关于上面 ekeren 的解决方案，在单线程中使用 PipedOutputStream/PipedInputStream 不是一个好主意，以免溢出缓冲区并导致死锁.ByteArrayOutputStream/ByteArrayInputStream 可以工作，但如果您的 JAXB 类生成多个模式(在不同的命名空间中)，您需要多个 StreamSource.

Regarding ekeren's solution above, it's not a good idea to use PipedOutputStream/PipedInputStream in a single thread, lest you overflow the buffer and cause a deadlock. ByteArrayOutputStream/ByteArrayInputStream works, but if your JAXB classes generate multiple schemas (in different namespaces) you need multiple StreamSources.

我最终得到了这个:

JAXBContext jc = JAXBContext.newInstance(Something.class);
final List<ByteArrayOutputStream> outs = new ArrayList<ByteArrayOutputStream>();
jc.generateSchema(new SchemaOutputResolver(){
    @Override
    public Result createOutput(String namespaceUri, String suggestedFileName) throws IOException {
        ByteArrayOutputStream out = new ByteArrayOutputStream();
        outs.add(out);
        StreamResult streamResult = new StreamResult(out);
        streamResult.setSystemId("");
        return streamResult;
    }});
StreamSource[] sources = new StreamSource[outs.size()];
for (int i=0; i<outs.size(); i++) {
    ByteArrayOutputStream out = outs.get(i);
    // to examine schema: System.out.append(new String(out.toByteArray()));
    sources[i] = new StreamSource(new ByteArrayInputStream(out.toByteArray()),"");
}
SchemaFactory sf = SchemaFactory.newInstance( XMLConstants.W3C_XML_SCHEMA_NS_URI );
m.setSchema(sf.newSchema(sources));
m.marshal(docs, new DefaultHandler());  // performs the schema validation