/// <summary>
/// Parses the current JSON token value from the source, unescaped, and transcoded as a <see cref="string"/>.
/// </summary>
/// <exception cref="InvalidOperationException">
/// Thrown if trying to get the value of the JSON token that is not a string
/// (i.e. other than <see cref="JsonTokenType.String"/> or <see cref="JsonTokenType.PropertyName"/>).
/// <seealso cref="TokenType" />
/// It will also throw when the JSON string contains invalid UTF-8 bytes, or invalid UTF-16 surrogates.
/// </exception>
public string GetString()
{
if (TokenType != JsonTokenType.String && TokenType != JsonTokenType.PropertyName)
{
throw ThrowHelper.GetInvalidOperationException_ExpectedString(TokenType);
}
ReadOnlySpan <byte> span = HasValueSequence ? ValueSequence.ToArray() : ValueSpan;
if (_stringHasEscaping)
{
int idx = span.IndexOf(JsonConstants.BackSlash);
Debug.Assert(idx != -1);
return(JsonReaderHelper.GetUnescapedString(span, idx));
}
Debug.Assert(span.IndexOf(JsonConstants.BackSlash) == -1);
return(JsonReaderHelper.TranscodeHelper(span));
}
private static TValue?Deserialize <TValue>(string json, Type returnType, JsonSerializerOptions?options)
{
const long ArrayPoolMaxSizeBeforeUsingNormalAlloc = 1024 * 1024;
if (options == null)
{
options = JsonSerializerOptions.s_defaultOptions;
}
byte[]? tempArray = null;
// For performance, avoid obtaining actual byte count unless memory usage is higher than the threshold.
Span <byte> utf8 = json.Length <= (ArrayPoolMaxSizeBeforeUsingNormalAlloc / JsonConstants.MaxExpansionFactorWhileTranscoding) ?
// Use a pooled alloc.
tempArray = ArrayPool <byte> .Shared.Rent(json.Length *JsonConstants.MaxExpansionFactorWhileTranscoding) :
// Use a normal alloc since the pool would create a normal alloc anyway based on the threshold (per current implementation)
// and by using a normal alloc we can avoid the Clear().
new byte[JsonReaderHelper.GetUtf8ByteCount(json.AsSpan())];
try
{
int actualByteCount = JsonReaderHelper.GetUtf8FromText(json.AsSpan(), utf8);
utf8 = utf8.Slice(0, actualByteCount);
var readerState = new JsonReaderState(options.GetReaderOptions());
var reader = new Utf8JsonReader(utf8, isFinalBlock: true, readerState);
TValue value = ReadCore <TValue>(ref reader, returnType, options);
// The reader should have thrown if we have remaining bytes.
Debug.Assert(reader.BytesConsumed == actualByteCount);
return(value);
}
finally
{
if (tempArray != null)
{
utf8.Clear();
ArrayPool <byte> .Shared.Return(tempArray);
}
}
}
/// <summary>
/// Parses the current JSON token value from the source and decodes the base 64 encoded JSON string as bytes.
/// Returns <see langword="true"/> if the entire token value is encoded as valid base 64 text and can be successfully
/// decoded to bytes.
/// Returns <see langword="false"/> otherwise.
/// </summary>
/// <exception cref="InvalidOperationException">
/// Thrown if trying to get the value of a JSON token that is not a <see cref="JsonTokenType.String"/>.
