bool Lexer::matchStr(T const * string, size_t len) {
for (size_t i = 0; i < len; ++i) {
if (!matchc(string[i])) {
return false;
}
}
return true;
}
void bx_boollit_tail(triple *t, boolean_t jmp_type_one, boolean_t jmp_to_next, boolean_t sense, oprtype *addr)
/* search the Boolean in t (recursively) for literal leaves; the logic is similar to bx_tail
* the rest of the arguments parallel those in bx_boolop and used primarily handling basic Boolean operations (ON, NOR, AND, NAND)
* to get the jump target and sense right for the left-hand operand of the operation
* jmp_type_one gives the sense of the jump associated with the first operand
* jmp_to_next gives whether we need a second jump to complete the operation
* sense gives the sense of the requested operation
* *addr points the operand for the jump and is eventually used by logic back in the invocation stack to fill in a target location
*/
{
boolean_t sin[ARRAYSIZE(t->operand)], tv[ARRAYSIZE(t->operand)];
int com, comval, dummy, j, neg, num, tvr;
mval *mv, *v[ARRAYSIZE(t->operand)];
opctype c;
oprtype *i, *p;
triple *cob[ARRAYSIZE(t->operand)], *ref0, *tl[ARRAYSIZE(t->operand)];
assert(OCT_BOOL & oc_tab[t->opcode].octype);
assert(TRIP_REF == t->operand[0].oprclass);
assert((OC_COBOOL != t->opcode) && (OC_COM != t->opcode) || (TRIP_REF == t->operand[1].oprclass));
for (i = t->operand, j = 0; i < ARRAYTOP(t->operand); i++, j++)
{ /* checkout an operand to see if we can simplify it */
p = i;
com = 0;
for (tl[j] = i->oprval.tref; OCT_UNARY & oc_tab[(c = tl[j]->opcode)].octype; tl[j] = p->oprval.tref)
{ /* find the real object of affection; WARNING assignment above */
assert((TRIP_REF == tl[j]->operand[0].oprclass) && (NO_REF == tl[j]->operand[1].oprclass));
com ^= (OC_COM == c); /* if we make a recursive call below, COM matters, but NEG and FORCENUM don't */
p = &tl[j]->operand[0];
}
if (OCT_ARITH & oc_tab[c].octype)
ex_tail(p); /* chained arithmetic */
else if (OCT_BOOL & oc_tab[c].octype)
{ /* recursively check an operand */
sin[j] = sense;
p = addr;
if (!j && !(OCT_REL & oc_tab[t->opcode].octype))
{ /* left hand operand of parent */
sin[j] = jmp_type_one;
if (jmp_to_next)
{ /* left operands need extra attention to decide between jump next or to the end */
p = (oprtype *)mcalloc(SIZEOF(oprtype));
*p = put_tjmp(t);
}
}
bx_boollit(tl[j], sin[j] ^ com, p);
}
if ((OC_JMPTRUE != tl[j]->opcode) && (OC_JMPFALSE != tl[j]->opcode) && (OC_LIT != tl[j]->opcode))
return; /* this operation doesn't qualify */
com = comval = neg = num = 0;
cob[j] = NULL;
for (ref0 = i->oprval.tref; OCT_UNARY & oc_tab[(c = ref0->opcode)].octype; ref0 = ref0->operand[0].oprval.tref)
{ /* we may be able to clean up this operand; WARNING assignment above */
assert((TRIP_REF == ref0->operand[0].oprclass) && (NO_REF == ref0->operand[1].oprclass));
num += (OC_FORCENUM == c);
com += (OC_COM == c);
if (!com) /* "outside" com renders neg mute */
neg ^= (OC_NEG == c);
if (!comval && (NULL == cob[j]))
{
if (comval = (OC_COMVAL == c)) /* WARNING assignment */
{
if (ref0 != t->operand[j].oprval.tref)
dqdel(t->operand[j].oprval.tref, exorder);
t->operand[j].oprval.tref = tl[j]; /* need mval: no COBOOL needed */
}
else if (OC_COBOOL == c)
{ /* the operand needs a COBOOL in case its operator remains unresolved */
cob[j] = t->operand[j].oprval.tref;
if (ref0 == cob[j])
