StringData::StringData(const char *data, int len, StringDataMode mode)
: m_data(NULL), _count(0), m_len(0) {
m_hash = 0;
assign(data, len, mode);
TAINT_OBSERVER_REGISTER_MUTATED(this);
}
void StringBuffer::absorb(StringBuffer &buf) {
if (empty()) {
TAINT_OBSERVER_REGISTER_ACCESSED(buf.getTaintDataRefConst());
TAINT_OBSERVER_REGISTER_MUTATED(m_taint_data, dataIgnoreTaint());
StringData* str = m_str;
m_str = buf.m_str;
m_buffer = buf.m_buffer;
m_len = buf.m_len;
m_cap = buf.m_cap;
buf.m_str = str;
if (str) {
buf.m_buffer = (char*)str->data();
buf.m_len = str->size();
buf.m_cap = str->capacity();
} else {
buf.m_buffer = 0;
buf.m_len = 0;
buf.m_cap = 0;
}
buf.reset();
} else {
// REGISTER_ACCESSED()/REGISTER_MUTATED() are called by append()/detach()
append(buf.detach());
}
}
CstrBuffer::CstrBuffer(const char *filename)
: m_buffer(NULL), m_len(0) {
struct stat sb;
if (stat(filename, &sb) == 0) {
if (sb.st_size > kMaxCap - 1) {
std::ostringstream out;
out << "file " << filename << " is too large";
throw StringBufferLimitException(kMaxCap,
String(out.str().c_str()));
}
m_cap = sb.st_size;
m_buffer = (char *)Util::safe_malloc(m_cap + 1);
int fd = ::open(filename, O_RDONLY);
if (fd != -1) {
while (m_len < m_cap) {
int buffer_size = m_cap - m_len;
int len = ::read(fd, m_buffer + m_len, buffer_size);
if (len == -1 && errno == EINTR) continue;
if (len <= 0) break;
m_len += len;
}
::close(fd);
}
}
TAINT_OBSERVER_REGISTER_MUTATED(m_taint_data, dataIgnoreTaint());
}
StringBuffer::StringBuffer(int initialSize /* = 63 */)
: m_initialCap(initialSize), m_maxBytes(kDefaultOutputLimit),
m_cap(initialSize), m_len(0) {
ASSERT(initialSize > 0);
m_str = NEW(StringData)(initialSize);
m_buffer = (char *)m_str->data();
TAINT_OBSERVER_REGISTER_MUTATED(m_taint_data, dataIgnoreTaint());
}
StringData::StringData(const char *data,
StringDataMode mode /* = AttachLiteral */)
: m_data(NULL), _count(0), m_len(0) {
m_hash = 0;
assign(data, mode);
TAINT_OBSERVER_REGISTER_MUTATED(this);
}
StringBuffer::StringBuffer(int initialSize /* = 63 */)
: m_initialCap(initialSize), m_maxBytes(kDefaultOutputLimit),
m_len(0) {
assert(initialSize > 0);
m_str = NEW(StringData)(initialSize);
MutableSlice s = m_str->mutableSlice();
m_buffer = s.ptr;
m_cap = s.len;
TAINT_OBSERVER_REGISTER_MUTATED(m_taint_data, dataIgnoreTaint());
}
HOT_FUNC
StringData::StringData(SharedVariant *shared)
: _count(0) {
ASSERT(shared);
shared->incRef();
m_hash = 0;
m_len = shared->stringLength();
m_cdata = shared->stringData();
m_big.shared = shared;
m_big.cap = m_len | IsShared;
TAINT_OBSERVER_REGISTER_MUTATED(m_taint_data, rawdata());
}
StringData::StringData(SharedVariant *shared)
: m_data(NULL), _count(0), m_len(0) {
m_hash = 0;
ASSERT(shared);
shared->incRef();
m_shared = shared;
m_data = m_shared->stringData();
m_len = m_shared->stringLength() | IsShared;
ASSERT(m_data);
TAINT_OBSERVER_REGISTER_MUTATED(this);
}
void StringData::initLiteral(const char* data, int len) {
if (uint32_t(len) > MaxSize) {
throw InvalidArgumentException("len>=2^30", len);
}
// Do not copy literals, this StringData can have a shorter lifetime than
// the literal, and the client can count on this->data() giving back
// the literal ptr with the longer lifetime. Sketchy!