/// <seealso cref="TokenType" />
/// </exception>
public bool TryGetBytesFromBase64(out byte[] value)
{
if (TokenType != JsonTokenType.String)
{
throw ThrowHelper.GetInvalidOperationException_ExpectedString(TokenType);
}
ReadOnlySpan <byte> span = HasValueSequence ? ValueSequence.ToArray() : ValueSpan;
if (_stringHasEscaping)
{
int idx = span.IndexOf(JsonConstants.BackSlash);
Debug.Assert(idx != -1);
return(JsonReaderHelper.TryGetUnescapedBase64Bytes(span, idx, out value));
}
Debug.Assert(span.IndexOf(JsonConstants.BackSlash) == -1);
return(JsonReaderHelper.TryDecodeBase64(span, out value));
}
internal double GetDoubleWithQuotes()
{
ReadOnlySpan <byte> span = GetUnescapedSpan();
if (JsonReaderHelper.TryGetFloatingPointConstant(span, out double value))
{
return(value);
}
// NETCOREAPP implementation of the TryParse method above permits case-insensitive variants of the
// float constants "NaN", "Infinity", "-Infinity". This differs from the NETFRAMEWORK implementation.
// The following logic reconciles the two implementations to enforce consistent behavior.
if (!(Utf8Parser.TryParse(span, out value, out int bytesConsumed) &&
span.Length == bytesConsumed &&
JsonHelpers.IsFinite(value)))
{
ThrowHelper.ThrowFormatException(NumericType.Double);
}
return(value);
}
/// <summary>
/// Parse text representing a single JSON value into a JsonDocument.
/// </summary>
/// <remarks>
/// The <see cref="ReadOnlyMemory{T}"/> value may be used for the entire lifetime of the
/// JsonDocument object, and the caller must ensure that the data therein does not change during
/// the object lifetime.
/// </remarks>
/// <param name="json">JSON text to parse.</param>
/// <param name="options">Options to control the reader behavior during parsing.</param>
/// <returns>
/// A JsonDocument representation of the JSON value.
/// </returns>
/// <exception cref="JsonException">
/// <paramref name="json"/> does not represent a valid single JSON value.
/// </exception>
/// <exception cref="ArgumentException">
/// <paramref name="options"/> contains unsupported options.
/// </exception>
public static JsonDocument Parse(ReadOnlyMemory <char> json, JsonDocumentOptions options = default)
{
ReadOnlySpan <char> jsonChars = json.Span;
int expectedByteCount = JsonReaderHelper.GetUtf8ByteCount(jsonChars);
byte[] utf8Bytes = ArrayPool <byte> .Shared.Rent(expectedByteCount);
try
{
int actualByteCount = JsonReaderHelper.GetUtf8FromText(jsonChars, utf8Bytes);
Debug.Assert(expectedByteCount == actualByteCount);
return(Parse(utf8Bytes.AsMemory(0, actualByteCount), options.GetReaderOptions(), utf8Bytes));
}
catch
{
// Holds document content, clear it before returning it.
utf8Bytes.AsSpan(0, expectedByteCount).Clear();
ArrayPool <byte> .Shared.Return(utf8Bytes);
throw;
}
}
private static ReadOnlySpan <byte> GetUnescapedString(ReadOnlySpan <byte> utf8Source, int idx)
{
// The escaped name is always longer than the unescaped, so it is safe to use escaped name for the buffer length.
int length = utf8Source.Length;
byte[]? pooledName = null;
Span <byte> unescapedName = length <= JsonConstants.StackallocThreshold ?
stackalloc byte[length] :
(pooledName = ArrayPool <byte> .Shared.Rent(length));
JsonReaderHelper.Unescape(utf8Source, unescapedName, idx, out int written);
ReadOnlySpan <byte> propertyName = unescapedName.Slice(0, written).ToArray();
if (pooledName != null)
{
// We clear the array because it is "user data" (although a property name).
new Span <byte>(pooledName, 0, written).Clear();
ArrayPool <byte> .Shared.Return(pooledName);
}
return(propertyName);
}
private static JsonEncodedText TranscodeAndEncode(ReadOnlySpan <char> value)
{
JsonWriterHelper.ValidateValue(value);
int expectedByteCount = JsonReaderHelper.GetUtf8ByteCount(value);
byte[] utf8Bytes = ArrayPool <byte> .Shared.Rent(expectedByteCount);
JsonEncodedText encodedText;
// Since GetUtf8ByteCount above already throws on invalid input, the transcoding
// to UTF-8 is guaranteed to succeed here. Therefore, there's no need for a try-catch-finally block.
