924060929 commented on PR #64160:
URL: https://github.com/apache/doris/pull/64160#issuecomment-5068566067

   Thanks for the fix. Heads-up before this goes further: it overlaps heavily 
with the external meta cache refactor in #65126, and the two currently conflict.
   
   I ran a 3-way merge (`git merge-tree`) of the two PR heads — it reports 
content conflicts in 4 files: `ExternalMetaCacheMgr.java`, 
`RefreshManager.java`, `ExternalCatalog.java`, `ExternalDatabase.java`. They're 
semantic, not just textual:
   
   - This PR renames `invalidateTableCache(ExternalTable)` → 
`invalidateTable(ExternalTable)` and reroutes its call sites; #65126 keeps 
`invalidateTableCache` and adds new methods next to it.
   - #65126 rewrites `replayRefreshTable` and the db/table invalidation call 
sites to use local-vs-remote name resolution (`localTableName`/`localDbName`, 
`getTableForReplay`). The row-count invalidation added here is inserted into 
the old structure.
   - This PR intentionally routes invalidation through the object-based 
`invalidateTable(ExternalTable)` so it has the real db/table IDs to drop the 
row-count entry; #65126 routes the same sites through the name-based 
`invalidateTable(catalogId, dbName, tableName)`.
   
   Since #65126 is the larger, foundational refactor and it doesn't touch 
`ExternalRowCountCache` (so this row-count invalidation is still needed), I'd 
suggest letting #65126 land first, then rebasing this on top and re-applying 
the row-count invalidation on the new call sites — keeping the invariant this 
PR establishes: name-based paths don't touch the row-count cache, object-based 
paths do.
   


-- 
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]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to