[features,parsers] move font request handler into luaotfload-parsers.lua

2 files changed, 268 insertions, 256 deletions

diff --git a/luaotfload-features.lua b/luaotfload-features.lua
index dc6b8a4..d786549 100644
--- a/luaotfload-features.lua
+++ b/luaotfload-features.lua
@@ -9,7 +9,12 @@ if not modules then modules = { } end modules ["features"] = {
 local type, next        = type, next
 local tonumber          = tonumber
 local tostring          = tostring
+
+local lpeg              = require "lpeg"
 local lpegmatch         = lpeg.match
+local P                 = lpeg.P
+local R                 = lpeg.R
+local C                 = lpeg.C
 
 ---[[ begin included font-ltx.lua ]]
 --- this appears to be based in part on luatex-fonts-def.lua
@@ -822,259 +827,6 @@ local set_default_features = function (speclist)
     return speclist
 end
 
------------------------------------------------------------------------
----                    request syntax parser 2.2
------------------------------------------------------------------------
---- the luaotfload font request syntax (see manual)
---- has a canonical form:
----
----     \font<csname>=<prefix>:<identifier>:<features>
----
---- where
----   <csname> is the control sequence that activates the font
----   <prefix> is either “file” or “name”, determining the lookup
----   <identifer> is either a file name (no path) or a font
----               name, depending on the lookup
----   <features> is a list of switches or options, separated by
----              semicolons or commas; a switch is of the form “+” foo
----              or “-” foo, options are of the form lhs “=” rhs
----
---- however, to ensure backward compatibility we also have
---- support for Xetex-style requests.
----
---- for the Xetex emulation see:
---- · The XeTeX Reference Guide by Will Robertson, 2011
---- · The XeTeX Companion by Michel Goosens, 2010
---- · About XeTeX by Jonathan Kew, 2005
----
----
---- caueat emptor.
----     the request is parsed into one of **four** different
----     lookup categories: the regular ones, file and name,
----     as well as the Xetex compatibility ones, path and anon.
----     (maybe a better choice of identifier would be “ambig”.)
----
----     according to my reconstruction, the correct chaining
----     of the lookups for each category is as follows:
----
----     | File -> ( db/filename lookup )
----
----     | Name -> ( db/name lookup,
----                 db/filename lookup )
----
----     | Path -> ( db/filename lookup,
----                 fullpath lookup )
----
----     | Anon -> ( kpse.find_file(),     // <- for tfm, ofm
----                 db/name lookup,
----                 db/filename lookup,
----                 fullpath lookup )
----
----     caching of successful lookups is essential. we now
----     as of v2.2 have an experimental lookup cache that is
----     stored in a separate file. it pertains only to name:
----     lookups, and is described in more detail in
----     luaotfload-database.lua.
----
------------------------------------------------------------------------
-
---[[doc--
-
-    One further incompatibility between Xetex and Luatex-Fonts consists
-    in their option list syntax: apparently, Xetex requires key-value
-    options to be prefixed by a "+" (ascii “plus”) character. We
-    silently accept this as well, dropping the first byte if it is a
-    plus or minus character.
-
-    Reference: https://github.com/lualatex/luaotfload/issues/79#issuecomment-18104483
-
---doc]]--
-
-local handle_normal_option = function (key, val)
-    val = stringlower(val)
-    --- the former “toboolean()” handler
-    if val == "true"  then
-        val = true
-    elseif val == "false" then
-        val = false
-    end
-    return key, val
-end
-
---[[doc--
-
-    Xetex style indexing begins at zero which we just increment before
-    passing it along to the font loader.  Ymmv.
-
---doc]]--
-
-local handle_xetex_option = function (key, val)
-    val = stringlower(val)
-    local numeric = tonumber(val) --- decimal only; keeps colors intact
-    if numeric then --- ugh
-        if  mathceil(numeric) == numeric then -- integer, possible index
-            val = tostring(numeric + 1)
-        end
-    elseif val == "true"  then
-        val = true
-    elseif val == "false" then
-        val = false
-    end
-    return key, val
-end
-
---[[doc--
-
-    Instead of silently ignoring invalid options we emit a warning to
-    the log.
-
-    Note that we have to return a pair to please rawset(). This creates
-    an entry on the resulting features hash which will later be removed
-    during set_default_features().
-
---doc]]--
-
-local handle_invalid_option = function (opt)
-    report("log", 0, "load", "font option %q unknown.", opt)
-    return "", false
-end
-
---[[doc--
-
-    Dirty test if a file: request is actually a path: lookup; don’t
-    ask! Note this fails on Windows-style absolute paths. These will
-    *really* have to use the correct request.
