BytevectorCell* StringCell::toUtf8Bytevector(World &world, SliceIndexType start, SliceIndexType end)
{
CharRange range = charRange(start, end);
if (range.isNull())
{
return nullptr;
}
ByteLengthType newLength = range.endPointer - range.startPointer;
SharedByteArray *byteArray;
if ((newLength == byteLength()) && !dataIsInline())
{
// Reuse our existing byte array
byteArray = static_cast<HeapStringCell*>(this)->heapByteArray()->ref();
}
else
{
// Create a new byte array and initialize it
byteArray = SharedByteArray::createInstance(newLength);
memcpy(byteArray->data(), range.startPointer, newLength);
}
return BytevectorCell::withByteArray(world, byteArray, newLength);
}
StringCell* StringCell::createUninitialized(World &world, ByteLengthType byteLength, CharLengthType charLength)
{
void *cellPlacement = alloc::allocateCells(world);
if (byteLength <= inlineDataSize())
{
// We can fit this string inline
auto newString = new (cellPlacement) InlineStringCell(byteLength, charLength);
#ifndef NDEBUG
if (byteLength < inlineDataSize())
{
// Explicitly terminate with non-NULL to catch users that assume we're NULL terminated internally
newString->inlineData()[byteLength] = 0xff;
}
#endif
return newString;
}
else
{
// Allocate a new shared byte array
SharedByteArray *newByteArray = SharedByteArray::createInstance(byteLength);
#ifndef NDEBUG
if (newByteArray->capacity(byteLength) > byteLength)
{
newByteArray->data()[byteLength] = 0xff;
}
#endif
return new (cellPlacement) HeapStringCell(newByteArray, byteLength, charLength);
}
}
SymbolCell* SymbolCell::fromUtf8Data(World &world, const std::uint8_t *data, ByteLengthType byteLength)
{
const std::uint8_t *scanPtr = data;
const std::uint8_t *endPtr = data + byteLength;
// Validate the UTF-8 data
const std::size_t charLength = utf8::validateData(scanPtr, endPtr);
void *cellPlacement = alloc::allocateCells(world);
if (byteLength <= inlineDataSize())
{
auto inlineSymbol = new (cellPlacement) InlineSymbolCell(byteLength, charLength);
memcpy(inlineSymbol->inlineData(), data, byteLength);
return inlineSymbol;
}
else
{
SharedByteArray *newByteArray = SharedByteArray::createUninitialised(byteLength);
memcpy(newByteArray->data(), data, byteLength);
return new (cellPlacement) HeapSymbolCell(newByteArray, byteLength, charLength);
}
}
BytevectorCell* BytevectorCell::fromAppended(World &world, const std::list<const BytevectorCell*> &byteVectors)
{
if (byteVectors.size() == 1)
{
// This allows implicit data sharing while the below always allocates
return byteVectors.front()->copy(world);
}
std::uint64_t totalLength = 0;
for(auto byteVector : byteVectors)
{
totalLength += byteVector->length();
}
if (totalLength > maximumLength())
{
return nullptr;
}
SharedByteArray *newByteArray = SharedByteArray::createInstance(totalLength);
std::uint8_t *copyPtr = newByteArray->data();
for(auto byteVector : byteVectors)
{
memcpy(copyPtr, byteVector->byteArray()->data(), byteVector->length());
copyPtr += byteVector->length();
}
return BytevectorCell::withByteArray(world, newByteArray, totalLength);
}
BytevectorCell* BytevectorCell::fromData(World &world, const std::uint8_t *data, LengthType length)
{
SharedByteArray *newByteArray = SharedByteArray::createInstance(length);
memcpy(newByteArray->data(), data, length);
return BytevectorCell::withByteArray(world, newByteArray, length);
}
BytevectorCell* BytevectorCell::fromFill(World &world, LengthType length, std::uint8_t fill)
{
SharedByteArray *newByteArray = SharedByteArray::createInstance(length);
if (newByteArray == nullptr)
{
return nullptr;
}
memset(newByteArray->data(), fill, length);
return BytevectorCell::withByteArray(world, newByteArray, length);
}
BytevectorCell* BytevectorCell::copy(World &world, SliceIndexType start, SliceIndexType end) const
{
if (!adjustSlice(start, end, length()))
{
return nullptr;
}
if ((start == 0) && (end == length()))
{
// We can do a copy-on-write here
return BytevectorCell::withByteArray(world, byteArray()->ref(), length());
}
const LengthType newLength = end - start;
SharedByteArray *newByteArray = SharedByteArray::createInstance(newLength);
