"] = "c", ["d"] = "d", ["e"] = "e",
+ ["f"] = "f", ["g"] = "g", ["h"] = "h", ["i"] = "i", ["j"] = "j",
+ ["k"] = "k", ["l"] = "l", ["m"] = "m", ["n"] = "n", ["o"] = "o",
+ ["p"] = "p", ["q"] = "q", ["r"] = "r", ["s"] = "s", ["t"] = "t",
+ ["u"] = "u", ["v"] = "v", ["w"] = "w", ["x"] = "x", ["y"] = "y",
+ ["z"] = "z", ["å"] = "å", ["ä"] = "ä", ["ö"] = "ö",
+ },
+ orders = {
+ "a", "b", "c", "d", "e", "f", "g", "h", "i", "j",
+ "k", "l", "m", "n", "o", "p", "q", "r", "s", "t",
+ "u", "v", "w", "x", "y", "z", "å", "ä", "ö",
+ }
+}
+
+-- slovenian by MM: this will change since we need to add accented vowels
+
+definitions['sl'] = {
+ entries = {
+ ["a"] = "a", ["b"] = "b", ["c"] = "c", ["č"] = "č", ["ć"] = "ć", ["d"] = "d",
+ ["đ"] = "đ", ["e"] = "e", ["f"] = "f", ["g"] = "g", ["h"] = "h", ["i"] = "i",
+ ["j"] = "j", ["k"] = "k", ["l"] = "l", ["m"] = "m", ["n"] = "n", ["o"] = "o",
+ ["p"] = "p", ["q"] = "q", ["r"] = "r", ["s"] = "s", ["š"] = "š", ["t"] = "t",
+ ["u"] = "u", ["v"] = "v", ["w"] = "w", ["x"] = "x", ["y"] = "y", ["z"] = "z",
+ ["ž"] = "ž",
+ },
+ orders = {
+ "a", "b", "c", "č", "ć", "d", "đ", "e", "f", "g", "h", "i",
+ "j", "k", "l", "m", "n", "o", "p", "q", "r", "s", "š", "t",
+ "u", "v", "w", "x", "y", "z", "ž",
+ }
+}
+
+-- The following data was provided by Philipp Gesang.
+
+definitions["ru"] = {
+ entries = {
+ ["а"] = "а", ["б"] = "б", ["в"] = "в", ["г"] = "г", ["д"] = "д",
+ ["е"] = "е", ["ё"] = "е", ["ж"] = "ж", ["з"] = "з", ["и"] = "и",
+ ["і"] = "и", ["й"] = "й", ["к"] = "к", ["л"] = "л", ["м"] = "м",
+ ["н"] = "н", ["о"] = "о", ["п"] = "п", ["р"] = "р", ["с"] = "с",
+ ["т"] = "т", ["у"] = "у", ["ф"] = "ф", ["х"] = "х", ["ц"] = "ц",
+ ["ч"] = "ч", ["ш"] = "ш", ["щ"] = "щ", ["ъ"] = "ъ", ["ы"] = "ы",
+ ["ь"] = "ь", ["ѣ"] = "ѣ", ["э"] = "э", ["ю"] = "ю", ["я"] = "я",
+ ["ѳ"] = "ѳ", ["ѵ"] = "ѵ",
+ },
+ orders = {
+ "а", "б", "в", "г", "д", "е", "ё", "ж", "з", "и",
+ "і", "й", "к", "л", "м", "н", "о", "п", "р", "с",
+ "т", "у", "ф", "х", "ц", "ч", "ш", "щ", "ъ", "ы",
+ "ь", "ѣ", "э", "ю", "я", "ѳ", "ѵ",
+ }
}
-mappings['fi'] = {
- ["a"] = 1, ["b"] = 3, ["c"] = 5, ["d"] = 7, ["e"] = 9,
- ["f"] = 11, ["g"] = 13, ["h"] = 15, ["i"] = 17, ["j"] = 19,
- ["k"] = 21, ["l"] = 23, ["m"] = 25, ["n"] = 27, ["o"] = 29,
- ["p"] = 31, ["q"] = 33, ["r"] = 35, ["s"] = 37, ["t"] = 39,
- ["u"] = 41, ["v"] = 43, ["w"] = 45, ["x"] = 47, ["y"] = 49,
- ["z"] = 51, ["å"] = 53, ["ä"] = 55, ["ö"] = 57,
-}
-
-adduppercaseentries ("fi")
-adduppercasemappings("fi")
-
--- slovenian
---
--- MM: this will change since we need to add accented vowels
-
-entries['sl'] = {
- ["a"] = "a", ["b"] = "b", ["c"] = "c", ["č"] = "č", ["ć"] = "ć", ["d"] = "d",
- ["đ"] = "đ", ["e"] = "e", ["f"] = "f", ["g"] = "g", ["h"] = "h", ["i"] = "i",
- ["j"] = "j", ["k"] = "k", ["l"] = "l", ["m"] = "m", ["n"] = "n", ["o"] = "o",
- ["p"] = "p", ["q"] = "q", ["r"] = "r", ["s"] = "s", ["š"] = "š", ["t"] = "t",
- ["u"] = "u", ["v"] = "v", ["w"] = "w", ["x"] = "x", ["y"] = "y", ["z"] = "z",
- ["ž"] = "ž",
-}
-
-mappings['sl'] = {
- ["a"] = 1, ["b"] = 3, ["c"] = 5, ["č"] = 7, ["ć"] = 9, ["d"] = 11,
- ["đ"] = 13, ["e"] = 15, ["f"] = 17, ["g"] = 19, ["h"] = 21, ["i"] = 23,
- ["j"] = 25, ["k"] = 27, ["l"] = 29, ["m"] = 31, ["n"] = 33, ["o"] = 35,
- ["p"] = 37, ["q"] = 39, ["r"] = 41, ["s"] = 43, ["š"] = 45, ["t"] = 47,
- ["u"] = 49, ["v"] = 51, ["w"] = 53, ["x"] = 55, ["y"] = 57, ["z"] = 59,
- ["ž"] = 61,
-}
-
-adduppercaseentries ("sl")
-adduppercasemappings("sl") -- cf. MM
-
--- The following (quite some) languages were provided by Philipp
--- Gesang (Phg), megas.kapaneus@gmail.com.
-
-replacements["ru"] = { --[[ None, do you miss any? ]] }
-
-entries["ru"] = {
- ["а"] = "а", ["б"] = "б", ["в"] = "в", ["г"] = "г", ["д"] = "д",
- ["е"] = "е", ["ё"] = "е", ["ж"] = "ж", ["з"] = "з", ["и"] = "и",
- ["і"] = "и", ["й"] = "й", ["к"] = "к", ["л"] = "л", ["м"] = "м",
- ["н"] = "н", ["о"] = "о", ["п"] = "п", ["р"] = "р", ["с"] = "с",
- ["т"] = "т", ["у"] = "у", ["ф"] = "ф", ["х"] = "х", ["ц"] = "ц",
- ["ч"] = "ч", ["ш"] = "ш", ["щ"] = "щ", ["ъ"] = "ъ", ["ы"] = "ы",
- ["ь"] = "ь", ["ѣ"] = "ѣ", ["э"] = "э", ["ю"] = "ю", ["я"] = "я",
- ["ѳ"] = "ѳ", ["ѵ"] = "ѵ",
-}
-
-mappings["ru"] = {
- ["а"] = 1, ["б"] = 2, ["в"] = 3, ["г"] = 4, ["д"] = 5,
- ["е"] = 6, ["ё"] = 6, ["ж"] = 7, ["з"] = 8, ["и"] = 9,
- ["і"] = 9, ["й"] = 10, ["к"] = 11, ["л"] = 12, ["м"] = 13,
- ["н"] = 14, ["о"] = 15, ["п"] = 16, ["р"] = 17, ["с"] = 18,
- ["т"] = 19, ["у"] = 20, ["ф"] = 21, ["х"] = 22, ["ц"] = 23,
- ["ч"] = 24, ["ш"] = 25, ["щ"] = 26, ["ъ"] = 27, ["ы"] = 28,
- ["ь"] = 29, ["ѣ"] = 30, ["э"] = 31, ["ю"] = 32, ["я"] = 33,
- ["ѳ"] = 34, ["ѵ"] = 35,
-}
-
-adduppercaseentries ("ru")
-adduppercasemappings("ru")
-
--- Basic Ukrainian
-replacements["uk"] = { --[[ None, do you miss any? ]] }
-
-entries["uk"] = {
- ["а"] = "а", ["б"] = "б", ["в"] = "в", ["г"] = "г", ["ґ"] = "ґ",
- ["д"] = "д", ["е"] = "е", ["є"] = "є", ["ж"] = "ж", ["з"] = "з",
- ["и"] = "и", ["і"] = "і", ["ї"] = "ї", ["й"] = "й", ["к"] = "к",
- ["л"] = "л", ["м"] = "м", ["н"] = "н", ["о"] = "о", ["п"] = "п",
- ["р"] = "р", ["с"] = "с", ["т"] = "т", ["у"] = "у", ["ф"] = "ф",
- ["х"] = "х", ["ц"] = "ц", ["ч"] = "ч", ["ш"] = "ш", ["щ"] = "щ",
- ["ь"] = "ь", ["ю"] = "ю", ["я"] = "я",
+definitions["uk"] = {
+ entries = {
+ ["а"] = "а", ["б"] = "б", ["в"] = "в", ["г"] = "г", ["ґ"] = "ґ",
+ ["д"] = "д", ["е"] = "е", ["є"] = "є", ["ж"] = "ж", ["з"] = "з",
+ ["и"] = "и", ["і"] = "і", ["ї"] = "ї", ["й"] = "й", ["к"] = "к",
+ ["л"] = "л", ["м"] = "м", ["н"] = "н", ["о"] = "о", ["п"] = "п",
+ ["р"] = "р", ["с"] = "с", ["т"] = "т", ["у"] = "у", ["ф"] = "ф",
+ ["х"] = "х", ["ц"] = "ц", ["ч"] = "ч", ["ш"] = "ш", ["щ"] = "щ",
+ ["ь"] = "ь", ["ю"] = "ю", ["я"] = "я",
+ },
+ orders = {
+ "а", "б", "в", "г", "ґ", "д", "е", "є", "ж", "з", "и", "і",
+ "ї", "й", "к", "л", "м", "н", "о", "п", "р", "с", "т", "у",
+ "ф", "х", "ц", "ч", "ш", "щ", "ь", "ю", "я",
+ }
}
-mappings["uk"] = {
- ["а"] = 1, ["б"] = 2, ["в"] = 3, ["г"] = 4, ["ґ"] = 5,
- ["д"] = 6, ["е"] = 7, ["є"] = 8, ["ж"] = 9, ["з"] = 10,
- ["и"] = 11, ["і"] = 12, ["ї"] = 13, ["й"] = 14, ["к"] = 15,
- ["л"] = 16, ["м"] = 17, ["н"] = 18, ["о"] = 19, ["п"] = 20,
- ["р"] = 21, ["с"] = 22, ["т"] = 23, ["у"] = 24, ["ф"] = 25,
- ["х"] = 26, ["ц"] = 27, ["ч"] = 28, ["ш"] = 29, ["щ"] = 30,
- ["ь"] = 31, ["ю"] = 32, ["я"] = 33,
-}
-
-adduppercaseentries ("uk")
-adduppercasemappings("uk")
-
--- Belarusian
-replacements["be"] = { --[[ None, do you miss any? ]] }
-
-entries["be"] = {
- ["а"] = "а", ["б"] = "б", ["в"] = "в", ["г"] = "г", ["д"] = "д",
- ["е"] = "е", ["ё"] = "е", ["ж"] = "ж", ["з"] = "з", ["і"] = "і",
- ["й"] = "й", ["к"] = "к", ["л"] = "л", ["м"] = "м", ["н"] = "н",
- ["о"] = "о", ["п"] = "п", ["р"] = "р", ["с"] = "с", ["т"] = "т",
- ["у"] = "у", ["ў"] = "ў", ["ф"] = "ф", ["х"] = "х", ["ц"] = "ц",
- ["ч"] = "ч", ["ш"] = "ш", ["ы"] = "ы", ["ь"] = "ь", ["э"] = "э",
- ["ю"] = "ю", ["я"] = "я",
+definitions["be"] = {
+ entries = {
+ ["а"] = "а", ["б"] = "б", ["в"] = "в", ["г"] = "г", ["д"] = "д",
+ ["е"] = "е", ["ё"] = "е", ["ж"] = "ж", ["з"] = "з", ["і"] = "і",
+ ["й"] = "й", ["к"] = "к", ["л"] = "л", ["м"] = "м", ["н"] = "н",
+ ["о"] = "о", ["п"] = "п", ["р"] = "р", ["с"] = "с", ["т"] = "т",
+ ["у"] = "у", ["ў"] = "ў", ["ф"] = "ф", ["х"] = "х", ["ц"] = "ц",
+ ["ч"] = "ч", ["ш"] = "ш", ["ы"] = "ы", ["ь"] = "ь", ["э"] = "э",
+ ["ю"] = "ю", ["я"] = "я",
+ },
+ orders = {
+ "а", "б", "в", "г", "д", "е", "ё", "ж", "з", "і",
+ "й", "к", "л", "м", "н", "о", "п", "р", "с", "т",
+ "у", "ў", "ф", "х", "ц", "ч", "ш", "ы", "ь", "э",
+ "ю", "я",
+ }
}
-mappings["be"] = {
- ["а"] = 1, ["б"] = 2, ["в"] = 3, ["г"] = 4, ["д"] = 5,
- ["е"] = 6, ["ё"] = 6, ["ж"] = 7, ["з"] = 8, ["і"] = 9,
- ["й"] = 10, ["к"] = 11, ["л"] = 12, ["м"] = 13, ["н"] = 14,
- ["о"] = 15, ["п"] = 16, ["р"] = 17, ["с"] = 18, ["т"] = 19,
- ["у"] = 20, ["ў"] = 21, ["ф"] = 22, ["х"] = 23, ["ц"] = 24,
- ["ч"] = 25, ["ш"] = 26, ["ы"] = 27, ["ь"] = 28, ["э"] = 29,
- ["ю"] = 30, ["я"] = 31,
-}
-
-adduppercaseentries ("be")
-adduppercasemappings("be")
-
--- Bulgarian
-replacements["bg"] = { --[[ None, do you miss any? ]] }
-
-entries["bg"] = {
- ["а"] = "а",
- ["б"] = "б",
- ["в"] = "в",
- ["г"] = "г",
- ["д"] = "д",
- ["е"] = "е",
- ["ж"] = "ж",
- ["з"] = "з",
- ["и"] = "и",
- ["й"] = "й",
- ["к"] = "к",
- ["a"] = "a",
- ["л"] = "л",
- ["a"] = "a",
- ["м"] = "м",
- ["н"] = "н",
- ["о"] = "о",
- ["п"] = "п",
- ["р"] = "р",
- ["с"] = "с",
- ["т"] = "т",
- ["у"] = "у",
- ["ф"] = "ф",
- ["х"] = "х",
- ["ц"] = "ц",
- ["ч"] = "ч",
- ["ш"] = "ш",
- ["щ"] = "щ",
- ["ъ"] = "ъ",
- ["ь"] = "ь",
- ["ю"] = "ю",
- ["я"] = "я",
-}
-
-mappings["bg"] = {
- ["а"] = 1,
- ["б"] = 2,
- ["в"] = 3,
- ["г"] = 4,
- ["д"] = 5,
- ["е"] = 6,
- ["ж"] = 7,
- ["з"] = 8,
- ["и"] = 9,
- ["й"] = 10,
- ["к"] = 11,
- ["a"] = 12,
- ["л"] = 13,
- ["a"] = 14,
- ["м"] = 15,
- ["н"] = 16,
- ["о"] = 17,
- ["п"] = 18,
- ["р"] = 19,
- ["с"] = 20,
- ["т"] = 21,
- ["у"] = 22,
- ["ф"] = 23,
- ["х"] = 24,
- ["ц"] = 25,
- ["ч"] = 26,
- ["ш"] = 27,
- ["щ"] = 28,
- ["ъ"] = 29,
- ["ь"] = 30,
- ["ю"] = 31,
- ["я"] = 32,
+definitions["bg"] = {
+ entries = {
+ ["а"] = "а", ["б"] = "б", ["в"] = "в", ["г"] = "г", ["д"] = "д",
+ ["е"] = "е", ["ж"] = "ж", ["з"] = "з", ["и"] = "и", ["й"] = "й",
+ ["к"] = "к", ["a"] = "a", ["л"] = "л", ["a"] = "a", ["м"] = "м",
+ ["н"] = "н", ["о"] = "о", ["п"] = "п", ["р"] = "р", ["с"] = "с",
+ ["т"] = "т", ["у"] = "у", ["ф"] = "ф", ["х"] = "х", ["ц"] = "ц",
+ ["ч"] = "ч", ["ш"] = "ш", ["щ"] = "щ", ["ъ"] = "ъ", ["ь"] = "ь",
+ ["ю"] = "ю", ["я"] = "я",
+ },
+ orders = {
+ "а", "б", "в", "г", "д", "е", "ж", "з","и", "й",
+ "к", "a", "л", "a", "м", "н", "о", "п", "р", "с",
+ "т", "у", "ф", "х", "ц", "ч", "ш", "щ", "ъ", "ь",
+ "ю", "я",
+ }
}
-adduppercaseentries ("bg")
-adduppercasemappings("bg")
-
--- Old Church Slavonic
-- The language symbol “cu” is taken from the Wikipedia subdomain
-- cu.wikipedia.org.
