refactor regex API again

This time making it easier to use from Lua side.
- regex is now a proper userdata instead of a table wrapped in userdata
- add regex:get_metadata()
- remove regex:nametable() option in favor of regex:get_metadata()
This commit is contained in:
takase1121 2021-10-09 12:32:36 +08:00
parent 09d6087df8
commit 1de4d0a4a3
1 changed files with 60 additions and 52 deletions

View File

@ -6,28 +6,10 @@
#include <string.h>
#include <pcre2.h>
// something similiar to luaL_checkudata() but for regex only
static pcre2_code* check_regex(lua_State* L, int arg) {
luaL_checktype(L, arg, LUA_TTABLE);
int hasmt = 0;
pcre2_code* re = NULL;
if (lua_getmetatable(L, arg)) {
luaL_getmetatable(L, API_TYPE_REGEX);
hasmt = lua_rawequal(L, -1, -2);
lua_pop(L, 2);
}
lua_rawgeti(L, arg, 1);
re = lua_touserdata(L, -1);
lua_pop(L, 1);
if (!hasmt || re == NULL)
luaL_argerror(L, arg, "invalid regex object");
return re;
}
typedef struct regex_t {
pcre2_code *re;
int metadata;
} regex_t;
static int pcre2_error(lua_State *L, int rc, const char *fmt, ...) {
va_list ap;
@ -45,9 +27,10 @@ static int pcre2_error(lua_State *L, int rc, const char *fmt, ...) {
}
static int f_pcre_gc(lua_State* L) {
pcre2_code* re = check_regex(L, 1);
if (re)
pcre2_code_free(re);
regex_t *re = luaL_checkudata(L, 1, API_TYPE_REGEX);
luaL_unref(L, LUA_REGISTRYINDEX, re->metadata);
if (re->re)
pcre2_code_free(re->re);
return 0;
}
@ -77,11 +60,11 @@ static int f_pcre_compile(lua_State *L) {
if (re == NULL)
return pcre2_error(L, error, "regex compilation failed at %d: ", error_offset);
lua_newtable(L);
regex_t *regex = (regex_t*) lua_newuserdata(L, sizeof(regex_t));
luaL_setmetatable(L, API_TYPE_REGEX);
regex->re = re;
lua_pushlightuserdata(L, (void*) re);
lua_rawseti(L, -2, 1);
lua_newtable(L);
lua_pushvalue(L, 1);
lua_setfield(L, -2, "source");
@ -89,12 +72,12 @@ static int f_pcre_compile(lua_State *L) {
lua_pushstring(L, optstr);
lua_setfield(L, -2, "flags");
regex->metadata = luaL_ref(L, LUA_REGISTRYINDEX);
return 1;
}
// get nametable (useful for named captures)
static int f_pcre_nametable(lua_State* L) {
pcre2_code* re = check_regex(L, 1);
static int get_nametable(lua_State *L, pcre2_code *re) {
int ret;
uint32_t namecount, entrysize;
PCRE2_SPTR nametable;
@ -107,7 +90,7 @@ static int f_pcre_nametable(lua_State* L) {
pcre2_pattern_info(re, PCRE2_INFO_NAMEENTRYSIZE, &entrysize);
lua_createtable(L, namecount, 0);
for (uint32_t i = 0; i < namecount; i++) {
for (uint32_t i = 1; i <= namecount; i++) {
uint16_t index = (nametable[1] << 0) | (nametable[0] << 8);
nametable += 2;
@ -116,7 +99,7 @@ static int f_pcre_nametable(lua_State* L) {
lua_setfield(L, -2, "index");
lua_pushstring(L, (const char *) nametable);
lua_setfield(L, -2, "name");
lua_rawseti(L, -2, i + 1);
lua_rawseti(L, -2, i);
nametable += (entrysize - 2);
}
@ -124,25 +107,51 @@ static int f_pcre_nametable(lua_State* L) {
return 1;
}
static int f_pcre_get_metadata(lua_State *L) {
regex_t *re = luaL_checkudata(L, 1, API_TYPE_REGEX);
lua_rawgeti(L, LUA_REGISTRYINDEX, re->metadata);
lua_getfield(L, -1, "nametable");
if (lua_type(L, -1) == LUA_TTABLE) {
lua_pop(L, 1);
} else {
lua_pop(L, 1);
get_nametable(L, re->re);
lua_setfield(L, -2, "nametable");
}
return 1;
}
static size_t check_bitfield(lua_State *L, int t) {
size_t size, bit = 0;
luaL_checktype(L, t, LUA_TTABLE);
size = lua_rawlen(L, t);
for (size_t i = 1; i <= size; i++) {
lua_rawgeti(L, t, i);
bit |= luaL_checkinteger(L, -1);
lua_pop(L, 1);
}
return bit;
}
// Takes string, compiled regex, returns list of indices of matched groups
// (including the whole match), if a match was found.
static int f_pcre_match(lua_State *L) {
size_t len, offset, options;
pcre2_code* re = check_regex(L, 1);
regex_t *re = luaL_checkudata(L, 1, API_TYPE_REGEX);
const char* str = luaL_checklstring(L, 2, &len);
offset = luaL_optnumber(L, 3, 1);
options = luaL_optnumber(L, 4, 0);
options = lua_gettop(L) == 4 ? check_bitfield(L, 4) : 0;
pcre2_match_data* md = pcre2_match_data_create_from_pattern(re, NULL);
int rc = pcre2_match(re, (PCRE2_SPTR)str, len, offset - 1, options, md, NULL);
pcre2_match_data* md = pcre2_match_data_create_from_pattern(re->re, NULL);
int rc = pcre2_match(re->re, (PCRE2_SPTR)str, len, offset - 1, options, md, NULL);
if (rc < 0) {
pcre2_match_data_free(md);
if (rc != PCRE2_ERROR_NOMATCH) {
PCRE2_UCHAR buffer[256];
pcre2_get_error_message(rc, buffer, sizeof(buffer));
luaL_error(L, "regex matching error: %s", buffer);
}
if (rc != PCRE2_ERROR_NOMATCH)
return pcre2_error(L, rc, "regex matching error: ");
else
return 0;
}
@ -152,9 +161,8 @@ static int f_pcre_match(lua_State *L) {
/* We must guard against patterns such as /(?=.\K)/ that use \K in an
assertion to set the start of a match later than its end. In the editor,
we just detect this case and give up. */
luaL_error(L, "regex matching error: \\K was used in an assertion to "
return luaL_error(L, "regex matching error: \\K was used in an assertion to "
" set the match start after its end");
return 0;
}
rc *= 2;
@ -167,7 +175,7 @@ static int f_pcre_match(lua_State *L) {
}
static const luaL_Reg lib[] = {
{ "nametable", f_pcre_nametable },
{ "get_metadata", f_pcre_get_metadata },
{ "compile", f_pcre_compile },
{ "cmatch", f_pcre_match },
{ "__gc", f_pcre_gc },