m_hash = 0;
_count = 0;
m_len = len;
m_cdata = data;
m_big.cap = len | IsLiteral;
ASSERT(checkSane());
TAINT_OBSERVER_REGISTER_MUTATED(m_taint_data, rawdata());
}
void StringData::initAttachDeprecated(const char* data, int len) {
if (uint32_t(len) > MaxSize) {
throw InvalidArgumentException("len>=2^30", len);
}
// Don't copy small strings here either because the caller sometimes
// assumes he can mess with data while this string is still alive,
// and we want to free it eagerly. Sketchy!
m_hash = 0;
_count = 0;
m_len = len;
m_cdata = data;
m_big.cap = len | IsMalloc;
ASSERT(checkSane());
TAINT_OBSERVER_REGISTER_MUTATED(m_taint_data, rawdata());
}
void StringData::append(const char *s, int len) {
if (len == 0) return;
if (len < 0 || (len & IsMask)) {
throw InvalidArgumentException("len: %d", len);
}
ASSERT(!isStatic()); // never mess around with static strings!
if (!isMalloced()) {
int newlen;
m_data = string_concat(data(), size(), s, len, newlen);
if (isShared()) {
m_shared->decRef();
}
m_len = newlen;
m_hash = 0;
} else if (m_data == s) {
int newlen;
char *newdata = string_concat(data(), size(), s, len, newlen);
releaseData();
m_data = newdata;
m_len = newlen;
} else {
int dataLen = size();
ASSERT((m_data > s && m_data - s > len) ||
(m_data < s && s - m_data > dataLen)); // no overlapping
m_len = len + dataLen;
m_data = (const char*)realloc((void*)m_data, m_len + 1);
memcpy((void*)(m_data + dataLen), s, len);
((char*)m_data)[m_len] = '\0';
m_hash = 0;
}
if (m_len & IsMask) {
int len = m_len;
m_len &= ~IsMask;
releaseData();
m_data = NULL;
throw FatalErrorException(0, "String length exceeded 2^29 - 1: %d", len);
}
TAINT_OBSERVER_REGISTER_MUTATED(this);
}
void StringBuffer::absorb(StringBuffer &buf) {
if (empty()) {
TAINT_OBSERVER_REGISTER_ACCESSED(buf.getTaintDataRefConst());
TAINT_OBSERVER_REGISTER_MUTATED(m_taint_data);
char *buffer = m_buffer;
int size = m_size;
m_buffer = buf.m_buffer;
m_size = buf.m_size;
m_pos = buf.m_pos;
buf.m_buffer = buffer;
buf.m_size = size;
buf.reset();
} else {
append(buf.detach());
}
}
StringBuffer::StringBuffer(const char *filename)
: m_buffer(NULL), m_initialSize(1024), m_maxBytes(0), m_size(0), m_pos(0) {
struct stat sb;
if (stat(filename, &sb) == 0) {
m_size = sb.st_size;
m_buffer = (char *)Util::safe_malloc(m_size + 1);
int fd = ::open(filename, O_RDONLY);
if (fd != -1) {
while (m_pos < m_size) {
int buffer_size = m_size - m_pos;
int len = ::read(fd, m_buffer + m_pos, buffer_size);
if (len == -1 && errno == EINTR) continue;
if (len <= 0) break;
m_pos += len;
}
::close(fd);
}
}
TAINT_OBSERVER_REGISTER_MUTATED(m_taint_data, dataIgnoreTaint());
}
HOT_FUNC
void StringData::initAttach(const char* data, int len) {
if (uint32_t(len) > MaxSize) {
throw InvalidArgumentException("len>=2^30", len);
}
m_hash = 0;
_count = 0;
if (uint32_t(len) <= MaxSmallSize) {
memcpy(m_small, data, len);
m_len = len;
m_data = m_small;
m_small[len] = 0;
m_small[MaxSmallSize] = 0;
free((void*)data);
} else {
m_len = len;
m_cdata = data;
m_big.cap = len | IsMalloc;
}
ASSERT(checkSane());
TAINT_OBSERVER_REGISTER_MUTATED(m_taint_data, rawdata());
