StringData* StringData::append(StringSlice range) {
assert(!hasMultipleRefs());
auto s = range.ptr;
auto const len = range.len;
if (len == 0) return this;
auto const newLen = size_t(m_len) + size_t(len);
if (UNLIKELY(newLen > MaxSize)) {
throw_string_too_large(newLen);
}
/*
* We may have an aliasing append. We don't allow appending with an
* interior pointer, although we may be asked to append less than
* the whole string in an aliasing situation.
*/
ALIASING_APPEND_ASSERT(s, len);
auto const requestLen = static_cast<uint32_t>(newLen);
auto const target = UNLIKELY(isShared()) ? escalate(requestLen)
: reserve(requestLen);
memcpy(target->mutableData() + m_len, s, len);
target->setSize(newLen);
assert(target->checkSane());
return target;
}
StringData* StringData::shrinkImpl(size_t len) {
assert(!isImmutable() && !hasMultipleRefs());
assert(isFlat());
assert(len <= capacity());
auto const sd = Make(len);
sd->m_len = len;
auto const src = slice();
auto const dst = sd->mutableData();
*memcpy8(dst, src.ptr, len) = 0;
assert(sd->checkSane());
return sd;
}
// State transition from Mode::Shared to Mode::Flat.
StringData* StringData::escalate(size_t cap) {
assert(isShared() && !isStatic() && cap >= m_len);
auto const sd = Make(cap);
auto const src = slice();
auto const dst = sd->mutableData();
sd->setSize(src.len);
auto const mcret = memcpy(dst, src.ptr, src.len);
auto const ret = static_cast<StringData*>(mcret) - 1;
// Recalculating ret from mcret avoids a spill.
assert(ret == sd);
assert(ret->checkSane());
return ret;
}
QPpsAttributeMap QPpsObjectPrivate::decode(const QByteArray &rawData, bool *ok)
{
QPpsAttributeMap attributeMap;
pps_decoder_t decoder;
QByteArray mutableData(rawData);
pps_decoder_error_t error = pps_decoder_initialize(&decoder, mutableData.data());
if (error == PPS_DECODER_OK) {
// no need to check ok in this case
attributeMap = decodeObject(&decoder, ok);
} else {
qWarning() << "QPpsObjectPrivate::decode: pps_decoder_initialize failed";
*ok = false;
}
pps_decoder_cleanup(&decoder);
return attributeMap;
}
StringData* StringData::shrinkImpl(size_t len) {
assert(!isImmutable() && !hasMultipleRefs());
assert(isFlat());
assert(len <= m_len);
assert(len <= capacity());
auto const sd = Make(len);
auto const src = slice();
auto const dst = sd->mutableData();
assert(len <= src.len);
sd->setSize(len);
auto const mcret = memcpy(dst, src.ptr, len);
auto const ret = static_cast<StringData*>(mcret) - 1;
// Recalculating ret from mcret avoids a spill.
assert(ret == sd);
assert(ret->checkSane());
return ret;
}
StringData* StringData::reserve(size_t cap) {
assert(!isImmutable() && !hasMultipleRefs());
assert(isFlat());
if (cap <= capacity()) return this;
cap = std::min(cap + cap/4, size_t(MaxSize) + 1);
auto const sd = Make(cap);
auto const src = slice();
auto const dst = sd->mutableData();
sd->setSize(src.len);
auto const mcret = memcpy(dst, src.ptr, src.len);
auto const ret = static_cast<StringData*>(mcret) - 1;
// Recalculating ret from mcret avoids a spill.
assert(ret == sd);
assert(ret->checkSane());
return ret;
}
StringData* StringData::append(StringSlice r1,
StringSlice r2,
StringSlice r3) {
assert(!hasMultipleRefs());
auto const len = r1.len + r2.len + r3.len;
if (len == 0) return this;
if (UNLIKELY(size_t(m_len) + size_t(len) > MaxSize)) {
throw_string_too_large(size_t(len) + size_t(m_len));
}
auto const newLen = m_len + len;
/*
* We may have an aliasing append. We don't allow appending with an
* interior pointer, although we may be asked to append less than
* the whole string in an aliasing situation.
*/
ALIASING_APPEND_ASSERT(r1.ptr, r1.len);
ALIASING_APPEND_ASSERT(r2.ptr, r2.len);
ALIASING_APPEND_ASSERT(r3.ptr, r3.len);
auto const target = UNLIKELY(isShared()) ? escalate(newLen)
: reserve(newLen);
/*
* memcpy is safe even if it's a self append---the regions will be
* disjoint, since rN.ptr can't point past the start of our source
* pointer, and rN.len is smaller than the old length.