int actualByteCount = JsonReaderHelper.GetUtf8FromText(value, utf8Bytes);
Debug.Assert(expectedByteCount == actualByteCount);
encodedText = EncodeHelper(utf8Bytes.AsSpan(0, actualByteCount));
// On the basis that this is user data, go ahead and clear it.
utf8Bytes.AsSpan(0, expectedByteCount).Clear();
ArrayPool <byte> .Shared.Return(utf8Bytes);
return(encodedText);
}
private bool TryGetNamedPropertyValue(
int startIndex,
int endIndex,
ReadOnlySpan <byte> propertyName,
out JsonElement value)
{
ReadOnlySpan <byte> documentSpan = _utf8Json.Span;
// Move to the row before the EndObject
int index = endIndex - DbRow.Size;
while (index > startIndex)
{
DbRow row = _parsedData.Get(index);
Debug.Assert(row.TokenType != JsonTokenType.PropertyName);
// Move before the value
if (row.IsSimpleValue)
{
index -= DbRow.Size;
}
else
{
Debug.Assert(row.NumberOfRows > 0);
index -= DbRow.Size * (row.NumberOfRows + 1);
}
row = _parsedData.Get(index);
Debug.Assert(row.TokenType == JsonTokenType.PropertyName);
ReadOnlySpan <byte> currentPropertyName = documentSpan.Slice(row.Location, row.SizeOrLength);
if (row.HasComplexChildren)
{
// An escaped property name will be longer than an unescaped candidate, so only unescape
// when the lengths are compatible.
if (currentPropertyName.Length > propertyName.Length)
{
int idx = currentPropertyName.IndexOf(JsonConstants.BackSlash);
Debug.Assert(idx >= 0);
// If everything up to where the property name has a backslash matches, keep going.
if (propertyName.Length > idx &&
currentPropertyName.Slice(0, idx).SequenceEqual(propertyName.Slice(0, idx)))
{
int remaining = currentPropertyName.Length - idx;
int written = 0;
byte[] rented = null;
try
{
Span <byte> utf8Unescaped = remaining <= JsonConstants.StackallocThreshold ?
stackalloc byte[remaining] :
(rented = ArrayPool <byte> .Shared.Rent(remaining));
// Only unescape the part we haven't processed.
JsonReaderHelper.Unescape(currentPropertyName.Slice(idx), utf8Unescaped, 0, out written);
// If the unescaped remainder matches the input remainder, it's a match.
if (utf8Unescaped.Slice(0, written).SequenceEqual(propertyName.Slice(idx)))
{
// If the property name is a match, the answer is the next element.
value = new JsonElement(this, index + DbRow.Size);
return(true);
}
}
finally
{
if (rented != null)
{
rented.AsSpan(0, written).Clear();
ArrayPool <byte> .Shared.Return(rented);
}
}
}
}
}
else if (currentPropertyName.SequenceEqual(propertyName))
{
// If the property name is a match, the answer is the next element.
value = new JsonElement(this, index + DbRow.Size);
return(true);
}
// Move to the previous value
index -= DbRow.Size;
}
value = default;
return(false);
}
internal bool TryGetNamedPropertyValue(int index, ReadOnlySpan <char> propertyName, out JsonElement value)
{
CheckNotDisposed();
DbRow row = _parsedData.Get(index);
CheckExpectedType(JsonTokenType.StartObject, row.TokenType);
// Only one row means it was EndObject.
if (row.NumberOfRows == 1)
{
value = default;
return(false);
}
int maxBytes = JsonReaderHelper.s_utf8Encoding.GetMaxByteCount(propertyName.Length);
int startIndex = index + DbRow.Size;
int endIndex = checked (row.NumberOfRows * DbRow.Size + index);
if (maxBytes < JsonConstants.StackallocThreshold)
{
Span <byte> utf8Name = stackalloc byte[JsonConstants.StackallocThreshold];
int len = JsonReaderHelper.GetUtf8FromText(propertyName, utf8Name);
utf8Name = utf8Name.Slice(0, len);
return(TryGetNamedPropertyValue(
startIndex,
endIndex,
utf8Name,
out value));
}
// Unescaping the property name will make the string shorter (or the same)
// So the first viable candidate is one whose length in bytes matches, or
// exceeds, our length in chars.
