SparkJava request.body() consumes payload in first defined before() filter

[balodza]

To help us debug your issue fill in the basic information below using the options provided

Serverless Java Container version: 1.5.1

Implementations: Spark

Framework version: 2.9.2

Frontend service: REST API

Deployment method: Lambda, Request is sent with AwsProxyRequest

Scenario

Previously we were using combination : SparkJava 2.9.2 + AwsServerless 1.3.1 and everything was working fine.
Just few days ago we tried to upgrade AwsServerless to 1.5.1 and noticed that POST works incorrectly.

Expected behavior

request.body() should be available for getting any number of times in filters, handlers.
It was working fine in 1.3.1

Actual behavior

request.body() returns value only in first specified filter. Then it returns "" (empty String)
It stopped working after we updated to 1.5.1

Steps to reproduce

To reproduce the issue, simply run RequestStreamLambdaHandler.main() with no any prerequisites.
Just check the version in build.gradle before debug.

apply plugin: 'java'
apply plugin: 'idea'

repositories {
    mavenCentral()
}

dependencies {
    compile 'com.sparkjava:spark-core:2.9.2'
    //compile 'com.amazonaws.serverless:aws-serverless-java-container-spark:1.3.1'
    compile 'com.amazonaws.serverless:aws-serverless-java-container-spark:1.5.1'
}

RequestStreamLambdaHandler.java

package hello;

import com.amazonaws.serverless.exceptions.ContainerInitializationException;
import com.amazonaws.serverless.proxy.model.AwsProxyRequest;
import com.amazonaws.serverless.proxy.model.AwsProxyResponse;
import com.amazonaws.serverless.proxy.spark.SparkLambdaContainerHandler;
import com.amazonaws.services.lambda.runtime.Context;
import com.amazonaws.services.lambda.runtime.RequestStreamHandler;

import java.io.ByteArrayInputStream;
import java.io.ByteArrayOutputStream;
import java.io.IOException;
import java.io.InputStream;
import java.io.OutputStream;

import spark.Request;
import spark.Response;
import spark.Spark;

import static spark.Spark.before;
import static spark.Spark.post;

public class RequestStreamLambdaHandler implements RequestStreamHandler {
    @Override
    public void handleRequest(InputStream inputStream, OutputStream outputStream, Context context) throws IOException {
        SparkLambdaContainerHandler<AwsProxyRequest, AwsProxyResponse> handler;
        try {
                handler = SparkLambdaContainerHandler.getAwsProxyHandler();
                defineBaseResources();                
                Spark.awaitInitialization();
            } catch (ContainerInitializationException e) {
                return;
            }

        handler.proxyStream(inputStream, outputStream, context);
    }

    public void defineBaseResources() {
        // define base routes
        before("/*", this::beforeFilter1);
        before("/*", this::beforeFilter2);
        before("/*", this::beforeFilter3);
        
        post("/post-request", this::handlePost);        
    }

    private Request handlePost(Request req, Response res) {
        System.out.println("POST Handler Request Body: " + req.body());
        return req;
    }

    protected void beforeFilter1(Request req, Response res) {
        System.out.println("Filter 1 Request Body: " + req.body());
    }

    protected void beforeFilter2(Request req, Response res) {
        System.out.println("Filter 2 Request Body: " + req.body());
    }

    protected void beforeFilter3(Request req, Response res) {
        System.out.println("Filter 3 Request Body: " + req.body());
    }

    public static void main(String[] args) throws Exception {
        String body = "{\"path\": \"/post-request\","
                + "    \"httpMethod\": \"POST\","
                + "    \"requestContext\": {"
                + "        \"identity\": {}"              
                + "    },"
                + "    \"body\":\"{\\\"field1\\\":\\\"value1\\\"}\""                
                + "}";
        
        
        RequestStreamLambdaHandler handler = new RequestStreamLambdaHandler();
        final ByteArrayOutputStream baos = new ByteArrayOutputStream();
        handler.handleRequest(new ByteArrayInputStream(body.getBytes()), baos, null);
    }
}

Full log output

Paste the full log output from the Lambda function's CloudWatch logs

Log while 1.3.1 is used :

> Task :RequestStreamLambdaHandler.main()
Filter 1 Request Body: {"field1":"value1"}
Filter 2 Request Body: {"field1":"value1"}
Filter 3 Request Body: {"field1":"value1"}
POST Handler Request Body: {"field1":"value1"}

Log while 1.5.1 is used :

> Task :RequestStreamLambdaHandler.main()
Filter 1 Request Body: {"field1":"value1"}
Filter 2 Request Body: 
Filter 3 Request Body: 
POST Handler Request Body: 

[sapessi]

Thanks for the report. We're digging into it. We did recently change the behavior to only allow access to the input stream once to follow the servlet specs (see javadoc: Either this method or getReader() may be called to read the body, not both).

To get around this, people tend to wrap the request themselves and cache the request data (see StackOverflow answer). Looks like what you are doing works with SparkJava because the embedded server is doing that for you and the older version of this library was creating a new input stream every time the getInputStream() method was called.

A reasonable solution would be to use the same request wrapper from the SparkJava embedded server to wrap the HttpServletRequest implementation provided by this library for SparkJava specifically since it relies on that behavior. Does that make sense?

[balodza]

Hello, @sapessi

I didn't dive too deep to understand how integration works with SparkJava in details.
But I know exact place, where behaviour is different :

class: spark.Request
...

private void readBodyAsBytes() {
	try {
		bodyAsBytes = IOUtils.toByteArray(servletRequest.getInputStream());
	} catch (Exception e) {
		LOG.warn("Exception when reading body", e);
	}
}

in case of aws-serverless 1.3.1 servletRequest.getInputStream() returns AwsProxyHttpServletRequest$AwsServletInputStream
but in case of 1.5.1 there is another implementation :
just a class AwsServletInputStream

and difference is that IOUtils.toByteArray(servletRequest.getInputStream()) from one of them could be called infinite times producing the same payload, and another one returns payload only one time.