-local cu_uk = uc(replacementoffset + 1)
-local cu_UK = uc(replacementoffset + 2)
-
-replacements["cu"] = {
- [1] = { "оу", cu_uk },
-}
-
-entries["cu"] = {
- ["а"] = "а",
- ["б"] = "б",
- ["в"] = "в",
- ["г"] = "г",
- ["д"] = "д",
- ["є"] = "є",
- ["ж"] = "ж",
- ["ѕ"] = "ѕ",
- ["ꙃ"] = "ѕ", -- Dzělo, U+0292, alternative: dz U+01f3
- ["з"] = "з",
- ["ꙁ"] = "з", -- Zemlja
- ["и"] = "и",
- ["і"] = "и",
- ["ї"] = "и",
- ["ћ"] = "ћ",
- ["к"] = "к",
- ["л"] = "л",
- ["м"] = "м",
- ["н"] = "н",
- ["о"] = "о",
- ["п"] = "п",
- ["р"] = "р",
- ["с"] = "с",
- ["т"] = "т",
- ["у"] = "у",
- ["ѹ"] = "у", -- U+0478 uk, horizontal ligature
- ["ꙋ"] = "у", -- U+0479 uk, vertical ligature
- [cu_uk] = "у",
- ["ф"] = "ф",
- ["х"] = "х",
- ["ѡ"] = "ѡ", --"ō"
- ["ѿ"] = "ѡ", -- U+047f \
- ["ѽ"] = "ѡ", -- U+047d > Omega variants
- ["ꙍ"] = "ѡ", -- U+064D /
- ["ц"] = "ц",
- ["ч"] = "ч",
- ["ш"] = "ш",
- ["щ"] = "щ",
- ["ъ"] = "ъ",
- ["ы"] = "ы",
- ["ꙑ"] = "ы", -- Old jery (U+a651) as used e.g. by the OCS Wikipedia.
- ["ь"] = "ь",
- ["ѣ"] = "ѣ",
- ["ю"] = "ю",
- ["ꙗ"] = "ꙗ", -- IOTIFIED A
- ["ѥ"] = "ѥ",
- ["ѧ"] = "ѧ",
- ["ѩ"] = "ѩ",
- ["ѫ"] = "ѫ",
- ["ѭ"] = "ѭ",
- ["ѯ"] = "ѯ",
- ["ѱ"] = "ѱ",
- ["ѳ"] = "ѳ",
- ["ѵ"] = "ѵ",
- ["ѷ"] = "ѵ", -- Why does this even have its own codepoint????
-}
-
-mappings["cu"] = {
- ["а"] = 1,
- ["б"] = 2,
- ["в"] = 3,
- ["г"] = 4,
- ["д"] = 5,
- ["є"] = 6,
- ["ж"] = 7,
- ["ѕ"] = 8,
- ["ꙃ"] = 8, -- Dzělo, U+0292, alternative: dz U+01f3
- ["з"] = 9,
- ["ꙁ"] = 9, -- Zemlja
- ["и"] = 10,
- ["і"] = 10,
- ["ї"] = 10,
- ["ћ"] = 11,
- ["к"] = 12,
- ["л"] = 13,
- ["м"] = 14,
- ["н"] = 15,
- ["о"] = 16,
- ["п"] = 17,
- ["р"] = 18,
- ["с"] = 19,
- ["т"] = 20,
- ["у"] = 21,
- ["ѹ"] = 21, -- U+0478 uk, horizontal ligature
- ["ꙋ"] = 21, -- U+0479 uk, vertical ligature
- [cu_uk] = 21,
- ["ф"] = 22,
- ["х"] = 23,
- ["ѡ"] = 24, --"ō"
- ["ѿ"] = 24, -- U+047f \
- ["ѽ"] = 24, -- U+047d > Omega variants
- ["ꙍ"] = 24, -- U+064D /
- ["ц"] = 25,
- ["ч"] = 26,
- ["ш"] = 27,
- ["щ"] = 28,
- ["ъ"] = 29,
- ["ы"] = 30,
- ["ꙑ"] = 30, -- Old jery (U+a651) as used e.g. by the OCS Wikipedia.
- ["ь"] = 31,
- ["ѣ"] = 32,
- ["ю"] = 33,
- ["ꙗ"] = 34, -- IOTIFIED A
- ["ѥ"] = 35,
- ["ѧ"] = 36,
- ["ѩ"] = 37,
- ["ѫ"] = 38,
- ["ѭ"] = 39,
- ["ѯ"] = 40,
- ["ѱ"] = 41,
- ["ѳ"] = 42,
- ["ѵ"] = 43,
- ["ѷ"] = 43, -- Why does this even have its own codepoint????
-}
-
-adduppercaseentries ("cu")
-adduppercasemappings("cu")
-
-entries ["cu"] [cu_UK] = entries ["cu"] [cu_uk]
-mappings["cu"] [cu_UK] = mappings["cu"] [cu_uk]
-
---- Polish (including the letters q, v, x)
-
--- Cf. ftp://ftp.gust.org.pl/pub/GUST/bulletin/03/02-bl.pdf.
-
-replacements["pl"] = {
- -- none
-}
-
-entries["pl"] = {
- ["a"] = "a", ["ą"] = "ą", ["b"] = "b", ["c"] = "c", ["ć"] = "ć",
- ["d"] = "d", ["e"] = "e", ["ę"] = "ę", ["f"] = "f", ["g"] = "g",
- ["h"] = "h", ["i"] = "i", ["j"] = "j", ["k"] = "k", ["l"] = "l",
- ["ł"] = "ł", ["m"] = "m", ["n"] = "n", ["ń"] = "ń", ["o"] = "o",
- ["ó"] = "ó", ["p"] = "p", ["q"] = "q", ["r"] = "r", ["s"] = "s",
- ["ś"] = "ś", ["t"] = "t", ["u"] = "u", ["v"] = "v", ["w"] = "w",
- ["x"] = "x", ["y"] = "y", ["z"] = "z", ["ź"] = "ź", ["ż"] = "ż",
-}
-
-mappings["pl"] = {
- ["a"] = 1, ["ą"] = 2, ["b"] = 3, ["c"] = 4, ["ć"] = 5,
- ["d"] = 6, ["e"] = 7, ["ę"] = 8, ["f"] = 9, ["g"] = 10,
- ["h"] = 11, ["i"] = 12, ["j"] = 13, ["k"] = 14, ["l"] = 15,
- ["ł"] = 16, ["m"] = 17, ["n"] = 18, ["ń"] = 19, ["o"] = 20,
- ["ó"] = 21, ["p"] = 22, ["q"] = 23, ["r"] = 24, ["s"] = 25,
- ["ś"] = 26, ["t"] = 27, ["u"] = 28, ["v"] = 29, ["w"] = 30,
- ["x"] = 31, ["y"] = 32, ["z"] = 33, ["ź"] = 34, ["ż"] = 35,
-}
-
-adduppercaseentries ("pl")
-adduppercasemappings("pl")
-
---- Czech
--- Modified to treat quantities and other secondary characteristics indifferently.
--- Cf. <http://racek.vlada.cz/usneseni/usneseni_webtest.nsf/WebGovRes/0AD8FEF4CC04B7A4C12571B6006D69D0?OpenDocument>
--- (2.4.3; via <http://cs.wikipedia.org/wiki/Abecední_řazení#.C4.8Ce.C5.A1tina>).
-
-local cz_ch = uc(replacementoffset + 1)
-local cz_CH = uc(replacementoffset + 2) -- Is this actually used somewhere (e.g. with “adduppercaseentries”)?
-
-replacements["cz"] = {
- [1] = { "ch", cz_ch }
-}
-
-entries["cz"] = {
- ["a"] = "a", -- a
- ["á"] = "a", -- aacute
- ["b"] = "b", -- b
- ["c"] = "c", -- c
- ["č"] = "č", -- ccaron
- ["d"] = "d", -- d
- ["ď"] = "d", -- dcaron
- ["e"] = "e", -- e
- ["é"] = "e", -- eacute
- ["ě"] = "e", -- ecaron
- ["f"] = "f", -- f
- ["g"] = "g", -- g
- ["h"] = "h", -- h
- [cz_ch] = "ch", -- ch
- ["i"] = "i", -- i
- ["í"] = "i", -- iacute
- ["j"] = "j", -- j
- ["k"] = "k", -- k
- ["l"] = "l", -- l
- ["m"] = "m", -- m
- ["n"] = "n", -- n
- ["ň"] = "n", -- ncaron
- ["o"] = "o", -- o
- ["ó"] = "o", -- ó
- ["p"] = "p", -- p
- ["q"] = "q", -- q
- ["r"] = "r", -- r
- ["ř"] = "ř", -- rcaron
- ["s"] = "s", -- s
- ["š"] = "š", -- scaron
- ["t"] = "t", -- t
- ["ť"] = "t", -- tcaron
- ["u"] = "u", -- u
- ["ú"] = "u", -- uacute
- ["ů"] = "u", -- uring
- ["v"] = "v", -- v
- ["w"] = "w", -- w
- ["x"] = "x", -- x
- ["y"] = "y", -- y
- ["ý"] = "y", -- yacute
- ["z"] = "z", -- z
- ["ž"] = "ž", -- zcaron
-}
-
-mappings["cz"] = {
- ["a"] = 1, -- a
- ["á"] = 1, -- aacute -> a
- ["b"] = 2, -- b
- ["c"] = 3, -- c
- ["č"] = 4, -- ccaron
- ["d"] = 5, -- d
- ["ď"] = 5, -- dcaron -> ď
- ["e"] = 6, -- e
- ["é"] = 6, -- eacute -> e
- ["ě"] = 6, -- ecaron -> e
- ["f"] = 7, -- f
- ["g"] = 8, -- g
- ["h"] = 9, -- h
- [cz_ch] = 10, -- ch
- ["i"] = 11, -- i
- ["í"] = 11, -- iacute -> i
- ["j"] = 12, -- j
- ["k"] = 13, -- k
- ["l"] = 14, -- l
- ["m"] = 15, -- m
- ["n"] = 16, -- n
- ["ň"] = 16, -- ncaron -> n
- ["o"] = 17, -- o
- ["ó"] = 17, -- o -> o
- ["p"] = 18, -- p
- ["q"] = 19, -- q
- ["r"] = 20, -- r
- ["ř"] = 21, -- rcaron
- ["s"] = 22, -- s
- ["š"] = 23, -- scaron
- ["t"] = 24, -- t
- ["ť"] = 24, -- tcaron -> t
- ["u"] = 25, -- u
- ["ú"] = 25, -- uacute -> u
- ["ů"] = 25, -- uring -> u
- ["v"] = 26, -- v
- ["w"] = 27, -- w
- ["x"] = 28, -- x
- ["y"] = 29, -- y
- ["ý"] = 29, -- yacute -> y
- ["z"] = 30, -- z
- ["ž"] = 31, -- zcaron Checksum: 42
-}
-
-adduppercaseentries ("cz")
-adduppercasemappings("cz") -- 1 can be option (but then we need a runtime variant)
-
-entries ["cz"][cz_CH] = entries ["cz"][cz_ch]
-mappings["cz"][cz_CH] = mappings["cz"][cz_ch]
-
-replacements["cs"] = replacements["cz"]
-entries ["cs"] = entries ["cz"]
-mappings ["cs"] = mappings ["cz"]
+local uk, UK = utfchar(replacementoffset + 1), utfchar(replacementoffset + 11)
+
+definitions["cu"] = {
+ replacements = {
+ { "оу", uk }, { "ОУ", UK },
+ },
+ entries = {
+ ["а"] = "а", ["б"] = "б", ["в"] = "в", ["г"] = "г", ["д"] = "д",
+ ["є"] = "є", ["ж"] = "ж", ["ѕ"] = "ѕ", ["ꙃ"] = "ѕ", ["з"] = "з",
+ ["ꙁ"] = "з", ["и"] = "и", ["і"] = "и", ["ї"] = "и", ["ћ"] = "ћ",
+ ["к"] = "к", ["л"] = "л", ["м"] = "м", ["н"] = "н", ["о"] = "о",
+ ["п"] = "п", ["р"] = "р", ["с"] = "с", ["т"] = "т", ["у"] = "у",
+ ["ѹ"] = "у", ["ꙋ"] = "у", [uk] = "у", ["ф"] = "ф", ["х"] = "х",
+ ["ѡ"] = "ѡ", ["ѿ"] = "ѡ", ["ѽ"] = "ѡ", ["ꙍ"] = "ѡ", ["ц"] = "ц",
+ ["ч"] = "ч", ["ш"] = "ш", ["щ"] = "щ", ["ъ"] = "ъ", ["ы"] = "ы",
+ ["ꙑ"] = "ы", ["ь"] = "ь", ["ѣ"] = "ѣ", ["ю"] = "ю", ["ꙗ"] = "ꙗ",
+ ["ѥ"] = "ѥ", ["ѧ"] = "ѧ", ["ѩ"] = "ѩ", ["ѫ"] = "ѫ", ["ѭ"] = "ѭ",
+ ["ѯ"] = "ѯ", ["ѱ"] = "ѱ", ["ѳ"] = "ѳ", ["ѵ"] = "ѵ", ["ѷ"] = "ѵ",
+ },
+ orders = {
+ "а", "б", "в", "г", "д", "є", "ж", "ѕ", "ꙃ", "з", -- Dzělo, U+0292, alternative: dz U+01f3
+ "ꙁ", "и", "і", "ї", "ћ", "к", "л", "м", "н", "о", -- Zemlja
+ "п", "р", "с", "т", "у", "ѹ", "ꙋ", uk, "ф", "х", -- U+0478 uk, horizontal ligature, U+0479 uk, vertical ligature
+ "ѡ", "ѿ", "ѽ", "ꙍ", "ц", "ч", "ш", "щ", "ъ", "ы", -- "ō", U+047f \, U+047d > Omega variants, U+064D /
+ "ꙑ", "ь", "ѣ", "ю", "ꙗ", "ѥ", "ѧ", "ѩ", "ѫ", "ѭ", -- Old jery (U+a651) as used e.g. by the OCS Wikipedia. IOTIFIED A
+ "ѯ", "ѱ", "ѳ", "ѵ", "ѷ",
+ },
+ upper = {
+ uk = UK,
+ },
+ lower = {
+ UK = uk,
+ }
+}
+
+--- Polish (including the letters q, v, x) Cf. ftp://ftp.gust.org.pl/pub/GUST/bulletin/03/02-bl.pdf.
+
+definitions["pl"] = {
+ entries = {
+ ["a"] = "a", ["ą"] = "ą", ["b"] = "b", ["c"] = "c", ["ć"] = "ć",
+ ["d"] = "d", ["e"] = "e", ["ę"] = "ę", ["f"] = "f", ["g"] = "g",
+ ["h"] = "h", ["i"] = "i", ["j"] = "j", ["k"] = "k", ["l"] = "l",
+ ["ł"] = "ł", ["m"] = "m", ["n"] = "n", ["ń"] = "ń", ["o"] = "o",
+ ["ó"] = "ó", ["p"] = "p", ["q"] = "q", ["r"] = "r", ["s"] = "s",
+ ["ś"] = "ś", ["t"] = "t", ["u"] = "u", ["v"] = "v", ["w"] = "w",
+ ["x"] = "x", ["y"] = "y", ["z"] = "z", ["ź"] = "ź", ["ż"] = "ż",
+ },
+ orders = {
+ "a", "ą", "b", "c", "ć", "d", "e", "ę", "f", "g",
+ "h", "i", "j", "k", "l", "ł", "m", "n", "ń", "o",
+ "ó", "p", "q", "r", "s", "ś", "t", "u", "v", "w",
+ "x", "y", "z", "ź", "ż",
+ },
+}
+
+-- Czech, modified to treat quantities and other secondary characteristics indifferently. Cf.
+-- http://racek.vlada.cz/usneseni/usneseni_webtest.nsf/WebGovRes/0AD8FEF4CC04B7A4C12571B6006D69D0?OpenDocument
+-- (2.4.3; via <http://cs.wikipedia.org/wiki/Abecední_řazení#.C4.8Ce.C5.A1tina>)
+
+local ch, CH = utfchar(replacementoffset + 1), utfchar(replacementoffset + 11)
+
+definitions["cz"] = {
+ replacements = {
+ { "ch", ch }, { "CH", CH }
+ },
+ entries = {
+ ["a"] = "a", ["á"] = "a", ["b"] = "b", ["c"] = "c", ["č"] = "č",
+ ["d"] = "d", ["ď"] = "d", ["e"] = "e", ["é"] = "e", ["ě"] = "e",
+ ["f"] = "f", ["g"] = "g", ["h"] = "h", [ch] = "ch", ["i"] = "i",
+ ["í"] = "i", ["j"] = "j", ["k"] = "k", ["l"] = "l", ["m"] = "m",
+ ["n"] = "n", ["ň"] = "n", ["o"] = "o", ["ó"] = "o", ["p"] = "p",
+ ["q"] = "q", ["r"] = "r", ["ř"] = "ř", ["s"] = "s", ["š"] = "š",
+ ["t"] = "t", ["ť"] = "t", ["u"] = "u", ["ú"] = "u", ["ů"] = "u",
+ ["v"] = "v", ["w"] = "w", ["x"] = "x", ["y"] = "y", ["ý"] = "y",
+ ["z"] = "z", ["ž"] = "ž",
+ },
+ orders = {
+ "a", "á", "b", "c", "č", "d", "ď", "e", "é", "ě",
+ "f", "g", "h", ch, "i", "í", "j", "k", "l", "m",
+ "n", "ň", "o", "ó", "p", "q", "r", "ř", "s", "š",
+ "t", "ť", "u", "ú", "ů", "v", "w", "x", "y", "ý",
+ "z", "ž",
+ },
+ upper = {
+ ch = CH,
+ },
+ lower = {
+ CH = ch,
+ }
+}
+
+definitions["cs"] = { parent = "cz" }
--- Slovak.
@@ -767,1695 +328,461 @@ mappings ["cs"] = mappings ["cz"]
-- to affect collation order, then use the values given in the comments. We could
-- define an additional vector for that.
