if not modules then modules = { } end modules ['font-afm'] = {
    version   = 1.001,
    comment   = "companion to font-ini.mkiv",
    author    = "Hans Hagen, PRAGMA-ADE, Hasselt NL",
    copyright = "PRAGMA ADE / ConTeXt Development Team",
    license   = "see context related readme files"
}

--[[ldx--
<p>Some code may look a bit obscure but this has to do with the
fact that we also use this code for testing and much code evolved
in the transition from <l n='tfm'/> to <l n='afm'/> to <l
n='otf'/>.</p>

<p>The following code still has traces of intermediate font support
where we handles font encodings. Eventually font encoding goes
away.</p>

<p>The embedding of a font involves creating temporary files and
depending on your system setup that can fail. It took more than a
day to figure out why sometimes embedding failed in mingw luatex
where running on a real path like c:\... failed while running on
say e:\... being a link worked well. The native windows binaries
don't have this issue.</p>

--ldx]]--

local fonts, logs, trackers, containers, resolvers = fonts, logs, trackers, containers, resolvers

local next, type, tonumber = next, type, tonumber
local match, gmatch, lower, gsub, strip, find = string.match, string.gmatch, string.lower, string.gsub, string.strip, string.find
local char, byte, sub = string.char, string.byte, string.sub
local abs, mod = math.abs, math.mod
local bxor, rshift = bit32.bxor, bit32.rshift
local P, S, R, Cmt, C, Ct, Cs, lpegmatch, patterns = lpeg.P, lpeg.S, lpeg.R, lpeg.Cmt, lpeg.C, lpeg.Ct, lpeg.Cs, lpeg.match, lpeg.patterns
local derivetable = table.derive

local trace_features     = false  trackers.register("afm.features",   function(v) trace_features = v end)
local trace_indexing     = false  trackers.register("afm.indexing",   function(v) trace_indexing = v end)
local trace_loading      = false  trackers.register("afm.loading",    function(v) trace_loading  = v end)
local trace_defining     = false  trackers.register("fonts.defining", function(v) trace_defining = v end)

local report_afm         = logs.reporter("fonts","afm loading")

local setmetatableindex  = table.setmetatableindex

local findbinfile        = resolvers.findbinfile

local definers           = fonts.definers
local readers            = fonts.readers
local constructors       = fonts.constructors

local afm                = constructors.newhandler("afm")
local pfb                = constructors.newhandler("pfb")

local afmfeatures        = constructors.newfeatures("afm")
local registerafmfeature = afmfeatures.register

afm.version              = 1.500 -- incrementing this number one up will force a re-cache
afm.cache                = containers.define("fonts", "afm", afm.version, true)
afm.autoprefixed         = true -- this will become false some day (catches texnansi-blabla.*)

afm.helpdata             = { }  -- set later on so no local for this
afm.syncspace            = true -- when true, nicer stretch values
afm.addligatures         = true -- best leave this set to true
afm.addtexligatures      = true -- best leave this set to true
afm.addkerns             = true -- best leave this set to true

local overloads          = fonts.mappings.overloads

local applyruntimefixes  = fonts.treatments and fonts.treatments.applyfixes

local function setmode(tfmdata,value)
    if value then
        tfmdata.properties.mode = lower(value)
    end
end

registerafmfeature {
    name         = "mode",
    description  = "mode",
    initializers = {
        base = setmode,
        node = setmode,
    }
}

--[[ldx--
<p>We start with the basic reader which we give a name similar to the
built in <l n='tfm'/> and <l n='otf'/> reader.</p>
--ldx]]--

--~ Comment FONTIDENTIFIER LMMATHSYMBOLS10
--~ Comment CODINGSCHEME TEX MATH SYMBOLS
--~ Comment DESIGNSIZE 10.0 pt
--~ Comment CHECKSUM O 4261307036
--~ Comment SPACE 0 plus 0 minus 0
--~ Comment QUAD 1000
--~ Comment EXTRASPACE 0
--~ Comment NUM 676.508 393.732 443.731
--~ Comment DENOM 685.951 344.841
--~ Comment SUP 412.892 362.892 288.889
--~ Comment SUB 150 247.217
--~ Comment SUPDROP 386.108
--~ Comment SUBDROP 50
--~ Comment DELIM 2390 1010
--~ Comment AXISHEIGHT 250

local comment = P("Comment")
local spacing = patterns.spacer  -- S(" \t")^1
local lineend = patterns.newline -- S("\n\r")
local words   = C((1 - lineend)^1)
local number  = C((R("09") + S("."))^1) / tonumber * spacing^0
local data    = lpeg.Carg(1)

