literal_t *parse_regexp(parser_ctx_t *ctx)
{
const WCHAR *re, *flags_ptr;
BOOL in_class = FALSE;
DWORD re_len, flags;
literal_t *ret;
HRESULT hres;
TRACE("\n");
while(*--ctx->ptr != '/');
/* Simple regexp pre-parser; '/' if used in char class does not terminate regexp literal */
re = ++ctx->ptr;
while(ctx->ptr < ctx->end) {
if(*ctx->ptr == '\\') {
if(++ctx->ptr == ctx->end)
break;
}else if(in_class) {
if(*ctx->ptr == '\n')
break;
if(*ctx->ptr == ']')
in_class = FALSE;
}else {
if(*ctx->ptr == '/')
break;
if(*ctx->ptr == '[')
in_class = TRUE;
}
ctx->ptr++;
}
if(ctx->ptr == ctx->end || *ctx->ptr != '/') {
WARN("pre-parsing failed\n");
return NULL;
}
re_len = ctx->ptr-re;
flags_ptr = ++ctx->ptr;
while(ctx->ptr < ctx->end && isalnumW(*ctx->ptr))
ctx->ptr++;
hres = parse_regexp_flags(flags_ptr, ctx->ptr-flags_ptr, &flags);
if(FAILED(hres))
return NULL;
ret = parser_alloc(ctx, sizeof(literal_t));
ret->type = LT_REGEXP;
ret->u.regexp.str = re;
ret->u.regexp.str_len = re_len;
ret->u.regexp.flags = flags;
return ret;
}
HRESULT create_regexp_var(script_ctx_t *ctx, jsval_t src_arg, jsval_t *flags_arg, jsdisp_t **ret)
{
unsigned flags, opt_len = 0;
const WCHAR *opt = NULL;
jsstr_t *src;
HRESULT hres;
if(is_object_instance(src_arg)) {
jsdisp_t *obj;
obj = iface_to_jsdisp((IUnknown*)get_object(src_arg));
if(obj) {
if(is_class(obj, JSCLASS_REGEXP)) {
RegExpInstance *regexp = (RegExpInstance*)obj;
hres = create_regexp(ctx, regexp->str, regexp->jsregexp->flags, ret);
jsdisp_release(obj);
return hres;
}
jsdisp_release(obj);
}
}
if(!is_string(src_arg)) {
FIXME("src_arg = %s\n", debugstr_jsval(src_arg));
return E_NOTIMPL;
}
src = get_string(src_arg);
if(flags_arg) {
jsstr_t *opt_str;
if(!is_string(*flags_arg)) {
FIXME("unimplemented for %s\n", debugstr_jsval(*flags_arg));
return E_NOTIMPL;
}
opt_str = get_string(*flags_arg);
opt = jsstr_flatten(opt_str);
if(!opt)
return E_OUTOFMEMORY;
opt_len = jsstr_length(opt_str);
}
hres = parse_regexp_flags(opt, opt_len, &flags);
if(FAILED(hres))
return hres;
return create_regexp(ctx, src, flags, ret);
}
literal_t *parse_regexp(parser_ctx_t *ctx)
{
const WCHAR *re, *flags_ptr;
DWORD re_len, flags;
literal_t *ret;
HRESULT hres;
TRACE("\n");
while(*ctx->ptr != '/')
ctx->ptr--;
re = ++ctx->ptr;
while(ctx->ptr < ctx->end && *ctx->ptr != '/') {
if(*ctx->ptr++ == '\\' && ctx->ptr < ctx->end)
ctx->ptr++;
}
if(ctx->ptr == ctx->end) {
WARN("unexpected end of file\n");
return NULL;
}
re_len = ctx->ptr-re;
flags_ptr = ++ctx->ptr;
while(ctx->ptr < ctx->end && isalnumW(*ctx->ptr))
ctx->ptr++;
hres = parse_regexp_flags(flags_ptr, ctx->ptr-flags_ptr, &flags);
if(FAILED(hres))
return NULL;
ret = parser_alloc(ctx, sizeof(literal_t));
ret->type = LT_REGEXP;
ret->u.regexp.str = re;
ret->u.regexp.str_len = re_len;
ret->u.regexp.flags = flags;
return ret;
}
void duk_regexp_compile(duk_hthread *thr) {
duk_context *ctx = (duk_context *) thr;
duk_re_compiler_ctx re_ctx;
duk_lexer_point lex_point;
duk_hstring *h_pattern;
duk_hstring *h_flags;
duk_hbuffer_dynamic *h_buffer;
DUK_ASSERT(thr != NULL);
DUK_ASSERT(ctx != NULL);
/*
* Args validation
*/
/* TypeError if fails */
h_pattern = duk_require_hstring(ctx, -2);
h_flags = duk_require_hstring(ctx, -1);
/*
* Create normalized 'source' property (E5 Section 15.10.3).
