Commit 3de2fced authored by Jon Skeet's avatar Jon Skeet

Handle JSON parsing for Any.

This required a rework of the tokenizer to allow for a "replaying" tokenizer, basically in case the @type value comes after the data itself. This rework is nice in some ways (all the pushback and object depth logic in one place) but is a little fragile in terms of token push-back when using the replay tokenizer. It'll be fine for the scenario we need it for, but we should be careful...
parent 567579b5
...@@ -30,6 +30,7 @@ ...@@ -30,6 +30,7 @@
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#endregion #endregion
using Google.Protobuf.Reflection;
using Google.Protobuf.TestProtos; using Google.Protobuf.TestProtos;
using Google.Protobuf.WellKnownTypes; using Google.Protobuf.WellKnownTypes;
using NUnit.Framework; using NUnit.Framework;
...@@ -717,6 +718,55 @@ namespace Google.Protobuf ...@@ -717,6 +718,55 @@ namespace Google.Protobuf
CollectionAssert.AreEqual(expectedPaths, parsed.Paths); CollectionAssert.AreEqual(expectedPaths, parsed.Paths);
} }
[Test]
public void Any_RegularMessage()
{
var registry = TypeRegistry.FromMessages(TestAllTypes.Descriptor);
var formatter = new JsonFormatter(new JsonFormatter.Settings(false, TypeRegistry.FromMessages(TestAllTypes.Descriptor)));
var message = new TestAllTypes { SingleInt32 = 10, SingleNestedMessage = new TestAllTypes.Types.NestedMessage { Bb = 20 } };
var original = Any.Pack(message);
var json = formatter.Format(original); // This is tested in JsonFormatterTest
var parser = new JsonParser(new JsonParser.Settings(10, registry));
Assert.AreEqual(original, parser.Parse<Any>(json));
string valueFirstJson = "{ \"singleInt32\": 10, \"singleNestedMessage\": { \"bb\": 20 }, \"@type\": \"type.googleapis.com/protobuf_unittest.TestAllTypes\" }";
Assert.AreEqual(original, parser.Parse<Any>(valueFirstJson));
}
[Test]
public void Any_UnknownType()
{
string json = "{ \"@type\": \"type.googleapis.com/bogus\" }";
Assert.Throws<InvalidOperationException>(() => Any.Parser.ParseJson(json));
}
[Test]
public void Any_WellKnownType()
{
var registry = TypeRegistry.FromMessages(Timestamp.Descriptor);
var formatter = new JsonFormatter(new JsonFormatter.Settings(false, registry));
var timestamp = new DateTime(1673, 6, 19, 12, 34, 56, DateTimeKind.Utc).ToTimestamp();
var original = Any.Pack(timestamp);
var json = formatter.Format(original); // This is tested in JsonFormatterTest
var parser = new JsonParser(new JsonParser.Settings(10, registry));
Assert.AreEqual(original, parser.Parse<Any>(json));
string valueFirstJson = "{ \"value\": \"1673-06-19T12:34:56Z\", \"@type\": \"type.googleapis.com/google.protobuf.Timestamp\" }";
Assert.AreEqual(original, parser.Parse<Any>(valueFirstJson));
}
[Test]
public void Any_Nested()
{
var registry = TypeRegistry.FromMessages(TestWellKnownTypes.Descriptor, TestAllTypes.Descriptor);
var formatter = new JsonFormatter(new JsonFormatter.Settings(false, registry));
var parser = new JsonParser(new JsonParser.Settings(10, registry));
var doubleNestedMessage = new TestAllTypes { SingleInt32 = 20 };
var nestedMessage = Any.Pack(doubleNestedMessage);
var message = new TestWellKnownTypes { AnyField = Any.Pack(nestedMessage) };
var json = formatter.Format(message);
// Use the descriptor-based parser just for a change.
Assert.AreEqual(message, parser.Parse(json, TestWellKnownTypes.Descriptor));
}
[Test] [Test]
public void DataAfterObject() public void DataAfterObject()
{ {
......
