dbt-labs/dbt-core · error

Salesforce column creation not implemented yet

Error message

Salesforce column creation not implemented yet

What it means

A `todo!()` panic in `ColumnBuilder::build_from_parts` (column_builder.rs:106) for the Salesforce adapter. The manual Column construction path (used when precision/scale are supplied explicitly) has no Salesforce arm, so constructing a Column for Salesforce panics with 'Salesforce column creation not implemented yet'.

Source

Thrown at crates/dbt-adapter/src/column/column_builder.rs:106

            // TODO: BigQuery fields
            Bigquery => Column::new_bigquery(name, dtype, &[], mode.unwrap_or_default()),
            Redshift => Column::new(
                Redshift,
                name,
                dtype,
                char_size,
                numeric_precision,
                numeric_scale,
            ),
            Databricks | Spark => Column::new(
                self.adapter_type,
                name,
                dtype,
                char_size,
                None, // numeric_precision
                None, // numeric_scale
            ),
            Salesforce => todo!("Salesforce column creation not implemented yet"),
            ClickHouse => Column::new(
                ClickHouse,
                name,
                dtype,
                char_size,
                numeric_precision,
                numeric_scale,
            ),
            Exasol => Column::new(
                Exasol,
                name,
                dtype,
                char_size,
                numeric_precision,
                numeric_scale,
            ),
            Starburst => todo!("Starburst"),
            Athena => todo!("Athena"),

View on GitHub (pinned to 0267ce9170)

Solutions

  1. Avoid the Salesforce adapter for workflows that rebuild columns from parts until implemented.
  2. Implement the Salesforce arm in column_builder.rs:106, choosing appropriate precision/scale semantics for Salesforce (SOQL decimal precision).
  3. Replace `todo!()` with a clear unsupported-adapter error so callers can handle it.

Example fix

// before
Salesforce => todo!("Salesforce column creation not implemented yet"),
// after
Salesforce => Ok(Column::new(
    Salesforce, name, dtype, char_size, numeric_precision, numeric_scale,
)),
Defensive patterns

Strategy: type-guard

Validate before calling

if adapter == AdapterType::Salesforce {
    return Err(anyhow!("Salesforce column creation is not implemented; cannot build_from_parts"));
}

Type guard

fn build_from_parts_supported(t: AdapterType) -> bool {
    !matches!(t, AdapterType::Salesforce | AdapterType::Starburst | AdapterType::Athena | AdapterType::Trino | AdapterType::Dremio | AdapterType::Oracle | AdapterType::Datafusion)
}

Try / catch

let res = std::panic::catch_unwind(AssertUnwindSafe(|| builder.build_from_parts(name, dtype, char_size, p, s)));
match res { Ok(c) => c, Err(_) => Err(anyhow!("unsupported adapter for build_from_parts")) }

Prevention

When it happens

Trigger: Calling the public `build_from_parts(name, dtype, char_size, numeric_precision, numeric_scale)` with adapter type Salesforce — e.g. when relation/column metadata is rebuilt with explicit numeric precision from a schema definition.

Common situations: Column metadata reconstruction on a Salesforce adapter profile where a parsed schema carries numeric_precision/scale; also from adapter-coverage tests that sweep all adapter variants through build_from_parts.

Related errors

AI-assisted analysis of dbt-labs/dbt-core@0267ce9170 (2026-09-07). Data as JSON: /api/errors/52bbff16e584099f. Report an issue: GitHub.