-
---doc]]--
-
-local check_garbage = function (_,i, garbage)
-    if stringfind(garbage, "/") then
-        report("log", 0, "load",  --- ffs use path!
-            "warning: path in file: lookups is deprecated; ")
-        report("log", 0, "load", "use bracket syntax instead!")
-        report("log", 0, "load",
-            "position: %d; full match: %q",
-            i, garbage)
-        return true
-    end
-    return false
-end
-
-local lpegmatch = lpeg.match
-local P, S, R   = lpeg.P, lpeg.S, lpeg.R
-local C, Cc, Cf, Cg, Cmt, Cs, Ct
-    = lpeg.C, lpeg.Cc, lpeg.Cf, lpeg.Cg, lpeg.Cmt, lpeg.Cs, lpeg.Ct
-
---- terminals and low-level classes -----------------------------------
---- note we could use the predefined ones from lpeg.patterns
-local dot         = P"."
-local colon       = P":"
-local featuresep  = S",;"
-local slash       = P"/"
-local equals      = P"="
-local lbrk, rbrk  = P"[", P"]"
-
-local spacing     = S" \t\v"
-local ws          = spacing^0
-
-local digit       = R"09"
-local alpha       = R("az", "AZ")
-local anum        = alpha + digit
-local decimal     = digit^1 * (dot * digit^0)^-1
-
---- modifiers ---------------------------------------------------------
---[[doc--
-    The slash notation: called “modifiers” (Kew) or “font options”
-    (Robertson, Goosens)
-    we only support the shorthands for italic / bold / bold italic
-    shapes, as well as setting optical size, the rest is ignored.
---doc]]--
-local style_modifier    = (P"BI" + P"IB" + P"bi" + P"ib" + S"biBI")
-                        / stringlower
-local size_modifier     = S"Ss" * P"="    --- optical size
-                        * Cc"optsize" * C(decimal)
-local other_modifier    = P"AAT" + P"aat" --- apple stuff;  unsupported
-                        + P"ICU" + P"icu" --- not applicable
-                        + P"GR"  + P"gr"  --- sil stuff;    unsupported
-local garbage_modifier  = ((1 - colon - slash)^0 * Cc(false))
-local modifier          = slash * (other_modifier      --> ignore
-                                 + Cs(style_modifier)  --> collect
-                                 + Ct(size_modifier)   --> collect
-                                 + garbage_modifier)   --> warn
-local modifier_list     = Cg(Ct(modifier^0), "modifiers")
-
---- lookups -----------------------------------------------------------
-local fontname          = C((1-S":(/")^1)  --- like luatex-fonts
-local unsupported       = Cmt((1-S":(")^1, check_garbage)
-local prefixed          = P"name:" * ws * Cg(fontname, "name")
---- initially we intended file: to emulate the behavior of
---- luatex-fonts, i.e. no paths allowed. after all, we do have XeTeX
---- emulation with the path lookup and it interferes with db lookups.
---- turns out fontspec and other widely used packages rely on file:
---- with paths already, so we’ll add a less strict rule here.  anyways,
---- we’ll emit a warning.
-                        + P"file:" * ws * Cg(unsupported, "path")
-                        + P"file:" * ws * Cg(fontname, "file")
---- EXPERIMENTAL: kpse lookup
-                        + P"kpse:" * ws * Cg(fontname, "kpse")
---- EXPERIMENTAL: custom lookup
-                        + P"my:" * ws * Cg(fontname, "my")
-local unprefixed        = Cg(fontname, "anon")
-local path_lookup       = lbrk * Cg(C((1-rbrk)^1), "path") * rbrk
-
---- features ----------------------------------------------------------
-local field_char        = anum + S"+-." --- sic!