...@@ -85,7 +85,7 @@ namespace Google.Protobuf ...@@ -85,7 +85,7 @@ namespace Google.Protobuf
public void ObjectDepth() public void ObjectDepth()
{ {
string json = "{ \"foo\": { \"x\": 1, \"y\": [ 0 ] } }"; string json = "{ \"foo\": { \"x\": 1, \"y\": [ 0 ] } }";
var tokenizer = new JsonTokenizer(new StringReader(json)); var tokenizer = JsonTokenizer.FromTextReader(new StringReader(json));
// If we had more tests like this, I'd introduce a helper method... but for one test, it's not worth it. // If we had more tests like this, I'd introduce a helper method... but for one test, it's not worth it.
Assert.AreEqual(0, tokenizer.ObjectDepth); Assert.AreEqual(0, tokenizer.ObjectDepth);
Assert.AreEqual(JsonToken.StartObject, tokenizer.Next()); Assert.AreEqual(JsonToken.StartObject, tokenizer.Next());
...@@ -118,7 +118,7 @@ namespace Google.Protobuf ...@@ -118,7 +118,7 @@ namespace Google.Protobuf
public void ObjectDepth_WithPushBack() public void ObjectDepth_WithPushBack()
{ {
string json = "{}"; string json = "{}";
var tokenizer = new JsonTokenizer(new StringReader(json)); var tokenizer = JsonTokenizer.FromTextReader(new StringReader(json));
Assert.AreEqual(0, tokenizer.ObjectDepth); Assert.AreEqual(0, tokenizer.ObjectDepth);
var token = tokenizer.Next(); var token = tokenizer.Next();
Assert.AreEqual(1, tokenizer.ObjectDepth); Assert.AreEqual(1, tokenizer.ObjectDepth);
...@@ -275,7 +275,7 @@ namespace Google.Protobuf ...@@ -275,7 +275,7 @@ namespace Google.Protobuf
// Note: we don't test that the earlier tokens are exactly as expected, // Note: we don't test that the earlier tokens are exactly as expected,
// partly because that's hard to parameterize. // partly because that's hard to parameterize.
var reader = new StringReader(json.Replace('\'', '"')); var reader = new StringReader(json.Replace('\'', '"'));
var tokenizer = new JsonTokenizer(reader); var tokenizer = JsonTokenizer.FromTextReader(reader);
for (int i = 0; i < expectedValidTokens; i++) for (int i = 0; i < expectedValidTokens; i++)
{ {
Assert.IsNotNull(tokenizer.Next()); Assert.IsNotNull(tokenizer.Next());
...@@ -334,7 +334,7 @@ namespace Google.Protobuf ...@@ -334,7 +334,7 @@ namespace Google.Protobuf
[Test] [Test]
public void NextAfterEndDocumentThrows() public void NextAfterEndDocumentThrows()
{ {
var tokenizer = new JsonTokenizer(new StringReader("null")); var tokenizer = JsonTokenizer.FromTextReader(new StringReader("null"));
Assert.AreEqual(JsonToken.Null, tokenizer.Next()); Assert.AreEqual(JsonToken.Null, tokenizer.Next());
Assert.AreEqual(JsonToken.EndDocument, tokenizer.Next()); Assert.AreEqual(JsonToken.EndDocument, tokenizer.Next());
Assert.Throws<InvalidOperationException>(() => tokenizer.Next()); Assert.Throws<InvalidOperationException>(() => tokenizer.Next());
...@@ -343,7 +343,7 @@ namespace Google.Protobuf ...@@ -343,7 +343,7 @@ namespace Google.Protobuf
[Test] [Test]
public void CanPushBackEndDocument() public void CanPushBackEndDocument()
{ {
var tokenizer = new JsonTokenizer(new StringReader("null")); var tokenizer = JsonTokenizer.FromTextReader(new StringReader("null"));
Assert.AreEqual(JsonToken.Null, tokenizer.Next()); Assert.AreEqual(JsonToken.Null, tokenizer.Next());
Assert.AreEqual(JsonToken.EndDocument, tokenizer.Next()); Assert.AreEqual(JsonToken.EndDocument, tokenizer.Next());
tokenizer.PushBack(JsonToken.EndDocument); tokenizer.PushBack(JsonToken.EndDocument);
...@@ -373,7 +373,7 @@ namespace Google.Protobuf ...@@ -373,7 +373,7 @@ namespace Google.Protobuf
private static void AssertTokensNoReplacement(string json, params JsonToken[] expectedTokens) private static void AssertTokensNoReplacement(string json, params JsonToken[] expectedTokens)