local pattern = ( -- needs testing ... not used anyway as we no longer need math afm's
    comment * spacing *
        (
            data * (
                ("CODINGSCHEME" * spacing * words                                      ) / function(fd,a)                                      end +
                ("DESIGNSIZE"   * spacing * number * words                             ) / function(fd,a)     fd[ 1]                 = a       end +
                ("CHECKSUM"     * spacing * number * words                             ) / function(fd,a)     fd[ 2]                 = a       end +
                ("SPACE"        * spacing * number * "plus" * number * "minus" * number) / function(fd,a,b,c) fd[ 3], fd[ 4], fd[ 5] = a, b, c end +
                ("QUAD"         * spacing * number                                     ) / function(fd,a)     fd[ 6]                 = a       end +
                ("EXTRASPACE"   * spacing * number                                     ) / function(fd,a)     fd[ 7]                 = a       end +
                ("NUM"          * spacing * number * number * number                   ) / function(fd,a,b,c) fd[ 8], fd[ 9], fd[10] = a, b, c end +
                ("DENOM"        * spacing * number * number                            ) / function(fd,a,b  ) fd[11], fd[12]         = a, b    end +
                ("SUP"          * spacing * number * number * number                   ) / function(fd,a,b,c) fd[13], fd[14], fd[15] = a, b, c end +
                ("SUB"          * spacing * number * number                            ) / function(fd,a,b)   fd[16], fd[17]         = a, b    end +
                ("SUPDROP"      * spacing * number                                     ) / function(fd,a)     fd[18]                 = a       end +
                ("SUBDROP"      * spacing * number                                     ) / function(fd,a)     fd[19]                 = a       end +
                ("DELIM"        * spacing * number * number                            ) / function(fd,a,b)   fd[20], fd[21]         = a, b    end +
                ("AXISHEIGHT"   * spacing * number                                     ) / function(fd,a)     fd[22]                 = a       end
            )
          + (1-lineend)^0
        )
  + (1-comment)^1
)^0

local function scan_comment(str)
    local fd = { }
    lpegmatch(pattern,str,1,fd)
    return fd
end

-- On a rainy day I will rewrite this in lpeg ... or we can use the (slower) fontloader
-- as in now supports afm/pfb loading but it's not too bad to have different methods
-- for testing approaches.

local keys = { }

function keys.FontName    (data,line) data.metadata.fontname    = strip    (line) -- get rid of spaces
                                      data.metadata.fullname    = strip    (line) end
function keys.ItalicAngle (data,line) data.metadata.italicangle = tonumber (line) end
function keys.IsFixedPitch(data,line) data.metadata.monospaced  = toboolean(line,true) end
function keys.CharWidth   (data,line) data.metadata.charwidth   = tonumber (line) end
function keys.XHeight     (data,line) data.metadata.xheight     = tonumber (line) end
function keys.Descender   (data,line) data.metadata.descender   = tonumber (line) end
function keys.Ascender    (data,line) data.metadata.ascender    = tonumber (line) end
function keys.Comment     (data,line)
 -- Comment DesignSize 12 (pts)
 -- Comment TFM designsize: 12 (in points)
    line = lower(line)
    local designsize = match(line,"designsize[^%d]*(%d+)")
    if designsize then data.metadata.designsize = tonumber(designsize) end
end

local function get_charmetrics(data,charmetrics,vector)
    local characters = data.characters
    local chr, ind = { }, 0
    for k, v in gmatch(charmetrics,"([%a]+) +(.-) *;") do
        if k == 'C'  then
            v = tonumber(v)
            if v < 0 then
                ind = ind + 1 -- ?
            else
                ind = v
            end
            chr = {
                index = ind
            }
        elseif k == 'WX' then
            chr.width = tonumber(v)
        elseif k == 'N'  then
            characters[v] = chr
        elseif k == 'B'  then
            local llx, lly, urx, ury = match(v,"^ *(.-) +(.-) +(.-) +(.-)$")
            chr.boundingbox = { tonumber(llx), tonumber(lly), tonumber(urx), tonumber(ury) }
        elseif k == 'L'  then
            local plus, becomes = match(v,"^(.-) +(.-)$")
            local ligatures = chr.ligatures
            if ligatures then
                ligatures[plus] = becomes
            else
                chr.ligatures = { [plus] = becomes }
            end
        end
    end
end

local function get_kernpairs(data,kernpairs)
    local characters = data.characters
    for one, two, value in gmatch(kernpairs,"KPX +(.-) +(.-) +(.-)\n") do
        local chr = characters[one]
        if chr then
            local kerns = chr.kerns
            if kerns then
                kerns[two] = tonumber(value)
            else
                chr.kerns = { [two] = tonumber(value) }
            end
        end
    end
end

