我正在尝试使用Gson将Guava Range对象序列化为JSON，但是默认序列化失败，并且我不确定如何为该泛型类型正确实现TypeAdapter。
Gson gson = new Gson();
Range<Integer> range = Range.closed(10, 20);
String json = gson.toJson(range);
System.out.println(json);
Range<Integer> range2 = gson.fromJson(json,
new TypeToken<Range<Integer>>(){}.getType());
System.out.println(range2);
assertEquals(range2, range);
这样失败：
{"lowerBound":{"endpoint":10},"upperBound":{"endpoint":20}}
PASSED: typeTokenInterface
FAILED: range
java.lang.RuntimeException: Unable to invoke no-args constructor for
com.google.common.collect.Cut<java.lang.Integer>. Register an
InstanceCreator with Gson for this type may fix this problem.
at com.google.gson.internal.ConstructorConstructor$12.construct(
ConstructorConstructor.java:210)
...
注意，默认序列化实际上会丢失信息-它无法报告端点是打开还是关闭。我希望看到它的序列化类似于toString()，例如[10‥20]然而，仅调用toString()不适用于通用Range实例，因为范围的元素可能不是基元（例如，Joda-Time LocalDate实例）。出于同样的原因，实现自定义TypeAdapter似乎很困难，因为我们不知道如何对端点进行反序列化。
我已经基于为TypeAdaptorFactory提供的模板实现了TypeAdaptorFactory的大部分内容，该模板应该可以使用，但是现在我被泛型卡住了。这是我到目前为止的内容：
Multimap但是public class RangeTypeAdapterFactory implements TypeAdapterFactory {
public <T> TypeAdapter<T> create(Gson gson, TypeToken<T> typeToken) {
Type type = typeToken.getType();
if (typeToken.getRawType() != Range.class
|| !(type instanceof ParameterizedType)) {
return null;
}
Type elementType = ((ParameterizedType) type).getActualTypeArguments()[0];
TypeAdapter<?> elementAdapter = (TypeAdapter<?>)gson.getAdapter(TypeToken.get(elementType));
// Bound mismatch: The generic method newRangeAdapter(TypeAdapter<E>) of type
// GsonUtils.RangeTypeAdapterFactory is not applicable for the arguments
// (TypeAdapter<capture#4-of ?>). The inferred type capture#4-of ? is not a valid
// substitute for the bounded parameter <E extends Comparable<?>>
return (TypeAdapter<T>) newRangeAdapter(elementAdapter);
}
private <E extends Comparable<?>> TypeAdapter<Range<E>> newRangeAdapter(final TypeAdapter<E> elementAdapter) {
return new TypeAdapter<Range<E>>() {
@Override
public void write(JsonWriter out, Range<E> value) throws IOException {
if (value == null) {
out.nullValue();
return;
}
String repr = (value.lowerBoundType() == BoundType.CLOSED ? "[" : "(") +
(value.hasLowerBound() ? elementAdapter.toJson(value.lowerEndpoint()) : "-\u221e") +
'\u2025' +
(value.hasLowerBound() ? elementAdapter.toJson(value.upperEndpoint()) : "+\u221e") +
(value.upperBoundType() == BoundType.CLOSED ? "]" : ")");
out.value(repr);
}
public Range<E> read(JsonReader in) throws IOException {
if (in.peek() == JsonToken.NULL) {
in.nextNull();
return null;
}
String[] endpoints = in.nextString().split("\u2025");
E lower = elementAdapter.fromJson(endpoints[0].substring(1));
E upper = elementAdapter.fromJson(endpoints[1].substring(0,endpoints[1].length()-1));
return Range.range(lower, endpoints[0].charAt(0) == '[' ? BoundType.CLOSED : BoundType.OPEN,
upper, endpoints[1].charAt(endpoints[1].length()-1) == '[' ? BoundType.CLOSED : BoundType.OPEN);
}
};
}
}
行出现编译错误，我现在不知所措。
解决此错误的最佳方法是什么？有没有更好的方法来序列化我所缺少的return (TypeAdapter<T>) newRangeAdapter(elementAdapter);对象？如果我要序列化Range，该怎么办？
令人沮丧的是，Google实用程序库和Google序列化库似乎需要太多胶水才能一起工作：（
感觉上有点像在重新发明轮子，但是比起使Gson表现出来所花费的时间，它的组装和测试要快得多，因此至少目前我将使用以下RangeSet进行序列化[ C0]和Converter *，而不是Gson。
Converter*对于进程间通信，由于两个类都实现了Range，因此Java序列化可能会正常工作。但是我正在序列化到磁盘上以提供更多永久性存储，这意味着我需要一种可以信赖的格式，它不会随着时间的推移而改变。番石榴的序列化RangeSet。
这里是Gson JsonSerializer和JsonDeserializer，它们通常支持Range：/**
* Converter between Range instances and Strings, essentially a custom serializer.