-local sk_dz = uc(replacementoffset + 1)
-local sk_DZ = uc(replacementoffset + 2)
-local sk_dzh = uc(replacementoffset + 3)
-local sk_DZH = uc(replacementoffset + 4)
-local sk_ch = uc(replacementoffset + 5)
-local sk_CH = uc(replacementoffset + 6)
-
-replacements["sk"] = {
- [1] = { "dz", sk_dz },
- [2] = { "dž", sk_dzh },
- [3] = { "ch", sk_ch },
+local dz, DZ = utfchar(replacementoffset + 1), utfchar(replacementoffset + 11)
+local dzh, DZH = utfchar(replacementoffset + 2), utfchar(replacementoffset + 12)
+local ch, CH = utfchar(replacementoffset + 3), utfchar(replacementoffset + 13)
+
+definitions["sk"] = {
+ replacements = {
+ { "dz", dz }, { "dz", DZ },
+ { "dž", dzh }, { "dž", DZH },
+ { "ch", ch }, { "ch", CH },
+ },
+ entries = {
+ ["a"] = "a", ["á"] = "a", ["ä"] = "a", ["b"] = "b", ["c"] = "c",
+ ["č"] = "č", ["d"] = "d", ["ď"] = "d", [dz] = "dz", [dzh] = "dž",
+ ["e"] = "e", ["é"] = "e", ["f"] = "f", ["g"] = "g", ["h"] = "h",
+ [ch] = "ch", ["i"] = "i", ["í"] = "i", ["j"] = "j", ["k"] = "k",
+ ["l"] = "l", ["ĺ"] = "l", ["ľ"] = "l", ["m"] = "m", ["n"] = "n",
+ ["ň"] = "n", ["o"] = "o", ["ó"] = "o", ["ô"] = "o", ["p"] = "p",
+ ["q"] = "q", ["r"] = "r", ["ŕ"] = "r", ["s"] = "s", ["š"] = "š",
+ ["t"] = "t", ["ť"] = "t", ["u"] = "u", ["ú"] = "u", ["v"] = "v",
+ ["w"] = "w", ["x"] = "x", ["y"] = "y", ["ý"] = "y", ["z"] = "z",
+ ["ž"] = "ž",
+ },
+ orders = {
+ "a", "á", "ä", "b", "c", "č", "d", "ď", dz, dzh,
+ "e", "é", "f", "g", "h", ch, "i", "í", "j", "k",
+ "l", "ĺ", "ľ", "m", "n", "ň", "o", "ó", "ô", "p",
+ "q", "r", "ŕ", "s", "š", "t", "ť", "u", "ú", "v",
+ "w", "x", "y", "ý", "z", "ž",
+ },
+ upper = {
+ dz = DZ, dzh = DZH, ch = CH,
+ },
+ lower = {
+ DZ = dz, DZH = dzh, CH = ch,
+ }
}
-entries["sk"] = {
- ["a"] = "a",
- ["á"] = "a", -- "á",
- ["ä"] = "a", -- "ä",
- ["b"] = "b",
- ["c"] = "c",
- ["č"] = "č",
- ["d"] = "d",
- ["ď"] = "d", -- "ď",
- [sk_dz] = "dz",
- [sk_dzh] = "dž",
- ["e"] = "e",
- ["é"] = "e", -- "é",
- ["f"] = "f",
- ["g"] = "g",
- ["h"] = "h",
- [sk_ch] = "ch",
- ["i"] = "i",
- ["í"] = "i", -- "í",
- ["j"] = "j",
- ["k"] = "k",
- ["l"] = "l",
- ["ĺ"] = "l", -- "ĺ",
- ["ľ"] = "l", -- "ľ",
- ["m"] = "m",
- ["n"] = "n",
- ["ň"] = "n", -- "ň",
- ["o"] = "o",
- ["ó"] = "o", -- "ó",
- ["ô"] = "o", -- "ô",
- ["p"] = "p",
- ["q"] = "q",
- ["r"] = "r",
- ["ŕ"] = "r", -- "ŕ",
- ["s"] = "s",
- ["š"] = "š",
- ["t"] = "t",
- ["ť"] = "t", -- "ť",
- ["u"] = "u",
- ["ú"] = "u", -- "ú",
- ["v"] = "v",
- ["w"] = "w",
- ["x"] = "x",
- ["y"] = "y",
- ["ý"] = "y", -- "ý",
- ["z"] = "z",
- ["ž"] = "ž",
-}
-
-mappings["sk"] = {
- ["a"] = 1,
- ["á"] = 1, -- 2,
- ["ä"] = 1, -- 3,
- ["b"] = 4,
- ["c"] = 5,
- ["č"] = 6,
- ["d"] = 7,
- ["ď"] = 7, -- 8,
- [sk_dz] = 9,
- [sk_dzh] = 10,
- ["e"] = 11,
- ["é"] = 11, -- 12,
- ["f"] = 13,
- ["g"] = 14,
- ["h"] = 15,
- [sk_ch] = 16,
- ["i"] = 17,
- ["í"] = 17, -- 18,
- ["j"] = 19,
- ["k"] = 20,
- ["l"] = 21,
- ["ĺ"] = 21, -- 22,
- ["ľ"] = 21, -- 23,
- ["m"] = 24,
- ["n"] = 25,
- ["ň"] = 25, -- 26,
- ["o"] = 27,
- ["ó"] = 27, -- 28,
- ["ô"] = 27, -- 29,
- ["p"] = 30,
- ["q"] = 31,
- ["r"] = 32,
- ["ŕ"] = 32, -- 33,
- ["s"] = 34,
- ["š"] = 35,
- ["t"] = 36,
- ["ť"] = 36, -- 37,
- ["u"] = 38,
- ["ú"] = 38, -- 39,
- ["v"] = 40,
- ["w"] = 41,
- ["x"] = 42,
- ["y"] = 43,
- ["ý"] = 43, -- 44,
- ["z"] = 45,
- ["ž"] = 46, -- Checksum: 46, přesně!
-}
-
-adduppercaseentries ("sk")
-adduppercasemappings("sk")
-
-entries ["sk"] [sk_DZ] = entries ["sk"] [sk_dz]
-mappings["sk"] [sk_DZ] = mappings["sk"] [sk_dz]
-entries ["sk"][sk_DZH] = entries ["sk"][sk_dzh]
-mappings["sk"][sk_DZH] = mappings["sk"][sk_dzh]
-entries ["sk"] [sk_CH] = entries ["sk"] [sk_ch]
-mappings["sk"] [sk_CH] = mappings["sk"] [sk_ch]
-
--- Croatian
-local hr_dzh = uc(replacementoffset + 1)
-local hr_DZH = uc(replacementoffset + 2)
-local hr_lj = uc(replacementoffset + 3)
-local hr_LJ = uc(replacementoffset + 4)
-local hr_nj = uc(replacementoffset + 5)
-local hr_NJ = uc(replacementoffset + 6)
-
-replacements["hr"] = {
- [1] = { "dž", hr_dzh },
- [2] = { "lj", hr_lj },
- [3] = { "nj", hr_nj },
-}
-
-entries["hr"] = {
- ["a"] = "a", -- Why do you sometimes encounter “â” (where Old Slavonic
- ["b"] = "b", -- has “ѣ”) and how does it collate?
- ["c"] = "c",
- ["č"] = "č",
- ["ć"] = "ć",
- ["d"] = "d",
- [hr_dzh] = "dž",
- ["đ"] = "đ",
- ["e"] = "e",
- ["f"] = "f",
- ["g"] = "g",
- ["h"] = "h",
- ["i"] = "i",
- ["j"] = "j",
- ["k"] = "k",
- ["l"] = "l",
- [hr_lj] = "lj",
- ["m"] = "m",
- ["n"] = "n",
- [hr_nj] = "nj",
- ["o"] = "o",
- ["p"] = "p",
- ["r"] = "r",
- ["s"] = "s",
- ["š"] = "š",
- ["t"] = "t",
- ["u"] = "u",
- ["v"] = "v",
- ["z"] = "z",
- ["ž"] = "ž",
-}
-
-mappings["hr"] = {
- ["a"] = 1,
- ["b"] = 2,
- ["c"] = 3,
- ["č"] = 4,
- ["ć"] = 5,
- ["d"] = 6,
- [hr_dzh] = 7,
- ["đ"] = 8,
- ["e"] = 9,
- ["f"] = 10,
- ["g"] = 11,
- ["h"] = 12,
- ["i"] = 13,
- ["j"] = 14,
- ["k"] = 15,
- ["l"] = 16,
- [hr_lj] = 17,
- ["m"] = 18,
- ["n"] = 19,
- [hr_nj] = 20,
- ["o"] = 21,
- ["p"] = 22,
- ["r"] = 23,
- ["s"] = 24,
- ["š"] = 25,
- ["t"] = 26,
- ["u"] = 27,
- ["v"] = 28,
- ["z"] = 29,
- ["ž"] = 30,
+local dzh, DZH = utfchar(replacementoffset + 1), utfchar(replacementoffset + 11)
+local lj, LJ = utfchar(replacementoffset + 2), utfchar(replacementoffset + 12)
+local nj, NJ = utfchar(replacementoffset + 3), utfchar(replacementoffset + 13)
+
+definitions["hr"] = {
+ replacements = {
+ { "dž", dzh }, { "DŽ", DZH },
+ { "lj", lj }, { "LJ", LJ },
+ { "nj", nj }, { "NJ", NJ },
+ },
+ entries = {
+ ["a"] = "a", ["b"] = "b", ["c"] = "c", ["č"] = "č", ["ć"] = "ć",
+ ["d"] = "d", [dzh] = "dž", ["đ"] = "đ", ["e"] = "e", ["f"] = "f",
+ ["g"] = "g", ["h"] = "h", ["i"] = "i", ["j"] = "j", ["k"] = "k",
+ ["l"] = "l", [lj] = "lj", ["m"] = "m", ["n"] = "n", [nj] = "nj",
+ ["o"] = "o", ["p"] = "p", ["r"] = "r", ["s"] = "s", ["š"] = "š",
+ ["t"] = "t", ["u"] = "u", ["v"] = "v", ["z"] = "z", ["ž"] = "ž",
+ },
+ orders = {
+ "a", "b", "c", "č", "ć", "d", dzh, "đ", "e", "f",
+ "g", "h", "i", "j", "k", "l", lj, "m", "n", nj,
+ "o", "p", "r", "s", "š", "t", "u", "v", "z", "ž",
+ },
+ upper = {
+ dzh = DZH, lj = LJ, nj = NJ,
+ },
+ lower = {
+ DZH = dzh, LJ = lj, NJ = nj,
+ }
}
-adduppercaseentries ("hr")
-adduppercasemappings("hr")
-
-entries ["hr"][hr_DZH] = entries ["hr"][hr_dzh]
-mappings["hr"][hr_DZH] = mappings["hr"][hr_dzh]
-entries ["hr"] [hr_LJ] = entries ["hr"] [hr_lj]
-mappings["hr"] [hr_LJ] = mappings["hr"] [hr_lj]
-entries ["hr"] [hr_NJ] = entries ["hr"] [hr_nj]
-mappings["hr"] [hr_NJ] = mappings["hr"] [hr_nj]
--- Serbian
-replacements["sr"] = {
- -- none
-}
-
-entries["sr"] = {
- ["а"] = "а",
- ["б"] = "б",
- ["в"] = "в",
- ["г"] = "г",
- ["д"] = "д",
- ["ђ"] = "ђ",
- ["е"] = "е",
- ["ж"] = "ж",
- ["з"] = "з",
- ["и"] = "и",
- ["ј"] = "ј",
- ["к"] = "к",
- ["л"] = "л",
- ["љ"] = "љ",
- ["м"] = "м",
- ["н"] = "н",
- ["њ"] = "њ",
- ["о"] = "о",
- ["п"] = "п",
- ["р"] = "р",
- ["с"] = "с",
- ["т"] = "т",
- ["ћ"] = "ћ",
- ["у"] = "у",
- ["ф"] = "ф",
- ["х"] = "х",
- ["ц"] = "ц",
- ["ч"] = "ч",
- ["џ"] = "џ",
- ["ш"] = "ш",
+definitions["sr"] = {
+ entries = {
+ ["а"] = "а", ["б"] = "б", ["в"] = "в", ["г"] = "г", ["д"] = "д",
+ ["ђ"] = "ђ", ["е"] = "е", ["ж"] = "ж", ["з"] = "з", ["и"] = "и",
+ ["ј"] = "ј", ["к"] = "к", ["л"] = "л", ["љ"] = "љ", ["м"] = "м",
+ ["н"] = "н", ["њ"] = "њ", ["о"] = "о", ["п"] = "п", ["р"] = "р",
+ ["с"] = "с", ["т"] = "т", ["ћ"] = "ћ", ["у"] = "у", ["ф"] = "ф",
+ ["х"] = "х", ["ц"] = "ц", ["ч"] = "ч", ["џ"] = "џ",
+ ["ш"] = "ш",
+ },
+ orders = {
+ "а", "б", "в", "г", "д", "ђ", "е", "ж", "з", "и",
+ "ј", "к", "л", "љ", "м", "н", "њ", "о", "п", "р",
+ "с", "т", "ћ", "у", "ф", "х", "ц", "ч", "џ", "ш",
+ }
}
-mappings["sr"] = {
- ["а"] = 1,
- ["б"] = 2,
- ["в"] = 3,
- ["г"] = 4,
- ["д"] = 5,
- ["ђ"] = 6,
- ["е"] = 7,
- ["ж"] = 8,
- ["з"] = 9,
- ["и"] = 10,
- ["ј"] = 11,
- ["к"] = 12,
- ["л"] = 13,
- ["љ"] = 14,
- ["м"] = 15,
- ["н"] = 16,
- ["њ"] = 17,
- ["о"] = 18,
- ["п"] = 19,
- ["р"] = 20,
- ["с"] = 21,
- ["т"] = 22,
- ["ћ"] = 23,
- ["у"] = 24,
- ["ф"] = 25,
- ["х"] = 26,
- ["ц"] = 27,
- ["ч"] = 28,
- ["џ"] = 29,
- ["ш"] = 30,
-}
-
-adduppercaseentries ("sr")
-adduppercasemappings("sr")
-
--- Transliteration: Russian|ISO9-1995
-- Keeping the same collation order as Russian (v.s.).
