bool String::isIdentifier (String str)
{
if (_len (str.buf) == 0)
return false;
if (!isIdStart (str.buf[0]))
return false;
for (int i = 1; i < _len (str.buf); i++)
if (!isIdChar (str.buf[i]))
return false;
return true;
}
void Macro::expand(OutBuffer *buf, size_t start, size_t *pend,
utf8_t *arg, size_t arglen)
{
#if 0
printf("Macro::expand(buf[%d..%d], arg = '%.*s')\n", start, *pend, arglen, arg);
printf("Buf is: '%.*s'\n", *pend - start, buf->data + start);
#endif
static int nest;
if (nest > 100) // limit recursive expansion
return;
nest++;
size_t end = *pend;
assert(start <= end);
assert(end <= buf->offset);
/* First pass - replace $0
*/
arg = memdup(arg, arglen);
for (size_t u = start; u + 1 < end; )
{
utf8_t *p = buf->data; // buf->data is not loop invariant
/* Look for $0, but not $$0, and replace it with arg.
*/
if (p[u] == '$' && (isdigit(p[u + 1]) || p[u + 1] == '+'))
{
if (u > start && p[u - 1] == '$')
{ // Don't expand $$0, but replace it with $0
buf->remove(u - 1, 1);
end--;
u += 1; // now u is one past the closing '1'
continue;
}
utf8_t c = p[u + 1];
int n = (c == '+') ? -1 : c - '0';
utf8_t *marg;
size_t marglen;
extractArgN(arg, arglen, &marg, &marglen, n);
if (marglen == 0)
{ // Just remove macro invocation
//printf("Replacing '$%c' with '%.*s'\n", p[u + 1], marglen, marg);
buf->remove(u, 2);
end -= 2;
}
else if (c == '+')
{
// Replace '$+' with 'arg'
//printf("Replacing '$%c' with '%.*s'\n", p[u + 1], marglen, marg);
buf->remove(u, 2);
buf->insert(u, marg, marglen);
end += marglen - 2;
// Scan replaced text for further expansion
size_t mend = u + marglen;
expand(buf, u, &mend, NULL, 0);
end += mend - (u + marglen);
u = mend;
}
else
{
// Replace '$1' with '\xFF{arg\xFF}'
//printf("Replacing '$%c' with '\xFF{%.*s\xFF}'\n", p[u + 1], marglen, marg);
buf->data[u] = 0xFF;
buf->data[u + 1] = '{';
buf->insert(u + 2, marg, marglen);
buf->insert(u + 2 + marglen, "\xFF}", 2);
end += -2 + 2 + marglen + 2;
// Scan replaced text for further expansion
size_t mend = u + 2 + marglen;
expand(buf, u + 2, &mend, NULL, 0);
end += mend - (u + 2 + marglen);
u = mend;
}
//printf("u = %d, end = %d\n", u, end);
//printf("#%.*s#\n", end, &buf->data[0]);
continue;
}
u++;
}
/* Second pass - replace other macros
*/
for (size_t u = start; u + 4 < end; )
{
utf8_t *p = buf->data; // buf->data is not loop invariant
/* A valid start of macro expansion is $(c, where c is
* an id start character, and not $$(c.
*/
if (p[u] == '$' && p[u + 1] == '(' && isIdStart(p+u+2))
{
//printf("\tfound macro start '%c'\n", p[u + 2]);
utf8_t *name = p + u + 2;
size_t namelen = 0;
utf8_t *marg;
size_t marglen;
size_t v;
/* Scan forward to find end of macro name and
* beginning of macro argument (marg).
*/
for (v = u + 2; v < end; v+=utfStride(p+v))
{ utf8_t c = p[v];
if (!isIdTail(p+v))
{ // We've gone past the end of the macro name.
