Merge pull request #18 from phi-gamma/master

sync

1 files changed, 241 insertions, 84 deletions

diff --git a/lualibs-unicode.lua b/lualibs-unicode.lua
index cf0c93d..3ce5bd3 100644
--- a/lualibs-unicode.lua
+++ b/lualibs-unicode.lua
@@ -25,7 +25,7 @@ utf.values     = utf.values     or string.utfvalues
 -- string.bytepairs
 
 local type = type
-local char, byte, format, sub = string.char, string.byte, string.format, string.sub
+local char, byte, format, sub, gmatch = string.char, string.byte, string.format, string.sub, string.gmatch
 local concat = table.concat
 local P, C, R, Cs, Ct, Cmt, Cc, Carg, Cp = lpeg.P, lpeg.C, lpeg.R, lpeg.Cs, lpeg.Ct, lpeg.Cmt, lpeg.Cc, lpeg.Carg, lpeg.Cp
 local lpegmatch, patterns = lpeg.match, lpeg.patterns
@@ -621,116 +621,273 @@ function utf.magic(f) -- not used
     return lpegmatch(p_utftype,str)
 end
 
-local function utf16_to_utf8_be(t)
-    if type(t) == "string" then
-        t = lpegmatch(utflinesplitter,t)
-    end
-    local result = { } -- we reuse result
-    for i=1,#t do
-        local r, more = 0, 0
-        for left, right in bytepairs(t[i]) do
-            if right then
-                local now = 256*left + right
-                if more > 0 then
-                    now = (more-0xD800)*0x400 + (now-0xDC00) + 0x10000 -- the 0x10000 smells wrong
-                    more = 0
-                    r = r + 1
-                    result[r] = utfchar(now)
-                elseif now >= 0xD800 and now <= 0xDBFF then
-                    more = now
-                else
-                    r = r + 1
-                    result[r] = utfchar(now)
+local utf16_to_utf8_be, utf16_to_utf8_le
+local utf32_to_utf8_be, utf32_to_utf8_le
+
+local utf_16_be_linesplitter = patterns.utfbom_16_be^-1 * lpeg.tsplitat(patterns.utf_16_be_nl)
+local utf_16_le_linesplitter = patterns.utfbom_16_le^-1 * lpeg.tsplitat(patterns.utf_16_le_nl)
+
+-- we have three possibilities:
+
+-- bytepairs: 0.048
+-- gmatch   : 0.069
+-- lpeg     : 0.089 (match time captures)
+
+if bytepairs then
+
+    -- with a little bit more code we could include the linesplitter
+
+    utf16_to_utf8_be = function(t)
+        if type(t) == "string" then
+            t = lpegmatch(utf_16_be_linesplitter,t)
+        end
+        local result = { } -- we reuse result
+        for i=1,#t do
+            local r, more = 0, 0
+            for left, right in bytepairs(t[i]) do
+                if right then
+                    local now = 256*left + right
+                    if more > 0 then
+                        now = (more-0xD800)*0x400 + (now-0xDC00) + 0x10000 -- the 0x10000 smells wrong
+                        more = 0
+                        r = r + 1
+                        result[r] = utfchar(now)
+                    elseif now >= 0xD800 and now <= 0xDBFF then
+                        more = now
+                    else
+                        r = r + 1
+                        result[r] = utfchar(now)
+                    end
                 end
             end
+            t[i] = concat(result,"",1,r) -- we reused tmp, hence t
         end
-        t[i] = concat(result,"",1,r) -- we reused tmp, hence t
+        return t
     end
-    return t
-end
 