-- Matches the tables from:
-- http://bitbucket.org/phg/transliterator/src/tip/tex/context/third/transliterator/trans_tables_iso9.lua
-local ru_iso9_yer = uc(replacementoffset + 1)
-
-replacements["ru-iso9"] = {
- [1] = { "''", ru_iso9_yer },
-}
-
-entries["ru-iso9"] = {
- ["a"] = "a",
- ["b"] = "b",
- ["v"] = "v",
- ["g"] = "g",
- ["d"] = "d",
- ["e"] = "e",
- ["ë"] = "ë",
- ["ž"] = "ž",
- ["z"] = "z",
- ["i"] = "i",
- ["ì"] = "ì",
- ["j"] = "j",
- ["k"] = "k",
- ["l"] = "l",
- ["m"] = "m",
- ["n"] = "n",
- ["o"] = "o",
- ["p"] = "p",
- ["r"] = "r",
- ["s"] = "s",
- ["t"] = "t",
- ["u"] = "u",
- ["f"] = "f",
- ["h"] = "h",
- ["c"] = "c",
- ["č"] = "č",
- ["š"] = "š",
- ["ŝ"] = "ŝ",
- ["ʺ"] = "ʺ",
- [ru_iso9_yer] = "ʺ",
- ["y"] = "y",
- ["ʹ"] = "ʹ",
- ["'"] = "ʹ",
- ["ě"] = "ě",
- ["è"] = "è",
- ["û"] = "û",
- ["â"] = "â",
- ["û"] = "û",
- ["â"] = "â",
-}
-
-mappings["ru-iso9"] = {
- ["a"] = 1,
- ["b"] = 2,
- ["v"] = 3,
- ["g"] = 4,
- ["d"] = 5,
- ["e"] = 6,
- ["ë"] = 6,
- ["ž"] = 7,
- ["z"] = 8,
- ["i"] = 9,
- ["ì"] = 9,
- ["j"] = 10,
- ["k"] = 11,
- ["l"] = 12,
- ["m"] = 13,
- ["n"] = 14,
- ["o"] = 15,
- ["p"] = 16,
- ["r"] = 17,
- ["s"] = 18,
- ["t"] = 19,
- ["u"] = 20,
- ["f"] = 21,
- ["h"] = 22,
- ["c"] = 23,
- ["č"] = 24,
- ["š"] = 25,
- ["ŝ"] = 26,
- ["ʺ"] = 27,
- [ru_iso9_yer] = 27,
- ["y"] = 28,
- ["ʹ"] = 29,
- ["'"] = 29,
- ["ě"] = 30,
- ["è"] = 31,
- ["û"] = 32,
- ["â"] = 33,
- ["û"] = 34,
- ["â"] = 35,
+local yer = utfchar(replacementoffset + 1)
+
+definitions["ru-iso9"] = {
+ replacements = {
+ { "''", yer },
+ },
+ entries = {
+ ["a"] = "a", ["b"] = "b", ["v"] = "v", ["g"] = "g", ["d"] = "d",
+ ["e"] = "e", ["ë"] = "ë", ["ž"] = "ž", ["z"] = "z", ["i"] = "i",
+ ["ì"] = "ì", ["j"] = "j", ["k"] = "k", ["l"] = "l", ["m"] = "m",
+ ["n"] = "n", ["o"] = "o", ["p"] = "p", ["r"] = "r", ["s"] = "s",
+ ["t"] = "t", ["u"] = "u", ["f"] = "f", ["h"] = "h", ["c"] = "c",
+ ["č"] = "č", ["š"] = "š", ["ŝ"] = "ŝ", ["ʺ"] = "ʺ", [yer] = "ʺ",
+ ["y"] = "y", ["ʹ"] = "ʹ", ["'"] = "ʹ", ["ě"] = "ě", ["è"] = "è",
+ ["û"] = "û", ["â"] = "â", ["û"] = "û", ["â"] = "â",
+ },
+ orders = {
+ "a", "b", "v", "g", "d", "e", "ë", "ž", "z", "i",
+ "ì", "j", "k", "l", "m", "n", "o", "p", "r", "s",
+ "t", "u", "f", "h", "c", "č", "š", "ŝ", "ʺ", yer,
+ "y", "ʹ", "'", "ě", "è", "û", "â", "û", "â",
+ }
}
-adduppercaseentries ("ru-iso9")
-adduppercasemappings("ru-iso9")
-
--- Transliteration: Old Slavonic|scientific
-- Matches the tables from:
-- http://bitbucket.org/phg/transliterator/src/tip/tex/context/third/transliterator/trans_tables_scntfc.lua
-local ocs_scn_uk = uc(replacementoffset + 1)
-local ocs_scn_tshe = uc(replacementoffset + 2)
-local ocs_scn_sht = uc(replacementoffset + 3)
-local ocs_scn_ju = uc(replacementoffset + 4)
-local ocs_scn_ja = uc(replacementoffset + 5)
-local ocs_scn_je = uc(replacementoffset + 6)
-local ocs_scn_ijus = uc(replacementoffset + 7)
-local ocs_scn_ibigjus = uc(replacementoffset + 8)
-local ocs_scn_xi = uc(replacementoffset + 9)
-local ocs_scn_psi = uc(replacementoffset + 10)
-local ocs_scn_theta = uc(replacementoffset + 11)
-local ocs_scn_shch = uc(replacementoffset + 12)
-
-local ocs_scn_UK = uc(replacementoffset + 13)
-local ocs_scn_TSHE = uc(replacementoffset + 14)
-local ocs_scn_SHT = uc(replacementoffset + 15)
-local ocs_scn_JU = uc(replacementoffset + 16)
-local ocs_scn_JA = uc(replacementoffset + 17)
-local ocs_scn_JE = uc(replacementoffset + 18)
-local ocs_scn_IJUS = uc(replacementoffset + 19)
-local ocs_scn_IBIGJUS = uc(replacementoffset + 20)
-local ocs_scn_XI = uc(replacementoffset + 21)
-local ocs_scn_PSI = uc(replacementoffset + 22)
-local ocs_scn_THETA = uc(replacementoffset + 23)
-local ocs_scn_SHCH = uc(replacementoffset + 24)
-
-replacements["ocs-scn"] = {
- [1] = { "ou", ocs_scn_uk },
- [2] = { "g’", ocs_scn_tshe },
- [3] = { "št", ocs_scn_sht },
- [4] = { "ju", ocs_scn_ju },
- [5] = { "ja", ocs_scn_ja },
- [6] = { "je", ocs_scn_je },
- [7] = { "ję", ocs_scn_ijus },
- [8] = { "jǫ", ocs_scn_ibigjus },
- [9] = { "ks", ocs_scn_xi },
- [10] = { "ps", ocs_scn_psi },
- [11] = { "th", ocs_scn_theta },
- [12] = { "šč", ocs_scn_shch },
-}
-
-entries["ocs-scn"] = {
- ["a"] = "a",
- ["b"] = "b",
- ["v"] = "v",
- ["g"] = "g",
- ["d"] = "d",
- ["e"] = "e",
- ["ž"] = "ž",
- ["ʒ"] = "ʒ",
- ["z"] = "z",
- ["i"] = "i",
- ["ï"] = "ï",
- [ocs_scn_tshe] = "g’",
- ["k"] = "k",
- ["l"] = "l",
- ["m"] = "m",
- ["n"] = "n",
- ["o"] = "o",
- ["p"] = "p",
- ["r"] = "r",
- ["s"] = "s",
- ["t"] = "t",
- ["u"] = "u",
- ["f"] = "f",
- ["x"] = "x",
- ["o"] = "o",
- ["c"] = "c",
- ["č"] = "č",
- ["š"] = "š",
- [ocs_scn_sht] = "št",
- [ocs_scn_shch] = "šč",
- ["ъ"] = "ъ",
- ["y"] = "y",
- [ocs_scn_uk] = "y",
- ["ь"] = "ь",
- ["ě"] = "ě",
- [ocs_scn_ju] = "ju",
- [ocs_scn_ja] = "ja",
- [ocs_scn_je] = "je",
- ["ę"] = "ę",
- [ocs_scn_ijus] = "ję",
- ["ǫ"] = "ǫ",
-[ocs_scn_ibigjus] = "jǫ",
- [ocs_scn_xi] = "ks",
- [ocs_scn_psi] = "ps",
- [ocs_scn_theta] = "th",
- ["ü"] = "ü",
+local uk, UK = utfchar(replacementoffset + 1), utfchar(replacementoffset + 21)
+local tshe, TSHE = utfchar(replacementoffset + 2), utfchar(replacementoffset + 22)
+local sht, SHT = utfchar(replacementoffset + 3), utfchar(replacementoffset + 23)
+local ju, JU = utfchar(replacementoffset + 4), utfchar(replacementoffset + 24)
+local ja, JA = utfchar(replacementoffset + 5), utfchar(replacementoffset + 25)
+local je, JE = utfchar(replacementoffset + 6), utfchar(replacementoffset + 26)
+local ijus, IJUS = utfchar(replacementoffset + 7), utfchar(replacementoffset + 27)
+local ibigjus, IBIGJUS = utfchar(replacementoffset + 8), utfchar(replacementoffset + 28)
+local xi, XI = utfchar(replacementoffset + 9), utfchar(replacementoffset + 29)
+local psi, PSI = utfchar(replacementoffset + 10), utfchar(replacementoffset + 30)
+local theta, THETA = utfchar(replacementoffset + 11), utfchar(replacementoffset + 31)
+local shch, SHCH = utfchar(replacementoffset + 12), utfchar(replacementoffset + 32)
+
+definitions["ocs-scn"] = {
+ replacements = {
+ { "ou", uk }, { "OU", UK },
+ { "g’", tshe }, { "G’", TSHE },
+ { "št", sht }, { "ŠT", SHT },
+ { "ju", ju }, { "JU", JU },
+ { "ja", ja }, { "JA", JA },
+ { "je", je }, { "JE", JE },
+ { "ję", ijus }, { "JĘ", IJUS },
+ { "jǫ", ibigjus }, { "JǪ", IBIGJUS },
+ { "ks", xi }, { "KS", XI },
+ { "ps", psi }, { "PS", PSI },
+ { "th", theta }, { "TH", THETA },
+ { "šč", shch }, { "ŠČ", SHCH },
+ },
+ entries = {
+ ["a"] = "a", ["b"] = "b", ["v"] = "v", ["g"] = "g", ["d"] = "d",
+ ["e"] = "e", ["ž"] = "ž", ["ʒ"] = "ʒ", ["z"] = "z", ["i"] = "i",
+ ["ï"] = "ï", [tshe] = "g’", ["k"] = "k", ["l"] = "l", ["m"] = "m",
+ ["n"] = "n", ["o"] = "o", ["p"] = "p", ["r"] = "r", ["s"] = "s",
+ ["t"] = "t", ["u"] = "u", ["f"] = "f", ["x"] = "x", ["o"] = "o",
+ ["c"] = "c", ["č"] = "č", ["š"] = "š", [sht] = "št", [shch] = "šč",
+ ["ъ"] = "ъ", ["y"] = "y", [uk] = "y", ["ь"] = "ь", ["ě"] = "ě",
+ [ju] = "ju", [ja] = "ja", [je] = "je", ["ę"] = "ę", [ijus] = "ję",
+ ["ǫ"] = "ǫ", [ibigjus] = "jǫ", [xi] = "ks", [psi] = "ps", [theta] = "th",
+ ["ü"] = "ü",
+ },
+ orders = {
+ "a", "b", "v", "g", "d", "e", "ž", "ʒ", "z", "i", "ï",
+ tshe, "k", "l", "m", "n", "o", "p", "r", "s", "t", "u",
+ "f", "x", "o", "c", "č", "š", sht, shch, "ъ", "y", uk,
+ "ь", "ě", ju, ja, je, "ę", ijus, "ǫ", ibigjus, xi, psi,
+ theta, "ü",
+ },
+ upper = {
+ uk = UK, tshe = TSHE, sht = SHT, ju = JU, ja = JA, je = JE, ijus = IJUS, ibigjus = IBIGJUS, xi = XI, psi = PSI, theta = THETA, shch = SHCH,
+ },
+ lower = {
+ UK = uk, TSHE = tshe, SHT = sht, JU = ju, JA = ja, JE = je, IJUS = ijus, IBIGJUS = ibigjus, XI = xi, PSI = psi, THETA = theta, SHCH = shch,
+ },
}
-mappings["ocs-scn"] = {
- ["a"] = 1,
- ["b"] = 2,
- ["v"] = 3,
- ["g"] = 4,
- ["d"] = 5,
- ["e"] = 6,
- ["ž"] = 7,
- ["ʒ"] = 8,
- ["z"] = 9,
- ["i"] = 10,
- ["ï"] = 10,
- [ocs_scn_tshe] = 11,
- ["k"] = 12,
- ["l"] = 13,
- ["m"] = 14,
- ["n"] = 15,
- ["o"] = 16,
- ["p"] = 17,
- ["r"] = 18,
- ["s"] = 19,
- ["t"] = 20,
- ["u"] = 21,
- ["f"] = 22,
- ["x"] = 23,
- ["o"] = 24,
- ["c"] = 25,
- ["č"] = 26,
- ["š"] = 27,
- [ocs_scn_sht] = 28,
- [ocs_scn_shch] = 28,
- ["ъ"] = 29,
- ["y"] = 30,
- [ocs_scn_uk] = 30,
- ["ь"] = 31,
- ["ě"] = 32,
- [ocs_scn_ju] = 33,
- [ocs_scn_ja] = 34,
- [ocs_scn_je] = 35,
- ["ę"] = 36,
- [ocs_scn_ijus] = 37,
- ["ǫ"] = 38,
-[ocs_scn_ibigjus] = 39,
- [ocs_scn_xi] = 40,
- [ocs_scn_psi] = 41,
- [ocs_scn_theta] = 42,
- ["ü"] = 43,
-}
-
-adduppercaseentries ("ocs-scn")
-adduppercasemappings("ocs-scn")
-
- entries["ocs-scn"][ocs_scn_UK ] = entries["ocs-scn"][ocs_scn_uk ]
-mappings["ocs-scn"][ocs_scn_UK ] = mappings["ocs-scn"][ocs_scn_uk ]
-
- entries["ocs-scn"][ocs_scn_TSHE ] = entries["ocs-scn"][ocs_scn_tshe ]
-mappings["ocs-scn"][ocs_scn_TSHE ] = mappings["ocs-scn"][ocs_scn_tshe ]
-
- entries["ocs-scn"][ocs_scn_SHT ] = entries["ocs-scn"][ocs_scn_sht ]
-mappings["ocs-scn"][ocs_scn_SHT ] = mappings["ocs-scn"][ocs_scn_sht ]
-
- entries["ocs-scn"][ocs_scn_JU ] = entries["ocs-scn"][ocs_scn_ju ]
-mappings["ocs-scn"][ocs_scn_JU ] = mappings["ocs-scn"][ocs_scn_ju ]
-
- entries["ocs-scn"][ocs_scn_JA ] = entries["ocs-scn"][ocs_scn_ja ]
-mappings["ocs-scn"][ocs_scn_JA ] = mappings["ocs-scn"][ocs_scn_ja ]
-
- entries["ocs-scn"][ocs_scn_JE ] = entries["ocs-scn"][ocs_scn_je ]
-mappings["ocs-scn"][ocs_scn_JE ] = mappings["ocs-scn"][ocs_scn_je ]
-
- entries["ocs-scn"][ocs_scn_IJUS ] = entries["ocs-scn"][ocs_scn_ijus ]
-mappings["ocs-scn"][ocs_scn_IJUS ] = mappings["ocs-scn"][ocs_scn_ijus ]
-
- entries["ocs-scn"][ocs_scn_IBIGJUS] = entries["ocs-scn"][ocs_scn_ibigjus]
-mappings["ocs-scn"][ocs_scn_IBIGJUS] = mappings["ocs-scn"][ocs_scn_ibigjus]
-
- entries["ocs-scn"][ocs_scn_XI ] = entries["ocs-scn"][ocs_scn_xi ]
-mappings["ocs-scn"][ocs_scn_XI ] = mappings["ocs-scn"][ocs_scn_xi ]
-
- entries["ocs-scn"][ocs_scn_PSI ] = entries["ocs-scn"][ocs_scn_psi ]
-mappings["ocs-scn"][ocs_scn_PSI ] = mappings["ocs-scn"][ocs_scn_psi ]
-
- entries["ocs-scn"][ocs_scn_THETA ] = entries["ocs-scn"][ocs_scn_theta ]
-mappings["ocs-scn"][ocs_scn_THETA ] = mappings["ocs-scn"][ocs_scn_theta ]
-
- entries["ocs-scn"][ocs_scn_SHCH ] = entries["ocs-scn"][ocs_scn_shch ]
-mappings["ocs-scn"][ocs_scn_SHCH ] = mappings["ocs-scn"][ocs_scn_shch ]
--- Norwegian (bokmål).
-replacements["no"] = { --[[ None, do you miss any? ]] }
-
-entries["no"] = {
- ["a"] = "a",
- ["b"] = "b",
- ["c"] = "c",
- ["d"] = "d",
- ["e"] = "e",
- ["f"] = "f",
- ["g"] = "g",
- ["h"] = "h",
- ["i"] = "i",
- ["j"] = "j",
- ["k"] = "k",
- ["l"] = "l",
- ["m"] = "m",
- ["n"] = "n",
- ["o"] = "o",
- ["p"] = "p",
- ["q"] = "q",
- ["r"] = "r",
- ["s"] = "s",
- ["t"] = "t",
- ["u"] = "u",
- ["v"] = "v",
- ["w"] = "w",
- ["x"] = "x",
- ["y"] = "y",
- ["z"] = "z",
- ["æ"] = "æ",
- ["ø"] = "ø",
- ["å"] = "å",
+definitions["no"] = {
+ entries = {
+ ["a"] = "a", ["b"] = "b", ["c"] = "c", ["d"] = "d", ["e"] = "e",
+ ["f"] = "f", ["g"] = "g", ["h"] = "h", ["i"] = "i", ["j"] = "j",
+ ["k"] = "k", ["l"] = "l", ["m"] = "m", ["n"] = "n", ["o"] = "o",
+ ["p"] = "p", ["q"] = "q", ["r"] = "r", ["s"] = "s", ["t"] = "t",
+ ["u"] = "u", ["v"] = "v", ["w"] = "w", ["x"] = "x", ["y"] = "y",
+ ["z"] = "z", ["æ"] = "æ", ["ø"] = "ø", ["å"] = "å",
+ },
+ orders = {
+ "a", "b", "c", "d", "e", "f", "g", "h", "i", "j",
+ "k", "l", "m", "n", "o", "p", "q", "r", "s", "t",
+ "u", "v", "w", "x", "y", "z", "æ", "ø", "å",
+ }
}
-mappings["no"] = {
- ["a"] = 1,
- ["b"] = 2,
- ["c"] = 3,
- ["d"] = 4,
- ["e"] = 5,
- ["f"] = 6,
- ["g"] = 7,
- ["h"] = 8,
- ["i"] = 9,
- ["j"] = 10,
- ["k"] = 11,
- ["l"] = 12,
- ["m"] = 13,
- ["n"] = 14,
- ["o"] = 15,
- ["p"] = 16,
- ["q"] = 17,
- ["r"] = 18,
- ["s"] = 19,
- ["t"] = 20,
- ["u"] = 21,
- ["v"] = 22,
- ["w"] = 23,
- ["x"] = 24,
- ["y"] = 25,
- ["z"] = 26,
- ["æ"] = 27,
- ["ø"] = 28,
- ["å"] = 29,
-}
-
-adduppercaseentries ("no")
-adduppercasemappings("no")
-
--- Danish (-> Norwegian).
-replacements["da"] = { --[[ None, do you miss any? ]] }
- entries["da"] = entries["no"]
- mappings["da"] = mappings["no"]
+definitions["da"] = { parent = "no" }
--- Swedish
-replacements["sv"] = { --[[ None, do you miss any? ]] }
-
-entries["sv"] = {
- ["a"] = "a",
- ["b"] = "b",
- ["c"] = "c",
- ["d"] = "d",
- ["e"] = "e",
- ["f"] = "f",
- ["g"] = "g",
- ["h"] = "h",
- ["i"] = "i",
- ["j"] = "j",
- ["k"] = "k",
- ["l"] = "l",
- ["m"] = "m",
- ["n"] = "n",
- ["o"] = "o",
- ["p"] = "p",
- ["q"] = "q",
- ["r"] = "r",
- ["s"] = "s",
- ["t"] = "t",
- ["u"] = "u",
- ["v"] = "v",
- ["w"] = "w",
- ["x"] = "x",
- ["y"] = "y",
- ["z"] = "z",
- ["å"] = "å",
- ["ä"] = "ä",
- ["ö"] = "ö",
-}
-
-mappings["sv"] = {
- ["a"] = 1,
- ["b"] = 2,
- ["c"] = 3,
- ["d"] = 4,
- ["e"] = 5,
- ["f"] = 6,
- ["g"] = 7,
- ["h"] = 8,
- ["i"] = 9,
- ["j"] = 10,
- ["k"] = 11,
- ["l"] = 12,
- ["m"] = 13,
- ["n"] = 14,
- ["o"] = 15,
- ["p"] = 16,
- ["q"] = 17,
- ["r"] = 18,
- ["s"] = 19,
- ["t"] = 20,
- ["u"] = 21,
- ["v"] = 22,
- ["w"] = 23,
- ["x"] = 24,
- ["y"] = 25,
- ["z"] = 26,
- ["å"] = 27,
- ["ä"] = 28,
- ["ö"] = 29,
+definitions["sv"] = {
+ entries = {
+ ["a"] = "a", ["b"] = "b", ["c"] = "c", ["d"] = "d", ["e"] = "e",
+ ["f"] = "f", ["g"] = "g", ["h"] = "h", ["i"] = "i", ["j"] = "j",
+ ["k"] = "k", ["l"] = "l", ["m"] = "m", ["n"] = "n", ["o"] = "o",
+ ["p"] = "p", ["q"] = "q", ["r"] = "r", ["s"] = "s", ["t"] = "t",
+ ["u"] = "u", ["v"] = "v", ["w"] = "w", ["x"] = "x", ["y"] = "y",
+ ["z"] = "z", ["å"] = "å", ["ä"] = "ä", ["ö"] = "ö",
+ },
+ orders = {
+ "a", "b", "c", "d", "e", "f", "g", "h", "i", "j",
+ "k", "l", "m", "n", "o", "p", "q", "r", "s", "t",
+ "u", "v", "w", "x", "y", "z", "å", "ä", "ö",
+ }
}
-adduppercaseentries ("sv")
-adduppercasemappings("sv")
-
--- Icelandic
-- Treating quantities as allographs.
