Changelog

New Features

• SQL Server supports manual transactions, allowing manual start, commit, and rollback of transactions, and is compatible with the legacy SQL Server 2000 driver
• XuguDB supports visual estimation of execution plans
• Execution plan results support export to file
• JSON-format execution plan results are displayed in a read-only editor with one-click copy support
• MySQL database right-click menu adds Clear All Table Data, executed in foreign key dependency order and reporting results for each table
• Redis key list supports mouse drag box-selection of multiple keys
• MongoDB query completion supports nested query operators and option suggestions
• Data grid adds a pinnable column finder sidebar to quickly locate target columns by name
• Data grid supports horizontal scrolling with the mouse wheel while holding Alt
• SQL editor supports selecting the encoding used when opening and saving SQL files
• Added a setting to go directly to the query page when opening a connection
• Plugins can be granted the unsafe-eval CSP permission required by the graphics engine on a per-plugin basis
• Cloud sync backup selection dialog supports batch selection and group control
• Query result toolbar adds a shortcut entry to view table DDL
• Table structure sidebar supports previewing the DDL statement of the selected column
• SQL library supports copy-paste of query files, and the input box automatically focuses when renaming
• SQL export adds a Keep Original Language option, decoding \uXXXX escapes during export
• SQL formatting supports configuring comma position as leading or trailing
• Structure-only transfer adds an execution plan summary, allowing a plan overview to be viewed before execution

Improvements

• Data grid "Load All" is no longer constrained by the single-request row limit and can load datasets exceeding the limit

Fixes

• Fixed the issue where legacy JDBC drivers (such as jTDS connecting to SQL Server 2000) throw AbstractMethodError when the getSchemas overload is missing, causing structure loading to fail
• Fixed the issue where Oracle q'' custom quoted strings are truncated in the middle during statement splitting
• Fixed the issue where dialects that support nested block comments (such as PostgreSQL) end comments prematurely during statement splitting, causing incorrect splitting
• Fixed the issue where recognition of the statement at the cursor does not follow dialect quoting and -- comment rules, causing incorrect execution range division
• Fixed the issue where statement splitting always treats backslashes as quote escapes according to MySQL rules; in dialects such as PostgreSQL this now only applies within E'' strings
• Fixed the issue where PostgreSQL array literals containing backslashes are not quoted as escaped strings (E'') when generating SQL
• Fixed the issue where row count estimates are incorrectly displayed as 0 after PostgreSQL statistics are reset (pg_stat_reset, pg_upgrade, etc.); in this case it now falls back to using reltuples
• Fixed the issue where the identity value of auto-increment columns and the current state of sequences are lost when exporting SQL from PostgreSQL
• Fixed the issue where the editor viewport is reset by scrolling after executing SQL; the viewport now remains unchanged when the cursor was originally visible
• Fixed multiple details on the homepage introduction page (logo size, import source selection, etc.), and improved highlight positioning in settings search results
• Fixed the issue where the total row count is displayed inaccurately when the data grid jumps to a page beyond the known total row count; it now rechecks the exact total
• Fixed the issue where SQL risk classification and AI execution policy scanning do not parse comments and string escapes according to the dialect, potentially misjudging write operations as read-only
• Fixed the issue where production environment protection scanning does not parse comments and escapes according to the connection dialect (such as SQL Server # temp tables and PostgreSQL backslashes), causing missed detection of write operations
• Fixed the issue where frontend production environment SQL scanning does not use the connection dialect lexical rules, making it inconsistent with backend security determination

---

<!-- dbx-cnb-mirror -->

Domestic download: If GitHub downloads are slow, you can download the desktop installation package from the CNB mirror, and pull the Docker image from docker.cnb.cool/dbxio.com/dbx:v0.6.35.