aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
-rw-r--r--src/lutf8lib.js44
1 files changed, 18 insertions, 26 deletions
diff --git a/src/lutf8lib.js b/src/lutf8lib.js
index de97bec..3c9147a 100644
--- a/src/lutf8lib.js
+++ b/src/lutf8lib.js
@@ -21,18 +21,17 @@ const u_posrelat = function(pos, len) {
/*
** Decode one UTF-8 sequence, returning NULL if byte sequence is invalid.
*/
-const utf8_decode = function(s) {
- let limits = [0xFF, 0x7F, 0x7FF, 0xFFFF];
- let c = s[0];
+const limits = [0xFF, 0x7F, 0x7FF, 0xFFFF];
+const utf8_decode = function(s, pos) {
+ let c = s[pos];
let res = 0; /* final result */
- let pos = 0;
if (c < 0x80) /* ascii? */
res = c;
else {
let count = 0; /* to count number of continuation bytes */
while (c & 0x40) { /* still have continuation bytes? */
- let cc = s[++count]; /* read next byte */
- if ((cc & 0xC0) != 0x80) /* not a continuation byte? */
+ let cc = s[pos + (++count)]; /* read next byte */
+ if ((cc & 0xC0) !== 0x80) /* not a continuation byte? */
return null; /* invalid byte sequence */
res = (res << 6) | (cc & 0x3F); /* add lower 6 bits from cont. byte */
c <<= 1; /* to test next bit */
@@ -40,12 +39,10 @@ const utf8_decode = function(s) {
res |= ((c & 0x7F) << (count * 5)); /* add first byte */
if (count > 3 || res > MAXUNICODE || res <= limits[count])
return null; /* invalid byte sequence */
- s = s.slice(count); /* skip continuation bytes read */
- pos += count;
+ pos += count; /* skip continuation bytes read */
}
return {
- string: s.slice(1), /* +1 to include first byte */
code: res,
pos: pos + 1
};
@@ -63,19 +60,17 @@ const utflen = function(L) {
let posi = u_posrelat(lauxlib.luaL_optinteger(L, 2, 1), len);
let posj = u_posrelat(lauxlib.luaL_optinteger(L, 3, -1), len);
- lauxlib.luaL_argcheck(L, 1 <= posi && --posi <= len, 2, "initial position out of string");
- lauxlib.luaL_argcheck(L, --posj < len, 3, "final position out of string");
+ lauxlib.luaL_argcheck(L, 1 <= posi && --posi <= len, 2, lua.to_luastring("initial position out of string"));
+ lauxlib.luaL_argcheck(L, --posj < len, 3, lua.to_luastring("final position out of string"));
while (posi <= posj) {
- let dec = utf8_decode(s.slice(posi));
- let s1 = dec ? dec.string : null;
- if (s1 === null) {
- /* conversion error? */
+ let dec = utf8_decode(s, posi);
+ if (dec === null) { /* conversion error? */
lua.lua_pushnil(L); /* return nil ... */
lua.lua_pushinteger(L, posi + 1); /* ... and current position */
return 2;
}
- posi = s.length - s1.length;
+ posi = dec.pos;
n++;
}
lua.lua_pushinteger(L, n);
@@ -170,13 +165,12 @@ const codepoint = function(L) {
let n = (pose - posi) + 1;
lauxlib.luaL_checkstack(L, n, lua.to_luastring("string slice too long", true));
n = 0;
- for (s = s.slice(posi - 1); n < pose - posi;) {
- let dec = utf8_decode(s);
+ for (posi -= 1; posi < pose;) {
+ let dec = utf8_decode(s, posi);
if (dec === null)
return lauxlib.luaL_error(L, lua.to_luastring("invalid UTF-8 code", true));
- s = dec.string;
- let code = dec.code;
- lua.lua_pushinteger(L, code);
+ lua.lua_pushinteger(L, dec.code);
+ posi = dec.pos;
n++;
}
return n;
@@ -197,13 +191,11 @@ const iter_aux = function(L) {
if (n >= len)
return 0; /* no more codepoints */
else {
- let dec = utf8_decode(s.slice(n));
- let code = dec ? dec.code : null;
- let next = dec ? dec.string : null;
- if (next === null || iscont(next[0]))
+ let dec = utf8_decode(s, n);
+ if (dec === null || iscont(s[dec.pos]))
return lauxlib.luaL_error(L, lua.to_luastring("invalid UTF-8 code", true));
lua.lua_pushinteger(L, n + 1);
- lua.lua_pushinteger(L, code);
+ lua.lua_pushinteger(L, dec.code);
return 2;
}
};