Merge pull request #43 from phi-gamma/master

modifications to color handling

3 files changed, 247 insertions, 122 deletions

diff --git a/luaotfload-colors.lua b/luaotfload-colors.lua
index 1525214..ec076c2 100644
--- a/luaotfload-colors.lua
+++ b/luaotfload-colors.lua
@@ -6,39 +6,105 @@ if not modules then modules = { } end modules ['luaotfload-colors'] = {
     license   = "GNU GPL v2"
 }
 
-local newnode            = node.new
-local nodetype           = node.id
-local traverse_nodes     = node.traverse
-local insert_node_before = node.insert_before
-local insert_node_after  = node.insert_after
-
-local stringformat = string.format
-local stringgsub   = string.gsub
-local stringfind   = string.find
-
-local otffeatures        = fonts.constructors.newfeatures("otf")
-local ids                = fonts.hashes.identifiers
-local registerotffeature = otffeatures.register
-
-local function setcolor(tfmdata,value)
-    local sanitized
-    local properties = tfmdata.properties
+--[[doc--
+buggy coloring with the pre_output_filter when expansion is enabled
+    · tfmdata for different expansion values is split over different objects
+    · in ``initializeexpansion()``, chr.expansion_factor is set, and only
+      those characters that have it are affected
+    · in constructors.scale: chr.expansion_factor = ve*1000 if commented out
+      makes the bug vanish
+--doc]]--
 
-    if value then
-        value = tostring(value)
-        if #value == 6 or #value == 8 then
-            sanitized = value
-        elseif #value == 7 then
-            _, _, sanitized = stringfind(value, "(......)")
-        elseif #value > 8 then
-            _, _, sanitized = stringfind(value, "(........)")
-        else
-            -- broken color code ignored, issue a warning?
-        end
+
+local color_callback = config.luaotfload.color_callback
+if not color_callback then
+    --- maybe this would be better as a method: "early" | "late"
+    color_callback = "pre_linebreak_filter"
+--  color_callback = "pre_output_filter" --- old behavior, breaks expansion
+end
+
+
+local newnode               = node.new
+local nodetype              = node.id
+local traverse_nodes        = node.traverse
+local insert_node_before    = node.insert_before
+local insert_node_after     = node.insert_after
+
+local stringformat          = string.format
+local stringgsub            = string.gsub
+local stringfind            = string.find
+local stringsub             = string.sub
+
+local otffeatures           = fonts.constructors.newfeatures("otf")
+local identifiers           = fonts.hashes.identifiers
+local registerotffeature    = otffeatures.register
+
+local add_color_callback --[[ this used to be a global‽ ]]
+
+--[[doc--
+This converts a single octet into a decimal with three digits of
+precision. The optional second argument limits precision to a single
+digit.
+--doc]]--
+
+--- string -> bool? -> string
+local hex_to_dec = function (hex,one) --- one isn’t actually used anywhere ...
+    if one then
+        return stringformat("%.1g", tonumber(hex, 16)/255)
+    else
+        return stringformat("%.3g", tonumber(hex, 16)/255)
+    end
+end
+
+--[[doc--
+Color string validator / parser.
+--doc]]--
+
+local lpeg           = require"lpeg"
+local lpegmatch      = lpeg.match
+local C, Cg, Ct, P, R, S = lpeg.C, lpeg.Cg, lpeg.Ct, lpeg.P, lpeg.R, lpeg.S
+
+local digit16        = R("09", "af", "AF")
+local octet          = C(digit16 * digit16)
+
+local p_rgb          = octet * octet * octet
+local p_rgba         = p_rgb * octet
+local valid_digits   = C(p_rgba + p_rgb) -- matches eight or six hex digits
+
+local p_Crgb         = Cg(octet/hex_to_dec, "red") --- for captures
+                     * Cg(octet/hex_to_dec, "green")
+                     * Cg(octet/hex_to_dec, "blue")
+local p_Crgba        = p_Crgb * Cg(octet/hex_to_dec, "alpha")
+local extract_color  = Ct(p_Crgba + p_Crgb)
+
+--- string -> (string | nil)
+local sanitize_color_expression = function (digits)
+    digits = tostring(digits)
+    local sanitized = lpegmatch(valid_digits, digits)
+    if not sanitized then
+        luaotfload.warning(
+            "“%s” is not a valid rgb[a] color expression", digits)
+        return nil
     end
+    return sanitized
+end
+
+--[[doc--
+``setcolor`` modifies tfmdata.properties.color in place
+--doc]]--
+
+--- fontobj -> string -> unit
+---
+---         (where “string” is a rgb value as three octet
+---         hexadecimal, with an optional fourth transparency
+---         value)
+---
+local setcolor = function (tfmdata, value)
+    local sanitized  = sanitize_color_expression(value)
+    local properties = tfmdata.properties
 
