private void SetStringBuffer(string str)
{
// max possible size - we avoid using GetByteCount because profiling showed it to take 2% of runtime
// the buffer might be a bit longer, but we'll reuse it, and it is better than the computing cost
int byteCount = Encodings.Utf8.GetMaxByteCount(str.Length);
int escapePositionsSize = JsonParserState.FindEscapePositionsMaxSize(str, out _);
// If we do not have a buffer or the buffer is too small, return the memory and get more.
var size = byteCount + escapePositionsSize;
if (_currentStateBuffer == null || _currentStateBuffer.SizeInBytes < size)
{
if (_currentStateBuffer != null)
{
_ctx.ReturnMemory(_currentStateBuffer);
}
_currentStateBuffer = _ctx.GetMemory(size);
Debug.Assert(_currentStateBuffer != null && _currentStateBuffer.Address != null);
}
_state.StringBuffer = _currentStateBuffer.Address;
fixed(char *pChars = str)
{
_state.StringSize = Encodings.Utf8.GetBytes(pChars, str.Length, _state.StringBuffer, byteCount);
_state.CompressedSize = null; // don't even try
_state.FindEscapePositionsIn(_state.StringBuffer, ref _state.StringSize, escapePositionsSize);
var escapePos = _state.StringBuffer + _state.StringSize;
_state.WriteEscapePositionsTo(escapePos);
}
}
private unsafe LazyStringValue GetLazyString(StringSegment field, bool longLived)
{
var state = new JsonParserState();
var maxByteCount = Encodings.Utf8.GetMaxByteCount(field.Length);
int escapePositionsSize = JsonParserState.FindEscapePositionsMaxSize(field);
int memorySize = maxByteCount + escapePositionsSize;
var memory = longLived ? GetLongLivedMemory(memorySize) : GetMemory(memorySize);
fixed(char *pField = field.Buffer)
{
var address = memory.Address;
var actualSize = Encodings.Utf8.GetBytes(pField + field.Offset, field.Length, address, memory.SizeInBytes);
state.FindEscapePositionsIn(address, actualSize, escapePositionsSize);
state.WriteEscapePositionsTo(address + actualSize);
LazyStringValue result = longLived == false?AllocateStringValue(field, address, actualSize) : new LazyStringValue(field, address, actualSize, this);
result.AllocatedMemoryData = memory;
if (state.EscapePositions.Count > 0)
{
result.EscapePositions = state.EscapePositions.ToArray();
}
return(result);
}
}
public unsafe LazyStringValue GetLazyString(char[] chars, int start, int count)
{
LazyStringValue value;
var state = new JsonParserState();
state.FindEscapePositionsIn(chars, start, count);
var maxByteCount = Encoding.GetMaxByteCount(count);
var memory = GetMemory(maxByteCount + state.GetEscapePositionsSize());
try
{
fixed(char *pChars = chars)
{
var address = (byte *)memory.Address;
var actualSize = Encoding.GetBytes(pChars + start, count, address, memory.SizeInBytes);
state.WriteEscapePositionsTo(address + actualSize);
value = new LazyStringValue(null, address, actualSize, this);
}
}
catch (Exception)
{
ReturnMemory(memory);
throw;
}
return(value);
}
public unsafe LazyStringValue GetLazyString(string field)
{
var state = new JsonParserState();
state.FindEscapePositionsIn(field);
var maxByteCount = Encoding.GetMaxByteCount(field.Length);
var memory = GetMemory(maxByteCount + state.GetEscapePositionsSize());
try
{
fixed(char *pField = field)
{
var address = (byte *)memory.Address;
var actualSize = Encoding.GetBytes(pField, field.Length, address, memory.SizeInBytes);
state.WriteEscapePositionsTo(address + actualSize);
return(new LazyStringValue(field, address, actualSize, this)
{
AllocatedMemoryData = memory
});
}
}
catch (Exception)
{
ReturnMemory(memory);
throw;
}
}
private unsafe LazyStringValue GetLazyString(StringSegment field, bool longLived)
{
var state = new JsonParserState();
state.FindEscapePositionsIn(field);
var maxByteCount = Encoding.GetMaxByteCount(field.Length);
var memory = GetMemory(maxByteCount + state.GetEscapePositionsSize(), longLived: longLived);
fixed(char *pField = field.String)
{
var address = memory.Address;
var actualSize = Encoding.GetBytes(pField + field.Start, field.Length, address, memory.SizeInBytes);
state.WriteEscapePositionsTo(address + actualSize);
var result = new LazyStringValue(field, address, actualSize, this)
{
AllocatedMemoryData = memory,
};
if (state.EscapePositions.Count > 0)
{
result.EscapePositions = state.EscapePositions.ToArray();
}
return(result);
}
}
public static ByteStringContext.InternalScope GetLowerIdSliceAndStorageKey <TTransaction>(
TransactionOperationContext <TTransaction> context, string str, out Slice lowerIdSlice, out Slice idSlice)
where TTransaction : RavenTransaction
{
// Because we need to also store escape positions for the key when we store it
// we need to store it as a lazy string value.
