private static void UnicodeGetLowerKeySliceAndStorageKey(JsonOperationContext context, string str,
out byte *lowerKey, out int lowerSize,
out byte *key, out int keySize)
{
// See comment in GetLowerKeySliceAndStorageKey for the format
_jsonParserState.Reset();
var byteCount = Utf8.GetMaxByteCount(str.Length);
_jsonParserState.FindEscapePositionsIn(str);
var maxKeyLenSize = JsonParserState.VariableSizeIntSize(byteCount);
var escapePositionsSize = _jsonParserState.GetEscapePositionsSize();
var buffer = context.GetNativeTempBuffer(
sizeof(char) * str.Length // for the lower calls
+ byteCount // lower key
+ maxKeyLenSize // the size of var int for the len of the key
+ byteCount // actual key
+ escapePositionsSize);
fixed(char *pChars = str)
{
var destChars = (char *)buffer;
for (var i = 0; i < str.Length; i++)
{
destChars[i] = char.ToLowerInvariant(pChars[i]);
}
lowerKey = buffer + str.Length * sizeof(char);
lowerSize = Utf8.GetBytes(destChars, str.Length, lowerKey, byteCount);
if (lowerSize > 512)
{
ThrowKeyTooBig(str);
}
key = buffer + str.Length * sizeof(char) + byteCount;
var writePos = key;
keySize = Utf8.GetBytes(pChars, str.Length, writePos + maxKeyLenSize, byteCount);
var actualKeyLenSize = JsonParserState.VariableSizeIntSize(keySize);
if (actualKeyLenSize < maxKeyLenSize)
{
var movePtr = maxKeyLenSize - actualKeyLenSize;
key += movePtr;
writePos += movePtr;
}
JsonParserState.WriteVariableSizeInt(ref writePos, keySize);
_jsonParserState.WriteEscapePositionsTo(writePos + keySize);
keySize += escapePositionsSize + maxKeyLenSize;
}
}
private unsafe LazyStringValue CreateLazyStringValueFromParserState(JsonParserState state)
{
int escapePositionsCount = state.EscapePositions.Count;
var maxSizeOfEscapePos = escapePositionsCount * 5 // max size of var int
+ JsonParserState.VariableSizeIntSize(escapePositionsCount);
var mem = _ctx.GetMemory(maxSizeOfEscapePos + state.StringSize);
_allocations.Add(mem);
Memory.Copy(mem.Address, state.StringBuffer, state.StringSize);
var lazyStringValueFromParserState = _ctx.AllocateStringValue(null, mem.Address, state.StringSize);
if (escapePositionsCount > 0)
{
lazyStringValueFromParserState.EscapePositions = state.EscapePositions.ToArray();
}
return(lazyStringValueFromParserState);
}
private static ByteStringContext.InternalScope UnicodeGetLowerIdAndStorageKey <TTransaction>(
TransactionOperationContext <TTransaction> context, string str,
out Slice lowerIdSlice, out Slice idSlice, int maxStrSize, int escapePositionsSize)
where TTransaction : RavenTransaction
{
// See comment in GetLowerIdSliceAndStorageKey for the format
var maxIdLenSize = JsonParserState.VariableSizeIntSize(maxStrSize);
int strLength = str.Length;
var scope = context.Allocator.Allocate(
sizeof(char) * strLength // for the lower calls
+ maxStrSize // lower ID
+ maxIdLenSize // the size of var int for the len of the ID
+ maxStrSize // actual ID
+ escapePositionsSize, out ByteString buffer);
fixed(char *pChars = str)
{
var size = Encoding.GetBytes(pChars, strLength, buffer.Ptr, maxStrSize);
_jsonParserState.FindEscapePositionsIn(buffer.Ptr, size, escapePositionsSize);
var destChars = (char *)buffer.Ptr;
for (var i = 0; i < strLength; i++)
{
destChars[i] = char.ToLowerInvariant(pChars[i]);
}
byte *lowerId = buffer.Ptr + strLength * sizeof(char);
int lowerSize = Encoding.GetBytes(destChars, strLength, lowerId, maxStrSize);
if (lowerSize > 512)
{
ThrowDocumentIdTooBig(str);
}
byte *id = buffer.Ptr + strLength * sizeof(char) + maxStrSize;
byte *writePos = id;
int idSize = Encoding.GetBytes(pChars, strLength, writePos + maxIdLenSize, maxStrSize);
var actualIdLenSize = JsonParserState.VariableSizeIntSize(idSize);
if (actualIdLenSize < maxIdLenSize)
{
var movePtr = maxIdLenSize - actualIdLenSize;
id += movePtr;
writePos += movePtr;
}
JsonParserState.WriteVariableSizeInt(ref writePos, idSize);
escapePositionsSize = _jsonParserState.WriteEscapePositionsTo(writePos + idSize);
idSize += escapePositionsSize + maxIdLenSize;
Slice.External(context.Allocator, id, idSize, out idSlice);
Slice.External(context.Allocator, lowerId, lowerSize, out lowerIdSlice);
return(scope);
}
}
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);
}