ad1happy2go opened a new issue, #20522:
URL: https://github.com/apache/hudi/issues/20522
The Trino connector (`hudi-trino`) fails snapshot reads of Merge-On-Read
tables whose log files hold parquet data blocks
(`hoodie.logfile.data.block.format=parquet`, table version 9 and below). Any
writer that uses a Spark record merger produces this layout. The query fails
with:
```
java.lang.UnsupportedOperationException: Inline log blocks are not supported
by the Hudi Trino connector: inlinefs://...
at
io.trino.plugin.hudi.reader.HudiTrinoReaderContext.getFileRecordIterator(HudiTrinoReaderContext.java)
```
Spark reads these tables fine, and read-optimized queries in Trino work.
Only the snapshot read path is missing. Proposal: read the parquet block in
place, through a view of its byte range in the log file, with the same parquet
page source the connector already uses for native log files.
--
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]