memcpy(newByteArray->data(), &byteArray()->data()[start], newLength);
return BytevectorCell::withByteArray(world, newByteArray, newLength);
}
bool StringCell::replaceBytes(const CharRange &range, const std::uint8_t *pattern, unsigned int patternBytes, unsigned int count)
{
assert(!isGlobalConstant());
const unsigned int requiredBytes = patternBytes * count;
const unsigned int replacedBytes = range.byteCount();
// If we have exclusive access to our data and we're not resizing the string we can use the fast path
if ((dataIsInline() || static_cast<HeapStringCell*>(this)->heapByteArray()->isExclusive()) &&
(requiredBytes == replacedBytes))
{
std::uint8_t *copyDest = const_cast<std::uint8_t*>(range.startPointer);
while(count--)
{
memmove(copyDest, pattern, patternBytes);
copyDest += patternBytes;
}
}
else
{
// Create a new string from pieces of the old string
const std::uint64_t newByteLength = byteLength() + requiredBytes - replacedBytes;
const auto newCharLength = charLength();
if (newByteLength > maximumByteLength())
{
return false;
}
const ByteLengthType initialBytes = range.startPointer - utf8Data();
const ByteLengthType finalBytes = newByteLength - initialBytes - requiredBytes;
const bool wasInline = dataIsInline();
const bool nowInline = newByteLength <= inlineDataSize();
SharedByteArray *oldByteArray = nullptr;
SharedByteArray *newByteArray = nullptr;
// Does this require a COW due to sharing our byte array?
const bool needsCow = (!wasInline && !nowInline) &&
!static_cast<HeapStringCell*>(this)->heapByteArray()->isExclusive();
// Determine if we exceeded our current capacity or if we're using less than half of our allocated space
// This will trigger a reallocation of our heap space
const auto currentCapacity = byteCapacity();
const bool needHeapRealloc = (newByteLength > currentCapacity) ||
((newByteLength < (currentCapacity / 2)) && !nowInline) ||
needsCow;
std::uint8_t* destString;
const std::uint8_t* copySource;
if (!wasInline && nowInline)
{
// We're converting to an inline string
destString = static_cast<InlineStringCell*>(this)->inlineData();
copySource = pattern;
// Store our old byte array so we can unref it later
// The code below will overwrite it with our new inline string
oldByteArray = static_cast<HeapStringCell*>(this)->heapByteArray();
// Fill the initial chunk of the string
memcpy(destString, utf8Data(), initialBytes);
}
else if (needHeapRealloc)
{
size_t byteArraySize = newByteLength;
newByteArray = SharedByteArray::createInstance(byteArraySize);
destString = newByteArray->data();
copySource = pattern;
// Fill the initial chunk of the string
memcpy(destString, utf8Data(), initialBytes);
if (!wasInline)
{
// Store our old byte array so we can unref it later
oldByteArray = static_cast<HeapStringCell*>(this)->heapByteArray();
}
}
else
{
destString = utf8Data();
// The initial chunk is already correct
// Are our pattern bytes in the range we're about to overwrite?
// We only need to check the end of the pattern because the pattern should only be completely inside our
// completely outside our string
if (((pattern + patternBytes) > (utf8Data() + initialBytes)) &&
((pattern + patternBytes) <= (utf8Data() + byteLength())))
{
// Create a temporary copy to work with
copySource = new std::uint8_t[patternBytes];
memcpy(const_cast<std::uint8_t*>(copySource), pattern, patternBytes);
}
else
{
copySource = pattern;
}
}
// Move the unchanged chunk at the end
// We need to do this now because if the pattern bytes are longer than the byte we're replacing then we might
// overwrite the beginning of the unchanged chunk
memmove(destString + initialBytes + requiredBytes, range.startPointer + replacedBytes, finalBytes);
std::uint8_t* copyDest = destString + initialBytes;
while(count--)
{
memcpy(copyDest, copySource, patternBytes);
copyDest += patternBytes;
}
if (copySource != pattern)
{
delete[] copySource;
}
// Update ourselves with our new string
setLengths(newByteLength, newCharLength);
if (newByteArray)
{
static_cast<HeapStringCell*>(this)->setHeapByteArray(newByteArray);
}
if (oldByteArray != nullptr)
{
// We can unref this now
oldByteArray->unref();
}
}
return true;
}