     if sanitized then
-        tfmdata.properties.color = sanitized
+        properties.color = sanitized
         add_color_callback()
     end
 end
@@ -52,133 +118,187 @@ registerotffeature {
     }
 }
 
-local function hex2dec(hex,one)
-    if one then
-        return stringformat("%.1g", tonumber(hex, 16)/255)
-    else
-        return stringformat("%.3g", tonumber(hex, 16)/255)
-    end
-end
 
-local res
+--- something is carried around in ``res``
+--- for later use by color_handler() --- but what?
+
+local res --- <- state of what?
 
-local function pageresources(a)
+--- float -> unit
+local function pageresources(alpha)
     local res2
     if not res then
        res = "/TransGs1<</ca 1/CA 1>>"
     end
-    res2 = stringformat("/TransGs%s<</ca %s/CA %s>>", a, a, a)
-    res  = stringformat("%s%s", res, stringfind(res, res2) and "" or res2)
+    res2 = stringformat("/TransGs%s<</ca %s/CA %s>>",
+                        alpha, alpha, alpha)
+    res  = stringformat("%s%s",
+                        res,
+                        stringfind(res, res2) and "" or res2)
 end
 
-local function hex_to_rgba(hex)
-    local r, g, b, a, push, pop, res3
-    if hex then
-        if #hex == 6 then
-            _, _, r, g, b    = stringfind(hex, '(..)(..)(..)')
-        elseif #hex == 8 then
-            _, _, r, g, b, a = stringfind(hex, '(..)(..)(..)(..)')
-            a                = hex2dec(a,true)
-            pageresources(a)
-        end
-    else
-        return nil
+--- we store results of below color handler as tuples of
+--- push/pop strings
+local color_cache = { } --- (string, (string * string)) hash_t
+
+--- string -> (string * string)
+local hex_to_rgba = function (digits)
+    if not digits then
+        return
     end
-    r = hex2dec(r)
-    g = hex2dec(g)
-    b = hex2dec(b)
-    if a then
-        push = stringformat('/TransGs%g gs %s %s %s rg', a, r, g, b)
-        pop  = '0 g /TransGs1 gs'
-    else
-        push = stringformat('%s %s %s rg', r, g, b)
-        pop  = '0 g'
+
+    --- this is called like a thousand times, so some
+    --- memoizing is in order.
+    local cached = color_cache[digits]
+    if not cached then
+        local push, pop
+        local rgb = lpegmatch(extract_color, digits)
+        if rgb.alpha then
+            pageresources(rgb.alpha)
+            push = stringformat(
+                        "/TransGs%g gs %s %s %s rg",
+                        rgb.alpha,
+                        rgb.red,
+                        rgb.green,
+                        rgb.blue)
+            pop  = "0 g /TransGs1 gs"
+        else
+            push = stringformat(
+                        "%s %s %s rg",
+                        rgb.red,
+                        rgb.green,
+                        rgb.blue)
+            pop  = "0 g"
+        end
+        color_cache[digits] = { push, pop }
+        return push, pop
     end
-    return push, pop
+
+    return cached[1], cached[2]
 end
 
-local glyph   = nodetype('glyph')
-local hlist   = nodetype('hlist')
-local vlist   = nodetype('vlist')
-local whatsit = nodetype('whatsit')
-local pgi     = nodetype('page_insert')
-local sbox    = nodetype('sub_box')
+--- Luatex internal types
 