local function get_variables(data,fontmetrics)
    for key, rest in gmatch(fontmetrics,"(%a+) *(.-)[\n\r]") do
        local keyhandler = keys[key]
        if keyhandler then
            keyhandler(data,rest)
        end
    end
end

local get_indexes

do

    -- old font loader

    local fontloader = fontloader

    if fontloader then

        local font_to_table = fontloader.to_table
        local open_font     = fontloader.open
        local close_font    = fontloader.close

        local function get_indexes_old(data,pfbname)
            local pfbblob = open_font(pfbname)
            if pfbblob then
                local characters = data.characters
                local pfbdata = font_to_table(pfbblob)
                if pfbdata then
                    local glyphs = pfbdata.glyphs
                    if glyphs then
                        if trace_loading then
                            report_afm("getting index data from %a",pfbname)
                        end
                        for index, glyph in next, glyphs do
                            local name = glyph.name
                            if name then
                                local char = characters[name]
                                if char then
                                    if trace_indexing then
                                        report_afm("glyph %a has index %a",name,index)
                                    end
                                    char.index = index
                                end
                            end
                        end
                    elseif trace_loading then
                        report_afm("no glyph data in pfb file %a",pfbname)
                    end
                elseif trace_loading then
                    report_afm("no data in pfb file %a",pfbname)
                end
                close_font(pfbblob)
            elseif trace_loading then
                report_afm("invalid pfb file %a",pfbname)
            end
        end

    end

    -- new (unfinished) font loader but i see no differences between
    -- old and new (one bad vector with old)

    local n, m

    local progress = function(str,position,name,size)
        local forward = position + tonumber(size) + 3 + 2
        n = n + 1
        if n >= m then
            return #str, name
        elseif forward < #str then
            return forward, name
        else
            return #str, name
        end
    end

    local initialize = function(str,position,size)
        n = 0
        m = tonumber(size)
        return position + 1
    end

    local charstrings = P("/CharStrings")
    local name        = P("/") * C((R("az")+R("AZ")+R("09")+S("-_."))^1)
    local size        = C(R("09")^1)
    local spaces      = P(" ")^1

    local p_filternames = Ct (
        (1-charstrings)^0 * charstrings * spaces * Cmt(size,initialize)
      * (Cmt(name * P(" ")^1 * C(R("09")^1), progress) + P(1))^1
    )

    -- if one of first 4 not 0-9A-F then binary else hex

    local decrypt

    do

        local r, c1, c2, n = 0, 0, 0, 0

        local function step(c)
            local cipher = byte(c)
            local plain  = bxor(cipher,rshift(r,8))
            r = mod((cipher + r) * c1 + c2,65536)
            return char(plain)
        end

        decrypt = function(binary)
            r, c1, c2, n = 55665, 52845, 22719, 4
            binary       = gsub(binary,".",step)
            return sub(binary,n+1)
        end

     -- local pattern = Cs((P(1) / step)^1)
     --
     -- decrypt = function(binary)
     --     r, c1, c2, n = 55665, 52845, 22719, 4
     --     binary = lpegmatch(pattern,binary)
     --     return sub(binary,n+1)
     -- end

    end

    local function loadpfbvector(filename)
        -- for the moment limited to encoding only

        local data = io.loaddata(resolvers.findfile(filename))

        if not find(data,"!PS%-AdobeFont%-") then
            print("no font",filename)
            return
        end

        if not data then
            print("no data",filename)
            return
        end

        local ascii, binary = match(data,"(.*)eexec%s+......(.*)")

        if not binary then
            print("no binary",filename)
            return
        end

        binary = decrypt(binary,4)

        local vector = lpegmatch(p_filternames,binary)

        vector[0] = table.remove(vector,1)

        if not vector then
            print("no vector",filename)
            return
        end

        return vector

    end

    get_indexes = function(data,pfbname)
        local vector = loadpfbvector(pfbname)
        if vector then
            local characters = data.characters
            if trace_loading then
                report_afm("getting index data from %a",pfbname)
            end
            for index=1,#vector do
                local name = vector[index]
                local char = characters[name]
                if char then
                    if trace_indexing then
                        report_afm("glyph %a has index %a",name,index)
                    end
                    char.index = index
                end
            end
        end
    end

    if fontloader then

        afm.use_new_indexer = true
        get_indexes_new     = get_indexes

        get_indexes = function(data,pfbname)
            if afm.use_new_indexer then
                return get_indexes_new(data,pfbname)
            else
                return get_indexes_old(data,pfbname)
            end
        end