-local field             = field_char^1
---- assignments are “lhs=rhs”
----              or “+lhs=rhs” (Xetex-style)
---- switches    are “+key” | “-key”
-local normal_option     = C(field) * ws * equals * ws * C(field) * ws
-local xetex_option      = P"+" * ws * normal_option
-local ignore_option     = (1 - equals - featuresep)^1
-                        * equals
-                        * (1 - featuresep)^1
-local assignment        = xetex_option  / handle_xetex_option
-                        + normal_option / handle_normal_option
-                        + ignore_option / handle_invalid_option
-local switch            = P"+" * ws * C(field) * Cc(true)
-                        + P"-" * ws * C(field) * Cc(false)
-                        +             C(field) * Cc(true)   --- default
-local feature_expr      = ws * Cg(assignment + switch) * ws
-local option            = feature_expr
-local feature_list      = Cf(Ct""
-                           * option
-                           * (featuresep * option^-1)^0
-                           , rawset)
-                        * featuresep^-1
-
---- other -------------------------------------------------------------
---- This rule is present in the original parser. It sets the “sub”
---- field of the specification which allows addressing a specific
---- font inside a TTC container. Neither in Luatex-Fonts nor in
---- Luaotfload is this documented, so we might as well silently drop
---- it. However, as backward compatibility is one of our prime goals we
---- just insert it here and leave it undocumented until someone cares
---- to ask. (Note: afair subfonts are numbered, but this rule matches a
---- string; I won’t mess with it though until someone reports a
---- problem.)
---- local subvalue   = P("(") * (C(P(1-S("()"))^1)/issub) * P(")") -- for Kim
---- Note to self: subfonts apparently start at index 0. Tested with
---- Cambria.ttc that includes “Cambria Math” at 0 and “Cambria” at 1.
---- Other values cause luatex to segfault.
-local subfont           = P"(" * Cg((1 - S"()")^1, "sub") * P")"
---- top-level rules ---------------------------------------------------
---- \font\foo=<specification>:<features>
-local features          = Cg(feature_list, "features")
-local specification     = (prefixed + unprefixed)
-                        * subfont^-1
-                        * modifier_list^-1
-local font_request      = Ct(path_lookup   * (colon^-1 * features)^-1
-                           + specification * (colon    * features)^-1)
-
---  lpeg.print(font_request)
---- new parser: 657 rules
---- old parser: 230 rules
-
 local import_values = {
     --- That’s what the 1.x parser did, not quite as graciously,
     --- with an array of branch expressions.
@@ -1138,7 +890,7 @@ end
 
 --- spec -> spec
 local handle_request = function (specification)
-    local request = lpegmatch(font_request,
+    local request = lpegmatch(luaotfload.parsers.font_request,
                               specification.specification)
     if not request then
         --- happens when called with an absolute path
diff --git a/luaotfload-parsers.lua b/luaotfload-parsers.lua
index 0789c42..a989722 100644
--- a/luaotfload-parsers.lua
+++ b/luaotfload-parsers.lua
@@ -1,12 +1,12 @@
 #!/usr/bin/env texlua
------------------------------------------------------------------------
+-------------------------------------------------------------------------------
 --         FILE:  luaotfload-parsers.lua
 --  DESCRIPTION:  various lpeg-based parsers used in Luaotfload
 -- REQUIREMENTS:  Luaotfload > 2.4
 --       AUTHOR:  Philipp Gesang (Phg), <phg42.2a@gmail.com>
 --      VERSION:  same as Luaotfload
 --      CREATED:  2014-01-14 10:15:20+0100
------------------------------------------------------------------------
+-------------------------------------------------------------------------------
 --
 
 if not modules then modules = { } end modules ['luaotfload-parsers'] = {
@@ -44,6 +44,7 @@ local report            = logs.report
 local string            = string
 local stringsub         = string.sub
 local stringfind        = string.find
+local stringlower       = string.lower
 
 local lfs               = lfs
 local lfsisfile         = lfs.isfile
@@ -316,3 +317,262 @@ local splitcomma        = Ct((C(noncomma^1) + comma)^1)
 parsers.splitcomma      = splitcomma
 
 
+
+-------------------------------------------------------------------------------
+---                          FONT REQUEST
+-------------------------------------------------------------------------------
+
+
+--[[doc------------------------------------------------------------------------
+
+    The luaotfload font request syntax (see manual)
+    has a canonical form:
+
+        \font<csname>=<prefix>:<identifier>:<features>
+
+    where
+      <csname> is the control sequence that activates the font
+      <prefix> is either “file” or “name”, determining the lookup
+      <identifer> is either a file name (no path) or a font
+                  name, depending on the lookup
+      <features> is a list of switches or options, separated by
+                 semicolons or commas; a switch is of the form “+” foo
+                 or “-” foo, options are of the form lhs “=” rhs
+
+    however, to ensure backward compatibility we also have
+    support for Xetex-style requests.
+
+    for the Xetex emulation see:
+    · The XeTeX Reference Guide by Will Robertson, 2011
+    · The XeTeX Companion by Michel Goosens, 2010
+    · About XeTeX by Jonathan Kew, 2005
+
+
+    caueat emptor.