continue; /* already where it belongs */
cob[j]->opcode = OC_COBOOL;
cob[j]->operand[0].oprval.tref = tl[j];
} else if (ref0 == t->operand[j].oprval.tref)
continue;
}
dqdel(ref0, exorder);
}
assert(ref0 == tl[j]);
if (!comval && (NULL == cob[j]) && (tl[j] != t->operand[j].oprval.tref))
{ /* left room for a COBOOL, but there's no need */
dqdel(t->operand[j].oprval.tref, exorder);
t->operand[j].oprval.tref = tl[j];
}
if ((OC_JMPTRUE == ref0->opcode) || (OC_JMPFALSE == ref0->opcode))
{ /* switch to a literal representation of TRUE / FALSE */
assert(INDR_REF == ref0->operand[0].oprclass);
ref0->operand[1] = ref0->operand[0]; /* track info as we switch opcode */
PUT_LITERAL_TRUTH((sin[j] ? OC_JMPFALSE : OC_JMPTRUE) == ref0->opcode, ref0);
ref0->opcode = OC_LIT;
com = 0; /* already accounted for by sin */
}
assert((OC_LIT == ref0->opcode) && (MLIT_REF == ref0->operand[0].oprclass));
v[j] = &ref0->operand[0].oprval.mlit->v;
if (com)
{ /* any complement reduces the literal value to [unsigned] 1 or 0 */
unuse_literal(v[j]);
tv[j] = (0 == v[j]->m[1]);
assert(ref0 == tl[j]);
PUT_LITERAL_TRUTH(tv[j], ref0);
v[j] = &ref0->operand[0].oprval.mlit->v;
num = 0; /* any complement trumps num */
}
if (neg || num)
{ /* get literal into uniform state */
unuse_literal(v[j]);
mv = (mval *)mcalloc(SIZEOF(mval));
*mv = *v[j];
if (neg)
{
if (MV_INT & mv->mvtype)
{
if (0 != mv->m[1])
mv->m[1] = -mv->m[1];
else
mv->sgn = 0;
} else if (MV_NM & mv->mvtype)
mv->sgn = !mv->sgn;
} else
s2n(mv);
n2s(mv);
v[j] = mv;
assert(ref0 == tl[j]);
put_lit_s(v[j], ref0);
}
}
assert(tl[0] != tl[1]); /* start processing a live one */
for (tvr = j, j = 0; j < tvr; j++)
{ /* both arguments are literals, so do the operation at compile time */
if (NULL != cob[j])
dqdel(cob[j], exorder);
v[j] = &tl[j]->operand[0].oprval.mlit->v;
tv[j] = (0 != v[j]->m[1]);
unuse_literal(v[j]);
tl[j]->opcode = OC_NOOP;
tl[j]->operand[0].oprclass = NO_REF;
}
t->operand[1].oprclass = NO_REF;
switch (c = t->opcode) /* WARNING assignment */
{ /* optimize the Boolean operations here */
case OC_NAND:
case OC_AND:
tvr = (tv[0] && tv[1]);
break;
case OC_NOR:
case OC_OR:
tvr = (tv[0] || tv[1]);
break;
case OC_NCONTAIN:
case OC_CONTAIN:
tvr = 1;
(void)matchc(v[1]->str.len, (unsigned char *)v[1]->str.addr, v[0]->str.len,
(unsigned char *)v[0]->str.addr, &dummy, &tvr);
tvr ^= 1;
break;
case OC_NEQU:
case OC_EQU:
tvr = is_equ(v[0], v[1]);
break;
case OC_NFOLLOW:
case OC_FOLLOW:
tvr = 0 < memvcmp(v[0]->str.addr, v[0]->str.len, v[1]->str.addr, v[1]->str.len);
break;
case OC_NGT:
case OC_GT:
tvr = 0 < numcmp(v[0], v[1]);
break;
case OC_NLT:
case OC_LT:
tvr = 0 > numcmp(v[0], v[1]);
break;
case OC_NPATTERN:
case OC_PATTERN:
tvr = !(*(uint4 *)v[1]->str.addr) ? do_pattern(v[0], v[1]) : do_patfixed(v[0], v[1]);
break;
case OC_NSORTS_AFTER:
case OC_SORTS_AFTER:
tvr = 0 < sorts_after(v[0], v[1]);
break;
default:
assertpro(FALSE);
}
tvr ^= !sense;
t->operand[0] = put_indr(addr);
t->opcode = tvr ? OC_JMPFALSE : OC_JMPTRUE;
return;
}
/*
* ----------------------------------------------------------
* Set piece procedure (unicode flavor).
* Set pieces first through last to expr.
*
* Arguments:
* src - source mval
* del - delimiter string mval
* expr - expression string mval
* first - starting index in source mval to be set
* last - last index
* dst - destination mval where the result is saved.