    end

end

local function readafm(filename)
    local ok, afmblob, size = resolvers.loadbinfile(filename) -- has logging
    if ok and afmblob then
        local data = {
            resources = {
                filename = resolvers.unresolve(filename),
                version  = afm.version,
                creator  = "context mkiv",
            },
            properties = {
                hasitalics = false,
            },
            goodies = {
            },
            metadata   = {
                filename = file.removesuffix(file.basename(filename))
            },
            characters = {
                -- a temporary store
            },
            descriptions = {
                -- the final store
            },
        }
        afmblob = gsub(afmblob,"StartCharMetrics(.-)EndCharMetrics", function(charmetrics)
            if trace_loading then
                report_afm("loading char metrics")
            end
            get_charmetrics(data,charmetrics,vector)
            return ""
        end)
        afmblob = gsub(afmblob,"StartKernPairs(.-)EndKernPairs", function(kernpairs)
            if trace_loading then
                report_afm("loading kern pairs")
            end
            get_kernpairs(data,kernpairs)
            return ""
        end)
        afmblob = gsub(afmblob,"StartFontMetrics%s+([%d%.]+)(.-)EndFontMetrics", function(version,fontmetrics)
            if trace_loading then
                report_afm("loading variables")
            end
            data.afmversion = version
            get_variables(data,fontmetrics)
            data.fontdimens = scan_comment(fontmetrics) -- todo: all lpeg, no time now
            return ""
        end)
        return data
    else
        if trace_loading then
            report_afm("no valid afm file %a",filename)
        end
        return nil
    end
end

--[[ldx--
<p>We cache files. Caching is taken care of in the loader. We cheat a bit
by adding ligatures and kern information to the afm derived data. That
way we can set them faster when defining a font.</p>
--ldx]]--

local addkerns, addligatures, addtexligatures, unify, normalize, fixnames -- we will implement these later

function afm.load(filename)
    -- hm, for some reasons not resolved yet
    filename = resolvers.findfile(filename,'afm') or ""
    if filename ~= "" and not fonts.names.ignoredfile(filename) then
        local name = file.removesuffix(file.basename(filename))
        local data = containers.read(afm.cache,name)
        local attr = lfs.attributes(filename)
        local size, time = attr.size or 0, attr.modification or 0
        --
        local pfbfile = file.replacesuffix(name,"pfb")
        local pfbname = resolvers.findfile(pfbfile,"pfb") or ""
        if pfbname == "" then
            pfbname = resolvers.findfile(file.basename(pfbfile),"pfb") or ""
        end
        local pfbsize, pfbtime = 0, 0
        if pfbname ~= "" then
            local attr = lfs.attributes(pfbname)
            pfbsize = attr.size or 0
            pfbtime = attr.modification or 0
        end
        if not data or data.size ~= size or data.time ~= time or data.pfbsize ~= pfbsize or data.pfbtime ~= pfbtime then
            report_afm("reading %a",filename)
            data = readafm(filename)
            if data then
                if pfbname ~= "" then
                    data.resources.filename = resolvers.unresolve(pfbname)
                    get_indexes(data,pfbname)
               elseif trace_loading then
                    report_afm("no pfb file for %a",filename)
                 -- data.resources.filename = "unset" -- better than loading the afm file
                end
                report_afm("unifying %a",filename)
                unify(data,filename)
                if afm.addligatures then
                    report_afm("add ligatures")
                    addligatures(data)
                end
                if afm.addtexligatures then
                    report_afm("add tex ligatures")
                    addtexligatures(data)
                end
                if afm.addkerns then
                    report_afm("add extra kerns")
                    addkerns(data)
                end
                normalize(data)
                fixnames(data)
                report_afm("add tounicode data")
                fonts.mappings.addtounicode(data,filename)
                data.size = size
                data.time = time
                data.pfbsize = pfbsize
                data.pfbtime = pfbtime
                report_afm("saving %a in cache",name)
                data.resources.unicodes = nil -- consistent with otf but here we save not much
                data = containers.write(afm.cache, name, data)
                data = containers.read(afm.cache,name)
            end
            if applyruntimefixes and data then
                applyruntimefixes(filename,data)
            end
        end
        return data
    else
        return nil
    end
end

local uparser = fonts.mappings.makenameparser()

