[client] Typed deletes drop partition keys on partitioned primary-key tables
- Dominant language
- Java
- Stars
- 2.1k
- Forks
- 625
- Avg merge
- 3d 14h
- Merged PRs (30d)
- 97
Description
### Search before asking
- [x] I searched in the [issues](https://github.com/apache/fluss/issues) and found nothing similar.
### Fluss version
main (development)
### Please describe the bug 🐞
`TypedUpsertWriter.delete(pojo)` fails on a partitioned primary-key table even when the POJO contains all primary-key and partition values.
Example:
```text
Columns: id INT, dt STRING, payload STRING
PRIMARY KEY: (id, dt)
PARTITION BY: dt
Delete input: {id: 42, dt: "2026-09-11", payload: null}
```
Expected: route the delete to partition `2026-09-11` for physical key `id=42`.
Actual: the POJO is converted to `[42, null, null]`, and partition resolution throws `NullPointerException: Partition value shouldn't be null.` before sending the delete.
`TypedUpsertWriterImpl` uses physical primary keys for both projection and row expansion. Physical primary keys exclude partition columns, so `dt` is lost.
### Solution
Use logical primary keys for the typed delete projection and row expansion, preserving partition values for routing.
### Are you willing to submit a PR?
- [x] I'm willing to submit a PR!
Contributor guide
No contributing guide indexed for this repository
Research direction
Start in TypedUpsertWriterImpl and trace how typed deletes project and expand rows before partition resolution. Reproduce the partitioned primary-key example from the issue; done means the delete preserves dt for routing while using id as the physical key and no longer raises a null partition-value error.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- java
- Domain
- databases
- Issue type
- Bug
- Difficulty
- 3/5
- Estimated time
- 1-2 days
- Activity status
- Active
- Clarity
- Clearly specified
- Newbie friendliness
- 75/100