-local function utf16_to_utf8_le(t)
-    if type(t) == "string" then
-        t = lpegmatch(utflinesplitter,t)
+    utf16_to_utf8_le = function(t)
+        if type(t) == "string" then
+            t = lpegmatch(utf_16_le_linesplitter,t)
+        end
+        local result = { } -- we reuse result
+        for i=1,#t do
+            local r, more = 0, 0
+            for left, right in bytepairs(t[i]) do
+                if right then
+                    local now = 256*right + left
+                    if more > 0 then
+                        now = (more-0xD800)*0x400 + (now-0xDC00) + 0x10000 -- the 0x10000 smells wrong
+                        more = 0
+                        r = r + 1
+                        result[r] = utfchar(now)
+                    elseif now >= 0xD800 and now <= 0xDBFF then
+                        more = now
+                    else
+                        r = r + 1
+                        result[r] = utfchar(now)
+                    end
+                end
+            end
+            t[i] = concat(result,"",1,r) -- we reused tmp, hence t
+        end
+        return t
     end
-    local result = { } -- we reuse result
-    for i=1,#t do
-        local r, more = 0, 0
-        for left, right in bytepairs(t[i]) do
-            if right then
-                local now = 256*right + left
-                if more > 0 then
-                    now = (more-0xD800)*0x400 + (now-0xDC00) + 0x10000 -- the 0x10000 smells wrong
-                    more = 0
-                    r = r + 1
-                    result[r] = utfchar(now)
-                elseif now >= 0xD800 and now <= 0xDBFF then
-                    more = now
+
+    utf32_to_utf8_be = function(t)
+        if type(t) == "string" then
+            t = lpegmatch(utflinesplitter,t)
+        end
+        local result = { } -- we reuse result
+        for i=1,#t do
+            local r, more = 0, -1
+            for a,b in bytepairs(t[i]) do
+                if a and b then
+                    if more < 0 then
+                        more = 256*256*256*a + 256*256*b
+                    else
+                        r = r + 1
+                        result[t] = utfchar(more + 256*a + b)
+                        more = -1
+                    end
                 else
-                    r = r + 1
-                    result[r] = utfchar(now)
+                    break
                 end
             end
+            t[i] = concat(result,"",1,r)
         end
-        t[i] = concat(result,"",1,r) -- we reused tmp, hence t
+        return t
     end
-    return t
-end
 
-local function utf32_to_utf8_be(t)
-    if type(t) == "string" then
-        t = lpegmatch(utflinesplitter,t)
-    end
-    local result = { } -- we reuse result
-    for i=1,#t do
-        local r, more = 0, -1
-        for a,b in bytepairs(t[i]) do
-            if a and b then
-                if more < 0 then
-                    more = 256*256*256*a + 256*256*b
+    utf32_to_utf8_le = function(t)
+        if type(t) == "string" then
+            t = lpegmatch(utflinesplitter,t)
+        end
+        local result = { } -- we reuse result
+        for i=1,#t do
+            local r, more = 0, -1
+            for a,b in bytepairs(t[i]) do
+                if a and b then
+                    if more < 0 then
+                        more = 256*b + a
+                    else
+                        r = r + 1
+                        result[t] = utfchar(more + 256*256*256*b + 256*256*a)
+                        more = -1
+                    end
                 else
-                    r = r + 1
-                    result[t] = utfchar(more + 256*a + b)
-                    more = -1
+                    break
                 end
-            else
-                break
             end
+            t[i] = concat(result,"",1,r)
         end
-        t[i] = concat(result,"",1,r)
+        return t
     end
-    return t
-end
 