namelen = v - (u + 2);
break;
}
}
v += extractArgN(p + v, end - v, &marg, &marglen, 0);
assert(v <= end);
if (v < end)
{ // v is on the closing ')'
if (u > start && p[u - 1] == '$')
{ // Don't expand $$(NAME), but replace it with $(NAME)
buf->remove(u - 1, 1);
end--;
u = v; // now u is one past the closing ')'
continue;
}
Macro *m = search(name, namelen);
if (m)
{
#if 0
if (m->textlen && m->text[0] == ' ')
{ m->text++;
m->textlen--;
}
#endif
if (m->inuse && marglen == 0)
{ // Remove macro invocation
buf->remove(u, v + 1 - u);
end -= v + 1 - u;
}
else if (m->inuse && arglen == marglen && memcmp(arg, marg, arglen) == 0)
{ // Recursive expansion; just leave in place
}
else
{
//printf("\tmacro '%.*s'(%.*s) = '%.*s'\n", m->namelen, m->name, marglen, marg, m->textlen, m->text);
#if 1
marg = memdup(marg, marglen);
// Insert replacement text
buf->spread(v + 1, 2 + m->textlen + 2);
buf->data[v + 1] = 0xFF;
buf->data[v + 2] = '{';
memcpy(buf->data + v + 3, m->text, m->textlen);
buf->data[v + 3 + m->textlen] = 0xFF;
buf->data[v + 3 + m->textlen + 1] = '}';
end += 2 + m->textlen + 2;
// Scan replaced text for further expansion
m->inuse++;
size_t mend = v + 1 + 2+m->textlen+2;
expand(buf, v + 1, &mend, marg, marglen);
end += mend - (v + 1 + 2+m->textlen+2);
m->inuse--;
buf->remove(u, v + 1 - u);
end -= v + 1 - u;
u += mend - (v + 1);
#else
// Insert replacement text
buf->insert(v + 1, m->text, m->textlen);
end += m->textlen;
// Scan replaced text for further expansion
m->inuse++;
size_t mend = v + 1 + m->textlen;
expand(buf, v + 1, &mend, marg, marglen);
end += mend - (v + 1 + m->textlen);
m->inuse--;
buf->remove(u, v + 1 - u);
end -= v + 1 - u;
u += mend - (v + 1);
#endif
mem.free(marg);
//printf("u = %d, end = %d\n", u, end);
//printf("#%.*s#\n", end - u, &buf->data[u]);
continue;
}
}
else
{
// Replace $(NAME) with nothing
buf->remove(u, v + 1 - u);
end -= (v + 1 - u);
continue;
}
}
}
u++;
}
mem.free(arg);
*pend = end;
nest--;
}
void Macro::expand(OutBuffer *buf, size_t start, size_t *pend,
const utf8_t *arg, size_t arglen)
{
#if 0
printf("Macro::expand(buf[%d..%d], arg = '%.*s')\n", start, *pend, arglen, arg);
printf("Buf is: '%.*s'\n", *pend - start, buf->data + start);
#endif
static int nest;
if (nest > 100) // limit recursive expansion
return;
nest++;
size_t end = *pend;
assert(start <= end);
assert(end <= buf->offset);
/* First pass - replace $0
*/
arg = memdup(arg, arglen);
for (size_t u = start; u + 1 < end; )
{
utf8_t *p = (utf8_t *)buf->data; // buf->data is not loop invariant
/* Look for $0, but not $$0, and replace it with arg.
*/
if (p[u] == '$' && (isdigit(p[u + 1]) || p[u + 1] == '+'))
{
if (u > start && p[u - 1] == '$')
{ // Don't expand $$0, but replace it with $0
buf->remove(u - 1, 1);
end--;
u += 1; // now u is one past the closing '1'
continue;
}
utf8_t c = p[u + 1];
int n = (c == '+') ? -1 : c - '0';
const utf8_t *marg;
size_t marglen;
if (n == 0)
{
marg = arg;
marglen = arglen;
}
else
extractArgN(arg, arglen, &marg, &marglen, n);
if (marglen == 0)
{ // Just remove macro invocation
//printf("Replacing '$%c' with '%.*s'\n", p[u + 1], marglen, marg);
buf->remove(u, 2);
end -= 2;
}
else if (c == '+')
{
// Replace '$+' with 'arg'
//printf("Replacing '$%c' with '%.*s'\n", p[u + 1], marglen, marg);
buf->remove(u, 2);
buf->insert(u, marg, marglen);
end += marglen - 2;
// Scan replaced text for further expansion
size_t mend = u + marglen;
expand(buf, u, &mend, NULL, 0);
end += mend - (u + marglen);
u = mend;
}
else
{
// Replace '$1' with '\xFF{arg\xFF}'
//printf("Replacing '$%c' with '\xFF{%.*s\xFF}'\n", p[u + 1], marglen, marg);
buf->data[u] = 0xFF;
buf->data[u + 1] = '{';
buf->insert(u + 2, marg, marglen);
buf->insert(u + 2 + marglen, (const char *)"\xFF}", 2);
end += -2 + 2 + marglen + 2;
// Scan replaced text for further expansion
size_t mend = u + 2 + marglen;
expand(buf, u + 2, &mend, NULL, 0);
end += mend - (u + 2 + marglen);
u = mend;
}
//printf("u = %d, end = %d\n", u, end);
//printf("#%.*s#\n", end, &buf->data[0]);
continue;
}
u++;
}
/* Second pass - replace other macros
*/
for (size_t u = start; u + 4 < end; )
{
utf8_t *p = (utf8_t *)buf->data; // buf->data is not loop invariant
/* A valid start of macro expansion is $(c, where c is
* an id start character, and not $$(c.