{ {
var reader = new StringReader(json); var reader = new StringReader(json);
var tokenizer = new JsonTokenizer(reader); var tokenizer = JsonTokenizer.FromTextReader(reader);
for (int i = 0; i < expectedTokens.Length; i++) for (int i = 0; i < expectedTokens.Length; i++)
{ {
var actualToken = tokenizer.Next(); var actualToken = tokenizer.Next();
...@@ -393,7 +393,7 @@ namespace Google.Protobuf ...@@ -393,7 +393,7 @@ namespace Google.Protobuf
private static void AssertThrowsAfter(string json, params JsonToken[] expectedTokens) private static void AssertThrowsAfter(string json, params JsonToken[] expectedTokens)
{ {
var reader = new StringReader(json); var reader = new StringReader(json);
var tokenizer = new JsonTokenizer(reader); var tokenizer = JsonTokenizer.FromTextReader(reader);
for (int i = 0; i < expectedTokens.Length; i++) for (int i = 0; i < expectedTokens.Length; i++)
{ {
var actualToken = tokenizer.Next(); var actualToken = tokenizer.Next();
......
...@@ -808,12 +808,17 @@ namespace Google.Protobuf ...@@ -808,12 +808,17 @@ namespace Google.Protobuf
/// </summary> /// </summary>
public sealed class Settings public sealed class Settings
{ {
private static readonly Settings defaultInstance = new Settings(false);
/// <summary> /// <summary>
/// Default settings, as used by <see cref="JsonFormatter.Default"/> /// Default settings, as used by <see cref="JsonFormatter.Default"/>
/// </summary> /// </summary>
public static Settings Default { get; } = new Settings(false); public static Settings Default { get; }
// Workaround for the Mono compiler complaining about XML comments not being on
// valid language elements.
static Settings()
{
Default = new Settings(false);
}
/// <summary> /// <summary>
/// Whether fields whose values are the default for the field type (e.g. 0 for integers) /// Whether fields whose values are the default for the field type (e.g. 0 for integers)
......
...@@ -77,6 +77,7 @@ namespace Google.Protobuf ...@@ -77,6 +77,7 @@ namespace Google.Protobuf
{ ListValue.Descriptor.FullName, (parser, message, tokenizer) => { ListValue.Descriptor.FullName, (parser, message, tokenizer) =>
parser.MergeRepeatedField(message, message.Descriptor.Fields[ListValue.ValuesFieldNumber], tokenizer) }, parser.MergeRepeatedField(message, message.Descriptor.Fields[ListValue.ValuesFieldNumber], tokenizer) },
{ Struct.Descriptor.FullName, (parser, message, tokenizer) => parser.MergeStruct(message, tokenizer) }, { Struct.Descriptor.FullName, (parser, message, tokenizer) => parser.MergeStruct(message, tokenizer) },
{ Any.Descriptor.FullName, (parser, message, tokenizer) => parser.MergeAny(message, tokenizer) },
{ FieldMask.Descriptor.FullName, (parser, message, tokenizer) => MergeFieldMask(message, tokenizer.Next()) }, { FieldMask.Descriptor.FullName, (parser, message, tokenizer) => MergeFieldMask(message, tokenizer.Next()) },
{ Int32Value.Descriptor.FullName, MergeWrapperField }, { Int32Value.Descriptor.FullName, MergeWrapperField },
{ Int64Value.Descriptor.FullName, MergeWrapperField }, { Int64Value.Descriptor.FullName, MergeWrapperField },
...@@ -128,7 +129,7 @@ namespace Google.Protobuf ...@@ -128,7 +129,7 @@ namespace Google.Protobuf
/// <param name="jsonReader">Reader providing the JSON to parse.</param> /// <param name="jsonReader">Reader providing the JSON to parse.</param>
internal void Merge(IMessage message, TextReader jsonReader) internal void Merge(IMessage message, TextReader jsonReader)
{ {
var tokenizer = new JsonTokenizer(jsonReader); var tokenizer = JsonTokenizer.FromTextReader(jsonReader);