-replacements["is"] = { --[[ None, do you miss any? ]] }
-
-entries["is"] = {
- ["a"] = "a",
- ["á"] = "a",
- ["b"] = "b",
- ["d"] = "d",
- ["ð"] = "ð",
- ["e"] = "e",
- ["é"] = "e",
- ["f"] = "f",
- ["g"] = "g",
- ["h"] = "h",
- ["i"] = "i",
- ["í"] = "i",
- ["j"] = "j",
- ["k"] = "k",
- ["l"] = "l",
- ["m"] = "m",
- ["n"] = "n",
- ["o"] = "o",
- ["ó"] = "o",
- ["p"] = "p",
- ["r"] = "r",
- ["s"] = "s",
- ["t"] = "t",
- ["u"] = "u",
- ["ú"] = "u",
- ["v"] = "v",
- ["x"] = "x",
- ["y"] = "y",
- ["ý"] = "y",
- ["þ"] = "þ",
- ["æ"] = "æ",
- ["ö"] = "ö",
-}
-
-mappings["is"] = {
- ["a"] = 1,
- ["á"] = 1,
- ["b"] = 2,
- ["d"] = 3,
- ["ð"] = 4,
- ["e"] = 5,
- ["é"] = 5,
- ["f"] = 6,
- ["g"] = 7,
- ["h"] = 8,
- ["i"] = 9,
- ["í"] = 9,
- ["j"] = 10,
- ["k"] = 11,
- ["l"] = 12,
- ["m"] = 13,
- ["n"] = 14,
- ["o"] = 15,
- ["ó"] = 15,
- ["p"] = 16,
- ["r"] = 17,
- ["s"] = 18,
- ["t"] = 19,
- ["u"] = 20,
- ["ú"] = 20,
- ["v"] = 21,
- ["x"] = 22,
- ["y"] = 23,
- ["ý"] = 23,
- ["þ"] = 24,
- ["æ"] = 25,
- ["ö"] = 26,
+definitions["is"] = {
+ entries = {
+ ["a"] = "a", ["á"] = "a", ["b"] = "b", ["d"] = "d", ["ð"] = "ð",
+ ["e"] = "e", ["é"] = "e", ["f"] = "f", ["g"] = "g", ["h"] = "h",
+ ["i"] = "i", ["í"] = "i", ["j"] = "j", ["k"] = "k", ["l"] = "l",
+ ["m"] = "m", ["n"] = "n", ["o"] = "o", ["ó"] = "o", ["p"] = "p",
+ ["r"] = "r", ["s"] = "s", ["t"] = "t", ["u"] = "u", ["ú"] = "u",
+ ["v"] = "v", ["x"] = "x", ["y"] = "y", ["ý"] = "y", ["þ"] = "þ",
+ ["æ"] = "æ", ["ö"] = "ö",
+ },
+ orders = {
+ "a", "á", "b", "d", "ð", "e", "é", "f", "g", "h",
+ "i", "í", "j", "k", "l", "m", "n", "o", "ó", "p",
+ "r", "s", "t", "u", "ú", "v", "x", "y", "ý", "þ",
+ "æ", "ö",
+ },
}
-adduppercaseentries ("is")
-adduppercasemappings("is")
-
--- Greek
-replacements["gr"] = { --[[ None, do you miss any? ]] }
-
-entries["gr"] = {
- ["α"] = "α",
- ["ά"] = "α",
- ["ὰ"] = "α",
- ["ᾶ"] = "α",
- ["ᾳ"] = "α",
- ["ἀ"] = "α",
- ["ἁ"] = "α",
- ["ἄ"] = "α",
- ["ἂ"] = "α",
- ["ἆ"] = "α",
- ["ἁ"] = "α",
- ["ἅ"] = "α",
- ["ἃ"] = "α",
- ["ἇ"] = "α",
- ["ᾁ"] = "α",
- ["ᾴ"] = "α",
- ["ᾲ"] = "α",
- ["ᾷ"] = "α",
- ["ᾄ"] = "α",
- ["ᾂ"] = "α",
- ["ᾅ"] = "α",
- ["ᾃ"] = "α",
- ["ᾆ"] = "α",
- ["ᾇ"] = "α",
- ["β"] = "β",
- ["γ"] = "γ",
- ["δ"] = "δ",
- ["ε"] = "ε",
- ["έ"] = "ε",
- ["ὲ"] = "ε",
- ["ἐ"] = "ε",
- ["ἔ"] = "ε",
- ["ἒ"] = "ε",
- ["ἑ"] = "ε",
- ["ἕ"] = "ε",
- ["ἓ"] = "ε",
- ["ζ"] = "ζ",
- ["η"] = "η",
- ["η"] = "η",
- ["ή"] = "η",
- ["ὴ"] = "η",
- ["ῆ"] = "η",
- ["ῃ"] = "η",
- ["ἠ"] = "η",
- ["ἤ"] = "η",
- ["ἢ"] = "η",
- ["ἦ"] = "η",
- ["ᾐ"] = "η",
- ["ἡ"] = "η",
- ["ἥ"] = "η",
- ["ἣ"] = "η",
- ["ἧ"] = "η",
- ["ᾑ"] = "η",
- ["ῄ"] = "η",
- ["ῂ"] = "η",
- ["ῇ"] = "η",
- ["ᾔ"] = "η",
- ["ᾒ"] = "η",
- ["ᾕ"] = "η",
- ["ᾓ"] = "η",
- ["ᾖ"] = "η",
- ["ᾗ"] = "η",
- ["θ"] = "θ",
- ["ι"] = "ι",
- ["ί"] = "ι",
- ["ὶ"] = "ι",
- ["ῖ"] = "ι",
- ["ἰ"] = "ι",
- ["ἴ"] = "ι",
- ["ἲ"] = "ι",
- ["ἶ"] = "ι",
- ["ἱ"] = "ι",
- ["ἵ"] = "ι",
- ["ἳ"] = "ι",
- ["ἷ"] = "ι",
- ["ϊ"] = "ι",
- ["ΐ"] = "ι",
- ["ῒ"] = "ι",
- ["ῗ"] = "ι",
- ["κ"] = "κ",
- ["λ"] = "λ",
- ["μ"] = "μ",
- ["ν"] = "ν",
- ["ξ"] = "ξ",
- ["ο"] = "ο",
- ["ό"] = "ο",
- ["ὸ"] = "ο",
- ["ὀ"] = "ο",
- ["ὄ"] = "ο",
- ["ὂ"] = "ο",
- ["ὁ"] = "ο",
- ["ὅ"] = "ο",
- ["ὃ"] = "ο",
- ["π"] = "π",
- ["ρ"] = "ρ",
- ["ῤ"] = "ῤ",
- ["ῥ"] = "ῥ",
- ["σ"] = "σ",
- ["ς"] = "ς",
- ["τ"] = "τ",
- ["υ"] = "υ",
- ["ύ"] = "υ",
- ["ὺ"] = "υ",
- ["ῦ"] = "υ",
- ["ὐ"] = "υ",
- ["ὔ"] = "υ",
- ["ὒ"] = "υ",
- ["ὖ"] = "υ",
- ["ὑ"] = "υ",
- ["ὕ"] = "υ",
- ["ὓ"] = "υ",
- ["ὗ"] = "υ",
- ["ϋ"] = "υ",
- ["ΰ"] = "υ",
- ["ῢ"] = "υ",
- ["ῧ"] = "υ",
- ["φ"] = "φ",
- ["χ"] = "χ",
- ["ψ"] = "ω",
- ["ω"] = "ω",
- ["ώ"] = "ω",
- ["ὼ"] = "ω",
- ["ῶ"] = "ω",
- ["ῳ"] = "ω",
- ["ὠ"] = "ω",
- ["ὤ"] = "ω",
- ["ὢ"] = "ω",
- ["ὦ"] = "ω",
- ["ᾠ"] = "ω",
- ["ὡ"] = "ω",
- ["ὥ"] = "ω",
- ["ὣ"] = "ω",
- ["ὧ"] = "ω",
- ["ᾡ"] = "ω",
- ["ῴ"] = "ω",
- ["ῲ"] = "ω",
- ["ῷ"] = "ω",
- ["ᾤ"] = "ω",
- ["ᾢ"] = "ω",
- ["ᾥ"] = "ω",
- ["ᾣ"] = "ω",
- ["ᾦ"] = "ω",
- ["ᾧ"] = "ω",
-}
-
-mappings["gr"] = {
- ["α"] = 1,
- ["ά"] = 1,
- ["ὰ"] = 1,
- ["ᾶ"] = 1,
- ["ᾳ"] = 1,
- ["ἀ"] = 1,
- ["ἁ"] = 1,
- ["ἄ"] = 1,
- ["ἂ"] = 1,
- ["ἆ"] = 1,
- ["ἁ"] = 1,
- ["ἅ"] = 1,
- ["ἃ"] = 1,
- ["ἇ"] = 1,
- ["ᾁ"] = 1,
- ["ᾴ"] = 1,
- ["ᾲ"] = 1,
- ["ᾷ"] = 1,
- ["ᾄ"] = 1,
- ["ᾂ"] = 1,
- ["ᾅ"] = 1,
- ["ᾃ"] = 1,
- ["ᾆ"] = 1,
- ["ᾇ"] = 1,
- ["β"] = 2,
- ["γ"] = 3,
- ["δ"] = 4,
- ["ε"] = 5,
- ["έ"] = 5,
- ["ὲ"] = 5,
- ["ἐ"] = 5,
- ["ἔ"] = 5,
- ["ἒ"] = 5,
- ["ἑ"] = 5,
- ["ἕ"] = 5,
- ["ἓ"] = 5,
- ["ζ"] = 6,
- ["η"] = 7,
- ["η"] = 7,
- ["ή"] = 7,
- ["ὴ"] = 7,
- ["ῆ"] = 7,
- ["ῃ"] = 7,
- ["ἠ"] = 7,
- ["ἤ"] = 7,
- ["ἢ"] = 7,
- ["ἦ"] = 7,
- ["ᾐ"] = 7,
- ["ἡ"] = 7,
- ["ἥ"] = 7,
- ["ἣ"] = 7,
- ["ἧ"] = 7,
- ["ᾑ"] = 7,
- ["ῄ"] = 7,
- ["ῂ"] = 7,
- ["ῇ"] = 7,
- ["ᾔ"] = 7,
- ["ᾒ"] = 7,
- ["ᾕ"] = 7,
- ["ᾓ"] = 7,
- ["ᾖ"] = 7,
- ["ᾗ"] = 7,
- ["θ"] = 8,
- ["ι"] = 9,
- ["ί"] = 9,
- ["ὶ"] = 9,
- ["ῖ"] = 9,
- ["ἰ"] = 9,
- ["ἴ"] = 9,
- ["ἲ"] = 9,
- ["ἶ"] = 9,
- ["ἱ"] = 9,
- ["ἵ"] = 9,
- ["ἳ"] = 9,
- ["ἷ"] = 9,
- ["ϊ"] = 9,
- ["ΐ"] = 9,
- ["ῒ"] = 9,
- ["ῗ"] = 9,
- ["κ"] = 10,
- ["λ"] = 11,
- ["μ"] = 12,
- ["ν"] = 13,
- ["ξ"] = 14,
- ["ο"] = 15,
- ["ό"] = 15,
- ["ὸ"] = 15,
- ["ὀ"] = 15,
- ["ὄ"] = 15,
- ["ὂ"] = 15,
- ["ὁ"] = 15,
- ["ὅ"] = 15,
- ["ὃ"] = 15,
- ["π"] = 16,
- ["ρ"] = 17,
- ["ῤ"] = 17,
- ["ῥ"] = 17,
- ["σ"] = 18,
- ["ς"] = 18,
- ["τ"] = 19,
- ["υ"] = 20,
- ["ύ"] = 20,
- ["ὺ"] = 20,
- ["ῦ"] = 20,
- ["ὐ"] = 20,
- ["ὔ"] = 20,
- ["ὒ"] = 20,
- ["ὖ"] = 20,
- ["ὑ"] = 20,
- ["ὕ"] = 20,
- ["ὓ"] = 20,
- ["ὗ"] = 20,
- ["ϋ"] = 20,
- ["ΰ"] = 20,
- ["ῢ"] = 20,
- ["ῧ"] = 20,
- ["φ"] = 21,
- ["χ"] = 22,
- ["ψ"] = 23,
- ["ω"] = 24,
- ["ώ"] = 24,
- ["ὼ"] = 24,
- ["ῶ"] = 24,
- ["ῳ"] = 24,
- ["ὠ"] = 24,
- ["ὤ"] = 24,
- ["ὢ"] = 24,
- ["ὦ"] = 24,
- ["ᾠ"] = 24,
- ["ὡ"] = 24,
- ["ὥ"] = 24,
- ["ὣ"] = 24,
- ["ὧ"] = 24,
- ["ᾡ"] = 24,
- ["ῴ"] = 24,
- ["ῲ"] = 24,
- ["ῷ"] = 24,
- ["ᾤ"] = 24,
- ["ᾢ"] = 24,
- ["ᾥ"] = 24,
- ["ᾣ"] = 24,
- ["ᾦ"] = 24,
- ["ᾧ"] = 24,
+definitions["gr"] = {
+ entries = {
+ ["α"] = "α", ["ά"] = "α", ["ὰ"] = "α", ["ᾶ"] = "α", ["ᾳ"] = "α",
+ ["ἀ"] = "α", ["ἁ"] = "α", ["ἄ"] = "α", ["ἂ"] = "α", ["ἆ"] = "α",
+ ["ἁ"] = "α", ["ἅ"] = "α", ["ἃ"] = "α", ["ἇ"] = "α", ["ᾁ"] = "α",
+ ["ᾴ"] = "α", ["ᾲ"] = "α", ["ᾷ"] = "α", ["ᾄ"] = "α", ["ᾂ"] = "α",
+ ["ᾅ"] = "α", ["ᾃ"] = "α", ["ᾆ"] = "α", ["ᾇ"] = "α", ["β"] = "β",
+ ["γ"] = "γ", ["δ"] = "δ", ["ε"] = "ε", ["έ"] = "ε", ["ὲ"] = "ε",
+ ["ἐ"] = "ε", ["ἔ"] = "ε", ["ἒ"] = "ε", ["ἑ"] = "ε", ["ἕ"] = "ε",
+ ["ἓ"] = "ε", ["ζ"] = "ζ", ["η"] = "η", ["η"] = "η", ["ή"] = "η",
+ ["ὴ"] = "η", ["ῆ"] = "η", ["ῃ"] = "η", ["ἠ"] = "η", ["ἤ"] = "η",
+ ["ἢ"] = "η", ["ἦ"] = "η", ["ᾐ"] = "η", ["ἡ"] = "η", ["ἥ"] = "η",
+ ["ἣ"] = "η", ["ἧ"] = "η", ["ᾑ"] = "η", ["ῄ"] = "η", ["ῂ"] = "η",
+ ["ῇ"] = "η", ["ᾔ"] = "η", ["ᾒ"] = "η", ["ᾕ"] = "η", ["ᾓ"] = "η",
+ ["ᾖ"] = "η", ["ᾗ"] = "η", ["θ"] = "θ", ["ι"] = "ι", ["ί"] = "ι",
+ ["ὶ"] = "ι", ["ῖ"] = "ι", ["ἰ"] = "ι", ["ἴ"] = "ι", ["ἲ"] = "ι",
+ ["ἶ"] = "ι", ["ἱ"] = "ι", ["ἵ"] = "ι", ["ἳ"] = "ι", ["ἷ"] = "ι",
+ ["ϊ"] = "ι", ["ΐ"] = "ι", ["ῒ"] = "ι", ["ῗ"] = "ι", ["κ"] = "κ",
+ ["λ"] = "λ", ["μ"] = "μ", ["ν"] = "ν", ["ξ"] = "ξ", ["ο"] = "ο",
+ ["ό"] = "ο", ["ὸ"] = "ο", ["ὀ"] = "ο", ["ὄ"] = "ο", ["ὂ"] = "ο",
+ ["ὁ"] = "ο", ["ὅ"] = "ο", ["ὃ"] = "ο", ["π"] = "π", ["ρ"] = "ρ",
+ ["ῤ"] = "ῤ", ["ῥ"] = "ῥ", ["σ"] = "σ", ["ς"] = "ς", ["τ"] = "τ",
+ ["υ"] = "υ", ["ύ"] = "υ", ["ὺ"] = "υ", ["ῦ"] = "υ", ["ὐ"] = "υ",
+ ["ὔ"] = "υ", ["ὒ"] = "υ", ["ὖ"] = "υ", ["ὑ"] = "υ", ["ὕ"] = "υ",
+ ["ὓ"] = "υ", ["ὗ"] = "υ", ["ϋ"] = "υ", ["ΰ"] = "υ", ["ῢ"] = "υ",
+ ["ῧ"] = "υ", ["φ"] = "φ", ["χ"] = "χ", ["ψ"] = "ω", ["ω"] = "ω",
+ ["ώ"] = "ω", ["ὼ"] = "ω", ["ῶ"] = "ω", ["ῳ"] = "ω", ["ὠ"] = "ω",
+ ["ὤ"] = "ω", ["ὢ"] = "ω", ["ὦ"] = "ω", ["ᾠ"] = "ω", ["ὡ"] = "ω",
+ ["ὥ"] = "ω", ["ὣ"] = "ω", ["ὧ"] = "ω", ["ᾡ"] = "ω", ["ῴ"] = "ω",
+ ["ῲ"] = "ω", ["ῷ"] = "ω", ["ᾤ"] = "ω", ["ᾢ"] = "ω", ["ᾥ"] = "ω",
+ ["ᾣ"] = "ω", ["ᾦ"] = "ω", ["ᾧ"] = "ω",
+ },
+ orders = {
+ "α", "ά", "ὰ", "ᾶ", "ᾳ", "ἀ", "ἁ", "ἄ", "ἂ", "ἆ",
+ "ἁ", "ἅ", "ἃ", "ἇ", "ᾁ", "ᾴ", "ᾲ", "ᾷ", "ᾄ", "ᾂ",
+ "ᾅ", "ᾃ", "ᾆ", "ᾇ", "β", "γ", "δ", "ε", "έ", "ὲ",
+ "ἐ", "ἔ", "ἒ", "ἑ", "ἕ", "ἓ", "ζ", "η", "η", "ή",
+ "ὴ", "ῆ", "ῃ", "ἠ", "ἤ", "ἢ", "ἦ", "ᾐ", "ἡ", "ἥ",
+ "ἣ", "ἧ", "ᾑ", "ῄ", "ῂ", "ῇ", "ᾔ", "ᾒ", "ᾕ", "ᾓ",
+ "ᾖ", "ᾗ", "θ", "ι", "ί", "ὶ", "ῖ", "ἰ", "ἴ", "ἲ",
+ "ἶ", "ἱ", "ἵ", "ἳ", "ἷ", "ϊ", "ΐ", "ῒ", "ῗ", "κ",
+ "λ", "μ", "ν", "ξ", "ο", "ό", "ὸ", "ὀ", "ὄ", "ὂ",
+ "ὁ", "ὅ", "ὃ", "π", "ρ", "ῤ", "ῥ", "σ", "ς", "τ",
+ "υ", "ύ", "ὺ", "ῦ", "ὐ", "ὔ", "ὒ", "ὖ", "ὑ", "ὕ",
+ "ὓ", "ὗ", "ϋ", "ΰ", "ῢ", "ῧ", "φ", "χ", "ψ", "ω",
+ "ώ", "ὼ", "ῶ", "ῳ", "ὠ", "ὤ", "ὢ", "ὦ", "ᾠ", "ὡ",
+ "ὥ", "ὣ", "ὧ", "ᾡ", "ῴ", "ῲ", "ῷ", "ᾤ", "ᾢ", "ᾥ",
+ "ᾣ", "ᾦ", "ᾧ",
+ },
}
-adduppercaseentries ("gr")
-adduppercasemappings("gr")
-
--- Latin
-- Treating the post-classical fricatives “j” and “v” as “i” and “u”
-- respectively.
