PosibErr<void> Dictionary::add_repl(ParmString mis, ParmString cor)
{
if (!invisible_soundslike) {
VARARRAY(char, sl, mis.size() + 1);
lang()->LangImpl::to_soundslike(sl, mis.str(), mis.size());
return add_repl(mis, cor, sl);
} else {
str
utf8_fix (const str &in, const str &repl)
{
if (!in) return in;
size_t repl_len = repl ? repl.len () : size_t (0);
mstr out (in.len () * max<size_t> (1, repl_len) + 1);
const char *ip = in.cstr ();
const char *endp = ip + in.len ();
char *op = out.cstr ();
const char *cps = NULL; // code point start
ssize_t expected_width = 0, tmp = 0;
for ( ; ip < endp; ip++) {
bool consumed = false;
// if 'cps' is true, that means we've previously started
// a code point
if (cps) {
// We're expecting a follow byte -- if there's none here, it's
// an error and we need to throw away our buffered code point.
if (!is_follow_byte (*ip)) {
cps = NULL;
expected_width = 0;
add_repl (op, repl);
// We do have a follow byte, and we've seen enough bytes,
// so we're all set.
} else if (ip - cps == expected_width - 1) {
while (cps <= ip) {
*op++ = *cps++;
}
cps = NULL;
expected_width = 0;
consumed = true;
} else {
// in the default case, we haven't seen enough data, so keep moving.
consumed = true;
}
}
// Need to handle the non-code-point case AFTER the code-point case,
// since in some cases (a busted code-point), we hit both this if
// and the one above.
if (!cps && !consumed) {
tmp = point_width (*ip);
// start a new code point
if (tmp > 1) {
cps = ip;
expected_width = tmp;
} else if (tmp == 1) {
*op++ = *ip;
} else {
// we hit a bad starting byte, or a follow byte, neither of which
// should be useful here.
add_repl (op, repl);
}
}
}
out.setlen (op - out.cstr ());
return out;
}