Merge(message, tokenizer); Merge(message, tokenizer);
var lastToken = tokenizer.Next(); var lastToken = tokenizer.Next();
if (lastToken != JsonToken.EndDocument) if (lastToken != JsonToken.EndDocument)
...@@ -338,6 +339,7 @@ namespace Google.Protobuf ...@@ -338,6 +339,7 @@ namespace Google.Protobuf
/// <exception cref="InvalidProtocolBufferException">The JSON does not represent a Protocol Buffers message correctly</exception> /// <exception cref="InvalidProtocolBufferException">The JSON does not represent a Protocol Buffers message correctly</exception>
public T Parse<T>(string json) where T : IMessage, new() public T Parse<T>(string json) where T : IMessage, new()
{ {
Preconditions.CheckNotNull(json, nameof(json));
return Parse<T>(new StringReader(json)); return Parse<T>(new StringReader(json));
} }
...@@ -350,11 +352,42 @@ namespace Google.Protobuf ...@@ -350,11 +352,42 @@ namespace Google.Protobuf
/// <exception cref="InvalidProtocolBufferException">The JSON does not represent a Protocol Buffers message correctly</exception> /// <exception cref="InvalidProtocolBufferException">The JSON does not represent a Protocol Buffers message correctly</exception>
public T Parse<T>(TextReader jsonReader) where T : IMessage, new() public T Parse<T>(TextReader jsonReader) where T : IMessage, new()
{ {
Preconditions.CheckNotNull(jsonReader, nameof(jsonReader));
T message = new T(); T message = new T();
Merge(message, jsonReader); Merge(message, jsonReader);
return message; return message;
} }
/// <summary>
/// Parses <paramref name="json"/> into a new message.
/// </summary>
/// <param name="json">The JSON to parse.</param>
/// <param name="descriptor">Descriptor of message type to parse.</param>
/// <exception cref="InvalidJsonException">The JSON does not comply with RFC 7159</exception>
/// <exception cref="InvalidProtocolBufferException">The JSON does not represent a Protocol Buffers message correctly</exception>
public IMessage Parse(string json, MessageDescriptor descriptor)
{
Preconditions.CheckNotNull(json, nameof(json));
Preconditions.CheckNotNull(descriptor, nameof(descriptor));
return Parse(new StringReader(json), descriptor);
}
/// <summary>
/// Parses JSON read from <paramref name="jsonReader"/> into a new message.
/// </summary>
/// <param name="jsonReader">Reader providing the JSON to parse.</param>
/// <param name="descriptor">Descriptor of message type to parse.</param>
/// <exception cref="InvalidJsonException">The JSON does not comply with RFC 7159</exception>
/// <exception cref="InvalidProtocolBufferException">The JSON does not represent a Protocol Buffers message correctly</exception>
public IMessage Parse(TextReader jsonReader, MessageDescriptor descriptor)
{
Preconditions.CheckNotNull(jsonReader, nameof(jsonReader));
Preconditions.CheckNotNull(descriptor, nameof(descriptor));
IMessage message = descriptor.Parser.CreateTemplate();
Merge(message, jsonReader);
return message;
}
private void MergeStructValue(IMessage message, JsonTokenizer tokenizer) private void MergeStructValue(IMessage message, JsonTokenizer tokenizer)
{ {
var firstToken = tokenizer.Next(); var firstToken = tokenizer.Next();
...@@ -410,6 +443,83 @@ namespace Google.Protobuf ...@@ -410,6 +443,83 @@ namespace Google.Protobuf
MergeMapField(message, field, tokenizer); MergeMapField(message, field, tokenizer);
} }
private void MergeAny(IMessage message, JsonTokenizer tokenizer)
{
// Record the token stream until we see the @type property. At that point, we can take the value, consult
// the type registry for the relevant message, and replay the stream, omitting the @type property.
