apache/flink · error · IllegalArgumentException

Unsupported data type in schema:

Error message

Unsupported data type in schema: 

What it means

PbCodegenUtils generates the RowData accessor call (getInt/getString/getRow/...) by switching on the element's type root. The default branch throws IllegalArgumentException for type roots with no RowData accessor mapping — effectively any type the protobuf codegen does not support appearing where an accessor must be generated.

Source

Thrown at flink-formats/flink-protobuf/src/main/java/org/apache/flink/formats/protobuf/util/PbCodegenUtils.java:71

            case DOUBLE:
                return flinkContainerCode + ".getDouble(" + index + ")";
            case BOOLEAN:
                return flinkContainerCode + ".getBoolean(" + index + ")";
            case VARCHAR:
            case CHAR:
                return flinkContainerCode + ".getString(" + index + ")";
            case VARBINARY:
            case BINARY:
                return flinkContainerCode + ".getBinary(" + index + ")";
            case ROW:
                int size = eleType.getChildren().size();
                return flinkContainerCode + ".getRow(" + index + ", " + size + ")";
            case MAP:
                return flinkContainerCode + ".getMap(" + index + ")";
            case ARRAY:
                return flinkContainerCode + ".getArray(" + index + ")";
            default:
                throw new IllegalArgumentException("Unsupported data type in schema: " + eleType);
        }
    }

    /**
     * Get java type str from {@link FieldDescriptor} which directly fetched from protobuf object.
     *
     * @return The returned code phrase will be used as java type str in codegen sections.
     * @throws PbCodegenException
     */
    public static String getTypeStrFromProto(FieldDescriptor fd, boolean isList)
            throws PbCodegenException {
        String typeStr;
        switch (fd.getJavaType()) {
            case MESSAGE:
                if (fd.isMapField()) {
                    // map
                    FieldDescriptor keyFd =
                            fd.getMessageType().findFieldByName(PbConstant.PB_MAP_KEY_NAME);

View on GitHub (pinned to 2f3c205e92)

Solutions

1. Flatten or cast the offending nested field to supported types (primitives, ROW, ARRAY, MAP).
2. Move computed/unsupported columns out of the protobuf-mapped part of the plan.
3. Search the exception's stack for the calling codegen class to identify which field triggered accessor generation.

void assertAccessorExists(LogicalType t) {
    switch (t.getTypeRoot()) {
        case MULTISET: case STRUCTURED_TYPE: case DISTINCT_TYPE: case RAW:
            throw new IllegalArgumentException("No RowData accessor codegen for " + t);
        default:
    }
}

When it happens

Trigger: Codegen needs to read a field whose LogicalTypeRoot is not in the switch (e.g. MULTISET, STRUCTURED_TYPE, DISTINCT_TYPE, RAW) inside a row/array/map being (de)serialized by the protobuf format.

Common situations: Deeply nested schemas with unsupported intermediate types; views that introduce multiset/structured types into a protobuf-mapped projection.

Related errors

• Do not support flink type:
• Do not support flink data type:
• Illegal type for protobuf enum, only char/vachar/int/bigint
• Unsupported protobuf simple type:
• Unsupported data type in schema:

AI-assisted analysis of apache/flink@2f3c205e92 (2026-08-14). Data as JSON: /api/errors/0af4480aa99a4cfb. Report an issue: GitHub.