unify = function(data, filename)
    local unicodevector = fonts.encodings.agl.unicodes -- loaded runtime in context
    local unicodes, names = { }, { }
    local private = constructors.privateoffset
    local descriptions = data.descriptions
    for name, blob in next, data.characters do
        local code = unicodevector[name] -- or characters.name_to_unicode[name]
        if not code then
            code = lpegmatch(uparser,name)
            if not code then
                code = private
                private = private + 1
                report_afm("assigning private slot %U for unknown glyph name %a",code,name)
            end
        end
        local index = blob.index
        unicodes[name] = code
        names[name] = index
        blob.name = name
        descriptions[code] = {
            boundingbox = blob.boundingbox,
            width       = blob.width,
            kerns       = blob.kerns,
            index       = index,
            name        = name,
        }
    end
    for unicode, description in next, descriptions do
        local kerns = description.kerns
        if kerns then
            local krn = { }
            for name, kern in next, kerns do
                local unicode = unicodes[name]
                if unicode then
                    krn[unicode] = kern
                else
                 -- print(unicode,name)
                end
            end
            description.kerns = krn
        end
    end
    data.characters = nil
    local resources = data.resources
    local filename = resources.filename or file.removesuffix(file.basename(filename))
    resources.filename = resolvers.unresolve(filename) -- no shortcut
    resources.unicodes = unicodes -- name to unicode
    resources.marks = { } -- todo
 -- resources.names = names -- name to index
    resources.private = private
end

normalize = function(data)
end

fixnames = function(data)
    for k, v in next, data.descriptions do
        local n = v.name
        local r = overloads[n]
        if r then
            local name = r.name
            if trace_indexing then
                report_afm("renaming characters %a to %a",n,name)
            end
            v.name    = name
            v.unicode = r.unicode
        end
    end
end


--[[ldx--
<p>These helpers extend the basic table with extra ligatures, texligatures
and extra kerns. This saves quite some lookups later.</p>
--ldx]]--

local addthem = function(rawdata,ligatures)
    if ligatures then
        local descriptions = rawdata.descriptions
        local resources    = rawdata.resources
        local unicodes     = resources.unicodes
     -- local names        = resources.names
        for ligname, ligdata in next, ligatures do
            local one = descriptions[unicodes[ligname]]
            if one then
                for _, pair in next, ligdata do
                    local two, three = unicodes[pair[1]], unicodes[pair[2]]
                    if two and three then
                        local ol = one.ligatures
                        if ol then
                            if not ol[two] then
                                ol[two] = three
                            end
                        else
                            one.ligatures = { [two] = three }
                        end
                    end
                end
            end
        end
    end
end

addligatures    = function(rawdata) addthem(rawdata,afm.helpdata.ligatures   ) end
addtexligatures = function(rawdata) addthem(rawdata,afm.helpdata.texligatures) end

--[[ldx--
<p>We keep the extra kerns in separate kerning tables so that we can use
them selectively.</p>
--ldx]]--

-- This is rather old code (from the beginning when we had only tfm). If
-- we unify the afm data (now we have names all over the place) then
-- we can use shcodes but there will be many more looping then. But we
-- could get rid of the tables in char-cmp then. Als, in the generic version
-- we don't use the character database. (Ok, we can have a context specific
-- variant).

addkerns = function(rawdata) -- using shcodes is not robust here
    local descriptions = rawdata.descriptions
    local resources    = rawdata.resources
    local unicodes     = resources.unicodes
    local function do_it_left(what)
        if what then
            for unicode, description in next, descriptions do
                local kerns = description.kerns
                if kerns then
                    local extrakerns
                    for complex, simple in next, what do
                        complex = unicodes[complex]
                        simple = unicodes[simple]
                        if complex and simple then
                            local ks = kerns[simple]
                            if ks and not kerns[complex] then
                                if extrakerns then
                                    extrakerns[complex] = ks
                                else
                                    extrakerns = { [complex] = ks }
                                end
                            end
                        end
                    end
                    if extrakerns then
                        description.extrakerns = extrakerns
                    end
                end
            end
        end
    end
    local function do_it_copy(what)
        if what then
            for complex, simple in next, what do
                complex = unicodes[complex]
                simple = unicodes[simple]
                if complex and simple then
                    local complexdescription = descriptions[complex]
                    if complexdescription then -- optional
                        local simpledescription = descriptions[complex]
                        if simpledescription then
                            local extrakerns
                            local kerns = simpledescription.kerns
                            if kerns then
                                for unicode, kern in next, kerns do
                                    if extrakerns then
                                        extrakerns[unicode] = kern
                                    else
                                        extrakerns = { [unicode] = kern }
                                    end
                                end
                            end
                            local extrakerns = simpledescription.extrakerns
                            if extrakerns then
                                for unicode, kern in next, extrakerns do
                                    if extrakerns then
                                        extrakerns[unicode] = kern
                                    else
                                        extrakerns = { [unicode] = kern }
                                    end
                                end
                            end
                            if extrakerns then
                                complexdescription.extrakerns = extrakerns
                            end
                        end
                    end
                end
            end
        end
    end
    -- add complex with values of simplified when present
    do_it_left(afm.helpdata.leftkerned)
    do_it_left(afm.helpdata.bothkerned)
    -- copy kerns from simple char to complex char unless set
    do_it_copy(afm.helpdata.bothkerned)
    do_it_copy(afm.helpdata.rightkerned)
end