*/
if (p[u] == '$' && p[u + 1] == '(' && isIdStart(p+u+2))
{
//printf("\tfound macro start '%c'\n", p[u + 2]);
utf8_t *name = p + u + 2;
size_t namelen = 0;
const utf8_t *marg;
size_t marglen;
size_t v;
/* Scan forward to find end of macro name and
* beginning of macro argument (marg).
*/
for (v = u + 2; v < end; v+=utfStride(p+v))
{
if (!isIdTail(p+v))
{ // We've gone past the end of the macro name.
namelen = v - (u + 2);
break;
}
}
v += extractArgN(p + v, end - v, &marg, &marglen, 0);
assert(v <= end);
if (v < end)
{ // v is on the closing ')'
if (u > start && p[u - 1] == '$')
{ // Don't expand $$(NAME), but replace it with $(NAME)
buf->remove(u - 1, 1);
end--;
u = v; // now u is one past the closing ')'
continue;
}
Macro *m = search(name, namelen);
if (!m)
{
static const char undef[] = "DDOC_UNDEFINED_MACRO";
m = search((const utf8_t *)undef, strlen(undef));
if (m)
{
// Macro was not defined, so this is an expansion of
// DDOC_UNDEFINED_MACRO. Prepend macro name to args.
// marg = name[ ] ~ "," ~ marg[ ];
if (marglen)
{
utf8_t *q = (utf8_t *)mem.xmalloc(namelen + 1 + marglen);
assert(q);
memcpy(q, name, namelen);
q[namelen] = ',';
memcpy(q + namelen + 1, marg, marglen);
marg = q;
marglen += namelen + 1;
}
else
{
marg = name;
marglen = namelen;
}
}
}
if (m)
{
if (m->inuse && marglen == 0)
{ // Remove macro invocation
buf->remove(u, v + 1 - u);
end -= v + 1 - u;
}
else if (m->inuse &&
((arglen == marglen && memcmp(arg, marg, arglen) == 0) ||
(arglen + 4 == marglen &&
marg[0] == 0xFF &&
marg[1] == '{' &&
memcmp(arg, marg + 2, arglen) == 0 &&
marg[marglen - 2] == 0xFF &&
marg[marglen - 1] == '}'
)
)
)
{
/* Recursive expansion:
* marg is same as arg (with blue paint added)
* Just leave in place.
*/
}
else
{
//printf("\tmacro '%.*s'(%.*s) = '%.*s'\n", m->namelen, m->name, marglen, marg, m->textlen, m->text);
marg = memdup(marg, marglen);
// Insert replacement text
buf->spread(v + 1, 2 + m->textlen + 2);
buf->data[v + 1] = 0xFF;
buf->data[v + 2] = '{';
memcpy(buf->data + v + 3, m->text, m->textlen);
buf->data[v + 3 + m->textlen] = 0xFF;
buf->data[v + 3 + m->textlen + 1] = '}';
end += 2 + m->textlen + 2;
// Scan replaced text for further expansion
m->inuse++;
size_t mend = v + 1 + 2+m->textlen+2;
expand(buf, v + 1, &mend, marg, marglen);
end += mend - (v + 1 + 2+m->textlen+2);
m->inuse--;
buf->remove(u, v + 1 - u);
end -= v + 1 - u;
u += mend - (v + 1);
mem.xfree((utf8_t *)marg);
//printf("u = %d, end = %d\n", u, end);
//printf("#%.*s#\n", end - u, &buf->data[u]);
continue;
}
}
else
{
// Replace $(NAME) with nothing
buf->remove(u, v + 1 - u);
end -= (v + 1 - u);
continue;
}
}
}
u++;
}
mem.xfree((utf8_t *)arg);
*pend = end;
nest--;
}