// But lazy string value has two lengths, one is the string length, and the other
// is the actual data size with the escape positions
// In order to resolve this, we process the key to find escape positions, then store it
// in the table using the following format:
//
// [var int - string len, string bytes, number of escape positions, escape positions]
//
// The total length of the string is stored in the actual table (and include the var int size
// prefix.
if (_jsonParserState == null)
{
_jsonParserState = new JsonParserState();
}
_jsonParserState.Reset();
int strLength = str.Length;
int maxStrSize = Encoding.GetMaxByteCount(strLength);
int idSize = JsonParserState.VariableSizeIntSize(strLength);
int escapePositionsSize = JsonParserState.FindEscapePositionsMaxSize(str);
var scope = context.Allocator.Allocate(maxStrSize // lower key
+ idSize // the size of var int for the len of the key
+ maxStrSize // actual key
+ escapePositionsSize, out ByteString buffer);
byte *ptr = buffer.Ptr;
fixed(char *pChars = str)
{
for (var i = 0; i < strLength; i++)
{
uint ch = pChars[i];
// PERF: Trick to avoid multiple compare instructions on hot loops.
// This is the same as (ch >= 65 && ch <= 90)
if (ch - 65 <= 90 - 65)
{
ptr[i] = (byte)(ch | 0x20);
}
else
{
if (ch > 127) // not ASCII, use slower mode
{
goto UnlikelyUnicode;
}
ptr[i] = (byte)ch;
}
ptr[i + idSize + maxStrSize] = (byte)ch;
}
_jsonParserState.FindEscapePositionsIn(ptr, strLength, escapePositionsSize);
}
var writePos = ptr + maxStrSize;
JsonParserState.WriteVariableSizeInt(ref writePos, strLength);
escapePositionsSize = _jsonParserState.WriteEscapePositionsTo(writePos + strLength);
idSize = escapePositionsSize + strLength + idSize;
Slice.External(context.Allocator, ptr + maxStrSize, idSize, out idSlice);
Slice.External(context.Allocator, ptr, str.Length, out lowerIdSlice);
return(scope);
UnlikelyUnicode:
scope.Dispose();
return(UnicodeGetLowerIdAndStorageKey(context, str, out lowerIdSlice, out idSlice, maxStrSize, escapePositionsSize));
}
public static ByteStringContext.InternalScope GetLowerIdSliceAndStorageKey <TTransaction>(
TransactionOperationContext <TTransaction> context, string str, out Slice lowerIdSlice, out Slice idSlice)
where TTransaction : RavenTransaction
{
// Because we need to also store escape positions for the key when we store it
// we need to store it as a lazy string value.
// But lazy string value has two lengths, one is the string length, and the other
// is the actual data size with the escape positions
// In order to resolve this, we process the key to find escape positions, then store it
// in the table using the following format:
//
// [var int - string len, string bytes, number of escape positions, escape positions]
//
// The total length of the string is stored in the actual table (and include the var int size
// prefix.
if (_jsonParserState == null)
{
_jsonParserState = new JsonParserState();
}
_jsonParserState.Reset();
int originalStrLength = str.Length;
int strLength = originalStrLength;
if (strLength > MaxIdSize)
{
ThrowDocumentIdTooBig(str);
}
int escapePositionsSize = JsonParserState.FindEscapePositionsMaxSize(str, out var escapedCount);
/*
* add the size of all control characters
* this is to treat case when we have 2+ control character in a row
* GetMaxByteCount returns smaller size than the actual size with escaped control characters
* For example: string with two control characters such as '\0\0' will be converted to '\u0000\u0000' (another example: '\b\b' => '\u000b\u000b')
* string size = 2, GetMaxByteCount = 9, converted string size = 12, maxStrSize = 19
*/
var maxStrSize = Encoding.GetMaxByteCount(strLength) + JsonParserState.ControlCharacterItemSize * escapedCount;
var originalMaxStrSize = maxStrSize;
int idSize = JsonParserState.VariableSizeIntSize(maxStrSize);
var scope = context.Allocator.Allocate(maxStrSize // lower key
+ idSize // the size of var int for the len of the key
+ maxStrSize // actual key
+ escapePositionsSize, out ByteString buffer);
byte *ptr = buffer.Ptr;
fixed(char *pChars = str)
{
for (var i = 0; i < strLength; i++)
{
uint ch = pChars[i];
// PERF: Trick to avoid multiple compare instructions on hot loops.