*
* Return:
* none
* ----------------------------------------------------------
*/
void op_setpiece(mval *src, mval *del, mval *expr, int4 first, int4 last, mval *dst)
{
size_t str_len, delim_cnt;
int match_res, len, src_len, first_src_ind, second_src_ind, numpcs;
unsigned char *match_ptr, *src_str, *str_addr, *tmp_str;
delimfmt unichar;
/* --- code start --- */
assert(gtm_utf8_mode);
if (--first < 0)
first = 0;
second_src_ind = last - first;
MV_FORCE_STR(del);
/* Null delimiter */
if (0 == del->str.len)
{
if (first && src->mvtype)
{ /* concat src & expr to dst */
op_cat(VARLSTCNT(3) dst, src, expr);
return;
}
MV_FORCE_STR(expr);
*dst = *expr;
return;
}
MV_FORCE_STR(expr);
if (!MV_DEFINED(src))
{
first_src_ind = 0;
second_src_ind = -1;
} else
{
/* Valid delimiter - See if we can take a short cut to op_fnp1. If so, delimiter value needs to be reformated */
if ((1 == second_src_ind) && (1 == MV_FORCE_LEN(del)))
{ /* Both valid chars of char_len=1 and single byte invalid chars get the fast path */
unichar.unichar_val = 0;
assert(SIZEOF(unichar.unibytes_val) >= del->str.len);
memcpy(unichar.unibytes_val, del->str.addr, del->str.len);
op_setp1(src, unichar.unichar_val, expr, last, dst); /* Use last since it has not been changed */
return;
}
/* We have a valid src with something in it */
MV_FORCE_STR(src);
src_str = (unsigned char *)src->str.addr;
src_len = src->str.len;
/* skip all pieces until start one */
if (first)
{
numpcs = first; /* copy int4 type "first" into "int" type numpcs for passing to matchc */
match_ptr = matchc(del->str.len, (uchar_ptr_t)del->str.addr, src_len, src_str, &match_res, &numpcs);
/* Note: "numpcs" is modified above by the function "matchc" to reflect the # of unmatched pieces */
first = numpcs; /* copy updated "numpcs" value back into "first" */
} else
{
match_ptr = src_str;
match_res = 1;
}
first_src_ind = INTCAST(match_ptr - (unsigned char *)src->str.addr);
if (0 == match_res) /* if match not found */
second_src_ind = -1;
else
{
src_len -= INTCAST(match_ptr - src_str);
src_str = match_ptr;
/* skip # delimiters this piece will replace, e.g. if we are setting
* pieces 2 - 4, then the pieces 2-4 will be replaced by one piece - expr.
*/
match_ptr = matchc(del->str.len, (uchar_ptr_t)del->str.addr, src_len, src_str, &match_res, &second_src_ind);
second_src_ind = (0 == match_res) ? -1 : INTCAST(match_ptr - (unsigned char *)src->str.addr - del->str.len);
}
}
delim_cnt = (size_t)first;
/* Calculate total string len. */
str_len = (size_t)expr->str.len + ((size_t)first_src_ind + ((size_t)del->str.len * delim_cnt));
/* add len. of trailing chars past insertion point */
if (0 <= second_src_ind)
str_len += (size_t)(src->str.len - second_src_ind);
if (MAX_STRLEN < str_len)
{
rts_error_csa(CSA_ARG(NULL) VARLSTCNT(1) ERR_MAXSTRLEN);
return;
}
ENSURE_STP_FREE_SPACE((int)str_len);
str_addr = stringpool.free;
/* copy prefix */
if (first_src_ind)
{
memcpy(str_addr, src->str.addr, first_src_ind);
str_addr += first_src_ind;
}
/* copy delimiters */
if (gtm_utf8_mode && (1 < del->str.len))
{ /* In this mode, delimiters can exceed 1 character so copy them this way */
while (0 < delim_cnt--)
{
memcpy(str_addr, del->str.addr, del->str.len);
str_addr += del->str.len;
}
} else
{ /* If delimiters are 1 byte (M mode always and perhaps UTF8 mode), use this simpler/faster method */
memset(str_addr, (char)*del->str.addr, delim_cnt);
str_addr += delim_cnt;
}
/* copy expression */
memcpy(str_addr, expr->str.addr, expr->str.len);
str_addr += expr->str.len;
/* copy trailing pieces */
if (0 <= second_src_ind)
{
len = src->str.len - second_src_ind;
tmp_str = (unsigned char *)src->str.addr + second_src_ind;
memcpy(str_addr, tmp_str, len);
str_addr += len;
}
assert(IS_AT_END_OF_STRINGPOOL(str_addr, -str_len));
dst->mvtype = MV_STR;
dst->str.len = INTCAST(str_addr - stringpool.free);
dst->str.addr = (char *)stringpool.free;
stringpool.free = str_addr;
return;
}