-local function utf32_to_utf8_le(t)
-    if type(t) == "string" then
-        t = lpegmatch(utflinesplitter,t)
+else
+
+    utf16_to_utf8_be = function(t)
+        if type(t) == "string" then
+            t = lpegmatch(utf_16_be_linesplitter,t)
+        end
+        local result = { } -- we reuse result
+        for i=1,#t do
+            local r, more = 0, 0
+            for left, right in gmatch(t[i],"(.)(.)") do
+                if left == "\000" then -- experiment
+                    r = r + 1
+                    result[r] = utfchar(byte(right))
+                elseif right then
+                    local now = 256*byte(left) + byte(right)
+                    if more > 0 then
+                        now = (more-0xD800)*0x400 + (now-0xDC00) + 0x10000 -- the 0x10000 smells wrong
+                        more = 0
+                        r = r + 1
+                        result[r] = utfchar(now)
+                    elseif now >= 0xD800 and now <= 0xDBFF then
+                        more = now
+                    else
+                        r = r + 1
+                        result[r] = utfchar(now)
+                    end
+                end
+            end
+            t[i] = concat(result,"",1,r) -- we reused tmp, hence t
+        end
+        return t
     end
-    local result = { } -- we reuse result
-    for i=1,#t do
-        local r, more = 0, -1
-        for a,b in bytepairs(t[i]) do
-            if a and b then
-                if more < 0 then
-                    more = 256*b + a
-                else
+
+    utf16_to_utf8_le = function(t)
+        if type(t) == "string" then
+            t = lpegmatch(utf_16_le_linesplitter,t)
+        end
+        local result = { } -- we reuse result
+        for i=1,#t do
+            local r, more = 0, 0
+            for left, right in gmatch(t[i],"(.)(.)") do
+                if right == "\000" then
                     r = r + 1
-                    result[t] = utfchar(more + 256*256*256*b + 256*256*a)
-                    more = -1
+                    result[r] = utfchar(byte(left))
+                elseif right then
+                    local now = 256*byte(right) + byte(left)
+                    if more > 0 then
+                        now = (more-0xD800)*0x400 + (now-0xDC00) + 0x10000 -- the 0x10000 smells wrong
+                        more = 0
+                        r = r + 1
+                        result[r] = utfchar(now)
+                    elseif now >= 0xD800 and now <= 0xDBFF then
+                        more = now
+                    else
+                        r = r + 1
+                        result[r] = utfchar(now)
+                    end
                 end
-            else
-                break
             end
+            t[i] = concat(result,"",1,r) -- we reused tmp, hence t
         end
-        t[i] = concat(result,"",1,r)
+        return t
     end
-    return t
+
+    utf32_to_utf8_le = function() return { } end -- never used anyway
+    utf32_to_utf8_be = function() return { } end -- never used anyway
+
+    -- the next one is slighty slower
+
+    -- local result, lines, r, more = { }, { }, 0, 0
+    --
+    -- local simple = Cmt(
+    --     C(1) * C(1), function(str,p,left,right)
+    --         local now = 256*byte(left) + byte(right)
+    --         if more > 0 then
+    --             now = (more-0xD800)*0x400 + (now-0xDC00) + 0x10000 -- the 0x10000 smells wrong
+    --             more = 0
+    --             r = r + 1
+    --             result[r] = utfchar(now)
+    --         elseif now >= 0xD800 and now <= 0xDBFF then
+    --             more = now
+    --         else
+    --             r = r + 1
+    --             result[r] = utfchar(now)
+    --         end
+    --         return p
+    --    end
+    -- )
+    --
+    -- local complex = Cmt(
+    --     C(1) * C(1), function(str,p,left,right)
+    --         local now = 256*byte(left) + byte(right)
+    --         if more > 0 then
+    --             now = (more-0xD800)*0x400 + (now-0xDC00) + 0x10000 -- the 0x10000 smells wrong
+    --             more = 0
+    --             r = r + 1
+    --             result[r] = utfchar(now)
+    --         elseif now >= 0xD800 and now <= 0xDBFF then
+    --             more = now
+    --         else
+    --             r = r + 1
+    --             result[r] = utfchar(now)
+    --         end
+    --         return p
+    --    end
+    -- )
+    --
+    -- local lineend = Cmt (
+    --     patterns.utf_16_be_nl, function(str,p)
+    --         lines[#lines+1] = concat(result,"",1,r)
+    --         r, more = 0, 0
+    --         return p
+    --     end
+    -- )
+    --
+    -- local be_1 = patterns.utfbom_16_be^-1 * (simple + complex)^0
+    -- local be_2 = patterns.utfbom_16_be^-1 * (lineend + simple + complex)^0
+    --
+    -- utf16_to_utf8_be = function(t)
+    --     if type(t) == "string" then
+    --         local s = t
+    --         lines, r, more = { }, 0, 0
+    --         lpegmatch(be_2,s)
+    --         if r > 0 then
+    --             lines[#lines+1] = concat(result,"",1,r)
+    --         end
+    --         result = { }
+    --         return lines
+    --     else
+    --         for i=1,#t do
+    --             r, more = 0, 0
+    --             lpegmatch(be_1,t[i])
+    --             t[i] = concat(result,"",1,r)
+    --         end
+    --         result = { }
+    --         return t
+    --     end
+    -- end
+
 end
 
-utf.utf32_to_utf8_be = utf32_to_utf8_be
-utf.utf32_to_utf8_le = utf32_to_utf8_le
-utf.utf16_to_utf8_be = utf16_to_utf8_be
 utf.utf16_to_utf8_le = utf16_to_utf8_le
+utf.utf16_to_utf8_be = utf16_to_utf8_be
+utf.utf32_to_utf8_le = utf32_to_utf8_le
+utf.utf32_to_utf8_be = utf32_to_utf8_be
 
 function utf.utf8_to_utf8(t)
     return type(t) == "string" and lpegmatch(utflinesplitter,t) or t