var tokens = new List<JsonToken>();
var token = tokenizer.Next();
if (token.Type != JsonToken.TokenType.StartObject)
{
throw new InvalidProtocolBufferException("Expected object value for Any");
}
int typeUrlObjectDepth = tokenizer.ObjectDepth;
// The check for the property depth protects us from nested Any values which occur before the type URL
// for *this* Any.
while (token.Type != JsonToken.TokenType.Name ||
token.StringValue != JsonFormatter.AnyTypeUrlField ||
tokenizer.ObjectDepth != typeUrlObjectDepth)
{
tokens.Add(token);
token = tokenizer.Next();
}
// Don't add the @type property or its value to the recorded token list
token = tokenizer.Next();
if (token.Type != JsonToken.TokenType.StringValue)
{
throw new InvalidProtocolBufferException("Expected string value for Any.@type");
}
string typeUrl = token.StringValue;
string typeName = JsonFormatter.GetTypeName(typeUrl);
MessageDescriptor descriptor = settings.TypeRegistry.Find(typeName);
if (descriptor == null)
{
throw new InvalidOperationException($"Type registry has no descriptor for type name '{typeName}'");
}
// Now replay the token stream we've already read and anything that remains of the object, just parsing it
// as normal. Our original tokenizer should end up at the end of the object.
var replay = JsonTokenizer.FromReplayedTokens(tokens, tokenizer);
var body = descriptor.Parser.CreateTemplate();
if (descriptor.IsWellKnownType)
{
MergeWellKnownTypeAnyBody(body, replay);
}
else
{
Merge(body, replay);
}
var data = body.ToByteString();
// Now that we have the message data, we can pack it into an Any (the message received as a parameter).
message.Descriptor.Fields[Any.TypeUrlFieldNumber].Accessor.SetValue(message, typeUrl);
message.Descriptor.Fields[Any.ValueFieldNumber].Accessor.SetValue(message, data);
}
// Well-known types end up in a property called "value" in the JSON. As there's no longer a @type property
// in the given JSON token stream, we should *only* have tokens of start-object, name("value"), the value
// itself, and then end-object.
private void MergeWellKnownTypeAnyBody(IMessage body, JsonTokenizer tokenizer)
{
var token = tokenizer.Next(); // Definitely start-object; checked in previous method
token = tokenizer.Next();
// TODO: What about an absent Int32Value, for example?
if (token.Type != JsonToken.TokenType.Name || token.StringValue != JsonFormatter.AnyWellKnownTypeValueField)
{
throw new InvalidProtocolBufferException($"Expected '{JsonFormatter.AnyWellKnownTypeValueField}' property for well-known type Any body");
}
Merge(body, tokenizer);
token = tokenizer.Next();
if (token.Type != JsonToken.TokenType.EndObject)
{
throw new InvalidProtocolBufferException($"Expected end-object token after @type/value for well-known type");
}
}
#region Utility methods which don't depend on the state (or settings) of the parser. #region Utility methods which don't depend on the state (or settings) of the parser.
private static object ParseMapKey(FieldDescriptor field, string keyText) private static object ParseMapKey(FieldDescriptor field, string keyText)
{ {
...@@ -789,29 +899,48 @@ namespace Google.Protobuf ...@@ -789,29 +899,48 @@ namespace Google.Protobuf
/// </summary> /// </summary>
public sealed class Settings public sealed class Settings
{ {
private static readonly Settings defaultInstance = new Settings(CodedInputStream.DefaultRecursionLimit);
private readonly int recursionLimit;
/// <summary> /// <summary>
/// Default settings, as used by <see cref="JsonParser.Default"/> /// Default settings, as used by <see cref="JsonParser.Default"/>. This has the same default
/// recursion limit as <see cref="CodedInputStream"/>, and an empty type registry.
