6.6 KiB
Org Tree Deepest Levels — Implementation Plan
For agentic workers: REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (
- [ ]) syntax for tracking. Required skills per agent:superpowers:vercel-react-best-practices,superpowers:ui-ux-pro-max
Goal: fetchOrgTreeWithUsers only returns departments at the deepest 2 levels of the global org tree.
Architecture: Add a getDeptDepthMap BFS method that mirrors getAllDeptIds but tracks level. Insert depth filtering in fetchOrgTreeWithUsers before the detail+user fetch loop, slashing API calls.
Tech Stack: NestJS + TypeScript, DingTalk Open API v2
Global Constraints
- Depth is computed from rootDeptId=1 globally, regardless of the user's
rootDeptIdparam fetchOrgTree(dept picker) is NOT modifiedsyncAllis NOT modified- Frontend receives the same
DingOrgTreeNodeWithUsers[]shape, zero frontend changes
Task 1: Add getDeptDepthMap method
Files:
- Modify:
apps/server/src/integration/dingtalk.service.ts(insert aftergetAllDeptIds)
Interfaces:
-
Consumes: nothing new (uses existing
rateLimit(),SubDeptIdListResponse, DingTalklistsubidAPI) -
Produces:
private async getDeptDepthMap(token: string): Promise<Map<number, number>>— key=deptId, value=1-based depth -
Step 1: Add
getDeptDepthMapaftergetAllDeptIds(after line 263)
/**
* BFS from rootDeptId=1, returns depth of every department.
* Depth is 1-based (root=1). Uses listsubid API only, no detail fetches.
*/
private async getDeptDepthMap(token: string): Promise<Map<number, number>> {
const depthMap = new Map<number, number>();
const queue: Array<{ id: number; depth: number }> = [{ id: 1, depth: 1 }];
while (queue.length > 0) {
const { id, depth } = queue.shift()!;
depthMap.set(id, depth);
try {
await this.rateLimit();
const res = await fetch(
`https://oapi.dingtalk.com/topapi/v2/department/listsubid?access_token=${token}`,
{
method: 'POST',
headers: { 'Content-Type': 'application/json' },
body: JSON.stringify({ dept_id: id }),
},
);
const body: SubDeptIdListResponse = await res.json();
if (body.errcode === 0 && body.result?.dept_id_list) {
for (const childId of body.result.dept_id_list) {
queue.push({ id: childId, depth: depth + 1 });
}
}
} catch (e) {
this.logger.error(`getDeptDepthMap 获取部门 ${id} 子部门失败: ${(e as Error).message}`);
}
}
return depthMap;
}
- Step 2: Verify it compiles
Run: cd apps/server && npx tsc --noEmit
Expected: no new errors (may have pre-existing ones in other files)
- Step 3: Commit
git add apps/server/src/integration/dingtalk.service.ts
git commit -m "feat: add getDeptDepthMap BFS method for org tree depth tracking"
Task 2: Filter fetchOrgTreeWithUsers to deepest 2 levels
Files:
- Modify:
apps/server/src/integration/dingtalk.service.ts:476-547
Interfaces:
-
Consumes:
getDeptDepthMap(token)from Task 1 -
Produces: same
Promise<DingOrgTreeNodeWithUsers[]>return type, now filtered -
Step 1: Insert depth filtering before the detail+user fetch loop
Replace lines 480-481 (const token = ...; const deptIds = ...;) with the depth-aware version, and wrap the loop to use filteredIds:
const token = await this.getAccessToken();
// Compute global depth map (always from rootDeptId=1)
const depthMap = await this.getDeptDepthMap(token);
const maxDepth = Math.max(...depthMap.values());
// Get subtree IDs for the user's selected root
const subtreeIds = await this.getAllDeptIds(token, rootDeptId);
// Filter: only keep departments at the deepest 2 levels of the global tree
const filteredIds = subtreeIds.filter((id) => {
const d = depthMap.get(id) ?? -1;
return d === maxDepth || d === maxDepth - 1;
});
if (filteredIds.length === 0) {
return [];
}
// 拉每个部门的详情(仅过滤后的)
const nodes: DingOrgTreeNodeWithUsers[] = [];
const userDeptMap = new Map<string, number[]>();
for (let i = 0; i < filteredIds.length; i++) {
if (i > 0) await this.delay(i);
const detail = await this.getDeptDetail(token, filteredIds[i]);
And update the deptId reference inside the loop body — line 494 uses deptIds[i]:
Change line 494:
// Before:
this.logger.log(`[dingtalk] dept ${deptIds[i]} (${detail.name}): ${dingUsers.length} users`);
// After:
this.logger.log(`[dingtalk] dept ${filteredIds[i]} (${detail.name}): ${dingUsers.length} users`);
And the getDeptUsers call on line 493 must also use filteredIds[i] instead of deptIds[i] — it already does since we're iterating filteredIds.
- Step 2: Update tree assembly to handle missing parent nodes
The tree assembly at lines 534-546 currently checks node.id !== rootDeptId to decide if a node is a root. Since intermediate levels are filtered out, a maxDepth-1 node's parent won't be in the map. Update the root detection:
// 组装成树(父节点可能已被过滤,缺失的父节点 → 节点提升为根)
const map = new Map<number, DingOrgTreeNodeWithUsers>();
nodes.forEach((n) => map.set(n.id, n));
const roots: DingOrgTreeNodeWithUsers[] = [];
for (const node of nodes) {
const parent = map.get(node.parentId);
if (parent) {
parent.children.push(node);
} else {
roots.push(node);
}
}
Note: removed the && node.id !== rootDeptId condition — it's redundant with the map.get check when intermediate levels are filtered. Nodes whose parent exists in map get attached; those whose parent was filtered out become roots. Same behavior, simpler logic.
- Step 3: Verify it compiles
Run: cd apps/server && npx tsc --noEmit
Expected: no new errors
- Step 4: Smoke test with a real DingTalk config (if available)
Run: curl -s http://localhost:3000/api/sync/dingtalk/org-tree-with-users | jq '. | length'
Expected: returns departments; check that the response depth is at most 2 levels (manual inspection of parentId chains, or verify response size is smaller than before).
- Step 5: Commit
git add apps/server/src/integration/dingtalk.service.ts
git commit -m "feat: filter org-tree-with-users to deepest 2 levels"