Changelog

3.1

3.1 focuses on reversible deletion, node copying, and multi-shape trees — adding the recycle bin (logical deletion), copyNode, cross-tree operations, and multi-root trees.

New Features

• Recycle bin (logical deletion): Enabled via the recyclebin config. deleteNode(node, { recycle: true }) moves the subtree into the recycle bin with its internal structure intact (reuses the move algorithm, atomic in a transaction); clearRecycleBin() empties it in one call. All read and write methods accept the includeRecyclebin view parameter — under the default view, nodes inside the bin logically do not exist; with true they behave as ordinary nodes (restore = read out in the bin view + moveNode out).
• Node copy copyNode: The whole operation runs in a transaction with a fixed number of set-based SQL statements — database access count is independent of descendant count. Supports copying only the node itself (includeDescendants: false), field filtering (fields), and cross-tree copying (options.treeId).
• Cross-tree move: moveNode accepts options.treeId to move a subtree across trees; when toNode is omitted, the node migrates out as the root of a new tree with that treeId; canMoveTo supports cross-tree checks as well.
• Multi-root tree MultiRootFlexTreeManager: Delivers a user-facing tree with multiple top-level nodes via a hidden root, with automatic level normalization — all single-tree operations work as-is.
• countField descendant count: All query methods and export methods (toJson/toList) support the countField parameter, computed directly by the database as (rightValue - leftValue - 1) / 2 — unaffected by level truncation; visible scope when the recycle bin is enabled.
• New sql.js adapter (flextree-sqljs-adapter): Based on sql.js, runs a full tree structure in the browser (wasm).

3.0

3.0 is a major upgrade focused on data safety, database compatibility, and developer experience.

New Features

• Transactional writes & concurrency dirty-read fix: All operations inside write are now wrapped in a database transaction—any failure rolls back the whole batch; a write-transaction context isolation and a read guard prevent concurrent reads from seeing intermediate states (dirty reads) during writes.
• Singleton pattern: FlexTreeManager.getInstance returns the same instance by table name, avoiding duplicate creation.
• Node events: Added node-level events such as node:added, node:deleted, node:cleared, node:updated, and node:moved.
• Tree traversal forEach: Supports both DFS / BFS modes, with interruption, level limiting, and a configurable start point.
• Tree repair repair: Rebuilds corrupted left/right value structures based on level and automatically verifies integrity.
• Nested batch add: addNodes supports adding a whole subtree at once via the children field.
• FlexTree lazy loading: Added a lazy option so nodes can be loaded on demand.
• Tree export: FlexTreeManager adds the getTree / toJson / toList convenience methods.
• Fine-grained verification: verify checks via pure SQL, covering node totals, value integrity, basic relations, uniqueness, and level relations.
• New Bun SQLite adapter (flextree-bun-sqlite-adapter).
• Type enhancements: treeId supports string; added the FlexTreeNodeInput nested input type.

Improvements

• Toolchain migrated from pnpm to Bun; tests migrated to Bun Test.
• Upgraded TypeScript to 6 and oxlint to 1.76.
• Test cases are now inlined into each package's __tests__/, along with a brand-new shared test toolkit.

Breaking Changes

• Adapter interface: ready → connected; added required method transaction and optional field type.
• FlexTreeManager: ready() → connected(), assertDriverReady() → assertConnected().
• Removed the sqlstring dependency in favor of a built-in SQL escaper.
• Now requires TypeScript 6 or above.

2.x and Earlier

See GitHub Releases.