*/
void* p = target->mutableData();
p = memcpy((char*)p + m_len, r1.ptr, r1.len);
p = memcpy((char*)p + r1.len, r2.ptr, r2.len);
memcpy((char*)p + r2.len, r3.ptr, r3.len);
target->setSize(newLen);
assert(target->checkSane());
return target;
}
void StringData::incrementHelper() {
m_hash = 0;
enum class CharKind {
UNKNOWN,
LOWER_CASE,
UPPER_CASE,
NUMERIC
};
auto const len = m_len;
auto const s = mutableData();
int carry = 0;
int pos = len - 1;
auto last = CharKind::UNKNOWN; // Shut up the compiler warning
int ch;
while (pos >= 0) {
ch = s[pos];
if (ch >= 'a' && ch <= 'z') {
if (ch == 'z') {
s[pos] = 'a';
carry=1;
} else {
s[pos]++;
carry=0;
}
last = CharKind::LOWER_CASE;
} else if (ch >= 'A' && ch <= 'Z') {
if (ch == 'Z') {
s[pos] = 'A';
carry=1;
} else {
s[pos]++;
carry=0;
}
last = CharKind::UPPER_CASE;
} else if (ch >= '0' && ch <= '9') {
if (ch == '9') {
s[pos] = '0';
carry=1;
} else {
s[pos]++;
carry=0;
}
last = CharKind::NUMERIC;
} else {
carry=0;
break;
}
if (carry == 0) {
break;
}
pos--;
}
if (carry) {
if (UNLIKELY(len + 1 > MaxSize)) {
throw_string_too_large(len);
}
assert(len + 1 <= capacity());
memmove(s + 1, s, len);
s[len + 1] = '\0';
m_len = len + 1;
switch (last) {
case CharKind::NUMERIC:
s[0] = '1';
break;
case CharKind::UPPER_CASE:
s[0] = 'A';
break;
case CharKind::LOWER_CASE:
s[0] = 'a';
break;
default:
break;
}
}
}
Variant HHVM_FUNCTION(wordwrap, const String& str, int64_t linewidth /* = 75 */,
const String& brk /* = s_nl */, bool cut /* = false */) {
const char* brkstr = brk.data();
size_t textlen = str.size();
size_t brklen = brk.size();
if (textlen == 0) {
return empty_string();
}
if (brklen == 0) {
raise_warning("Break string cannot be empty");
return false;
}
if (linewidth == 0 && cut) {
raise_warning("Can't force cut when width is zero");
return false;
}
size_t w = linewidth >= 0 ? linewidth : 0;
// If the string's length is less than or equal to the specified
// width, there's nothing to do and we can just return the string.
if (textlen <= w) return str;
// Special case for a single-character break as it needs no
// additional storage space
if (brklen == 1 && !cut) {
auto new_sd = StringData::Make(str.get(), CopyString);
new_sd->invalidateHash();
auto ret = Variant::attach(new_sd);
char* newtext = new_sd->mutableData();
auto bc = brkstr[0];
size_t current = 0, laststart = 0, lastspace = 0;
for (; current < textlen; current++) {
if (newtext[current] == bc) {
laststart = lastspace = current + 1;
} else if (newtext[current] == ' ') {
if (current - laststart >= w) {
newtext[current] = bc;
laststart = current + 1;
}
lastspace = current;
} else if (current - laststart >= w && laststart != lastspace) {
newtext[lastspace] = bc;
laststart = lastspace + 1;
}
}
return ret;
}
// Multiple character line break or forced cut
// Estimate how big the output string will be. It's okay if this estimate
// is wrong as we will grow or shrink as needed. The goals here are two-
// fold: (1) avoid the need to grow or shrink in the common case, and
// (2) for extreme cases where it's hard to make an accurate estimate
// (ex. when w is very small or brk is very large) we should be careful
// to avoid making huge over-estimations.
StringBuffer strbuf(
textlen +
textlen / (std::max<size_t>(w, 16) - 8) * std::min<size_t>(brklen, 8));
const char* text = str.data();
size_t current = 0, laststart = 0, lastspace = 0;
for (; current < textlen; current++) {
// when we hit an existing break, copy to new buffer, and
// fix up laststart and lastspace
if (text[current] == brkstr[0] && current + brklen < textlen &&
!strncmp(text + current, brkstr, brklen)) {
strbuf.append(text + laststart, current - laststart + brklen);
current += brklen - 1;
laststart = lastspace = current + 1;
}
// if it is a space, check if it is at the line boundary,
// copy and insert a break, or just keep track of it
else if (text[current] == ' ') {
if (current - laststart >= w) {
strbuf.append(text + laststart, current - laststart);
strbuf.append(brkstr, brklen);
laststart = current + 1;
}
lastspace = current;
}
// if we are cutting, and we've accumulated enough
// characters, and we haven't see a space for this line,
// copy and insert a break.
else if (current - laststart >= w && cut && laststart >= lastspace) {
strbuf.append(text + laststart, current - laststart);
strbuf.append(brkstr, brklen);
laststart = lastspace = current;
}
// if the current word puts us over width w, copy back up
// until the last space, insert a break, and move up the
// laststart
else if (current - laststart >= w && laststart < lastspace) {
strbuf.append(text + laststart, lastspace - laststart);
strbuf.append(brkstr, brklen);
laststart = lastspace = lastspace + 1;
}
}
// copy over any stragglers
if (laststart != current) {
strbuf.append(text + laststart, current - laststart);
}
auto s = strbuf.detach();
// if it's not possible to reduce the output string's capacity by more
// than 25%, then we can just return the string as is.
size_t estShrinkCap =
MemoryManager::estimateCap(sizeof(StringData) + s.size() + 1);
if (estShrinkCap * 4 >= (size_t)s.capacity() * 3) {
return s;
}
// reallocate into a smaller buffer so that we don't waste memory
return Variant::attach(StringData::Make(s.get(), CopyString));
}