+
+        the request is parsed into one of **four** different lookup
+        categories: the regular ones, file and name, as well as the
+        Xetex compatibility ones, path and anon. (maybe a better choice
+        of identifier would be “ambig”.)
+
+        according to my reconstruction, the correct chaining of the
+        lookups for each category is as follows:
+
+        | File -> ( db/filename lookup )
+
+        | Name -> ( db/name lookup,
+                    db/filename lookup )
+
+        | Path -> ( db/filename lookup,
+                    fullpath lookup )
+
+        | Anon -> ( kpse.find_file(),     // <- for tfm, ofm
+                    db/name lookup,
+                    db/filename lookup,
+                    fullpath lookup )
+
+        caching of successful lookups is essential. we now as of v2.2
+        have a lookup cache that is stored in a separate file. it
+        pertains only to name: lookups, and is described in more detail
+        in luaotfload-database.lua.
+
+-------------------------------------------------------------------------------
+
+    One further incompatibility between Xetex and Luatex-Fonts consists
+    in their option list syntax: apparently, Xetex requires key-value
+    options to be prefixed by a "+" (ascii “plus”) character. We
+    silently accept this as well, dropping the first byte if it is a
+    plus or minus character.
+
+    Reference: https://github.com/lualatex/luaotfload/issues/79#issuecomment-18104483
+
+--doc]]------------------------------------------------------------------------
+
+
+local handle_normal_option = function (key, val)
+    val = stringlower(val)
+    --- the former “toboolean()” handler
+    if val == "true"  then
+        val = true
+    elseif val == "false" then
+        val = false
+    end
+    return key, val
+end
+
+--[[doc--
+
+    Xetex style indexing begins at zero which we just increment before
+    passing it along to the font loader.  Ymmv.
+
+--doc]]--
+
+local handle_xetex_option = function (key, val)
+    val = stringlower(val)
+    local numeric = tonumber(val) --- decimal only; keeps colors intact
+    if numeric then --- ugh
+        if  mathceil(numeric) == numeric then -- integer, possible index
+            val = tostring(numeric + 1)
+        end
+    elseif val == "true"  then
+        val = true
+    elseif val == "false" then
+        val = false
+    end
+    return key, val
+end
+
+--[[doc--
+
+    Instead of silently ignoring invalid options we emit a warning to
+    the log.
+
+    Note that we have to return a pair to please rawset(). This creates
+    an entry on the resulting features hash which will later be removed
+    during set_default_features().
+
+--doc]]--
+
+local handle_invalid_option = function (opt)
+    report("log", 0, "load", "font option %q unknown.", opt)
+    return "", false
+end
+
+--[[doc--
+
+    Dirty test if a file: request is actually a path: lookup; don’t
+    ask! Note this fails on Windows-style absolute paths. These will
+    *really* have to use the correct request.
+
+--doc]]--
+
+local check_garbage = function (_,i, garbage)
+    if stringfind(garbage, "/") then
+        report("log", 0, "load",  --- ffs use path!
+            "warning: path in file: lookups is deprecated; ")
+        report("log", 0, "load", "use bracket syntax instead!")
+        report("log", 0, "load",
+            "position: %d; full match: %q",
+            i, garbage)
+        return true
+    end
+    return false
+end
+
+local lpegmatch = lpeg.match
+local P, S, R   = lpeg.P, lpeg.S, lpeg.R
+local C, Cc, Cf, Cg, Cmt, Cs, Ct
+    = lpeg.C, lpeg.Cc, lpeg.Cf, lpeg.Cg, lpeg.Cmt, lpeg.Cs, lpeg.Ct
+
+--- terminals and low-level classes -----------------------------------
+--- note we could use the predefined ones from lpeg.patterns
+local dot         = P"."
+local colon       = P":"
+local featuresep  = S",;"
+local slash       = P"/"
+local equals      = P"="
+local lbrk, rbrk  = P"[", P"]"
+
+local spacing     = S" \t\v"
+local ws          = spacing^0
+
+local digit       = R"09"
+local alpha       = R("az", "AZ")
+local anum        = alpha + digit
+local decimal     = digit^1 * (dot * digit^0)^-1
+
+--- modifiers ---------------------------------------------------------
+--[[doc--
+    The slash notation: called “modifiers” (Kew) or “font options”
+    (Robertson, Goosens)
+    we only support the shorthands for italic / bold / bold italic
+    shapes, as well as setting optical size, the rest is ignored.