-replacements["la"] = {
- [1] = { "æ", "ae" },
-}
-
-entries["la"] = {
- ["a"] = "a",
- ["ā"] = "a",
- ["ă"] = "a",
- ["b"] = "b",
- ["c"] = "c",
- ["d"] = "d",
- ["e"] = "e",
- ["ē"] = "e",
- ["ĕ"] = "e",
- ["f"] = "f",
- ["g"] = "g",
- ["h"] = "h",
- ["i"] = "i",
- ["ī"] = "i",
- ["ĭ"] = "i",
- ["j"] = "i",
- ["k"] = "k",
- ["l"] = "l",
- ["m"] = "m",
- ["n"] = "n",
- ["o"] = "o",
- ["ō"] = "o",
- ["ŏ"] = "o",
- ["p"] = "p",
- ["q"] = "q",
- ["r"] = "r",
- ["s"] = "s",
- ["t"] = "t",
- ["u"] = "u",
- ["ū"] = "u",
- ["ŭ"] = "u",
- ["v"] = "u",
- ["w"] = "w",
- ["x"] = "x",
- ["y"] = "y",
- ["ȳ"] = "y", -- Should exist in Greek words.
- ["y̆"] = "y", -- Should exist in Greek words.
- ["z"] = "z",
-}
-
-mappings["la"] = {
- ["a"] = 1,
- ["ā"] = 1,
- ["ă"] = 1,
- ["b"] = 2,
- ["c"] = 3,
- ["d"] = 4,
- ["e"] = 5,
- ["ē"] = 5,
- ["ĕ"] = 5,
- ["f"] = 6,
- ["g"] = 7,
- ["h"] = 8,
- ["i"] = 9,
- ["ī"] = 9,
- ["ĭ"] = 9,
- ["j"] = 9,
- ["k"] = 10,
- ["l"] = 11,
- ["m"] = 12,
- ["n"] = 13,
- ["o"] = 14,
- ["ō"] = 14,
- ["ŏ"] = 14,
- ["p"] = 15,
- ["q"] = 16,
- ["r"] = 17,
- ["s"] = 18,
- ["t"] = 19,
- ["u"] = 20,
- ["ū"] = 20,
- ["ŭ"] = 20,
- ["v"] = 20,
- ["w"] = 21,
- ["x"] = 22,
- ["y"] = 23,
- ["ȳ"] = 23,
- ["y̆"] = 23,
- ["z"] = 24,
+definitions["la"] = {
+ replacements = {
+ { "æ", "ae" }, { "Æ", "AE" },
+ },
+ entries = {
+ ["a"] = "a", ["ā"] = "a", ["ă"] = "a", ["b"] = "b", ["c"] = "c",
+ ["d"] = "d", ["e"] = "e", ["ē"] = "e", ["ĕ"] = "e", ["f"] = "f",
+ ["g"] = "g", ["h"] = "h", ["i"] = "i", ["ī"] = "i", ["ĭ"] = "i",
+ ["j"] = "i", ["k"] = "k", ["l"] = "l", ["m"] = "m", ["n"] = "n",
+ ["o"] = "o", ["ō"] = "o", ["ŏ"] = "o", ["p"] = "p", ["q"] = "q",
+ ["r"] = "r", ["s"] = "s", ["t"] = "t", ["u"] = "u", ["ū"] = "u",
+ ["ŭ"] = "u", ["v"] = "u", ["w"] = "w", ["x"] = "x", ["y"] = "y",
+ ["ȳ"] = "y", ["y̆"] = "y", ["z"] = "z",
+ },
+ orders = {
+ "a", "ā", "ă", "b", "c", "d", "e", "ē", "ĕ", "f",
+ "g", "h", "i", "ī", "ĭ", "j", "k", "l", "m", "n",
+ "o", "ō", "ŏ", "p", "q", "r", "s", "t", "u", "ū",
+ "ŭ", "v", "w", "x", "y", "ȳ", "y̆", "z",
+ }
}
-adduppercaseentries ("la")
-adduppercasemappings("la")
-
--- Italian
-replacements["it"] = { --[[ None, do you miss any? ]] }
-
-entries["it"] = {
- ["a"] = "a",
- ["á"] = "a",
- ["b"] = "b",
- ["c"] = "c",
- ["d"] = "d",
- ["e"] = "e",
- ["é"] = "e",
- ["è"] = "e",
- ["f"] = "f",
- ["g"] = "g",
- ["h"] = "h",
- ["i"] = "i",
- ["í"] = "i",
- ["ì"] = "i",
- ["j"] = "i",
- ["k"] = "k",
- ["l"] = "l",
- ["m"] = "m",
- ["n"] = "n",
- ["o"] = "o",
- ["ó"] = "o",
- ["ò"] = "o",
- ["p"] = "p",
- ["q"] = "q",
- ["r"] = "r",
- ["s"] = "s",
- ["t"] = "t",
- ["u"] = "u",
- ["ú"] = "u",
- ["ù"] = "u",
- ["v"] = "u",
- ["w"] = "w",
- ["x"] = "x",
- ["y"] = "y",
- ["z"] = "z",
-}
-
-mappings["it"] = {
- ["a"] = 1,
- ["á"] = 1,
- ["b"] = 2,
- ["c"] = 3,
- ["d"] = 4,
- ["e"] = 5,
- ["é"] = 5,
- ["è"] = 5,
- ["f"] = 6,
- ["g"] = 7,
- ["h"] = 8,
- ["i"] = 9,
- ["í"] = 9,
- ["ì"] = 9,
- ["j"] = 10,
- ["k"] = 11,
- ["l"] = 12,
- ["m"] = 13,
- ["n"] = 14,
- ["o"] = 15,
- ["ó"] = 15,
- ["ò"] = 15,
- ["p"] = 16,
- ["q"] = 17,
- ["r"] = 18,
- ["s"] = 19,
- ["t"] = 20,
- ["u"] = 21,
- ["ú"] = 21,
- ["ù"] = 21,
- ["v"] = 22,
- ["w"] = 23,
- ["x"] = 24,
- ["y"] = 25,
- ["z"] = 26,
+definitions["it"] = {
+ entries = {
+ ["a"] = "a", ["á"] = "a", ["b"] = "b", ["c"] = "c", ["d"] = "d",
+ ["e"] = "e", ["é"] = "e", ["è"] = "e", ["f"] = "f", ["g"] = "g",
+ ["h"] = "h", ["i"] = "i", ["í"] = "i", ["ì"] = "i", ["j"] = "i",
+ ["k"] = "k", ["l"] = "l", ["m"] = "m", ["n"] = "n", ["o"] = "o",
+ ["ó"] = "o", ["ò"] = "o", ["p"] = "p", ["q"] = "q", ["r"] = "r",
+ ["s"] = "s", ["t"] = "t", ["u"] = "u", ["ú"] = "u", ["ù"] = "u",
+ ["v"] = "u", ["w"] = "w", ["x"] = "x", ["y"] = "y", ["z"] = "z",
+ },
+ orders = {
+ "a", "á", "b", "c", "d", "e", "é", "è", "f", "g",
+ "h", "i", "í", "ì", "j", "k", "l", "m", "n", "o",
+ "ó", "ò", "p", "q", "r", "s", "t", "u", "ú", "ù",
+ "v", "w", "x", "y", "z",
+ }
}
-adduppercaseentries ("it")
-adduppercasemappings("it")
-
--- Romanian
-
-replacements["ro"] = { --[[ None, do you miss any? ]] }
-
-entries["ro"] = {
- ["a"] = "a",
- ["ă"] = "ă",
- ["â"] = "â",
- ["b"] = "b",
- ["c"] = "c",
- ["d"] = "d",
- ["e"] = "e",
- ["f"] = "f",
- ["g"] = "g",
- ["h"] = "h",
- ["i"] = "i",
- ["î"] = "î",
- ["j"] = "j",
- ["k"] = "k",
- ["l"] = "l",
- ["m"] = "m",
- ["n"] = "n",
- ["o"] = "o",
- ["p"] = "p",
- ["q"] = "q",
- ["r"] = "r",
- ["s"] = "s",
- ["ș"] = "ș",
- ["t"] = "t",
- ["ț"] = "ț",
- ["u"] = "u",
- ["v"] = "v",
- ["w"] = "w",
- ["x"] = "x",
- ["y"] = "y",
- ["z"] = "z",
+definitions["ro"] = {
+ entries = {
+ ["a"] = "a", ["ă"] = "ă", ["â"] = "â", ["b"] = "b", ["c"] = "c",
+ ["d"] = "d", ["e"] = "e", ["f"] = "f", ["g"] = "g", ["h"] = "h",
+ ["i"] = "i", ["î"] = "î", ["j"] = "j", ["k"] = "k", ["l"] = "l",
+ ["m"] = "m", ["n"] = "n", ["o"] = "o", ["p"] = "p", ["q"] = "q",
+ ["r"] = "r", ["s"] = "s", ["ș"] = "ș", ["t"] = "t", ["ț"] = "ț",
+ ["u"] = "u", ["v"] = "v", ["w"] = "w", ["x"] = "x", ["y"] = "y",
+ ["z"] = "z",
+ },
+ orders = {
+ "a", "ă", "â", "b", "c", "d", "e", "f", "g", "h",
+ "i", "î", "j", "k", "l", "m", "n", "o", "p", "q",
+ "r", "s", "ș", "t", "ț", "u", "v", "w", "x", "y",
+ "z",
+ }
}
-mappings["ro"] = {
- ["a"] = 1,
- ["ă"] = 2,
- ["â"] = 3,
- ["b"] = 4,
- ["c"] = 5,
- ["d"] = 6,
- ["e"] = 7,
- ["f"] = 8,
- ["g"] = 9,
- ["h"] = 10,
- ["i"] = 11,
- ["î"] = 12,
- ["j"] = 13,
- ["k"] = 14,
- ["l"] = 15,
- ["m"] = 16,
- ["n"] = 17,
- ["o"] = 18,
- ["p"] = 19,
- ["q"] = 20,
- ["r"] = 21,
- ["s"] = 22,
- ["ș"] = 23,
- ["t"] = 24,
- ["ț"] = 25,
- ["u"] = 26,
- ["v"] = 27,
- ["w"] = 28,
- ["x"] = 29,
- ["y"] = 30,
- ["z"] = 31,
-}
-
-adduppercaseentries ("ro")
-adduppercasemappings("ro")
-
--- Spanish
-replacements["es"] = { --[[ None, do you miss any? ]] }
-
-entries["es"] = {
- ["a"] = "a",
- ["á"] = "a",
- ["b"] = "b",
- ["c"] = "c",
- ["d"] = "d",
- ["e"] = "e",
- ["é"] = "e",
- ["f"] = "f",
- ["g"] = "g",
- ["h"] = "h",
- ["i"] = "i",
- ["í"] = "i",
- ["j"] = "j",
- ["k"] = "k",
- ["l"] = "l",
- ["m"] = "m",
- ["n"] = "n",
- ["ñ"] = "ñ",
- ["o"] = "o",
- ["ó"] = "o",
- ["p"] = "p",
- ["q"] = "q",
- ["r"] = "r",
- ["s"] = "s",
- ["t"] = "t",
- ["u"] = "u",
- ["ú"] = "u",
- ["ü"] = "u",
- ["v"] = "v",
- ["w"] = "w",
- ["x"] = "x",
- ["y"] = "y",
- ["z"] = "z",
-}
-
-mappings["es"] = {
- ["a"] = 1,
- ["á"] = 1,
- ["b"] = 2,
- ["c"] = 3,
- ["d"] = 4,
- ["e"] = 5,
- ["é"] = 5,
- ["f"] = 6,
- ["g"] = 7,
- ["h"] = 8,
- ["i"] = 9,
- ["í"] = 9,
- ["j"] = 10,
- ["k"] = 11,
- ["l"] = 12,
- ["m"] = 13,
- ["n"] = 14,
- ["ñ"] = 15,
- ["o"] = 16,
- ["ó"] = 16,
- ["p"] = 17,
- ["q"] = 18,
- ["r"] = 19,
- ["s"] = 20,
- ["t"] = 21,
- ["u"] = 22,
- ["ú"] = 22,
- ["ü"] = 22,
- ["v"] = 23,
- ["w"] = 24,
- ["x"] = 25,
- ["y"] = 26,
- ["z"] = 27,
+definitions["es"] = {
+ entries = {
+ ["a"] = "a", ["á"] = "a", ["b"] = "b", ["c"] = "c", ["d"] = "d",
+ ["e"] = "e", ["é"] = "e", ["f"] = "f", ["g"] = "g", ["h"] = "h",
+ ["i"] = "i", ["í"] = "i", ["j"] = "j", ["k"] = "k", ["l"] = "l",
+ ["m"] = "m", ["n"] = "n", ["ñ"] = "ñ", ["o"] = "o", ["ó"] = "o",
+ ["p"] = "p", ["q"] = "q", ["r"] = "r", ["s"] = "s", ["t"] = "t",
+ ["u"] = "u", ["ú"] = "u", ["ü"] = "u", ["v"] = "v", ["w"] = "w",
+ ["x"] = "x", ["y"] = "y", ["z"] = "z",
+ },
+ orders = {
+ "a", "á", "b", "c", "d", "e", "é", "f", "g", "h",
+ "i", "í", "j", "k", "l", "m", "n", "ñ", "o", "ó",
+ "p", "q", "r", "s", "t", "u", "ú", "ü", "v", "w",
+ "x", "y", "z",
+ }
}
-adduppercaseentries ("es")
-adduppercasemappings("es")
-
--- Portuguese
-replacements["pt"] = { --[[ None, do you miss any? ]] }
-
-entries["pt"] = {
- ["a"] = "a",
- ["á"] = "a",
- ["â"] = "a",
- ["ã"] = "a",
- ["à"] = "a",
- ["b"] = "b",
- ["c"] = "c",
- ["ç"] = "c",
- ["d"] = "d",
- ["e"] = "e",
- ["é"] = "e",
- ["ê"] = "e",
- ["f"] = "f",
- ["g"] = "g",
- ["h"] = "h",
- ["i"] = "i",
- ["í"] = "i",
- ["j"] = "j",
- ["k"] = "k",
- ["l"] = "l",
- ["m"] = "m",
- ["n"] = "n",
- ["o"] = "o",
- ["ó"] = "o",
- ["ô"] = "o",
- ["õ"] = "o",
- ["p"] = "p",
- ["q"] = "q",
- ["r"] = "r",
- ["s"] = "s",
- ["t"] = "t",
- ["u"] = "u",
- ["ú"] = "u",
- ["ü"] = "u", -- qüinqüelíngüe
- ["v"] = "v",
- ["w"] = "w",
- ["x"] = "x",
- ["y"] = "y",
- ["z"] = "z",
-}
-
-mappings["pt"] = {
- ["a"] = 1,
- ["á"] = 1,
- ["â"] = 1,
- ["ã"] = 1,
- ["à"] = 1,
- ["b"] = 2,
- ["c"] = 3,
- ["ç"] = 3,
- ["d"] = 4,
- ["e"] = 5,
- ["é"] = 5,
- ["ê"] = 5,
- ["f"] = 6,
- ["g"] = 7,
- ["h"] = 8,
- ["i"] = 9,
- ["í"] = 9,
- ["j"] = 10,
- ["k"] = 11,
- ["l"] = 12,
- ["m"] = 13,
- ["n"] = 14,
- ["o"] = 15,
- ["ó"] = 15,
- ["ô"] = 15,
- ["õ"] = 15,
- ["p"] = 16,
- ["q"] = 17,
- ["r"] = 18,
- ["s"] = 19,
- ["t"] = 20,
- ["u"] = 21,
- ["ú"] = 21,
- ["ü"] = 21,
- ["v"] = 22,
- ["w"] = 23,
- ["x"] = 24,
- ["y"] = 25,
- ["z"] = 26,
+definitions["pt"] = {
+ entries = {
+ ["a"] = "a", ["á"] = "a", ["â"] = "a", ["ã"] = "a", ["à"] = "a",
+ ["b"] = "b", ["c"] = "c", ["ç"] = "c", ["d"] = "d", ["e"] = "e",
+ ["é"] = "e", ["ê"] = "e", ["f"] = "f", ["g"] = "g", ["h"] = "h",
+ ["i"] = "i", ["í"] = "i", ["j"] = "j", ["k"] = "k", ["l"] = "l",
+ ["m"] = "m", ["n"] = "n", ["o"] = "o", ["ó"] = "o", ["ô"] = "o",
+ ["õ"] = "o", ["p"] = "p", ["q"] = "q", ["r"] = "r", ["s"] = "s",
+ ["t"] = "t", ["u"] = "u", ["ú"] = "u", ["ü"] = "u", ["v"] = "v",
+ ["w"] = "w", ["x"] = "x", ["y"] = "y", ["z"] = "z",
+ },
+ orders = {
+ "a", "á", "â", "ã", "à", "b", "c", "ç", "d", "e",
+ "é", "ê", "f", "g", "h", "i", "í", "j", "k", "l",
+ "m", "n", "o", "ó", "ô", "õ", "p", "q", "r", "s",
+ "t", "u", "ú", "ü", "v", "w", "x", "y", "z",
+ }
}
-adduppercaseentries ("pt")
-adduppercasemappings("pt")
-
-
--- Lithuanian
-local lt_ch = uc(replacementoffset + 1)
-local lt_CH = uc(replacementoffset + 2)
-
-replacements["lt"] = {
- [1] = { "ch", lt_ch }
+local ch, CH = utfchar(replacementoffset + 1), utfchar(replacementoffset + 11)
+
+definitions["lt"] = {
+ replacements = {
+ { "ch", ch }, { "CH", CH}
+ },
+ entries = {
+ ["a"] = "a", ["ą"] = "a", ["b"] = "b", ["c"] = "c", [ch ] = "c",
+ ["č"] = "č", ["d"] = "d", ["e"] = "e", ["ę"] = "e", ["ė"] = "e",
+ ["f"] = "f", ["g"] = "g", ["h"] = "h", ["i"] = "i", ["į"] = "i",
+ ["y"] = "i", ["j"] = "j", ["k"] = "k", ["l"] = "l", ["m"] = "m",
+ ["n"] = "n", ["o"] = "o", ["p"] = "p", ["r"] = "r", ["s"] = "s",
+ ["š"] = "š", ["t"] = "t", ["u"] = "u", ["ų"] = "u", ["ū"] = "u",
+ ["v"] = "v", ["z"] = "z", ["ž"] = "ž",
+ },
+ orders = {
+ "a", "ą", "b", "c", ch, "č", "d", "e", "ę", "ė",
+ "f", "g", "h", "i", "į", "y", "j", "k", "l", "m",
+ "n", "o", "p", "r", "s", "š", "t", "u", "ų", "ū",
+ "v", "z", "ž",
+ },
+ lower = {
+ ch = CH,
+ },
+ upper = {
+ CH = ch,
+ },
}
-entries["lt"] = {
- ["a"] = "a",
- ["ą"] = "a",
- ["b"] = "b",
- ["c"] = "c",
- [lt_ch] = "c",
- ["č"] = "č",
- ["d"] = "d",
- ["e"] = "e",
- ["ę"] = "e",
- ["ė"] = "e",
- ["f"] = "f",
- ["g"] = "g",
- ["h"] = "h",
- ["i"] = "i",
- ["į"] = "i",
- ["y"] = "i",
- ["j"] = "j",
- ["k"] = "k",
- ["l"] = "l",
- ["m"] = "m",
- ["n"] = "n",
- ["o"] = "o",
- ["p"] = "p",