}
CstrBuffer::CstrBuffer(int cap)
: m_buffer((char*)Util::safe_malloc(cap + 1)), m_len(0), m_cap(cap) {
ASSERT(unsigned(cap) <= kMaxCap);
TAINT_OBSERVER_REGISTER_MUTATED(m_taint_data, dataIgnoreTaint());
}
StringBuffer::StringBuffer(int initialSize /* = 1024 */)
: m_initialSize(initialSize), m_maxBytes(0), m_size(initialSize), m_pos(0) {
ASSERT(initialSize > 0);
m_buffer = (char *)Util::safe_malloc(initialSize + 1);
TAINT_OBSERVER_REGISTER_MUTATED(m_taint_data, dataIgnoreTaint());
}
StringBuffer::StringBuffer(char *data, int len)
: m_buffer(data), m_initialSize(1024), m_maxBytes(0), m_size(len),
m_pos(len) {
TAINT_OBSERVER_REGISTER_MUTATED(m_taint_data, dataIgnoreTaint());
}
void StringData::append(const char *s, int len) {
ASSERT(!isStatic()); // never mess around with static strings!
if (len == 0) return;
if (UNLIKELY(uint32_t(len) > MaxSize)) {
throw InvalidArgumentException("len>=2^30", len);
}
if (UNLIKELY(len + m_len > MaxSize)) {
throw FatalErrorException(0, "String length exceeded 2^30 - 1: %u",
len + m_len);
}
int newlen;
// TODO: t1122987: in any of the cases below where we need a bigger buffer,
// we can probably assume we're in a concat-loop and pick a good buffer
// size to avoid O(N^2) copying cost.
if (isShared() || isLiteral()) {
// buffer is immutable, don't modify it.
// We are mutating, so we don't need to repropagate our own taint
StringSlice r = slice();
char* newdata = string_concat(r.ptr, r.len, s, len, newlen);
if (isShared()) m_big.shared->decRef();
m_len = newlen;
m_data = newdata;
m_big.cap = newlen | IsMalloc;
m_hash = 0;
} else if (rawdata() == s) {
// appending ourself to ourself, be conservative.
// We are mutating, so we don't need to repropagate our own taint
StringSlice r = slice();
char *newdata = string_concat(r.ptr, r.len, s, len, newlen);
releaseData();
m_len = newlen;
m_data = newdata;
m_big.cap = newlen | IsMalloc;
m_hash = 0;
} else if (isSmall()) {
// we're currently small but might not be after append.
// We are mutating, so we don't need to repropagate our own taint
int oldlen = m_len;
newlen = oldlen + len;
if (unsigned(newlen) <= MaxSmallSize) {
// win.
memcpy(&m_small[oldlen], s, len);
m_small[newlen] = 0;
m_small[MaxSmallSize] = 0;
m_len = newlen;
m_data = m_small;
m_hash = 0;
} else {
// small->big string transition.
char *newdata = string_concat(m_small, oldlen, s, len, newlen);
m_len = newlen;
m_data = newdata;
m_big.cap = newlen | IsMalloc;
m_hash = 0;
}
} else {
// generic "big string concat" path. realloc buffer.
int oldlen = m_len;
char* oldp = m_data;
ASSERT((oldp > s && oldp - s > len) ||
(oldp < s && s - oldp > oldlen)); // no overlapping
newlen = oldlen + len;
char* newdata = (char*) realloc(oldp, newlen + 1);
memcpy(newdata + oldlen, s, len);
newdata[newlen] = 0;
m_len = newlen;
m_data = newdata;
m_big.cap = newlen | IsMalloc;
m_hash = 0;
}
ASSERT(uint32_t(newlen) <= MaxSize);
TAINT_OBSERVER_REGISTER_MUTATED(m_taint_data, rawdata());
ASSERT(checkSane());
}
void StringBuffer::append(CStrRef s) {
append(s.data(), s.size());
TAINT_OBSERVER_REGISTER_MUTATED(m_taint_data);
}