#8 "output path is required" error when loading data to SQL Server or Oracle

[Anonymous]

Originally created by: gmacc00

When loading CSV or PARQUET data into any of these engines, the following error may occur:

[SQL Server, Oracle]: output path is required
(“sink-without-path” on the right side)

This issue occurs for both newly created and existing tables in both of them.

[Anonymous]

Ticket changed by: SouravRoy-ETL

• status: open --> closed

[Anonymous]

Ticket changed by: SouravRoy-ETL

• status: closed --> open

[Anonymous]

Originally posted by: SouravRoy-ETL

This had two distinct causes, both now fixed on main.

1. The "output path is required" block (the error you saw)

The frontend validator required a path on every snk.* component, so a run was blocked before it even started when the sink was a database / warehouse / broker / HTTP target (snk.sqlserver, snk.oracle, snk.postgres, snk.mongodb, snk.kafka, ...). Those sinks write to a connection + table + topic and have no path. The path check is now restricted to the file / object-store formats that actually need one (csv, tsv, parquet, json, jsonl, excel, xml, yaml, toml, avro, spatial). Per-connector required fields (host, table, ...) are still validated separately from the component manifest.

2. Loading into a not-yet-created table

You noted it failed "for both newly created and existing tables." Existing tables worked once cause [#1] was cleared, but a brand-new table still failed deeper in the run: the SQL Server and Oracle sinks only issued INSERTs, so a missing target errored with Invalid object name (SQL Server) / ORA-00942: table or view does not exist (Oracle).

The driver sinks now auto-create the target table when it is absent, inferring column types from the upstream schema. An existing table is left untouched and simply appended to (SQL Server guards with IF OBJECT_ID(...) IS NULL; Oracle wraps the CREATE TABLE in a PL/SQL block that swallows ORA-00955 only).

Verified on live servers

• SQL Server 2022: src.csv -> snk.sqlserver into a non-existent table auto-creates id BIGINT, name NVARCHAR, amt FLOAT and lands all rows; re-running appends to the existing table.
• Oracle 21: src.csv -> snk.oracle into a non-existent table auto-creates id NUMBER, name VARCHAR2, amt BINARY_DOUBLE and lands all rows.

Both fixes are on main and will ship in the next build. Leaving this open for you to confirm against your own setup before it is marked done.

[Anonymous]

Originally posted by: SouravRoy-ETL

Please wait for the next hotfix v2 release.

[Anonymous]

Originally posted by: SouravRoy-ETL

Update: this fix is shipping in the re-rolled v0.1.0-hotfix2 binaries (same release link, rebuilt 2026-05-29 - the tag is unchanged).

Both parts of the report are addressed:

• The "output path is required" block is gone for database / warehouse / broker sinks (they write to a connection + table, not a path).
• The SQL Server and Oracle sinks now auto-create the target table when it does not exist yet (column types inferred from the upstream), so loading into a brand-new table works as well as an existing one. Verified on live SQL Server 2022 and Oracle 21 for both cases.

Once the refreshed binary for your platform is on the release, please re-download it and try the SQL Server / Oracle load again. Leaving this open for you to confirm.

[Anonymous]

Originally posted by: SouravRoy-ETL

Download -https://github.com/SouravRoy-ETL/duckle/releases/tag/v0.1.0-hotfix2

[Anonymous]

Originally posted by: SouravRoy-ETL

The fix is now live in the published v0.1.0-hotfix2 binaries. The release was rebuilt today (2026-05-30, assets dated ~17:43-17:50 UTC) from the latest main, same release link, tag unchanged.

What is fixed:

• The "output path is required" error is gone for database / warehouse / broker sinks (these write to a connection + table, not a file path).
• The SQL Server and Oracle sinks auto-create the target table when it does not exist yet, inferring column types from the upstream, so loading into a brand-new table works the same as loading into an existing one.

Both were verified on live SQL Server 2022 and Oracle 21.

Download (re-download to pick up the refreshed binary for your platform): https://github.com/SouravRoy-ETL/duckle/releases/tag/v0.1.0-hotfix2

Leaving this open for you to confirm on your setup, @gmacc00 - let me know how the SQL Server / Oracle load goes.

[Anonymous]

Ticket changed by: gmacc00

• status: open --> closed