DB Tsai created SPARK-60686:
-------------------------------
Summary: Use columnar Arrow for supported pandas timestamp input
in Spark Connect Rust
Key: SPARK-60686
URL: https://issues.apache.org/jira/browse/SPARK-60686
Project: Spark
Issue Type: Improvement
Components: Connect
Affects Versions: connect-rust-4.3.0
Reporter: DB Tsai
Assignee: DB Tsai
Fix For: connect-rust-4.3.0
The frozen pandas timestamp 10000x3 client case takes60.83ms Rust vs1.23ms
Python. The signed-int64 columnar input implementation in SPARK-60676/PR121
intentionally leaves timestamps on the row path. Extend a narrow supported
timestamp dtype/schema path through the same Arrow importer, preserving session
timezone, aware/naive behavior, NaT/null handling, precision and TimestampNTZ
distinctions. Unsupported conversions should retain the fallback. Compare
baseline and candidate with the identical PR121 importer; include independent
emitted-IPC correctness checks across timezones/DST. Publish only measured
benefit with explicit limitations.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
