apache/flink · error · java.lang.IllegalStateException

Illegal Json Data...

Error message

Illegal Json Data...

What it means

IllegalStateException from RowNestedConverter.convertNotNull (the projected/nested-row path used when only selected nested fields are extracted) when the token is not START_OBJECT. It is the nested-projection counterpart of error 1375: the JSON fragment being projected must be an object, but a scalar/array was found.

Source

Thrown at flink-formats/flink-json/src/main/java/org/apache/flink/formats/json/JsonParserToRowDataConverters.java:608

                path[i] = fieldIndex;
            }
            outputPosToPath.put(pos, path);
        }
    }

    private class RowNestedConverter extends ProjectedConverter {

        private static final long serialVersionUID = 1L;

        private final Map<String, ProjectedConverter> fieldConverters = new HashMap<>();

        // Keep path here for fallback to get nested field from a converted row field.
        private final Map<Integer, String[]> outputPosToPath = new HashMap<>();

        @Override
        public void convertNotNull(JsonParser jp, GenericRowData outputRow) throws IOException {
            if (jp.currentToken() != JsonToken.START_OBJECT) {
                throw new IllegalStateException("Illegal Json Data...");
            }
            int arity = fieldConverters.size();
            int cnt = 0;
            jp.nextToken();
            while (jp.currentToken() != JsonToken.END_OBJECT) {
                if (cnt >= arity) {
                    skipToNextField(jp);
                    continue;
                }
                String fieldName = jp.getText();
                jp.nextToken();
                ProjectedConverter converter = fieldConverters.get(fieldName);
                if (converter != null) {
                    converter.convert(jp, outputRow);
                    jp.nextToken();
                    cnt++;
                } else {
                    skipToNextField(jp);

View on GitHub (pinned to 2f3c205e92)

Solutions

1. Align DDL nested types with the real payload shape (nested ROW fields must be JSON objects)
2. Fix producer output for that nested field
3. With ignore-parse-errors=true the record is skipped rather than failing (verify policy fit first)

Example fix

-- before: info ROW<name STRING> but data is "info": "n/a"
SELECT info.name FROM t

-- after: model it as what it is
info STRING

JsonNode n = mapper.readTree(sample).get("info");
if (n != null && !n.isObject()) throw new IllegalStateException("projected nested field source must be an object");

catch (IllegalStateException e) {
    if (e.getMessage().contains("Illegal Json Data")) {
        // nested projection hit a non-object; align nested schema or skip row
    }
}

When it happens

Trigger: Using nested-field projection (e.g., SELECT info.name from a ROW-typed or json_derived column) where the projected source in the payload is not a JSON object — e.g., "info": "N/A" or "info": [..].

Common situations: Same shape-mismatch family as 1372/1374/1375, hit specifically when the optimizer prunes to nested fields and the nested converter runs instead of the full row converter.

Related errors

• Illegal JSON object data...
• Failed to deserialize JSON '%s'.
• Illegal JSON array data...
• Illegal JSON map data...
• Fail to deserialize at field: %s.

AI-assisted analysis of apache/flink@2f3c205e92 (2026-08-14). Data as JSON: /api/errors/c6110fc90461fb4b. Report an issue: GitHub.