/// </summary> /// </summary>
public static Settings Default { get { return defaultInstance; } } public static Settings Default { get; }
// Workaround for the Mono compiler complaining about XML comments not being on
// valid language elements.
static Settings()
{
Default = new Settings(CodedInputStream.DefaultRecursionLimit);
}
/// <summary> /// <summary>
/// The maximum depth of messages to parse. Note that this limit only applies to parsing /// The maximum depth of messages to parse. Note that this limit only applies to parsing
/// messages, not collections - so a message within a collection within a message only counts as /// messages, not collections - so a message within a collection within a message only counts as
/// depth 2, not 3. /// depth 2, not 3.
/// </summary> /// </summary>
public int RecursionLimit { get { return recursionLimit; } } public int RecursionLimit { get; }
/// <summary>
/// The type registry used to parse <see cref="Any"/> messages.
/// </summary>
public TypeRegistry TypeRegistry { get; }
/// <summary> /// <summary>
/// Creates a new <see cref="Settings"/> object with the specified recursion limit. /// Creates a new <see cref="Settings"/> object with the specified recursion limit.
/// </summary> /// </summary>
/// <param name="recursionLimit">The maximum depth of messages to parse</param> /// <param name="recursionLimit">The maximum depth of messages to parse</param>
public Settings(int recursionLimit) public Settings(int recursionLimit) : this(recursionLimit, TypeRegistry.Empty)
{
}
/// <summary>
/// Creates a new <see cref="Settings"/> object with the specified recursion limit and type registry.
/// </summary>
/// <param name="recursionLimit">The maximum depth of messages to parse</param>
/// <param name="typeRegistry">The type registry used to parse <see cref="Any"/> messages</param>
public Settings(int recursionLimit, TypeRegistry typeRegistry)
{ {
this.recursionLimit = recursionLimit; RecursionLimit = recursionLimit;
TypeRegistry = Preconditions.CheckNotNull(typeRegistry, nameof(typeRegistry));
} }
} }
} }
......
...@@ -47,32 +47,38 @@ namespace Google.Protobuf ...@@ -47,32 +47,38 @@ namespace Google.Protobuf
/// between values. It validates the token stream as it goes - so callers can assume that the /// between values. It validates the token stream as it goes - so callers can assume that the
/// tokens it produces are appropriate. For example, it would never produce "start object, end array." /// tokens it produces are appropriate. For example, it would never produce "start object, end array."
/// </para> /// </para>
/// <para>Implementation details: the base class handles single token push-back and </para>
/// <para>Not thread-safe.</para> /// <para>Not thread-safe.</para>
/// </remarks> /// </remarks>
internal sealed class JsonTokenizer internal abstract class JsonTokenizer
{ {
// The set of states in which a value is valid next token.
private static readonly State ValueStates = State.ArrayStart | State.ArrayAfterComma | State.ObjectAfterColon | State.StartOfDocument;
private readonly Stack<ContainerType> containerStack = new Stack<ContainerType>();
private readonly PushBackReader reader;
private JsonToken bufferedToken; private JsonToken bufferedToken;
private State state;
private int objectDepth = 0;
/// <summary> /// <summary>
/// Returns the depth of the stack, purely in objects (not collections). /// Creates a tokenizer that reads from the given text reader.
/// Informally, this is the number of remaining unclosed '{' characters we have.
