Spring WebClient：如何将 HTTP 响应主体转换为特定编码的字符串

Question

I am using WebClient of the Spring webflux module for the first time.

The idea is to query a certain URL which gives me a CSV in shift-jis encoding.

String content = webClient
    .method(HttpMethod.GET)
    .acceptCharset(Charset.forName("shift-jis")) 
    .uri(uri)
    .accept(MediaType.TEXT_PLAIN, new MediaType("text", "csv"))
    .retrieve()
    .toEntity(String.class)
    .mapNotNull(HttpEntity::getBody)
    .block();

This works somewhat, but the content String looks like it is in a wrong encoding - probably UTF-8.

Alternatively, an Apache HttpClient version seems to work:

HttpClient client = HttpClientBuilder.create().build();
HttpGet get = new HttpGet(uri);
String content = IOUtils.toString(
    client.execute(get).getEntity().getContent(), 
    Charset.forName("shift-jis"));

So I am wondering how one could interfere with the conversion happening in toEntity . I tried adding this to the webClient call:

.codecs(configurer -> {
   StringDecoder decoder = StringDecoder.textPlainOnly();
   decoder.setDefaultCharset(Charset.forName("shift-jis"));
   configurer.customCodecs().registerWithDefaultConfig(decoder);
})

Alas, this does not seem to be picked up. Setting a breakpoint in StringDecoder#decode and manually overwriting the defaultCharset field during debugging, however, does help.

Answer 1

Ah... Just after posting I realized: Maybe my custom text/csv mime type is not really interpreted as textPlainOnly .

And, indeed:

StringDecoder decoder = StringDecoder.allMimeTypes();

apparently does the trick.

Leaving the question open in case my approach is awfully hacky and there are better/correct ways to tackle this issue.