--[[ldx--
<p>The copying routine looks messy (and is indeed a bit messy).</p>
--ldx]]--

local function adddimensions(data) -- we need to normalize afm to otf i.e. indexed table instead of name
    if data then
        for unicode, description in next, data.descriptions do
            local bb = description.boundingbox
            if bb then
                local ht, dp = bb[4], -bb[2]
                if ht == 0 or ht < 0 then
                    -- no need to set it and no negative heights, nil == 0
                else
                    description.height = ht
                end
                if dp == 0 or dp < 0 then
                    -- no negative depths and no negative depths, nil == 0
                else
                    description.depth  = dp
                end
            end
        end
    end
end

local function copytotfm(data)
    if data and data.descriptions then
        local metadata     = data.metadata
        local resources    = data.resources
        local properties   = derivetable(data.properties)
        local descriptions = derivetable(data.descriptions)
        local goodies      = derivetable(data.goodies)
        local characters   = { }
        local parameters   = { }
        local unicodes     = resources.unicodes
        --
        for unicode, description in next, data.descriptions do -- use parent table
            characters[unicode] = { }
        end
        --
        local filename   = constructors.checkedfilename(resources)
        local fontname   = metadata.fontname or metadata.fullname
        local fullname   = metadata.fullname or metadata.fontname
        local endash     = 0x0020 -- space
        local emdash     = 0x2014
        local spacer     = "space"
        local spaceunits = 500
        --
        local monospaced  = metadata.monospaced
        local charwidth   = metadata.charwidth
        local italicangle = metadata.italicangle
        local charxheight = metadata.xheight and metadata.xheight > 0 and metadata.xheight
        properties.monospaced  = monospaced
        parameters.italicangle = italicangle
        parameters.charwidth   = charwidth
        parameters.charxheight = charxheight
        -- same as otf
        if properties.monospaced then
            if descriptions[endash] then
                spaceunits, spacer = descriptions[endash].width, "space"
            end
            if not spaceunits and descriptions[emdash] then
                spaceunits, spacer = descriptions[emdash].width, "emdash"
            end
            if not spaceunits and charwidth then
                spaceunits, spacer = charwidth, "charwidth"
            end
        else
            if descriptions[endash] then
                spaceunits, spacer = descriptions[endash].width, "space"
            end
            if not spaceunits and charwidth then
                spaceunits, spacer = charwidth, "charwidth"
            end
        end
        spaceunits = tonumber(spaceunits)
        if spaceunits < 200 then
            -- todo: warning
        end
        --
        parameters.slant         = 0
        parameters.space         = spaceunits
        parameters.space_stretch = 500
        parameters.space_shrink  = 333
        parameters.x_height      = 400
        parameters.quad          = 1000
        --
        if italicangle and italicangle ~= 0 then
            parameters.italicangle  = italicangle
            parameters.italicfactor = math.cos(math.rad(90+italicangle))
            parameters.slant        = - math.tan(italicangle*math.pi/180)
        end
        if monospaced then
            parameters.space_stretch = 0
            parameters.space_shrink  = 0
        elseif afm.syncspace then
            parameters.space_stretch = spaceunits/2
            parameters.space_shrink  = spaceunits/3
        end
        parameters.extra_space = parameters.space_shrink
        if charxheight then
            parameters.x_height = charxheight
        else
            -- same as otf
            local x = 0x0078 -- x
            if x then
                local x = descriptions[x]
                if x then
                    parameters.x_height = x.height
                end
            end
            --
        end
        local fd = data.fontdimens
        if fd and fd[8] and fd[9] and fd[10] then -- math
            for k,v in next, fd do
                parameters[k] = v
            end
        end
        --
        parameters.designsize = (metadata.designsize or 10)*65536
        parameters.ascender   = abs(metadata.ascender  or 0)
        parameters.descender  = abs(metadata.descender or 0)
        parameters.units      = 1000
        --
        properties.spacer        = spacer
        properties.encodingbytes = 2
        properties.format        = fonts.formats[filename] or "type1"
        properties.filename      = filename
        properties.fontname      = fontname
        properties.fullname      = fullname
        properties.psname        = fullname
        properties.name          = filename or fullname or fontname
        --
        if next(characters) then
            return {
                characters   = characters,
                descriptions = descriptions,
                parameters   = parameters,
                resources    = resources,
                properties   = properties,
                goodies      = goodies,
            }
        end
    end
    return nil
end