+--doc]]--
+local style_modifier    = (P"BI" + P"IB" + P"bi" + P"ib" + S"biBI")
+                        / stringlower
+local size_modifier     = S"Ss" * P"="    --- optical size
+                        * Cc"optsize" * C(decimal)
+local other_modifier    = P"AAT" + P"aat" --- apple stuff;  unsupported
+                        + P"ICU" + P"icu" --- not applicable
+                        + P"GR"  + P"gr"  --- sil stuff;    unsupported
+local garbage_modifier  = ((1 - colon - slash)^0 * Cc(false))
+local modifier          = slash * (other_modifier      --> ignore
+                                 + Cs(style_modifier)  --> collect
+                                 + Ct(size_modifier)   --> collect
+                                 + garbage_modifier)   --> warn
+local modifier_list     = Cg(Ct(modifier^0), "modifiers")
+
+--- lookups -----------------------------------------------------------
+local fontname          = C((1-S":(/")^1)  --- like luatex-fonts
+local unsupported       = Cmt((1-S":(")^1, check_garbage)
+local prefixed          = P"name:" * ws * Cg(fontname, "name")
+--- initially we intended file: to emulate the behavior of
+--- luatex-fonts, i.e. no paths allowed. after all, we do have XeTeX
+--- emulation with the path lookup and it interferes with db lookups.
+--- turns out fontspec and other widely used packages rely on file:
+--- with paths already, so we’ll add a less strict rule here.  anyways,
+--- we’ll emit a warning.
+                        + P"file:" * ws * Cg(unsupported, "path")
+                        + P"file:" * ws * Cg(fontname, "file")
+--- EXPERIMENTAL: kpse lookup
+                        + P"kpse:" * ws * Cg(fontname, "kpse")
+--- EXPERIMENTAL: custom lookup
+                        + P"my:" * ws * Cg(fontname, "my")
+local unprefixed        = Cg(fontname, "anon")
+local path_lookup       = lbrk * Cg(C((1-rbrk)^1), "path") * rbrk
+
+--- features ----------------------------------------------------------
+local field_char        = anum + S"+-." --- sic!
+local field             = field_char^1
+--- assignments are “lhs=rhs”
+---              or “+lhs=rhs” (Xetex-style)
+--- switches    are “+key” | “-key”
+local normal_option     = C(field) * ws * equals * ws * C(field) * ws
+local xetex_option      = P"+" * ws * normal_option
+local ignore_option     = (1 - equals - featuresep)^1
+                        * equals
+                        * (1 - featuresep)^1
+local assignment        = xetex_option  / handle_xetex_option
+                        + normal_option / handle_normal_option
+                        + ignore_option / handle_invalid_option
+local switch            = P"+" * ws * C(field) * Cc(true)
+                        + P"-" * ws * C(field) * Cc(false)
+                        +             C(field) * Cc(true)   --- default
+local feature_expr      = ws * Cg(assignment + switch) * ws
+local option            = feature_expr
+local feature_list      = Cf(Ct""
+                           * option
+                           * (featuresep * option^-1)^0
+                           , rawset)
+                        * featuresep^-1
+
+--- other -------------------------------------------------------------
+--- This rule is present in the original parser. It sets the “sub”
+--- field of the specification which allows addressing a specific
+--- font inside a TTC container. Neither in Luatex-Fonts nor in
+--- Luaotfload is this documented, so we might as well silently drop
+--- it. However, as backward compatibility is one of our prime goals we
+--- just insert it here and leave it undocumented until someone cares
+--- to ask. (Note: afair subfonts are numbered, but this rule matches a
+--- string; I won’t mess with it though until someone reports a
+--- problem.)
+--- local subvalue   = P("(") * (C(P(1-S("()"))^1)/issub) * P(")") -- for Kim
+--- Note to self: subfonts apparently start at index 0. Tested with
+--- Cambria.ttc that includes “Cambria Math” at 0 and “Cambria” at 1.
+--- Other values cause luatex to segfault.
+local subfont           = P"(" * Cg((1 - S"()")^1, "sub") * P")"
+--- top-level rules ---------------------------------------------------
+--- \font\foo=<specification>:<features>
+local features          = Cg(feature_list, "features")
+local specification     = (prefixed + unprefixed)
+                        * subfont^-1
+                        * modifier_list^-1
+local font_request      = Ct(path_lookup   * (colon^-1 * features)^-1
+                           + specification * (colon    * features)^-1)
+
+--  lpeg.print(font_request)
+--- v2.5 parser: 1065 rules
+--- v1.2 parser:  230 rules
+
+luaotfload.parsers.font_request = font_request
+