*/
/* [ ... pattern flags ] */
create_escaped_source(thr, -2);
/* [ ... pattern flags escaped_source ] */
/*
* Init compilation context
*/
duk_push_dynamic_buffer(ctx, 0);
h_buffer = (duk_hbuffer_dynamic *) duk_require_hbuffer(ctx, -1);
DUK_ASSERT(DUK_HBUFFER_HAS_DYNAMIC(h_buffer));
/* [ ... pattern flags escaped_source buffer ] */
DUK_MEMSET(&re_ctx, 0, sizeof(re_ctx));
DUK_LEXER_INITCTX(&re_ctx.lex); /* duplicate zeroing, expect for (possible) NULL inits */
re_ctx.thr = thr;
re_ctx.lex.thr = thr;
re_ctx.lex.input = DUK_HSTRING_GET_DATA(h_pattern);
re_ctx.lex.input_length = DUK_HSTRING_GET_BYTELEN(h_pattern);
re_ctx.buf = h_buffer;
re_ctx.recursion_limit = DUK_RE_COMPILE_RECURSION_LIMIT;
re_ctx.re_flags = parse_regexp_flags(thr, h_flags);
DUK_DDPRINT("regexp compiler ctx initialized, flags=0x%08x, recursion_limit=%d",
(unsigned int) re_ctx.re_flags, (int) re_ctx.recursion_limit);
/*
* Init lexer
*/
lex_point.offset = 0; /* expensive init, just want to fill window */
lex_point.line = 1;
DUK_LEXER_SETPOINT(&re_ctx.lex, &lex_point);
/*
* Compilation
*/
DUK_DPRINT("starting regexp compilation");
append_u32(&re_ctx, DUK_REOP_SAVE);
append_u32(&re_ctx, 0);
(void) parse_disjunction(&re_ctx, 1); /* 1 = expect eof */
append_u32(&re_ctx, DUK_REOP_SAVE);
append_u32(&re_ctx, 1);
append_u32(&re_ctx, DUK_REOP_MATCH);
DUK_DPRINT("regexp bytecode size (before header) is %d bytes",
(int) DUK_HBUFFER_GET_SIZE(re_ctx.buf));
/*
* Check for invalid backreferences; note that it is NOT an error
* to back-reference a capture group which has not yet been introduced
* in the pattern (as in /\1(foo)/); in fact, the backreference will
* always match! It IS an error to back-reference a capture group
* which will never be introduced in the pattern. Thus, we can check
* for such references only after parsing is complete.
*/
if (re_ctx.highest_backref > re_ctx.captures) {
DUK_ERROR(thr, DUK_ERR_SYNTAX_ERROR, "invalid backreference(s)");
}
/*
* Emit compiled regexp header: flags, ncaptures
* (insertion order inverted on purpose)
*/
insert_u32(&re_ctx, 0, (re_ctx.captures + 1) * 2);
insert_u32(&re_ctx, 0, re_ctx.re_flags);
DUK_DPRINT("regexp bytecode size (after header) is %d bytes",
(int) DUK_HBUFFER_GET_SIZE(re_ctx.buf));
DUK_DDDPRINT("compiled regexp: %!xO", re_ctx.buf);
/* [ ... pattern flags escaped_source buffer ] */
duk_to_string(ctx, -1); /* coerce to string */
/* [ ... pattern flags escaped_source bytecode ] */
/*
* Finalize stack
*/
duk_remove(ctx, -4); /* -> [ ... flags escaped_source bytecode ] */
duk_remove(ctx, -3); /* -> [ ... escaped_source bytecode ] */
DUK_DPRINT("regexp compilation successful, bytecode: %!T, escaped source: %!T",
duk_get_tval(ctx, -1), duk_get_tval(ctx, -2));
}