-local function lookup_next_color(head)
+local glyph_t           = nodetype("glyph")
+local hlist_t           = nodetype("hlist")
+local vlist_t           = nodetype("vlist")
+local whatsit_t         = nodetype("whatsit")
+local page_insert_t     = nodetype("page_insert")
+local sub_box_t         = nodetype("sub_box")
+
+--- node -> nil | -1 | color‽
+local lookup_next_color
+lookup_next_color = function (head) --- paragraph material
     for n in traverse_nodes(head) do
-        if n.id == glyph then
-            if ids[n.font] and ids[n.font].properties and ids[n.font].properties.color then
-                return ids[n.font].properties.color
+        local n_id = n.id
+
+        if n_id == glyph_t then
+            local n_font
+            if  identifiers[n_font]
+            and identifiers[n_font].properties
+            and identifiers[n_font].properties.color
+            then
+                return identifiers[n.font].properties.color
             else
                 return -1
             end
-        elseif n.id == vlist or n.id == hlist or n.id == sbox then
+
+        elseif n_id == vlist_t or n_id == hlist_t or n_id == sub_box_t then
             local r = lookup_next_color(n.list)
-            if r == -1 then
-                return -1
-            elseif r then
+            if r then
                 return r
             end
-        elseif n.id == whatsit or n.id == pgi then
+
+        elseif n_id == whatsit_t or n_id == page_insert_t then
             return -1
         end
     end
     return nil
 end
 
-local function node_colorize(head, current_color, next_color)
+--[[doc--
+While the second argument and second returned value are apparently
+always nil when the function is called, they temporarily take string
+values during the node list traversal.
+--doc]]--
+
+local cnt = 0
+--- node -> string -> int -> (node * string)
+local node_colorize
+node_colorize = function (head, current_color, next_color)
     for n in traverse_nodes(head) do
-        if n.id == hlist or n.id == vlist or n.id == sbox then
-            local next_color_in = lookup_next_color(n.next) or next_color
+        local n_id      = n.id
+        local nextnode  = n.next
+
+        if n_id == hlist_t or n_id == vlist_t or n_id == sub_box_t then
+            local next_color_in = lookup_next_color(nextnode) or next_color
             n.list, current_color = node_colorize(n.list, current_color, next_color_in)
-        elseif n.id == glyph then
-            local tfmdata = ids[n.font]
+
+        elseif n_id == glyph_t then
+            cnt = cnt + 1
+            local tfmdata = identifiers[n.font]
+
+            --- colorization is restricted to those fonts
+            --- that received the “color” property upon
+            --- loading (see ``setcolor()`` above)
             if tfmdata and tfmdata.properties  and tfmdata.properties.color then
-                if tfmdata.properties.color ~= current_color then
-                    local pushcolor = hex_to_rgba(tfmdata.properties.color)
-                    local push = newnode(whatsit, 8)
-                    push.mode  = 1
-                    push.data  = pushcolor
-                    head       = insert_node_before(head, n, push)
-                    current_color = tfmdata.properties.color
+                local font_color = tfmdata.properties.color
+--                luaotfload.info(
+--                    "n: %d; %s; %d %s, %s",
+--                    cnt, utf.char(n.char), n.font, "<TRUE>", font_color)
+                if font_color ~= current_color then
+                    local pushcolor = hex_to_rgba(font_color)
+                    local push      = newnode(whatsit_t, 8)
+                    push.mode       = 1
+                    push.data       = pushcolor
+                    head            = insert_node_before(head, n, push)
+                    current_color   = font_color
                 end
-                local next_color_in = lookup_next_color (n.next) or next_color
-                if next_color_in ~= tfmdata.properties.color then
-                    local _, popcolor = hex_to_rgba(tfmdata.properties.color)
-                    local pop  = newnode(whatsit, 8)
-                    pop.mode   = 1
-                    pop.data   = popcolor
-                    head       = insert_node_after(head, n, pop)
-                    current_color = nil
+                local next_color_in = lookup_next_color (nextnode) or next_color
+                if next_color_in ~= font_color then
+                    local _, popcolor = hex_to_rgba(font_color)
+                    local pop         = newnode(whatsit_t, 8)
+                    pop.mode          = 1
+                    pop.data          = popcolor
+                    head              = insert_node_after(head, n, pop)
+                    current_color     = nil
                 end
+
+--            else
+--                luaotfload.info(
+--                    "n: %d; %s; %d %s",
+--                    cnt, utf.char(n.char), n.font, "<FALSE>")
             end
         end
     end
     return head, current_color
 end
 
