SEPURI-SAI-KRISHNA opened a new issue, #19666:
URL: https://github.com/apache/hudi/issues/19666
## Bug Description
**What happened:**
`hoodie.datasource.write.slash.separated.date.partitioning=true` is accepted
on a table with
more than one partition field, but the write and the query-side path
composition disagree, so
prefix-pruned queries silently return zero rows.
For a `(datestr, city)` table written by the Avro path,
`KeyGenUtils#getRecordPartitionPath`
guards the substitution on `partitionPathFields.size() == 1`, so the
directory is
`2026-01-05/san-francisco`.
On the query side, `SparkHoodieTableFileIndex#composeRelativePartitionPath`
builds a
`StringPartitionPathFormatter` from table config and combines only the bound
*prefix* of the
partition columns. A query filtering on `datestr` alone composes the
single-part value, the
formatter's single-field branch applies the dash-to-slash substitution, and
the resulting
`2026/01/05` does not exist on disk. `exists()` fails and the query returns
nothing.
`ShowHoodieTablePartitionsCommand.scala:60-61` already asserts the invariant
this configuration
violates:
```scala
ValidationUtils.checkState(partitionColumnNamesOpt.get().length == 1,
"Only one partition field is allowed for SlashEncodedPartitioning")
```
so one code path treats single-field as a hard requirement while table
creation accepts any
number of fields.
**What you expected:**
Either the combination is rejected at table creation and in writer config
validation, or all
paths agree on the layout for a multi-field slash table.
**Steps to reproduce:**
1. Create a COW table partitioned by `(datestr, city)` with
`hoodie.datasource.write.slash.separated.date.partitioning=true`.
2. Insert a row with `datestr='2026-01-05'`, `city='san-francisco'`.
3. `SELECT * FROM t WHERE datestr = '2026-01-05'` -- returns zero rows,
while a query binding
both partition columns returns the row.
**Suggested fix:**
Reject `slash.separated.date.partitioning=true` together with more than one
partition field, next
to the existing hive-style check in `HoodieCatalogTable#extraTableConfig`
and in
`HoodieWriterUtils.validateTableConfig` so non-SQL writers are covered too.
## Environment
**Hudi version:** master (1.3.0-SNAPSHOT)
**Query engine:** Spark
**Relevant configs:**
`hoodie.datasource.write.slash.separated.date.partitioning=true` with two or
more partition fields
## Logs and Stack Trace
No failure -- the query silently returns an empty result.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