--[[ldx--
<p>Originally we had features kind of hard coded for <l n='afm'/>
files but since I expect to support more font formats, I decided
to treat this fontformat like any other and handle features in a
more configurable way.</p>
--ldx]]--

function afm.setfeatures(tfmdata,features)
    local okay = constructors.initializefeatures("afm",tfmdata,features,trace_features,report_afm)
    if okay then
        return constructors.collectprocessors("afm",tfmdata,features,trace_features,report_afm)
    else
        return { } -- will become false
    end
end

local function addtables(data)
    local resources  = data.resources
    local lookuptags = resources.lookuptags
    local unicodes   = resources.unicodes
    if not lookuptags then
        lookuptags = { }
        resources.lookuptags = lookuptags
    end
    setmetatableindex(lookuptags,function(t,k)
        local v = type(k) == "number" and ("lookup " .. k) or k
        t[k] = v
        return v
    end)
    if not unicodes then
        unicodes = { }
        resources.unicodes = unicodes
        setmetatableindex(unicodes,function(t,k)
            setmetatableindex(unicodes,nil)
            for u, d in next, data.descriptions do
                local n = d.name
                if n then
                    t[n] = u
                end
            end
            return rawget(t,k)
        end)
    end
    constructors.addcoreunicodes(unicodes) -- do we really need this?
end

local function afmtotfm(specification)
    local afmname = specification.filename or specification.name
    if specification.forced == "afm" or specification.format == "afm" then -- move this one up
        if trace_loading then
            report_afm("forcing afm format for %a",afmname)
        end
    else
        local tfmname = findbinfile(afmname,"ofm") or ""
        if tfmname ~= "" then
            if trace_loading then
                report_afm("fallback from afm to tfm for %a",afmname)
            end
            return -- just that
        end
    end
    if afmname ~= "" then
        -- weird, isn't this already done then?
        local features = constructors.checkedfeatures("afm",specification.features.normal)
        specification.features.normal = features
        constructors.hashinstance(specification,true) -- also weird here
        --
        specification = definers.resolve(specification) -- new, was forgotten
        local cache_id = specification.hash
        local tfmdata  = containers.read(constructors.cache, cache_id) -- cache with features applied
        if not tfmdata then
            local rawdata = afm.load(afmname)
            if rawdata and next(rawdata) then
                addtables(rawdata)
                adddimensions(rawdata)
                tfmdata = copytotfm(rawdata)
                if tfmdata and next(tfmdata) then
                    local shared = tfmdata.shared
                    if not shared then
                        shared         = { }
                        tfmdata.shared = shared
                    end
                    shared.rawdata   = rawdata
                    shared.features  = features
                    shared.processes = afm.setfeatures(tfmdata,features)
                end
            elseif trace_loading then
                report_afm("no (valid) afm file found with name %a",afmname)
            end
            tfmdata = containers.write(constructors.cache,cache_id,tfmdata)
        end
        return tfmdata
    end
end

--[[ldx--
<p>As soon as we could intercept the <l n='tfm'/> reader, I implemented an
<l n='afm'/> reader. Since traditional <l n='pdftex'/> could use <l n='opentype'/>
fonts with <l n='afm'/> companions, the following method also could handle
those cases, but now that we can handle <l n='opentype'/> directly we no longer
need this features.</p>
--ldx]]--

local function read_from_afm(specification)
    local tfmdata = afmtotfm(specification)
    if tfmdata then
        tfmdata.properties.name = specification.name
        tfmdata = constructors.scale(tfmdata, specification)
        local allfeatures = tfmdata.shared.features or specification.features.normal
        constructors.applymanipulators("afm",tfmdata,allfeatures,trace_features,report_afm)
        fonts.loggers.register(tfmdata,'afm',specification)
    end
    return tfmdata
end