-local function font_colorize(head)
-   -- check if our page resources existed in the previous run
-   -- and remove it to avoid duplicating it later
-   if res then
-      local r = "/ExtGState<<"..res..">>"
-      tex.pdfpageresources = stringgsub(tex.pdfpageresources, r, "")
-   end
-   local h = node_colorize(head, nil, nil)
-   -- now append our page resources
-   if res and stringfind(res, "%S") then -- test for non-empty string
-      local r = "/ExtGState<<"..res..">>"
-      tex.pdfpageresources = tex.pdfpageresources..r
-   end
-   return h
+--- node -> node
+local color_handler = function (head)
+    -- check if our page resources existed in the previous run
+    -- and remove it to avoid duplicating it later
+    if res then
+        local r = "/ExtGState<<" .. res .. ">>"
+        tex.pdfpageresources = stringgsub(tex.pdfpageresources, r, "")
+    end
+    local new_head = node_colorize(head, nil, nil)
+    -- now append our page resources
+    if res and stringfind(res, "%S") then -- test for non-empty string
+        local r = "/ExtGState<<" .. res .. ">>"
+        tex.pdfpageresources = tex.pdfpageresources..r
+    end
+    return new_head
 end
 
 local color_callback_activated = 0
 
-function add_color_callback()
+--- unit -> unit
+add_color_callback = function ( )
     if color_callback_activated == 0 then
-        luatexbase.add_to_callback(
-          "pre_output_filter", font_colorize, "luaotfload.colorize")
+        luatexbase.add_to_callback(color_callback,
+                                   color_handler,
+                                   "luaotfload.color_handler")
         color_callback_activated = 1
     end
 end
diff --git a/luaotfload.dtx b/luaotfload.dtx
index fb57462..d8d1863 100644
--- a/luaotfload.dtx
+++ b/luaotfload.dtx
@@ -1182,11 +1182,12 @@ and the derived files
 luaotfload                  = luaotfload or {}
 local luaotfload            = luaotfload
 
-config                      = config or { }
-config.luaotfload           = config.luaotfload or { }
-config.luaotfload.resolver  = config.luaotfload.resolver  or "normal"
-config.luaotfload.definer   = config.luaotfload.definer   or "patch"
-config.luaotfload.loglevel  = config.luaotfload.loglevel  or 1
+config                            = config or { }
+config.luaotfload                 = config.luaotfload or { }
+config.luaotfload.resolver        = config.luaotfload.resolver  or "normal"
+config.luaotfload.definer         = config.luaotfload.definer   or "patch"
+config.luaotfload.loglevel        = config.luaotfload.loglevel  or 1
+config.luaotfload.color_callback  = config.luaotfload.color_callback  or "pre_linebreak_filter"
 --luaotfload.prefer_merge     = config.luaotfload.prefer_merge or true
 
 luaotfload.module = {
diff --git a/mkglyphlist b/mkglyphlist
index d73a608..281c736 100755
--- a/mkglyphlist
+++ b/mkglyphlist
@@ -27,9 +27,13 @@ require"lpeg"
 require"socket"
 
 kpse.set_program_name"luatex"
-dofile(kpse.find_file("lualibs-lua.lua",   "lua"))
-dofile(kpse.find_file("lualibs-lpeg.lua",  "lua"))
-dofile(kpse.find_file("lualibs-table.lua", "lua")) --- for serialization
+for _, lib in next, { "lualibs-lua.lua",
+                      "lualibs-lpeg.lua",
+                      "lualibs-table.lua", } do
+  local found = assert(kpse.find_file(lib, "lua"),
+                       "Could not locate " .. lib)
+  require(found)
+end
 
 local C, Cf, Cg, Ct, P, R =
   lpeg.C, lpeg.Cf, lpeg.Cg, lpeg.Ct, lpeg.P, lpeg.R