Formats
Format docs explain how SeaTunnel maps between its internal row model and external data encodings such as Avro, Debezium JSON, or Protobuf. Read this section when the connector itself is not enough and you also need to control the payload shape.
When You Should Read Format Docs
- your source or sink exchanges schema-based payloads
- your CDC pipeline depends on a specific envelope format
- you need to align external serialization with downstream consumers
DECIMAL Precision in JSON
When a schema declares a DECIMAL field, including one nested in an array, the JSON reader parses floating-point numbers as BigDecimal to avoid losing numeric precision. This covers ARRAY<DECIMAL>, ARRAY<ARRAY<DECIMAL>> and ARRAY<ROW<DECIMAL>>. The setting applies to the whole document: a bare floating-point number assigned to a STRING sibling is rendered from its decimal value and may retain more digits. It does not preserve the original token spelling; for example, 0.0000001234 can become 1.234E-7. The same applies to a JSON object or array captured as STRING: the floating-point numbers inside it are re-rendered from their decimal values, so their text can change too, while numbers quoted inside that object or array keep the original text. Quoted strings and schemas without DECIMAL are unchanged.