thswlsqls · Nov 5, 2026 · JingsongLi · Nov 5, 2026
diff --git a/docs/docs/append-table/bucketed.mdx b/docs/docs/append-table/bucketed.mdx
@@ -91,10 +91,11 @@ set of buckets through this optimization.
 
 ## Bucketed Join
 
-Spark can use compatible bucket distributions to avoid a shuffle in a batch join. Enable V2 bucketing and join on the
-distribution keys:
+Spark can use compatible bucket distributions to avoid a shuffle in a batch join. Enable
+`scan.preserve-data-grouping` and V2 bucketing, then join on the distribution keys:
 
 ```sql
+SET spark.paimon.scan.preserve-data-grouping = true;
 SET spark.sql.sources.v2.bucketing.enabled = true;
 
 CREATE TABLE fact_table (order_id INT, f1 STRING) USING paimon

diff --git a/docs/docs/flink/action-jars.md b/docs/docs/flink/action-jars.md
@@ -44,7 +44,9 @@ The following command is used to compact a table.
 <FLINK_HOME>/bin/flink run \
  /path/to/paimon-flink-action-@@VERSION@@.jar \
  compact \
- --path <TABLE_PATH>
+ --warehouse <warehouse-path> \
+ --database <database-name> \
+ --table <table-name>
 ```
 
 Use `<action> --help` to inspect its arguments. Partition specifications use comma-separated

diff --git a/docs/docs/flink/sql-ddl.md b/docs/docs/flink/sql-ddl.md
@@ -123,7 +123,7 @@ You can set location in the properties of table/database by the config of `locat
 
 By using the Paimon JDBC catalog, changes to the catalog will be directly stored in relational databases such as SQLite, MySQL, postgres, etc.
 
-Currently, lock configuration is only supported for MySQL and SQLite. If you are using a different type of database for catalog storage, please do not configure `lock.enabled`.
+Currently, lock configuration is only supported for MySQL, PostgreSQL, and SQLite. If you are using a different type of database for catalog storage, please do not configure `lock.enabled`.
 
 Paimon JDBC Catalog in Flink needs to correctly add the corresponding jar package for connecting to the database. You should first download JDBC  connector bundled jar and add it to classpath. such as MySQL, postgres
 

diff --git a/docs/docs/learn-paimon/scenario-guide.mdx b/docs/docs/learn-paimon/scenario-guide.mdx
@@ -227,7 +227,7 @@ An equality or `IN` predicate on the complete bucket key can prune unrelated buc
 not guarantee reading exactly one eighth of the bytes: bucket sizes can be skewed, and matching
 buckets still contain other keys. A composite bucket key requires constraints on every key column.
 
-Spark can exploit compatible bucket distributions in joins with V2 bucketing enabled. Confirm
+Spark can exploit compatible bucket distributions in joins with `scan.preserve-data-grouping` and V2 bucketing enabled. Confirm
 shuffle removal with `EXPLAIN`; see [Bucketed Join](../append-table/bucketed#bucketed-join).
 
 ### Ordered Streaming {#scenario-7-queue-like-ordered-streaming-bucketed-append}

diff --git a/docs/docs/maintenance/filesystems.mdx b/docs/docs/maintenance/filesystems.mdx
@@ -306,7 +306,7 @@ SELECT COUNT(1) FROM test_table;
 
 From version 0.8, paimon-trino uses trino filesystem as basic file read and write system. We strongly recommend you to use jindo-sdk in trino.
 
-You can find [How to config jindo sdk on trino](https://github.com/aliyun/alibabacloud-jindodata/blob/master/docs/user/4.x/4.6.x/4.6.12/oss/presto/jindosdk_on_presto) here.
+You can find [How to config jindo sdk on trino](https://github.com/aliyun/alibabacloud-jindodata/blob/master/docs/user/4.x/4.6.x/4.6.12/oss/presto/jindosdk_on_presto.md) here.
 Please note that:
 * Use paimon to replace hive-hadoop2 when you decompress the plugin jar and find location to put in.
 * You can specify the `core-site.xml` in `paimon.properties` on configuration [hive.config.resources](https://trino.io/docs/current/connector/hive.html#hdfs-configuration).