- ["r"] = "r",
- ["s"] = "s",
- ["š"] = "š",
- ["t"] = "t",
- ["u"] = "u",
- ["ų"] = "u",
- ["ū"] = "u",
- ["v"] = "v",
- ["z"] = "z",
- ["ž"] = "ž",
-}
-
-mappings["lt"] = {
- ["a"] = 1,
- ["ą"] = 1,
- ["b"] = 2,
- ["c"] = 3,
- [lt_ch] = 3,
- ["č"] = 4,
- ["d"] = 5,
- ["e"] = 6,
- ["ę"] = 6,
- ["ė"] = 6,
- ["f"] = 7,
- ["g"] = 8,
- ["h"] = 9,
- ["i"] = 10,
- ["į"] = 10,
- ["y"] = 10,
- ["j"] = 11,
- ["k"] = 12,
- ["l"] = 13,
- ["m"] = 14,
- ["n"] = 15,
- ["o"] = 16,
- ["p"] = 17,
- ["r"] = 18,
- ["s"] = 19,
- ["š"] = 20,
- ["t"] = 21,
- ["u"] = 22,
- ["ų"] = 22,
- ["ū"] = 22,
- ["v"] = 23,
- ["z"] = 24,
- ["ž"] = 25,
-}
-
-adduppercaseentries ("lt")
-adduppercasemappings("lt")
-
-entries ["lt"][lt_CH] = entries ["lt"][lt_ch]
-mappings["lt"][lt_CH] = mappings["lt"][lt_ch]
-
--- Latvian
-replacements["lv"] = { --[[ None, do you miss any? ]] }
-
-entries["lv"] = {
- ["a"] = "a",
- ["ā"] = "a",
- ["b"] = "b",
- ["c"] = "c",
- ["č"] = "č",
- ["d"] = "d",
- ["e"] = "e",
- ["ē"] = "e",
- ["f"] = "f",
- ["g"] = "g",
- ["ģ"] = "ģ",
- ["h"] = "h",
- ["i"] = "i",
- ["ī"] = "i",
- ["j"] = "j",
- ["k"] = "k",
- ["ķ"] = "ķ",
- ["l"] = "l",
- ["ļ"] = "ļ",
- ["m"] = "m",
- ["n"] = "n",
- ["ņ"] = "ņ",
- ["o"] = "o",
- ["ō"] = "o",
- ["p"] = "p",
- ["r"] = "r",
- ["ŗ"] = "ŗ",
- ["s"] = "s",
- ["š"] = "š",
- ["t"] = "t",
- ["u"] = "u",
- ["ū"] = "u",
- ["v"] = "v",
- ["z"] = "z",
- ["ž"] = "ž",
-}
-
-mappings["lv"] = {
- ["a"] = 1,
- ["ā"] = 1,
- ["b"] = 2,
- ["c"] = 3,
- ["č"] = 4,
- ["d"] = 5,
- ["e"] = 6,
- ["ē"] = 6,
- ["f"] = 7,
- ["g"] = 8,
- ["ģ"] = 9,
- ["h"] = 10,
- ["i"] = 11,
- ["ī"] = 11,
- ["j"] = 12,
- ["k"] = 13,
- ["ķ"] = 14,
- ["l"] = 15,
- ["ļ"] = 16,
- ["m"] = 17,
- ["n"] = 18,
- ["ņ"] = 19,
- ["o"] = 20,
- ["ō"] = 20,
- ["p"] = 21,
- ["r"] = 22,
- ["ŗ"] = 23,
- ["s"] = 24,
- ["š"] = 25,
- ["t"] = 26,
- ["u"] = 27,
- ["ū"] = 27,
- ["v"] = 28,
- ["z"] = 29,
- ["ž"] = 30,
+definitions["lv"] = {
+ entries = {
+ ["a"] = "a", ["ā"] = "a", ["b"] = "b", ["c"] = "c", ["č"] = "č",
+ ["d"] = "d", ["e"] = "e", ["ē"] = "e", ["f"] = "f", ["g"] = "g",
+ ["ģ"] = "ģ", ["h"] = "h", ["i"] = "i", ["ī"] = "i", ["j"] = "j",
+ ["k"] = "k", ["ķ"] = "ķ", ["l"] = "l", ["ļ"] = "ļ", ["m"] = "m",
+ ["n"] = "n", ["ņ"] = "ņ", ["o"] = "o", ["ō"] = "o", ["p"] = "p",
+ ["r"] = "r", ["ŗ"] = "ŗ", ["s"] = "s", ["š"] = "š", ["t"] = "t",
+ ["u"] = "u", ["ū"] = "u", ["v"] = "v", ["z"] = "z", ["ž"] = "ž",
+ },
+ orders = {
+ "a", "ā", "b", "c", "č", "d", "e", "ē", "f", "g",
+ "ģ", "h", "i", "ī", "j", "k", "ķ", "l", "ļ", "m",
+ "n", "ņ", "o", "ō", "p", "r", "ŗ", "s", "š", "t",
+ "u", "ū", "v", "z", "ž",
+ }
}
-adduppercaseentries ("lv")
-adduppercasemappings("lv")
-
--- Hungarian
-- Helpful but disturbing:
@@ -2464,230 +791,68 @@ adduppercasemappings("lv")
-- for sequences like “nny”, “ssz”, and “zsz”. This is left as an exercise to
-- the reader…)
-local hu_cs = uc(replacementoffset + 1)
-local hu_CS = uc(replacementoffset + 2)
-
-local hu_dz = uc(replacementoffset + 3)
-local hu_DZ = uc(replacementoffset + 4)
-
-local hu_dzs = uc(replacementoffset + 5)
-local hu_DZS = uc(replacementoffset + 6)
-
-local hu_gy = uc(replacementoffset + 7)
-local hu_GY = uc(replacementoffset + 8)
-
-local hu_ly = uc(replacementoffset + 9)
-local hu_LY = uc(replacementoffset + 10)
-
-local hu_ny = uc(replacementoffset + 11)
-local hu_NY = uc(replacementoffset + 12)
-
-local hu_sz = uc(replacementoffset + 13)
-local hu_SZ = uc(replacementoffset + 14)
-
-local hu_ty = uc(replacementoffset + 15)
-local hu_TY = uc(replacementoffset + 16)
-
-local hu_zs = uc(replacementoffset + 17)
-local hu_ZS = uc(replacementoffset + 18)
-
-replacements["hu"] = {
- [1] = { "cs", hu_cs },
- [2] = { "dz", hu_dz },
- [3] = { "dzs", hu_dzs },
- [4] = { "gy", hu_gy },
- [5] = { "ly", hu_ly },
- [6] = { "ny", hu_ny },
- [7] = { "sz", hu_sz },
- [8] = { "ty", hu_ty },
- [9] = { "zs", hu_zs },
-}
-
-entries["hu"] = {
- ["a"] = "a",
- ["á"] = "a",
- ["b"] = "b",
- ["c"] = "c",
- [hu_cs] = "cs",
- ["d"] = "d",
- [hu_dz] = "dz",
- [hu_dzs] = "dzs",
- ["e"] = "e",
- ["é"] = "e",
- ["f"] = "f",
- ["g"] = "g",
- [hu_gy] = "gy",
- ["h"] = "h",
- ["i"] = "i",
- ["í"] = "i",
- ["j"] = "j",
- ["k"] = "k",
- ["l"] = "l",
- [hu_ly] = "ly",
- ["m"] = "m",
- ["n"] = "n",
- [hu_ny] = "ny",
- ["o"] = "o",
- ["ó"] = "o",
- ["ö"] = "ö",
- ["ő"] = "ö",
- ["p"] = "p",
- ["q"] = "q",
- ["r"] = "r",
- ["s"] = "s",
- [hu_sz] = "sz",
- ["t"] = "t",
- [hu_ty] = "ty",
- ["u"] = "u",
- ["ú"] = "u",
- ["ü"] = "ü",
- ["ű"] = "ü",
- ["v"] = "v",
- ["w"] = "w",
- ["x"] = "x",
- ["y"] = "y",
- ["z"] = "z",
- [hu_zs] = "zs",
-}
-
-mappings["hu"] = {
- ["a"] = 1,
- ["á"] = 1, -- -> a
- ["b"] = 2,
- ["c"] = 3,
- [hu_cs] = 4,
- ["d"] = 5,
- [hu_dz] = 6,
- [hu_dzs] = 7,
- ["e"] = 8,
- ["é"] = 8, -- -> e
- ["f"] = 9,
- ["g"] = 10,
- [hu_gy] = 11,
- ["h"] = 12,
- ["i"] = 13,
- ["í"] = 13, -- -> i
- ["j"] = 14,
- ["k"] = 15,
- ["l"] = 16,
- [hu_ly] = 17,
- ["m"] = 18,
- ["n"] = 19,
- [hu_ny] = 20,
- ["o"] = 21,
- ["ó"] = 21, -- -> o
- ["ö"] = 22,
- ["ő"] = 22, -- -> ö
- ["p"] = 23,
- ["q"] = 24,
- ["r"] = 25,
- ["s"] = 26,
- [hu_sz] = 27,
- ["t"] = 28,
- [hu_ty] = 29,
- ["u"] = 30,
- ["ú"] = 30, -- -> u
- ["ü"] = 31,
- ["ű"] = 31, -- -> ü
- ["v"] = 32,
- ["w"] = 33,
- ["x"] = 34,
- ["y"] = 35,
- ["z"] = 36,
- [hu_zs] = 37,
+local cs, CS = utfchar(replacementoffset + 1), utfchar(replacementoffset + 11)
+local dz, DZ = utfchar(replacementoffset + 2), utfchar(replacementoffset + 12)
+local dzs, DZS = utfchar(replacementoffset + 3), utfchar(replacementoffset + 13)
+local gy, GY = utfchar(replacementoffset + 4), utfchar(replacementoffset + 14)
+local ly, LY = utfchar(replacementoffset + 5), utfchar(replacementoffset + 15)
+local ny, NY = utfchar(replacementoffset + 6), utfchar(replacementoffset + 16)
+local sz, SZ = utfchar(replacementoffset + 7), utfchar(replacementoffset + 17)
+local ty, TY = utfchar(replacementoffset + 8), utfchar(replacementoffset + 18)
+local zs, ZS = utfchar(replacementoffset + 9), utfchar(replacementoffset + 19)
+
+definitions["hu"] = {
+ replacements = {
+ { "cs", cs }, { "CS", CS },
+ { "dz", dz }, { "DZ", DZ },
+ { "dzs", dzs }, { "DZS", DZS },
+ { "gy", gy }, { "GY", GY },
+ { "ly", ly }, { "LY", LY },
+ { "ny", ny }, { "NY", NY },
+ { "sz", sz }, { "SZ", SZ },
+ { "ty", ty }, { "TY", TY },
+ { "zs", zs }, { "ZS", ZS },
+ },
+ entries = {
+ ["a"] = "a", ["á"] = "a", ["b"] = "b", ["c"] = "c", [cs ] = "cs",
+ ["d"] = "d", [dz ] = "dz", [dzs] = "dzs", ["e"] = "e", ["é"] = "e",
+ ["f"] = "f", ["g"] = "g", [gy ] = "gy", ["h"] = "h", ["i"] = "i",
+ ["í"] = "i", ["j"] = "j", ["k"] = "k", ["l"] = "l", [ly ] = "ly",
+ ["m"] = "m", ["n"] = "n", [ny ] = "ny", ["o"] = "o", ["ó"] = "o",
+ ["ö"] = "ö", ["ő"] = "ö", ["p"] = "p", ["q"] = "q", ["r"] = "r",
+ ["s"] = "s", [sz ] = "sz", ["t"] = "t", [ty ] = "ty", ["u"] = "u",
+ ["ú"] = "u", ["ü"] = "ü", ["ű"] = "ü", ["v"] = "v", ["w"] = "w",
+ ["x"] = "x", ["y"] = "y", ["z"] = "z", [zs ] = "zs",
+ },
+ orders = {
+ "a", "á", "b", "c", cs, "d", dz, dzs, "e", "é",
+ "f", "g", gy, "h", "i", "í", "j", "k", "l", ly,
+ "m", "n", ny, "o", "ó", "ö", "ő", "p", "q", "r",
+ "s", sz, "t", ty, "u", "ú", "ü", "ű", "v", "w",
+ "x", "y", "z", zs,
+ },
+ lower = {
+ CS = cs, DZ = dz, DZS = dzs, GY = gy, LY = ly, NY = ny, SZ = sz, TY = ty, ZS = zs,
+ },
+ upper = {
+ cs = CS, dz = DZ, dzs = DZS, gy = GY, ly = LY, ny = NY, sz = SZ, ty = TY, zs = ZS,
+ },
}
-adduppercaseentries ("hu")
-adduppercasemappings("hu")
-
-entries ["hu"] [hu_CS] = entries ["hu"] [hu_cs]
-mappings["hu"] [hu_CS] = mappings["hu"] [hu_cs]
-entries ["hu"] [hu_DZ] = entries ["hu"] [hu_dz]
-mappings["hu"] [hu_DZ] = mappings["hu"] [hu_dz]
-entries ["hu"][hu_DZS] = entries ["hu"][hu_dzs]
-mappings["hu"][hu_DZS] = mappings["hu"][hu_dzs]
-entries ["hu"] [hu_GY] = entries ["hu"] [hu_gy]
-mappings["hu"] [hu_GY] = mappings["hu"] [hu_gy]
-entries ["hu"] [hu_LY] = entries ["hu"] [hu_ly]
-mappings["hu"] [hu_LY] = mappings["hu"] [hu_ly]
-entries ["hu"] [hu_NY] = entries ["hu"] [hu_ny]
-mappings["hu"] [hu_NY] = mappings["hu"] [hu_ny]
-entries ["hu"] [hu_SZ] = entries ["hu"] [hu_sz]
-mappings["hu"] [hu_SZ] = mappings["hu"] [hu_sz]
-entries ["hu"] [hu_TY] = entries ["hu"] [hu_ty]
-mappings["hu"] [hu_TY] = mappings["hu"] [hu_ty]
-entries ["hu"] [hu_ZS] = entries ["hu"] [hu_zs]
-mappings["hu"] [hu_ZS] = mappings["hu"] [hu_zs]
-
--- Estonian
-replacements["et"] = { --[[ None, do you miss any? ]] }
-
-entries["et"] = {
- ["a"] = "a",
- ["b"] = "b",
- ["d"] = "d",
- ["e"] = "e",
- ["f"] = "f",
- ["g"] = "g",
- ["h"] = "h",
- ["i"] = "i",
- ["j"] = "j",
- ["k"] = "k",
- ["l"] = "l",
- ["m"] = "m",
- ["n"] = "n",
- ["o"] = "o",
- ["p"] = "p",
- ["r"] = "r",
- ["s"] = "s",
- ["š"] = "š",
- ["z"] = "z",
- ["ž"] = "ž",
- ["t"] = "t",
- ["u"] = "u",
- ["v"] = "v",
- ["w"] = "v", -- foreign words only
- ["õ"] = "õ",
- ["ä"] = "ä",
- ["ö"] = "ö",
- ["ü"] = "ü",
- ["x"] = "x", --foreign words only
- ["y"] = "y", --foreign words only
+definitions["et"] = {
+ entries = { -- w x y are used for foreign words only
+ ["a"] = "a", ["b"] = "b", ["d"] = "d", ["e"] = "e", ["f"] = "f",
+ ["g"] = "g", ["h"] = "h", ["i"] = "i", ["j"] = "j", ["k"] = "k",
+ ["l"] = "l", ["m"] = "m", ["n"] = "n", ["o"] = "o", ["p"] = "p",
+ ["r"] = "r", ["s"] = "s", ["š"] = "š", ["z"] = "z", ["ž"] = "ž",
+ ["t"] = "t", ["u"] = "u", ["v"] = "v", ["w"] = "v", ["õ"] = "õ",
+ ["ä"] = "ä", ["ö"] = "ö", ["ü"] = "ü", ["x"] = "x", ["y"] = "y",
+ },
+ orders = {
+ "a", "b", "d", "e", "f", "g", "h", "i", "j", "k",
+ "l", "m", "n", "o", "p", "r", "s", "š", "z", "ž",
+ "t", "u", "v", "w", "õ", "ä", "ö", "ü", "x", "y",
+ }
}
-
-mappings["et"] = {
- ["a"] = 1,
- ["b"] = 2,
- ["d"] = 3,
- ["e"] = 4,
- ["f"] = 5,
- ["g"] = 6,
- ["h"] = 7,
- ["i"] = 8,
- ["j"] = 9,
- ["k"] = 10,
- ["l"] = 11,
- ["m"] = 12,
- ["n"] = 13,
- ["o"] = 14,
- ["p"] = 15,
- ["r"] = 16,
- ["s"] = 17,
- ["š"] = 18,
- ["z"] = 19,
- ["ž"] = 20,
- ["t"] = 21,
- ["u"] = 22,
- ["v"] = 23,
- ["w"] = 23,
- ["õ"] = 24,
- ["ä"] = 25,
- ["ö"] = 26,
- ["ü"] = 27,
- ["x"] = 28,
- ["y"] = 29,
-}
-
-adduppercaseentries ("et")
-adduppercasemappings("et")
diff --git a/tex/context/base/strc-ref.lua b/tex/context/base/strc-ref.lua
index 9345dc3bf..bced781fd 100644
--- a/tex/context/base/strc-ref.lua
+++ b/tex/context/base/strc-ref.lua
@@ -7,7 +7,7 @@ if not modules then modules = { } end modules ['strc-ref'] = {
}
local format, find, gmatch, match, concat = string.format, string.find, string.gmatch, string.match, table.concat
-local lpegmatch = lpeg.match
+local lpegmatch, lpegP, lpegCs = lpeg.match, lpeg.P, lpeg.Cs
local texsprint, texwrite, texcount, texsetcount = tex.sprint, tex.write, tex.count, tex.setcount
local allocate, mark = utilities.storage.allocate, utilities.storage.mark
@@ -80,7 +80,7 @@ function references.registerfinalizer(func) -- we could use a token register ins
finalizers[#finalizers+1] = func
end
-local function initializer()
+local function initializer() -- can we use a tobesaved as metatable for collected?