/// </summary> /// </summary>
internal int ObjectDepth { get { return objectDepth; } } internal static JsonTokenizer FromTextReader(TextReader reader)
{
return new JsonTextTokenizer(reader);
}
internal JsonTokenizer(TextReader reader) /// <summary>
/// Creates a tokenizer that first replays the given list of tokens, then continues reading
/// from another tokenizer. Note that if the returned tokenizer is "pushed back", that does not push back
/// on the continuation tokenizer, or vice versa. Care should be taken when using this method - it was
/// created for the sake of Any parsing.
/// </summary>
internal static JsonTokenizer FromReplayedTokens(IList<JsonToken> tokens, JsonTokenizer continuation)
{ {
this.reader = new PushBackReader(reader); return new JsonReplayTokenizer(tokens, continuation);
state = State.StartOfDocument;
containerStack.Push(ContainerType.Document);
} }
/// <summary>
/// Returns the depth of the stack, purely in objects (not collections).
/// Informally, this is the number of remaining unclosed '{' characters we have.
/// </summary>
internal int ObjectDepth { get; private set; }
// TODO: Why do we allow a different token to be pushed back? It might be better to always remember the previous // TODO: Why do we allow a different token to be pushed back? It might be better to always remember the previous
// token returned, and allow a parameterless Rewind() method (which could only be called once, just like the current PushBack). // token returned, and allow a parameterless Rewind() method (which could only be called once, just like the current PushBack).
internal void PushBack(JsonToken token) internal void PushBack(JsonToken token)
...@@ -84,11 +90,11 @@ namespace Google.Protobuf ...@@ -84,11 +90,11 @@ namespace Google.Protobuf
bufferedToken = token; bufferedToken = token;
if (token.Type == JsonToken.TokenType.StartObject) if (token.Type == JsonToken.TokenType.StartObject)
{ {
objectDepth--; ObjectDepth--;
} }
else if (token.Type == JsonToken.TokenType.EndObject) else if (token.Type == JsonToken.TokenType.EndObject)
{ {
objectDepth++; ObjectDepth++;
} }
} }
...@@ -96,32 +102,95 @@ namespace Google.Protobuf ...@@ -96,32 +102,95 @@ namespace Google.Protobuf
/// Returns the next JSON token in the stream. An EndDocument token is returned to indicate the end of the stream, /// Returns the next JSON token in the stream. An EndDocument token is returned to indicate the end of the stream,
/// after which point <c>Next()</c> should not be called again. /// after which point <c>Next()</c> should not be called again.
/// </summary> /// </summary>
/// <remarks> /// <remarks>This implementation provides single-token buffering, and calls <see cref="NextImpl"/> if there is no buffered token.</remarks>
/// This method essentially just loops through characters skipping whitespace, validating and
/// changing state (e.g. from ObjectBeforeColon to ObjectAfterColon)
/// until it reaches something which will be a genuine token (e.g. a start object, or a value) at which point
/// it returns the token. Although the method is large, it would be relatively hard to break down further... most
/// of it is the large switch statement, which sometimes returns and sometimes doesn't.
/// </remarks>
/// <returns>The next token in the stream. This is never null.</returns> /// <returns>The next token in the stream. This is never null.</returns>
/// <exception cref="InvalidOperationException">This method is called after an EndDocument token has been returned</exception> /// <exception cref="InvalidOperationException">This method is called after an EndDocument token has been returned</exception>
/// <exception cref="InvalidJsonException">The input text does not comply with RFC 7159</exception> /// <exception cref="InvalidJsonException">The input text does not comply with RFC 7159</exception>
internal JsonToken Next() internal JsonToken Next()
{ {
JsonToken tokenToReturn;
if (bufferedToken != null) if (bufferedToken != null)
{ {
var ret = bufferedToken; tokenToReturn = bufferedToken;
bufferedToken = null; bufferedToken = null;
if (ret.Type == JsonToken.TokenType.StartObject) }
else
{ {
objectDepth++; tokenToReturn = NextImpl();
} }
else if (ret.Type == JsonToken.TokenType.EndObject) if (tokenToReturn.Type == JsonToken.TokenType.StartObject)
{ {
objectDepth--; ObjectDepth++;
} }
return ret; else if (tokenToReturn.Type == JsonToken.TokenType.EndObject)
{
ObjectDepth--;
}
return tokenToReturn;
} }
/// <summary>
/// Returns the next JSON token in the stream, when requested by the base class. (The <see cref="Next"/> method delegates
/// to this if it doesn't have a buffered token.)