* Ideally we'd let Gson or Guava do this for us, but presently this is cleaner.
*/
public static <T extends Comparable<? super T>> Converter<Range<T>, String> rangeConverter(final Converter<T, String> elementConverter) {
final String NEG_INFINITY = "-\u221e";
final String POS_INFINITY = "+\u221e";
final String DOTDOT = "\u2025";
return new Converter<Range<T>, String>() {
@Override
protected String doForward(Range<T> range) {
return (range.hasLowerBound() && range.lowerBoundType() == BoundType.CLOSED ? "[" : "(") +
(range.hasLowerBound() ? elementConverter.convert(range.lowerEndpoint()) : NEG_INFINITY) +
DOTDOT +
(range.hasUpperBound() ? elementConverter.convert(range.upperEndpoint()) : POS_INFINITY) +
(range.hasUpperBound() && range.upperBoundType() == BoundType.CLOSED ? "]" : ")");
}
@Override
protected Range<T> doBackward(String range) {
String[] endpoints = range.split(DOTDOT);
Range<T> ret = Range.all();
if(!endpoints[0].substring(1).equals(NEG_INFINITY)) {
T lower = elementConverter.reverse().convert(endpoints[0].substring(1));
ret = ret.intersection(Range.downTo(lower, endpoints[0].charAt(0) == '[' ? BoundType.CLOSED : BoundType.OPEN));
}
if(!endpoints[1].substring(0,endpoints[1].length()-1).equals(POS_INFINITY)) {
T upper = elementConverter.reverse().convert(endpoints[1].substring(0,endpoints[1].length()-1));
ret = ret.intersection(Range.upTo(upper, endpoints[1].charAt(endpoints[1].length()-1) == ']' ? BoundType.CLOSED : BoundType.OPEN));
}
return ret;
}
};
}
/**
* Converter between RangeSet instances and Strings, essentially a custom serializer.
* Ideally we'd let Gson or Guava do this for us, but presently this is cleaner.
*/
public static <T extends Comparable<? super T>> Converter<RangeSet<T>, String> rangeSetConverter(final Converter<T, String> elementConverter) {
return new Converter<RangeSet<T>, String>() {
private final Converter<Range<T>, String> rangeConverter = rangeConverter(elementConverter);
@Override
protected String doForward(RangeSet<T> rs) {
ArrayList<String> ls = new ArrayList<>();
for(Range<T> range : rs.asRanges()) {
ls.add(rangeConverter.convert(range));
}
return Joiner.on(", ").join(ls);
}
@Override
protected RangeSet<T> doBackward(String rs) {
Iterable<String> parts = Splitter.on(",").trimResults().split(rs);
ImmutableRangeSet.Builder<T> build = ImmutableRangeSet.builder();
for(String range : parts) {
build.add(rangeConverter.reverse().convert(range));
}
return build.build();
}
};
}
Serializable这里是直接解决方案。效果很好
doesn't provide that guarantee