//
// The maximal escaping seems to be 6 -> 1 ("\u0030" => "0"), but just transcode
// and switch once one viable long property is found.
int minBytes = propertyName.Length;
// Move to the row before the EndObject
int candidateIndex = endIndex - DbRow.Size;
while (candidateIndex > index)
{
int passedIndex = candidateIndex;
row = _parsedData.Get(candidateIndex);
Debug.Assert(row.TokenType != JsonTokenType.PropertyName);
// Move before the value
if (row.IsSimpleValue)
{
candidateIndex -= DbRow.Size;
}
else
{
Debug.Assert(row.NumberOfRows > 0);
candidateIndex -= DbRow.Size * (row.NumberOfRows + 1);
}
row = _parsedData.Get(candidateIndex);
Debug.Assert(row.TokenType == JsonTokenType.PropertyName);
if (row.SizeOrLength >= minBytes)
{
byte[] tmpUtf8 = ArrayPool <byte> .Shared.Rent(maxBytes);
Span <byte> utf8Name = default;
try
{
int len = JsonReaderHelper.GetUtf8FromText(propertyName, tmpUtf8);
utf8Name = tmpUtf8.AsSpan(0, len);
return(TryGetNamedPropertyValue(
startIndex,
passedIndex + DbRow.Size,
utf8Name,
out value));
}
finally
{
// While property names aren't usually a secret, they also usually
// aren't long enough to end up in the rented buffer transcode path.
//
// On the basis that this is user data, go ahead and clear it.
utf8Name.Clear();
ArrayPool <byte> .Shared.Return(tmpUtf8);
}
}
// Move to the previous value
candidateIndex -= DbRow.Size;
}
// None of the property names were within the range that the UTF-8 encoding would have been.
value = default;
return(false);
}
/// <summary>
/// Parse the text representing a single JSON value into a <paramref name="returnType"/>.
/// </summary>
/// <returns>A <paramref name="returnType"/> representation of the JSON value.</returns>
/// <param name="json">JSON text to parse.</param>
/// <param name="returnType">The type of the object to convert to and return.</param>
/// <param name="options">Options to control the behavior during parsing.</param>
/// <exception cref="System.ArgumentNullException">
/// Thrown if <paramref name="json"/> or <paramref name="returnType"/> is null.
/// </exception>
/// <exception cref="JsonException">
/// Thrown when the JSON is invalid,
/// the <paramref name="returnType"/> is not compatible with the JSON,
/// or when there is remaining data in the Stream.
/// </exception>
/// <remarks>Using a <see cref="string"/> is not as efficient as using the
/// UTF-8 methods since the implementation natively uses UTF-8.
/// </remarks>
public static object Deserialize(string json, Type returnType, JsonSerializerOptions options = null)
{
if (json == null)
{
throw new ArgumentNullException(nameof(json));
}
if (returnType == null)
{
throw new ArgumentNullException(nameof(returnType));
}
if (options == null)
{
options = JsonSerializerOptions.s_defaultOptions;
}
object result;
byte[] tempArray = null;
int maxBytes;
// For performance, avoid asking for the actual byte count unless necessary.
if (json.Length > MaxArrayLengthBeforeCalculatingSize)
{
// Get the actual byte count in order to handle large input.
