[rust/python] Writing to datalake-enabled Iceberg tables fails: "KeyEncoder for Iceberg format is not yet implemented" #3818

Search before asking

• I searched in the issues and found nothing similar.

Fluss version

0.9.1-incubating (server), pyfluss 0.1.0 (client). The gap is still present on current main.

Please describe the bug 🐞

Any write to a primary-key table with table.datalake.enabled = 'true' and Iceberg as the
cluster's datalake format fails from the Python client (pyfluss, backed by fluss-rust)
with:

KeyEncoder for Iceberg format is not yet implemented

The same write against the same table succeeds with the Java client / Flink connector, so
the table itself is fine — the limitation is client-side.

Root cause: in fluss-rust/crates/fluss/src/row/encode/mod.rs, the key-encoder factory
implements Paimon and Lance but explicitly rejects Iceberg:

match data_lake_format {
    Some(DataLakeFormat::Paimon) => {
        Ok(Box::new(PaimonKeyEncoder::new(row_type, key_fields)?))
    }
    Some(DataLakeFormat::Lance) => Ok(Box::new(CompactedKeyEncoder::create_key_encoder(
        row_type, key_fields,
    )?)),
    Some(DataLakeFormat::Iceberg) => Err(Error::UnsupportedOperation {
        message: "KeyEncoder for Iceberg format is not yet implemented".to_string(),
    }),
    None => Ok(Box::new(CompactedKeyEncoder::create_key_encoder(
        row_type, key_fields,
    )?)),
}

The Java client encodes keys for Iceberg-enabled tables (bucket-id computation compatible
with Iceberg's bucket transform), so the Rust client needs the equivalent encoder to reach
parity.

Impact

This blocks Python-based pipelines (e.g. CDC ingestion via pyfluss) from writing to any
datalake-enabled table on Iceberg clusters — the datalake feature and the Python client are
mutually exclusive today. Reported by a production user running Fluss 0.9.1-incubating with
pyfluss 0.1.0.

Steps to reproduce

1. Run a Fluss cluster with datalake.format: iceberg.
2. Create a primary-key table with 'table.datalake.enabled' = 'true'.
3. From Python: connect with pyfluss 0.1.0 and attempt any upsert into the table.
4. The write fails with UnsupportedOperation: KeyEncoder for Iceberg format is not yet implemented.

Expected behavior

The Rust/Python client can write to Iceberg datalake-enabled tables, matching the Java
client's behavior — i.e. an Iceberg-compatible KeyEncoder implementation in
fluss-rust/crates/fluss/src/row/encode/.

Are you willing to submit a PR?

• I'm willing to submit a PR!