Unsupported to query directly from this {} source, please cr

Error message

Unsupported to query directly from this {} source, please create a table or streaming job from it

What it means

`SourceScanInfo::complete` (src/frontend/src/scheduler/plan_fragmenter.rs:475) rejects batch queries that scan a source connector directly. Only sources with partitionable batch split support (e.g. Kafka with offsets, Iceberg/Posix FS files) can be queried in batch mode; other connector kinds must first be materialized into a table or streaming job.

Source

Thrown at src/frontend/src/scheduler/plan_fragmenter.rs:475

                Ok(SourceScanInfo::Complete(res))
            }
            (ConnectorProperties::BatchPosixFs(prop), SourceFetchParameters::Empty) => {
                use risingwave_connector::source::SplitEnumerator;
                let mut enumerator = BatchPosixFsEnumerator::new(
                    *prop,
                    risingwave_connector::source::SourceEnumeratorContext::dummy().into(),
                )
                .await?;
                let splits = enumerator.list_splits().await?;
                let res = splits
                    .into_iter()
                    .map(SplitImpl::BatchPosixFs)
                    .collect_vec();

                Ok(SourceScanInfo::Complete(res))
            }
            (connector, _) => Err(SchedulerError::Internal(anyhow!(
                "Unsupported to query directly from this {} source, \
                 please create a table or streaming job from it",
                connector.kind()
            ))),
        }
    }
}

#[derive(Clone, Debug)]
pub struct TableScanInfo {
    /// The name of the table to scan.
    name: String,

    /// Indicates the table partitions to be read by scan tasks. Unnecessary partitions are already
    /// pruned.
    ///
    /// For singleton table, this field is still `Some` and only contains a single partition with
    /// full vnode bitmap, since we need to know where to schedule the singleton scan task.

View on GitHub (pinned to 6469eb736d)

Solutions

Create a table or materialized view over the source (`CREATE MATERIALIZED VIEW mv AS SELECT * FROM source;`) and query that.
Use `CREATE TABLE ... WITH (connector=...)` to ingest into a table, then SELECT from the table.
If the connector should support batch queries, verify the source kind string is correct (e.g. `kafka`, `iceberg`) or add a split implementation for the kind.

Example fix

// before: SELECT * FROM my_cdc_source;  -- fails
// after
CREATE MATERIALIZED VIEW mv_cdc AS SELECT * FROM my_cdc_source;
SELECT * FROM mv_cdc;

Defensive patterns

Strategy: validation

Validate before calling

-- Only query tables/MVs, or sources with batch split support
SELECT name, connector FROM rw_catalog.rw_sources WHERE name = 'my_source';
-- if connector is not batch-queryable (e.g. cdc), query a MV over it instead

Try / catch

match scheduler_result {
    Err(e) if e.to_string().contains("Unsupported to query directly") => {
        eprintln!("Create a materialized view over the source before querying it.");
    }
    r => r?,
}

Prevention

Materialize sources into tables/MVs before ad-hoc SELECTs.
Remember only Kafka/Iceberg/FS-style sources support direct batch queries.
Validate source kind before enabling direct query paths in tools.

When it happens

Trigger: Running `SELECT * FROM <source>` where `<source>` is a connector source whose kind has no batch split implementation (the catch-all `(connector, _)` match arm).

Common situations: Querying directly from CDC or other non-batch-queryable sources created with `CREATE SOURCE`, expecting ad-hoc SELECT to work like it does on tables or MVs.

Understand the failure class

Background: "is not a compatible type" / "cannot merge" errors: when a value's type doesn't match what the library requires — this error's family across 65 libraries.

Related errors

AI-assisted analysis of risingwavelabs/risingwave@6469eb736d (2026-09-11). Data as JSON: /api/errors/d1ae6bfb5434bfd8. Report an issue: GitHub.