Sotirios Kougiouris created SPARK-57421:
-------------------------------------------
Summary: Support @-syntax version and timestamp time travel on
table name
Key: SPARK-57421
URL: https://issues.apache.org/jira/browse/SPARK-57421
Project: Spark
Issue Type: New Feature
Components: Connect, SQL
Affects Versions: 5.0.0
Reporter: Sotirios Kougiouris
h2. What
Add the "@" time-travel shorthand on table names so that
{code:sql}
SELECT * FROM name@vN
SELECT * FROM name@<TS>
{code}
and
{code:java}
spark.read.format("delta").table("name@vN")
spark.read.format("delta").table("name@<ts>")
{code}
parse into the same plan as the existing {{VERSION AS OF}} / {{TIMESTAMP AS
OF}} and {{option("versionAsOf")}} / {{option("timestampAsOf")}} clauses.
Support for SQL, the classic DataFrame reader, and Spark Connect.
h2. Why
Spark already supports {{VERSION AS OF}} / {{TIMESTAMP AS OF}} and the
{{{}versionAsOf{}}}/{{{}timestampAsOf{}}} reader options, but not the compact
"@" suffix. Resolving it at parse time also simplifies the time travel entry
points pipeline.
h2. Proposed change
* Lexer/grammar: add @-suffix on table identifiers.
* AstBuilder converts the suffix into {{RelationTimeTravel}} at parse time.
* New conf {{spark.sql.timeTravel.atSyntax.enabled}} (default true) to gate it.
* Conflicts (e.g. "@" together with {{AS OF}}) raise error.
h2. Notes
* Behavior is opt-out via the conf.
* Backquoted identifiers like {{`name@v1`}} remain literal table names.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