diff --git a/docs/docs/maintenance/manage-snapshots.mdx b/docs/docs/maintenance/manage-snapshots.mdx
@@ -280,9 +280,9 @@ Run the following command:
     /path/to/paimon-flink-action-@@VERSION@@.jar \
     expire_snapshots \
     --warehouse <warehouse-path> \
-    --identifier <identifier> \
+    --database <database-name> \
+    --table <table-name> \
     --older_than <timestamp> \
-    --version <snapshot-id> \
     --max_deletes <max-deletes> \
     --retain_max <retain-max> \
     --retain_min <retain-min> \

diff --git a/docs/docs/multimodal-table/variant.mdx b/docs/docs/multimodal-table/variant.mdx
@@ -32,7 +32,7 @@ suitable for event attributes, model output, tool parameters, and other data tha
 frequently to model every field as a table column.
 
 Paimon stores `VARIANT` with the
-[Parquet Variant binary encoding](https://github.com/apache/parquet-format/blob/master/VariantEncoding),
+[Parquet Variant binary encoding](https://github.com/apache/parquet-format/blob/master/VariantEncoding.md),
 instead of storing its JSON text. The data is parsed when it is converted to `VARIANT`, and queries
 can extract a path as a requested SQL type.
 

diff --git a/docs/docs/primary-key-table/query-performance.md b/docs/docs/primary-key-table/query-performance.md
@@ -103,6 +103,7 @@ coverage rules, especially for Vector and Full Text search.
 Spark can use compatible fixed-bucket layouts to avoid a join shuffle. For example:
 
 ```sql
+SET spark.paimon.scan.preserve-data-grouping = true;
 SET spark.sql.sources.v2.bucketing.enabled = true;
 
 CREATE TABLE fact_table (order_id INT, f1 STRING) USING paimon
@@ -115,6 +116,7 @@ SELECT * FROM fact_table AS fact
 JOIN dim_table AS dim ON fact.order_id = dim.order_id;
 ```
 
-The setting enables Spark to use partitioning reported by V2 data sources. The join keys, bucket
+These settings let Paimon report its bucket layout to Spark (see
+[Scan Layout and Storage Partition Joins](../spark/sql-query#scan-layout-and-storage-partition-joins)). The join keys, bucket
 keys, and bucket counts must be compatible; the same bucket count alone is not sufficient.
 Check `EXPLAIN` to confirm that Spark avoids the shuffle for the actual query.
diff --git a/docs/docs/pypaimon/sql.md b/docs/docs/pypaimon/sql.md
@@ -118,7 +118,7 @@ batches = ctx.sql("SELECT * FROM paimon.default.my_temp")
 
 ## Supported SQL Syntax
 
-The SQL engine is powered by Apache DataFusion, which supports a rich set of SQL syntax. For the full SQL reference, see the [paimon-rust SQL documentation](https://paimon-rust.apache.org/sql.html) which covers:
+The SQL engine is powered by Apache DataFusion, which supports a rich set of SQL syntax. For the full SQL reference, see the [paimon-rust SQL documentation](https://paimon.apache.org/docs/rust/sql/) which covers:
 
 - **DDL**: `CREATE SCHEMA`, `CREATE TABLE` (with `PARTITIONED BY`, `PRIMARY KEY`, `WITH` options), `DROP TABLE`, `ALTER TABLE`, `CREATE TEMPORARY TABLE/VIEW`
 - **DML**: `INSERT INTO`, `INSERT OVERWRITE` (dynamic/static partitions), `UPDATE`, `DELETE`, `MERGE INTO`, `TRUNCATE TABLE`

diff --git a/docs/docs/spark/catalogs.md b/docs/docs/spark/catalogs.md
@@ -94,7 +94,7 @@ If you want to see a partitioned table in Hive and also synchronize newly create
 
 The JDBC catalog stores catalog metadata in a relational database such as SQLite, MySQL, or PostgreSQL.
 
-Currently, lock configuration is only supported for MySQL and SQLite. If you are using a different type of database for catalog storage, please do not configure `lock.enabled`.
+Currently, lock configuration is only supported for MySQL, PostgreSQL, and SQLite. If you are using a different type of database for catalog storage, please do not configure `lock.enabled`.
 
 Add the JDBC driver for your database to Spark's classpath along with the Paimon connector.