maxBytes = JsonReaderHelper.GetUtf8ByteCount(json.AsSpan());
}
else
{
maxBytes = json.Length * JsonConstants.MaxExpansionFactorWhileTranscoding;
}
Span <byte> utf8 = maxBytes <= JsonConstants.StackallocThreshold ?
stackalloc byte[maxBytes] :
(tempArray = ArrayPool <byte> .Shared.Rent(maxBytes));
try
{
int actualByteCount = JsonReaderHelper.GetUtf8FromText(json.AsSpan(), utf8);
utf8 = utf8.Slice(0, actualByteCount);
var readerState = new JsonReaderState(options.GetReaderOptions());
var reader = new Utf8JsonReader(utf8, isFinalBlock: true, readerState);
result = ReadCore(returnType, options, ref reader);
if (reader.BytesConsumed != actualByteCount)
{
ThrowHelper.ThrowJsonException_DeserializeDataRemaining(
actualByteCount, actualByteCount - reader.BytesConsumed);
}
}
finally
{
if (tempArray != null)
{
utf8.Clear();
ArrayPool <byte> .Shared.Return(tempArray);
}
}
return(result);
}
internal static JsonPropertyInfo LookupProperty(
object obj,
ref Utf8JsonReader reader,
JsonSerializerOptions options,
ref ReadStack state,
out bool useExtensionProperty)
{
Debug.Assert(state.Current.JsonClassInfo.ClassType == ClassType.Object);
JsonPropertyInfo jsonPropertyInfo;
ReadOnlySpan <byte> unescapedPropertyName;
ReadOnlySpan <byte> propertyName = reader.GetSpan();
if (reader._stringHasEscaping)
{
int idx = propertyName.IndexOf(JsonConstants.BackSlash);
Debug.Assert(idx != -1);
unescapedPropertyName = JsonReaderHelper.GetUnescapedSpan(propertyName, idx);
}
else
{
unescapedPropertyName = propertyName;
}
if (options.ReferenceHandling.ShouldReadPreservedReferences())
{
if (propertyName.Length > 0 && propertyName[0] == '$')
{
ThrowHelper.ThrowUnexpectedMetadataException(propertyName, ref reader, ref state);
}
}
jsonPropertyInfo = state.Current.JsonClassInfo.GetProperty(unescapedPropertyName, ref state.Current);
// Increment PropertyIndex so GetProperty() starts with the next property the next time this function is called.
state.Current.PropertyIndex++;
// Determine if we should use the extension property.
if (jsonPropertyInfo == JsonPropertyInfo.s_missingProperty)
{
JsonPropertyInfo?dataExtProperty = state.Current.JsonClassInfo.DataExtensionProperty;
if (dataExtProperty != null)
{
state.Current.JsonPropertyNameAsString = JsonHelpers.Utf8GetString(unescapedPropertyName);
CreateDataExtensionProperty(obj, dataExtProperty);
jsonPropertyInfo = dataExtProperty;
}
state.Current.JsonPropertyInfo = jsonPropertyInfo;
useExtensionProperty = true;
return(jsonPropertyInfo);
}
// Support JsonException.Path.
Debug.Assert(
jsonPropertyInfo.JsonPropertyName == null ||
options.PropertyNameCaseInsensitive ||
unescapedPropertyName.SequenceEqual(jsonPropertyInfo.JsonPropertyName));
state.Current.JsonPropertyInfo = jsonPropertyInfo;
if (jsonPropertyInfo.JsonPropertyName == null)
{
byte[] propertyNameArray = unescapedPropertyName.ToArray();
if (options.PropertyNameCaseInsensitive)
{
// Each payload can have a different name here; remember the value on the temporary stack.
state.Current.JsonPropertyName = propertyNameArray;
}
else
{
// Prevent future allocs by caching globally on the JsonPropertyInfo which is specific to a Type+PropertyName
// so it will match the incoming payload except when case insensitivity is enabled (which is handled above).
state.Current.JsonPropertyInfo.JsonPropertyName = propertyNameArray;
}
}
state.Current.JsonPropertyInfo = jsonPropertyInfo;
useExtensionProperty = false;
return(jsonPropertyInfo);
}
private static void EscapeString(ReadOnlySpan <char> value, Span <char> destination, JavaScriptEncoder encoder, ref int written)
{
// todo: issue #39523: add an Encode(ReadOnlySpan<char>) decode API to System.Text.Encodings.Web.TextEncoding to avoid utf16->utf8->utf16 conversion.
