apache/flink · error · NullPointerException

Reducer may not be null.

Error message

Reducer may not be null.

What it means

Thrown by the HadoopReducerWrappedFunction constructor when wrapping a Hadoop mapred Reducer into a Flink GroupReduceFunction. The wrapper requires a non-null Reducer because the whole point of the class is to delegate reduce() calls to it; a null reducer has no behavior to invoke. It surfaces a caller bug (passing null) rather than a runtime state problem.

Source

Thrown at flink-connectors/flink-hadoop-compatibility/src/main/java/org/apache/flink/hadoopcompatibility/mapred/HadoopReducerWrappedFunction.java:90

    /**
     * Maps a Hadoop Reducer (mapred API) to a non-combinable Flink GroupReduceFunction.
     *
     * @param hadoopReducer The Hadoop Reducer to wrap.
     */
    public HadoopReducerWrappedFunction(Reducer<KEYIN, VALUEIN, KEYOUT, VALUEOUT> hadoopReducer) {
        this(hadoopReducer, new JobConf());
    }

    /**
     * Maps a Hadoop Reducer (mapred API) to a non-combinable Flink GroupReduceFunction.
     *
     * @param hadoopReducer The Hadoop Reducer to wrap.
     * @param conf The JobConf that is used to configure the Hadoop Reducer.
     */
    public HadoopReducerWrappedFunction(
            Reducer<KEYIN, VALUEIN, KEYOUT, VALUEOUT> hadoopReducer, JobConf conf) {
        if (hadoopReducer == null) {
            throw new NullPointerException("Reducer may not be null.");
        }
        if (conf == null) {
            throw new NullPointerException("JobConf may not be null.");
        }

        this.reducer = hadoopReducer;
        this.jobConf = conf;
    }

    @SuppressWarnings("unchecked")
    @PublicEvolving
    @Override
    public void open(OpenContext openContext) throws Exception {
        super.open(openContext);
        this.reducer.configure(jobConf);

        this.reporter = new HadoopDummyReporter();
        this.reduceCollector = new HadoopOutputCollector<KEYOUT, VALUEOUT>();

View on GitHub (pinned to 2f3c205e92)

Solutions

1. Ensure the org.apache.hadoop.mapred.Reducer instance is instantiated and non-null before passing it to the constructor.
2. If loading the reducer class from config, validate the configured class name resolves and instantiate it inside a try/catch, propagating a clear error before reaching the wrapper.
3. Prefer using HadoopRedistributeFunction.wrap/reduce helpers from flink-hadoop-compatibility which build the wrapper for you.

Example fix

// before
new HadoopReducerWrappedFunction<>(nullReducer, jobConf);
// after
Reducer<KIn,VIn,KOut,VOut> reducer =
    (Reducer<KIn,VIn,KOut,VOut>) jobConf.getClass("mapred.reducer.class", null, Reducer.class)
        .getDeclaredConstructor().newInstance();
if (reducer == null) {
    throw new IllegalArgumentException("mapred.reducer.class not set in JobConf");
}
new HadoopReducerWrappedFunction<>(reducer, jobConf);

Objects.requireNonNull(hadoopReducer, "Reducer must not be null");
new HadoopReducerWrappedFunction<>(hadoopReducer, conf);

// java: null check is the guard
Reducer<KIn,VIn,KOut,VOut> r = hadoopReducer;
if (r == null) throw new IllegalArgumentException("reducer is null");

When it happens

Trigger: Constructing new HadoopReducerWrappedFunction(null, conf) or new HadoopReducerWrappedFunction(null) (the one-arg overload delegates to the two-arg with a new JobConf). Happens when a Reducer instance is created via reflection/config that silently returned null.

Common situations: HadoopCompatibility wrappers built dynamically from a JobConf where conf.getClass(...).newInstance() failed or the configured reducer class name was wrong and resolved to null; migration code that conditionally supplies a reducer.

Related errors

• JobConf may not be null.
• The collection contains a null element
• The collection contains a null element
• Hadoop input split must not be null
• Hadoop JobConf must not be null when input split is configur

AI-assisted analysis of apache/flink@2f3c205e92 (2026-08-14). Data as JSON: /api/errors/0fc9a525080b2196. Report an issue: GitHub.