rangareddy commented on issue #17364:
URL: https://github.com/apache/hudi/issues/17364#issuecomment-5351239598

   This issue was reviewed as part of the JIRA-migrated backlog triage 
(HUDI-8871).
   
   **Findings: needs the reproduction run against a current build with Hive 
sync enabled - I could not settle it statically.**
   
   This one is worth prioritising because the report claims a **data** issue 
rather than only a schema mismatch, and it comes with a complete PySpark 
reproduction (write `trips_table`, add a column in the middle of the column 
list rather than appending, then read through Hive).
   
   On `master` the sync path still funnels through 
`hudi-sync/hudi-hive-sync/src/main/java/org/apache/hudi/hive/HiveSyncTool.java:492`:
   
   ```java
   syncClient.updateTableSchema(tableName, schema, schemaDiff);
   ```
   
   Whether inserting a column mid-list produces a correct mapping in the Hive 
table depends on how `schemaDiff` is computed and on whether Hive resolves the 
underlying parquet columns by position or by name. That interaction is not 
decidable from the source, and getting it wrong in either direction would be a 
bad outcome to assert without evidence.
   
   For whoever picks this up, the questions worth answering in order:
   
   1. Does the Hive table's column list after sync match the Hudi schema's 
order, or does the new column land at the end?
   2. If the order differs, do reads return values from the wrong column (data 
issue) or just fail (schema issue)?
   3. Does it reproduce on both COW and MOR, and on both the snapshot and 
read-optimized Hive tables?
   
   Keeping this open.
   


-- 
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]

Reply via email to