--[[ldx--
<p>Here comes the implementation of a few features. We only implement
those that make sense for this format.</p>
--ldx]]--

local function prepareligatures(tfmdata,ligatures,value)
    if value then
        local descriptions = tfmdata.descriptions
        local hasligatures = false
        for unicode, character in next, tfmdata.characters do
            local description = descriptions[unicode]
            local dligatures = description.ligatures
            if dligatures then
                local cligatures = character.ligatures
                if not cligatures then
                    cligatures = { }
                    character.ligatures = cligatures
                end
                for unicode, ligature in next, dligatures do
                    cligatures[unicode] = {
                        char = ligature,
                        type = 0
                    }
                end
                hasligatures = true
            end
        end
        tfmdata.properties.hasligatures = hasligatures
    end
end

local function preparekerns(tfmdata,kerns,value)
    if value then
        local rawdata      = tfmdata.shared.rawdata
        local resources    = rawdata.resources
        local unicodes     = resources.unicodes
        local descriptions = tfmdata.descriptions
        local haskerns     = false
        for u, chr in next, tfmdata.characters do
            local d = descriptions[u]
            local newkerns = d[kerns]
            if newkerns then
                local kerns = chr.kerns
                if not kerns then
                    kerns = { }
                    chr.kerns = kerns
                end
                for k,v in next, newkerns do
                    local uk = unicodes[k]
                    if uk then
                        kerns[uk] = v
                    end
                end
                haskerns = true
            end
        end
        tfmdata.properties.haskerns = haskerns
    end
end

local list = {
 -- [0x0022] = 0x201D,
    [0x0027] = 0x2019,
 -- [0x0060] = 0x2018,
}

local function texreplacements(tfmdata,value)
    local descriptions = tfmdata.descriptions
    local characters   = tfmdata.characters
    for k, v in next, list do
        characters  [k] = characters  [v] -- we forget about kerns
        descriptions[k] = descriptions[v] -- we forget about kerns
    end
end

local function ligatures   (tfmdata,value) prepareligatures(tfmdata,'ligatures',   value) end
local function texligatures(tfmdata,value) prepareligatures(tfmdata,'texligatures',value) end
local function kerns       (tfmdata,value) preparekerns    (tfmdata,'kerns',       value) end
local function extrakerns  (tfmdata,value) preparekerns    (tfmdata,'extrakerns',  value) end

registerafmfeature {
    name         = "liga",
    description  = "traditional ligatures",
    initializers = {
        base = ligatures,
        node = ligatures,
    }
}

registerafmfeature {
    name         = "kern",
    description  = "intercharacter kerning",
    initializers = {
        base = kerns,
        node = kerns,
    }
}

registerafmfeature {
    name         = "extrakerns",
    description  = "additional intercharacter kerning",
    initializers = {
        base = extrakerns,
        node = extrakerns,
    }
}

registerafmfeature {
    name         = 'tlig',
    description  = 'tex ligatures',
    initializers = {
        base = texligatures,
        node = texligatures,
    }
}

registerafmfeature {
    name         = 'trep',
    description  = 'tex replacements',
    initializers = {
        base = texreplacements,
        node = texreplacements,
    }
}

-- readers

local check_tfm   = readers.check_tfm

fonts.formats.afm = "type1"
fonts.formats.pfb = "type1"

local function check_afm(specification,fullname)
    local foundname = findbinfile(fullname, 'afm') or "" -- just to be sure
    if foundname == "" then
        foundname = fonts.names.getfilename(fullname,"afm") or ""
    end
    if foundname == "" and afm.autoprefixed then
        local encoding, shortname = match(fullname,"^(.-)%-(.*)$") -- context: encoding-name.*
        if encoding and shortname and fonts.encodings.known[encoding] then
            shortname = findbinfile(shortname,'afm') or "" -- just to be sure
            if shortname ~= "" then
                foundname = shortname
                if trace_defining then
                    report_afm("stripping encoding prefix from filename %a",afmname)
                end
            end
        end
    end
    if foundname ~= "" then
        specification.filename = foundname
        specification.format   = "afm"
        return read_from_afm(specification)
    end
end

function readers.afm(specification,method)
    local fullname, tfmdata = specification.filename or "", nil
    if fullname == "" then
        local forced = specification.forced or ""
        if forced ~= "" then
            tfmdata = check_afm(specification,specification.name .. "." .. forced)
        end
        if not tfmdata then
            method = method or definers.method or "afm or tfm"
            if method == "tfm" then
                tfmdata = check_tfm(specification,specification.name)
            elseif method == "afm" then
                tfmdata = check_afm(specification,specification.name)
            elseif method == "tfm or afm" then
                tfmdata = check_tfm(specification,specification.name) or check_afm(specification,specification.name)
            else -- method == "afm or tfm" or method == "" then
                tfmdata = check_afm(specification,specification.name) or check_tfm(specification,specification.name)
            end
        end
    else
        tfmdata = check_afm(specification,fullname)
    end
    return tfmdata
end

function readers.pfb(specification,method) -- only called when forced
    local original = specification.specification
    if trace_defining then
        report_afm("using afm reader for %a",original)
    end
    specification.specification = gsub(original,"%.pfb",".afm")
    specification.forced = "afm"
    return readers.afm(specification,method)
end