// This is the same as (ch >= 65 && ch <= 90)
if (ch - 65 <= 90 - 65)
{
ptr[i] = (byte)(ch | 0x20);
}
else
{
if (ch > 127) // not ASCII, use slower mode
{
goto UnlikelyUnicode;
}
ptr[i] = (byte)ch;
}
ptr[i + idSize + maxStrSize] = (byte)ch;
}
_jsonParserState.FindEscapePositionsIn(ptr, ref strLength, escapePositionsSize);
if (strLength != originalStrLength)
{
var anotherStrLength = originalStrLength;
_jsonParserState.FindEscapePositionsIn(ptr + idSize + maxStrSize, ref anotherStrLength, escapePositionsSize);
#if DEBUG
if (strLength != anotherStrLength)
{
throw new InvalidOperationException($"String length mismatch between Id ({str}) and it's lowercased counterpart after finding escape positions. Original: {anotherStrLength}. Lowercased: {strLength}");
}
#endif
}
}
var writePos = ptr + maxStrSize;
Debug.Assert(strLength <= originalMaxStrSize, $"Calculated {nameof(originalMaxStrSize)} value {originalMaxStrSize}, was smaller than actually {nameof(strLength)} value {strLength}");
// in case there were no control characters the idSize could be smaller
var sizeDifference = idSize - JsonParserState.VariableSizeIntSize(strLength);
writePos += sizeDifference;
idSize -= sizeDifference;
JsonParserState.WriteVariableSizeInt(ref writePos, strLength);
escapePositionsSize = _jsonParserState.WriteEscapePositionsTo(writePos + strLength);
idSize = escapePositionsSize + strLength + idSize;
Slice.External(context.Allocator, ptr + maxStrSize + sizeDifference, idSize, out idSlice);
Slice.External(context.Allocator, ptr, strLength, out lowerIdSlice);
return(scope);
UnlikelyUnicode:
scope.Dispose();
return(UnicodeGetLowerIdAndStorageKey(context, str, out lowerIdSlice, out idSlice, maxStrSize, escapePositionsSize));
}
public static void GetLowerKeySliceAndStorageKey(JsonOperationContext context, string str, out byte *lowerKey,
out int lowerSize,
out byte *key,
out int keySize)
{
// Because we need to also store escape positions for the key when we store it
// we need to store it as a lazy string value.
// But lazy string value has two lengths, one is the string length, and the other
// is the actual data size with the escape positions
// In order to resolve this, we process the key to find escape positions, then store it
// in the table using the following format:
//
// [var int - string len, string bytes, number of escape positions, escape positions]
//
// The total length of the string is stored in the actual table (and include the var int size
// prefix.
if (_jsonParserState == null)
{
_jsonParserState = new JsonParserState();
}
_jsonParserState.Reset();
keySize = JsonParserState.VariableSizeIntSize(str.Length);
_jsonParserState.FindEscapePositionsIn(str);
var escapePositionsSize = _jsonParserState.GetEscapePositionsSize();
var buffer = context.GetNativeTempBuffer(
str.Length // lower key
+ keySize // the size of var int for the len of the key
+ str.Length // actual key
+ escapePositionsSize);
for (var i = 0; i < str.Length; i++)
{
var ch = str[i];
if (ch > 127) // not ASCII, use slower mode
{
goto UnlikelyUnicode;
}
if ((ch >= 65) && (ch <= 90))
{
buffer[i] = (byte)(ch | 0x20);
}
else
{
buffer[i] = (byte)ch;
}
buffer[i + keySize + str.Length] = (byte)ch;
}
var writePos = buffer + str.Length;
JsonParserState.WriteVariableSizeInt(ref writePos, str.Length);
_jsonParserState.WriteEscapePositionsTo(writePos + str.Length);
keySize = escapePositionsSize + str.Length + keySize;
key = buffer + str.Length;
lowerKey = buffer;
lowerSize = str.Length;
return;
UnlikelyUnicode:
UnicodeGetLowerKeySliceAndStorageKey(context, str, out lowerKey, out lowerSize, out key, out keySize);
}