Debug.Assert(encoder != null);
// Convert char to byte.
byte[] utf8DestinationArray = null;
Span <byte> utf8Destination;
int length = checked ((value.Length) * JsonConstants.MaxExpansionFactorWhileTranscoding);
if (length > JsonConstants.StackallocThreshold)
{
utf8DestinationArray = ArrayPool <byte> .Shared.Rent(length);
utf8Destination = utf8DestinationArray;
}
else
{
unsafe
{
byte *ptr = stackalloc byte[JsonConstants.StackallocThreshold];
utf8Destination = new Span <byte>(ptr, JsonConstants.StackallocThreshold);
}
}
ReadOnlySpan <byte> utf16Value = MemoryMarshal.AsBytes(value);
OperationStatus toUtf8Status = ToUtf8(utf16Value, utf8Destination, out int bytesConsumed, out int bytesWritten);
Debug.Assert(toUtf8Status != OperationStatus.DestinationTooSmall);
Debug.Assert(toUtf8Status != OperationStatus.NeedMoreData);
if (toUtf8Status != OperationStatus.Done)
{
if (utf8DestinationArray != null)
{
utf8Destination.Slice(0, bytesWritten).Clear();
ArrayPool <byte> .Shared.Return(utf8DestinationArray);
}
ThrowHelper.ThrowArgumentException_InvalidUTF8(utf16Value.Slice(bytesWritten));
}
Debug.Assert(toUtf8Status == OperationStatus.Done);
Debug.Assert(bytesConsumed == utf16Value.Length);
// Escape the bytes.
byte[] utf8ConvertedDestinationArray = null;
Span <byte> utf8ConvertedDestination;
length = checked (bytesWritten * JsonConstants.MaxExpansionFactorWhileEscaping);
if (length > JsonConstants.StackallocThreshold)
{
utf8ConvertedDestinationArray = ArrayPool <byte> .Shared.Rent(length);
utf8ConvertedDestination = utf8ConvertedDestinationArray;
}
else
{
unsafe
{
byte *ptr = stackalloc byte[JsonConstants.StackallocThreshold];
utf8ConvertedDestination = new Span <byte>(ptr, JsonConstants.StackallocThreshold);
}
}
EscapeString(utf8Destination.Slice(0, bytesWritten), utf8ConvertedDestination, indexOfFirstByteToEscape: 0, encoder, out int convertedBytesWritten);
if (utf8DestinationArray != null)
{
utf8Destination.Slice(0, bytesWritten).Clear();
ArrayPool <byte> .Shared.Return(utf8DestinationArray);
}
// Convert byte to char.
#if BUILDING_INBOX_LIBRARY
OperationStatus toUtf16Status = Utf8.ToUtf16(utf8ConvertedDestination.Slice(0, convertedBytesWritten), destination, out int bytesRead, out int charsWritten);
Debug.Assert(toUtf16Status == OperationStatus.Done);
Debug.Assert(bytesRead == convertedBytesWritten);
#else
string utf16 = JsonReaderHelper.GetTextFromUtf8(utf8ConvertedDestination.Slice(0, convertedBytesWritten));
utf16.AsSpan().CopyTo(destination);
int charsWritten = utf16.Length;
#endif
written += charsWritten;
if (utf8ConvertedDestinationArray != null)
{
utf8ConvertedDestination.Slice(0, written).Clear();
ArrayPool <byte> .Shared.Return(utf8ConvertedDestinationArray);
}
}
