Measures against garbled characters in Multipart Request with Quarkus

I got a little fit, so make a note.

TL;DR

This will fix the garbled characters

<dependency>
  <groupId>io.quarkus</groupId>
  <artifactId>quarkus-undertow</artifactId>
</dependency>

@Provider
public class ContentTypeSetterPreProcessorInterceptor implements ContainerRequestFilter {
    private @Context HttpServletRequest request;

    @Override
    public void filter(ContainerRequestContext requestContext) {
        request.setAttribute(InputPart.DEFAULT_CHARSET_PROPERTY, StandardCharsets.UTF_8.name());
    }
}

background

I made a file upload function with Vue.js and Quarkus, but there was a phenomenon that the titles linked in the same form were garbled. The code I wrote is as follows.

Client side. I'm using new FormData () to use multipart / form-data.

let data = new FormData();
data.append("title", this.item.title);
data.append("myfile", this.item.files[0]);

this.axios.post(uri, data).then(() => {
    this.$swal({
        icon: "success",
        text: "Upload Success!"
    });
})

Server side. Use JAX-RS @ MultipartForm. Both titles and files are automatically mapped to beans.

public class MyFileFormBean {
    @FormParam("title")
    @PartType(MediaType.TEXT_PLAIN)
    private String title = null;

    @FormParam("myfile")
    @PartType(MediaType.APPLICATION_OCTET_STREAM)
    private byte[] myfile = null;
....

JAX-RS side.

@POST
@Consumes({MediaType.MULTIPART_FORM_DATA})
@Produces(MediaType.APPLICATION_JSON)
public Response upload(@MultipartForm MyFileFormBean form) throws IOException {
    System.out.println("title: " + form.getTitle());
...
}

I was planning to do this, but for some reason the Japanese characters were garbled. .. ..

title: ????????????

Cause

It seems that RESTEasy converts to ASCII (ISO-8859-1) by default when the character code is not included. It seems that you should specify ʻaccept-charset` in Form, but it seems a little troublesome to do it with JS, so it seems better to forcibly convert it on the server.

After googled there was a sample implementation in JBoss Community.

@Provider
@ServerInterceptor
public class ContentTypeSetterPreProcessorInterceptor implements PreProcessInterceptor {
    public ServerResponse preProcess(HttpRequest request, ResourceMethod method)
            throws Failure, WebApplicationException {
        request.setAttribute(InputPart.DEFAULT_CHARSET_PROPERTY, "UTF-8");
        return null;
    }
}

However, RestEasy used in Quarkus is JAX-RS 2.0 compliant, so the above writing method is not possible. So you need to use ContainerRequestFilter instead of PreProcessInterceptor.

Countermeasures

Rewrite according to JAX-RS 2.0 while referring to the sample. By default, Quarkus uses Vert.x HTTP server as described in Servlet compatibility, so HttpServletRequest cannot be used. So add quarkus-undertow as a dependency.

<dependency>
  <groupId>io.quarkus</groupId>
  <artifactId>quarkus-undertow</artifactId>
</dependency>

The following code hooks a request based on JAX-RS 2.0 and forcibly converts DEFAULT_CHARSET_PROPERTY to ʻUTF-8`.

@Provider
public class ContentTypeSetterPreProcessorInterceptor implements ContainerRequestFilter {
    private @Context HttpServletRequest request;

    @Override
    public void filter(ContainerRequestContext requestContext) {
        request.setAttribute(InputPart.DEFAULT_CHARSET_PROPERTY, StandardCharsets.UTF_8.name());
    }
}

If you set @Provider, you don't need to add a configuration file. This solved the garbled characters.

Summary

I used to write a filter because it is garbled in JavaServlet, but I didn't think I would do it even in modern times, so I got into it a little. File uploads are still full of traps. Although it is different from RFC, UTF-8 has the same code in the ASCII area, and I feel that UTF-8 may be the default.

The issue on GitHub also mentions "UTF-8 encoding problem with MultiPart / RestEasy", so you can easily switch with the configuration file. It may be in the future.

Then Happy Hacking!