tobesaved = mark(references.tobesaved)
collected = mark(references.collected)
for i=1,#initializers do
@@ -163,9 +163,9 @@ end
local result = { }
-local lparent, rparent, lbrace, rbrace, dcolon, backslash = lpeg.P("("), lpeg.P(")"), lpeg.P("{"), lpeg.P("}"), lpeg.P("::"), lpeg.P("\\")
+local lparent, rparent, lbrace, rbrace, dcolon, backslash = lpegP("("), lpegP(")"), lpegP("{"), lpegP("}"), lpegP("::"), lpegP("\\")
-local reset = lpeg.P("") / function() result = { } end
+local reset = lpegP("") / function() result = { } end
local b_token = backslash / function(s) result.has_tex = true return s end
local o_token = 1 - rparent - rbrace - lparent - lbrace
@@ -174,11 +174,11 @@ local s_token = 1 - lparent - lbrace - lparent - lbrace
local i_token = 1 - lparent - lbrace
local f_token = 1 - lparent - lbrace - dcolon
-local outer = (f_token )^1 / function (s) result.outer = s end
-local operation = lpeg.Cs((b_token + o_token)^1) / function (s) result.operation = s end
-local arguments = lpeg.Cs((b_token + a_token)^0) / function (s) result.arguments = s end
-local special = (s_token )^1 / function (s) result.special = s end
-local inner = (i_token )^1 / function (s) result.inner = s end
+local outer = (f_token )^1 / function (s) result.outer = s end
+local operation = lpegCs((b_token + o_token)^1) / function (s) result.operation = s end
+local arguments = lpegCs((b_token + a_token)^0) / function (s) result.arguments = s end
+local special = (s_token )^1 / function (s) result.special = s end
+local inner = (i_token )^1 / function (s) result.inner = s end
local outer_reference = (outer * dcolon)^0
@@ -244,6 +244,7 @@ function references.resolvers.reference(var)
end
local function register_from_lists(collected,derived)
+ local g = derived[""] if not g then g = { } derived[""] = g end -- global
for i=1,#collected do
local entry = collected[i]
local m, r = entry.metadata, entry.references
@@ -258,7 +259,8 @@ local function register_from_lists(collected,derived)
if trace_referencing then
report_references("list entry %s provides %s reference '%s' on realpage %s",i,kind,s,realpage)
end
- d[s] = t -- share them
+ d[s] = d[s] or t -- share them
+ g[s] = g[s] or t -- first wins
end
end
end
@@ -632,7 +634,9 @@ end
-- t.special t.operation t.arguments t.outer t.inner
-local prefixsplitter = lpeg.splitat(":")
+-- to what extend do we check the non prefixed variant
+
+local strict = false
local function resolve(prefix,reference,args,set) -- we start with prefix,reference
texcount.referencehastexstate = 0
@@ -646,29 +650,39 @@ local function resolve(prefix,reference,args,set) -- we start with prefix,refere
local r = settings_to_array(reference)
for i=1,#r do
local ri = r[i]
- local dp = defined[prefix] or defined[""]
- local d = dp[ri]
+ local d
+ if strict then
+ d = defined[prefix] or defined[""]
+ d = d and d[ri]
+ else
+ d = defined[prefix]
+ d = d and d[ri]
+ if not d then
+ d = defined[""]
+ d = d and d[ri]
+ end
+ end
if d then
resolve(prefix,d[2],nil,set)
else
local var = lpegmatch(scanner,ri)
if var then
var.reference = ri
- if not var.outer and var.inner then
- local d = defined[prefix]
- d = d and d[var.inner]
---~ if not d then
---~ local p, r = lpegmatch(prefixsplitter,var.inner)
---~ d = defined[p]
---~ d = d and d[r]
---~ print(p,r,d)
---~ table.print(defined)
---~ end
- if not d then
- d = defined[""]
- d = d and d[var.inner]
+ local vo, vi = var.outer, var.inner
+ if not vo and vi then
+ -- to be checked
+ if strict then
+ d = defined[prefix] or defined[""]
+ d = d and d[vi]
+ else
+ d = defined[prefix]
+ d = d and d[vi]
+ if not d then
+ d = defined[""]
+ d = d and d[vi]
+ end
end
- d = d and d[var.inner]
+ --
if d then
resolve(prefix,d[2],var.arguments,set) -- args can be nil
else
@@ -690,6 +704,7 @@ local function resolve(prefix,reference,args,set) -- we start with prefix,refere
if set.has_tex then
texcount.referencehastexstate = 1
end
+--~ table.print(set)
return set
else
return { }
@@ -746,6 +761,8 @@ end
--~ end
--~ end
+local prefixsplitter = lpegCs(lpegP((1-lpegP(":"))^1 * lpegP(":"))) * lpegCs(lpegP(1)^1)
+
local function identify(prefix,reference)
local set = resolve(prefix,reference)
local bug = false
@@ -899,47 +916,91 @@ local function identify(prefix,reference)
var.error = "unknown inner or special"
end
else
- -- inner
---~ local i = tobesaved[prefix]
- local i = collected[prefix]
- i = i and i[inner]
+ -- inner ... we could move the prefix logic into the parser so that we have 'm for each entry
+ -- foo:bar -> foo == prefix (first we try the global one)
+ -- -:bar -> ignore prefix
+ local p, i = prefix, nil
+ local splitprefix, splitinner = lpegmatch(prefixsplitter,inner)
+ -- these are taken from other anonymous references
+ if splitprefix and splitinner then
+ if splitprefix == "-" then
+ i = collected[""]
+ i = i and i[splitinner]
+ if i then
+ p = ""
+ end
+ else
+ i = collected[splitprefix]
+ i = i and i[splitinner]
+ if i then
+ p = splitprefix
+ end
+ end
+ else
+ i = collected[prefix]
+ i = i and i[inner]
+ if i then
+ p = prefix
+ end
+ end
if i then
var.i = { "reference", i }
references.resolvers.reference(var)
var.kind = "inner"
- var.p = prefix
+ var.p = p
else
- i = derived[prefix]
- i = i and i[inner]
+ -- these are taken from other data structures (like lists)
+--~ print("!!!!!!!!!!!!!!",splitprefix,splitinner)
+--~ table.print(derived)
+ if splitprefix and splitinner then
+ if splitprefix == "-" then
+ i = derived[""]
+ i = i and i[splitinner]
+ if i then
+ p = ""
+ end
+ else
+ i = derived[splitprefix]
+ i = i and i[splitinner]
+ if i then
+ p = splitprefix
+ end
+ end
+ else
+ i = derived[prefix]
+ i = i and i[inner]
+ if i then
+ p = prefix
+ end
+ end
+ if not i and prefix ~= "" then
+ i = derived[""]
+ i = i and i[inner]
+ if i then
+ p = ""
+ end
+ end
if i then
var.kind = "inner"
var.i = i
references.resolvers[i[1]](var)
- var.p = prefix
+ var.p = p
else
- i = collected[prefix]
- i = i and i[inner]
- if i then
- var.kind = "inner"
- var.i = { "reference", i }
- references.resolvers.reference(var)
- var.p = prefix
+ -- no prefixes here
+ local s = specials[inner]
+ if s then
+ var.kind = "special"
else
- local s = specials[inner]
- if s then
- var.kind = "special"
+ i = (collected[""] and collected[""][inner]) or
+ (derived [""] and derived [""][inner]) or
+ (tobesaved[""] and tobesaved[""][inner])
+ if i then
+ var.kind = "inner"
+ var.i = { "reference", i }
+ references.resolvers.reference(var)
+ var.p = ""
else
- i = (collected[""] and collected[""][inner]) or
- (derived [""] and derived [""][inner]) or
- (tobesaved[""] and tobesaved[""][inner])
- if i then
- var.kind = "inner"
- var.i = { "reference", i }
- references.resolvers.reference(var)
- var.p = ""
- else
- var.error = "unknown inner or special"
- end
+ var.error = "unknown inner or special"
end
end
end
diff --git a/tex/context/base/strc-ref.mkii b/tex/context/base/strc-ref.mkii
index 1df5d09d8..4550f7144 100644
--- a/tex/context/base/strc-ref.mkii
+++ b/tex/context/base/strc-ref.mkii
@@ -44,22 +44,6 @@
\unprotect
-% messages moved
-
-% messages moved
-
-% messages moved
-
-% messages moved
-
-% messages moved
-
-% messages moved
-
-% messages moved
-
-% messages moved
-
%D This module deals with referencing. In \CONTEXT\ referencing
%D is one of the core features, although at a first glance
%D probably nobody will notice. This is good, because
@@ -786,7 +770,7 @@
%D \starttyping
%D \pagereference[important]
%D \pagereference[unimportant]
-%D \setupreferencing[prefixprefix=sidetrack]
+%D \setupreferencing[prefix=sidetrack]
%D \pagereference[important]
%D \stoptyping
%D
diff --git a/tex/context/base/strc-reg.lua b/tex/context/base/strc-reg.lua
index 1e9781b38..23ce4d221 100644
--- a/tex/context/base/strc-reg.lua
+++ b/tex/context/base/strc-reg.lua
@@ -418,7 +418,7 @@ function registers.analyzed(class,options)
local data = collected[class]
if data and data.entries then
options = options or { }
- sorters.setlanguage(options.language)
+ sorters.setlanguage(options.language,options.method)
registers.filter(data,options) -- filter entries into results (criteria)
registers.prepare(data,options) -- adds split table parallel to list table
registers.sort(data,options) -- sorts results
diff --git a/tex/context/base/strc-reg.mkiv b/tex/context/base/strc-reg.mkiv
index 118ba2a35..37c849790 100644
--- a/tex/context/base/strc-reg.mkiv
+++ b/tex/context/base/strc-reg.mkiv
@@ -122,6 +122,7 @@
\c!pagesegments=2:2,
\c!file=\jobname,
%\c!deeptextcommand=, % undefined by default !
+ \c!method=, % no default as we have them in the module, maybe some day in lang-*
\s!language=\currentmainlanguage]%
%D Definition:
@@ -443,6 +444,8 @@
%D Rendering:
+% todo: c!language ipv s!language
+
\let\utilityregisterlength\!!zerocount
\def\determineregistercharacteristics
@@ -454,6 +457,7 @@
\setupregister[\currentregister][#2]%
\normalexpanded{\endgroup\noexpand\xdef\noexpand\utilityregisterlength{\ctxlua{structures.registers.analyze('\currentregister',{
language = "\registerparameter\s!language",
+ method = "\registerparameter\c!method",
compress = "\registerparameter\c!compress",
criterium = "\registerparameter\c!criterium",
})}}}% brrr
@@ -487,6 +491,7 @@
\startpacked[\v!blank]%
\ctxlua{structures.registers.process('\currentregister',{
language = "\registerparameter\s!language",
+ method = "\registerparameter\c!method",
compress = "\registerparameter\c!compress",
criterium = "\registerparameter\c!criterium",
},
diff --git a/tex/context/base/strc-sec.mkiv b/tex/context/base/strc-sec.mkiv
index b2b962183..2a6d25631 100644
--- a/tex/context/base/strc-sec.mkiv
+++ b/tex/context/base/strc-sec.mkiv
@@ -287,22 +287,35 @@
\let\structureheadprefix\empty \def\structureheadprefixplus{+}
% When do we reset the referenceprefix? This needs to be checked. Does it work
-% at all?
+% at all? Todo: pushpop.
+
+% \def\setstructureheadreference
+% {\edef\structureheadreference {\structureheadparameter\c!reference}%
+% \edef\structureheadreferenceprefix{\structureheadparameter\c!prefix}%
+% \ifx\structureheadreferenceprefix\empty
+% \ifx\currentrefenceprefix\structureheadreferenceprefixplus
+% \setupreferenceprefix[]% yes or no? only when set by structure
+% \fi
+% \else\ifx\structureheadreferenceprefix\structureheadreferenceprefixplus
+% \ifx\structureheadreference\empty
+% \expanded{\setupreferenceprefix[\structureheadreferenceprefixplus]}%
+% \else
+% \expanded{\setupreferenceprefix[\structureheadreference]}% we assume just one reference
+% \fi
+% \else
+% \expanded{\setupreferenceprefix[\structureheadreferenceprefix]}%
+% \fi\fi}
+
+% pop in stopsection commands unless empty
\def\setstructureheadreference
{\edef\structureheadreference {\structureheadparameter\c!reference}%
\edef\structureheadreferenceprefix{\structureheadparameter\c!prefix}%
\ifx\structureheadreferenceprefix\empty
- \setupreferenceprefix[]% yes or no?
- \else\ifx\structureheadreferenceprefix\structureheadreferenceprefixplus
- \ifx\structureheadreference\empty
- \expanded{\setupreferenceprefix[\structureheadreferenceprefixplus]}%
- \else
- \expanded{\setupreferenceprefix[\structureheadreference]}% we assume just one reference
- \fi
+ \let\structureheadreferenceprefix\referenceprefix
\else
- \expanded{\setupreferenceprefix[\structureheadreferenceprefix]}%
- \fi\fi}
+ \setupreferenceprefix[\structureheadreferenceprefix]% currenty no pop
+ \fi}
\setvalue{\??nh:\c!incrementnumber:\v!yes }{\settrue \structureheaddoincrement\settrue \structureheadtolist}
\setvalue{\??nh:\c!incrementnumber:\v!no }{\setfalse\structureheaddoincrement\setfalse\structureheadtolist}
diff --git a/tex/context/base/strc-syn.mkiv b/tex/context/base/strc-syn.mkiv
index 1c59ad7a4..d46edaa39 100644
--- a/tex/context/base/strc-syn.mkiv
+++ b/tex/context/base/strc-syn.mkiv
@@ -96,6 +96,7 @@
%\c!after=,
\c!indentnext=\v!no,
%\c!expansion=,
+ \c!method=,
\s!language=\currentmainlanguage]
\unexpanded\def\definesynonyms
@@ -200,7 +201,11 @@
\c!color=.
#2]%
\startpacked
- \ctxlua{structures.synonyms.process('#1',{ criterium = "\synonymparameter\c!criterium" })}%
+ \ctxlua{structures.synonyms.process('#1',{
+ criterium = "\synonymparameter\c!criterium",
+ language = "\synonymparameter\s!language",
+ method = "\synonymparameter\c!method",
+ })}%
\stoppacked
\endgroup}
@@ -263,6 +268,7 @@
%\c!before=,
\c!after=\endgraf,
%\c!expansion=,
+ \c!method=,
\s!language=\currentmainlanguage]
\unexpanded\def\definesorting
@@ -344,7 +350,11 @@
\def\currentsorting{#1}%
\getparameters[\??so#1][#2]%
\startpacked
- \ctxlua{structures.synonyms.process('#1',{ criterium = "\sortingparameter\c!criterium" })}%
+ \ctxlua{structures.synonyms.process('#1',{
+ criterium = "\sortingparameter\c!criterium",
+ language = "\sortingparameter\s!language",
+ method = "\sortingparameter\c!method",
+ })}%
\stoppacked
\endgroup}
diff --git a/tex/generic/context/luatex-fonts-merged.lua b/tex/generic/context/luatex-fonts-merged.lua
index 102f0db3b..4c221ae12 100644
--- a/tex/generic/context/luatex-fonts-merged.lua
+++ b/tex/generic/context/luatex-fonts-merged.lua
@@ -1,6 +1,6 @@
-- merged file : luatex-fonts-merged.lua
-- parent file : luatex-fonts.lua
--- merge date : 09/16/10 10:46:45
+-- merge date : 09/22/10 09:35:53
do -- begin closure to overcome local limits and interference
@@ -736,7 +736,7 @@ local function sortedkeys(tab)
elseif tkey == "number" then
-- if kind == 1 then kind = 3 else kind = 2 end
kind = (kind == 1 and 3) or 2
- else
+ else -- if tkey then
kind = 3
end
end
@@ -752,7 +752,9 @@ end
local function sortedhashkeys(tab) -- fast one
local srt = { }
for key,_ in next, tab do
- srt[#srt+1] = key
+ if key then
+ srt[#srt+1] = key
+ end
end
sort(srt)
return srt