Version 5.10.1¶

Released on 2025-02-13.

Note

If you are upgrading a cluster, you must be running CrateDB 4.0.2 or higher before you upgrade to 5.10.1.

We recommend that you upgrade to the latest 5.9 release before moving to 5.10.1.

A rolling upgrade from 5.9.x to 5.10.1 is supported. Before upgrading, you should back up your data.

Warning

Tables that were created before CrateDB 4.x will not function with 5.x and must be recreated before moving to 5.x.x.

You can recreate tables using COPY TO and COPY FROM or by inserting the data into a new table.

Table of contents

See the Version 5.10.0 release notes for a full list of changes in the 5.10 series.

Fixes ¶

Fixed a recovery issue that led to records of a table with an OBJECT sub-column being part of a standalone FULLTEXT index to be never visible for MATCH queries. This issue cannot be mitigated and a full re-index of the data is required.
Fixed an issue that would lead to errors when attempting to select a sub-column from a virtual table or view participating in a join, e.g.:
```
SELECT vw1.col2['arr']
FROM (SELECT * FROM tbl1) as vw1
INNER JOIN tbl2 on vw1.col1 = tbl2.col1
LIMIT 10;
```
Fixed a regression from 5.10.0 that could get COPY FROM statements to get stuck if used with RETURN SUMMARY
Fixed a regression from 5.10.0 that caused a ColumnUnknownException when accessing a child of an object literal from an aliased table.
Fixed an issue that could lead to a Scale of numeric must be less than the precision error when using the AVG aggregation in a way that involved values of type NUMERIC.
Fixed an issue that could cause data loss if increasing the number of shards using ALTER TABLE ... SET (number_of_shards = ?) while having other allocation settings in place that prevent the shard allocation.
Fixed an issue that could cause duplicate records to be returned when filtering on a table by PRIMARY KEY, using query parameters, and binding the same value to them, e.g.:
```
SELECT * FROM tbl WHERE pk_col = ? OR pk_col = ?
-- Bind the same value to both query parameters
```
Fixed a handshake version compatibility issue that causes a node with a higher major version (e.g. 6.x) to fail to join a cluster of version < Version 5.10.1 (rolling upgrade scenario).
Fixed an issue that caused an error when using a PreparedStatement and selecting an expression, involving a parameter and a column used in ORDER BY and query had a LIMIT clause. Example.:
```
SELECT a, b + ? as sum FROM tbl ORDER BY sum LIMIT 10
```

Fixed an issue that caused an error when using a correlated sub-query and a normal subquery on the same level together with an additional sub-query on an outer level. Example:

WITH tbl AS (SELECT 1 as x)
SELECT (
  SELECT x FROM tbl
  WHERE t.x = tbl.x
    AND tbl.x IN (SELECT x FROM unnest([1]))
),
(SELECT y FROM unnest([2]) as t2(y))
FROM tbl t

Fixed an issue that caused a planner exception when using a query containing an ORDER BY clause on top of a UNION clause, while the ordering column is not part of the most-top select items. Example:
```
SELECT id FROM (
    SELECT id, other_id, name FROM users
    UNION ALL
    SELECT id, other_id, name FROM users
    ) u
ORDER BY name
```
Fixed an issue that would lead to wrong results when using MIN or MAX aggregations on numeric values. Previously, the values where implicitly casted to STRING and the aggregations where using alphanumeric ordering, instead of numeric ordering.
Fixed an issue that led to ArithmeticException when using AVG with NUMERIC type and result was an infinite fraction, like 1/3.
Fixed an issue that would lead to returning a default name for PRIMARY KEY constraint in key_column_usage and pg_class tables even if a custom name was explicitly provided during table creation.
Fixed an issue that would lead to returning a different default name for PRIMARY KEY constraint in key_column_usage and pg_class tables, <table_name>_pk and <table_name>_pkey respectively, when a custom name is not explicitly provided during table creation.
Fixed an issue that would cause array_position to return wrong results when used on a column with NULL values in the WHERE combined with a NOT predicate. e.g.:
```
SELECT * FROM tbl WHERE NOT array_position(string_array_col, 'foo');
```
Fixed a regression introduced with Version 5.6.0 that caused any partitioned table contained inside the snapshot to be restored, if not exists, by the RESTORE SNAPSHOT statement when only a concrete table was specified as to be restored. Only the partitioned table definition was falsely restored, but not the the actual data.
Fixed a regression introduced with Version 5.10.0 that would lead to wrong results for LEFT JOIN queries when the join condition has a CASE with an equality condition inside. Previously, the optimizer would use Hash Join instead of Nested Loop leading to wrong results. e.g.:
```
SELECT * FROM tlb
LEFT JOIN (SELECT 2 AS col2 FROM tbl) AS tbl2
  ON (CASE t0.c0 WHEN sub0.col2 THEN t0.c1 ELSE t0.c1 END)
```

Version 5.10.1¶

Fixes¶

Fixes ¶