/// </summary>
/// <exception cref="InvalidOperationException">This method is called after an EndDocument token has been returned</exception>
/// <exception cref="InvalidJsonException">The input text does not comply with RFC 7159</exception>
protected abstract JsonToken NextImpl();
/// <summary>
/// Tokenizer which first exhausts a list of tokens, then consults another tokenizer.
/// </summary>
private class JsonReplayTokenizer : JsonTokenizer
{
private readonly IList<JsonToken> tokens;
private readonly JsonTokenizer nextTokenizer;
private int nextTokenIndex;
internal JsonReplayTokenizer(IList<JsonToken> tokens, JsonTokenizer nextTokenizer)
{
this.tokens = tokens;
this.nextTokenizer = nextTokenizer;
}
// FIXME: Object depth not maintained...
protected override JsonToken NextImpl()
{
if (nextTokenIndex >= tokens.Count)
{
return nextTokenizer.Next();
}
return tokens[nextTokenIndex++];
}
}
/// <summary>
/// Tokenizer which does all the *real* work of parsing JSON.
/// </summary>
private sealed class JsonTextTokenizer : JsonTokenizer
{
// The set of states in which a value is valid next token.
private static readonly State ValueStates = State.ArrayStart | State.ArrayAfterComma | State.ObjectAfterColon | State.StartOfDocument;
private readonly Stack<ContainerType> containerStack = new Stack<ContainerType>();
private readonly PushBackReader reader;
private State state;
internal JsonTextTokenizer(TextReader reader)
{
this.reader = new PushBackReader(reader);
state = State.StartOfDocument;
containerStack.Push(ContainerType.Document);
}
/// <remarks>
/// This method essentially just loops through characters skipping whitespace, validating and
/// changing state (e.g. from ObjectBeforeColon to ObjectAfterColon)
/// until it reaches something which will be a genuine token (e.g. a start object, or a value) at which point
/// it returns the token. Although the method is large, it would be relatively hard to break down further... most
/// of it is the large switch statement, which sometimes returns and sometimes doesn't.
/// </remarks>
protected override JsonToken NextImpl()
{
if (state == State.ReaderExhausted) if (state == State.ReaderExhausted)
{ {
throw new InvalidOperationException("Next() called after end of document"); throw new InvalidOperationException("Next() called after end of document");
...@@ -167,12 +236,10 @@ namespace Google.Protobuf ...@@ -167,12 +236,10 @@ namespace Google.Protobuf
ValidateState(ValueStates, "Invalid state to read an open brace: "); ValidateState(ValueStates, "Invalid state to read an open brace: ");
state = State.ObjectStart; state = State.ObjectStart;
containerStack.Push(ContainerType.Object); containerStack.Push(ContainerType.Object);
objectDepth++;
return JsonToken.StartObject; return JsonToken.StartObject;
case '}': case '}':
ValidateState(State.ObjectAfterProperty | State.ObjectStart, "Invalid state to read a close brace: "); ValidateState(State.ObjectAfterProperty | State.ObjectStart, "Invalid state to read a close brace: ");
PopContainer(); PopContainer();
objectDepth--;
return JsonToken.EndObject; return JsonToken.EndObject;
case '[': case '[':
ValidateState(ValueStates, "Invalid state to read an open square bracket: "); ValidateState(ValueStates, "Invalid state to read an open square bracket: ");
...@@ -667,4 +734,5 @@ namespace Google.Protobuf ...@@ -667,4 +734,5 @@ namespace Google.Protobuf
} }
} }
} }
}
} }
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment