refactor: build asset tree and perm tree, do not need add node_key into assets_asset_nodes table

apps/assets/migrations/0020_assetnodebackup.py

@@ -0,0 +1,25 @@
+# Generated by Django 4.1.13 on 2025-12-13 02:18
+
+from django.db import migrations, models
+
+
+class Migration(migrations.Migration):
+
+    dependencies = [
+        ('assets', '0019_alter_asset_connectivity'),
+    ]
+
+    operations = [
+        migrations.CreateModel(
+            name='AssetNodeBackup',
+            fields=[
+                ('id', models.AutoField(auto_created=True, primary_key=True, serialize=False, verbose_name='ID')),
+                ('asset_id', models.CharField(max_length=1024, verbose_name='Asset ID')),
+                ('node_id', models.CharField(max_length=1024, verbose_name='Node ID')),
+            ],
+            options={
+                'verbose_name': 'Asset Node Backup',
+                'db_table': 'assets_asset_nodes_backup',
+            },
+        ),
+    ]

apps/assets/migrations/0021_assetnode_alter_asset_nodes_and_more.py

@@ -0,0 +1,343 @@
+# Generated by Django 4.1.13 on 2025-12-12 03:55
+
+"""
+【数据迁移流程】
+
+本迁移将 Asset.nodes 从自动生成的 M2M through 表迁移到自定义的 AssetNode 模型，并添加 node_key 字段。
+
+五阶段迁移流程：
+
+【阶段1】读取 through 表并备份到 AssetNodeBackup 中
+  - 创建 AssetNodeBackup 表用于备份
+  - 读取原 Asset.nodes.through 表中所有 (asset_id, node_id) 数据
+  - 将数据保存到全局变量 through_old_data 中（内存缓存）
+  - 同时将数据备份到 AssetNodeBackup 表中（持久化，支持重试）
+  - 好处：即使迁移中断，再次执行时也可以从 backup 恢复
+
+【阶段2】数据库表结构修改
+  - 删除 Asset.nodes 的 M2M 关系字段
+  - 创建新的 AssetNode 自定义 through 模型
+  - 重新添加 Asset.nodes M2M 字段，指向新的 AssetNode
+  - 创建优化后的索引和修改一个联合唯一索引：
+    * idx_node_key_asset_id: 支持按 node_key 范围查询资产
+    * idx_node_id_asset_id: 支持按 node_id 查询资产
+    * idx_asset_id_node_id_key: 支持按 asset_id 反向查询节点
+    * unique_together (asset, node): 保证每个资产和节点组合唯一
+
+【阶段3】恢复数据并填充 node_key
+  - 优先使用内存缓存中的数据（through_old_data）
+  - 如果内存为空，说明之前可能迁移失败过，则从 AssetNodeBackup 表加载数据
+  - 预加载 Node.key 映射，为每条数据填充 node_key 字段
+  - 预加载已存在的 (asset_id, node_id) 对，避免重复插入
+  - 批量插入到 AssetNode 表中（50k/batch）
+  - 如果批插入失败，降级为单条插入
+  - 统计插入和跳过的记录数
+
+【阶段4】清理 through 表中的重复数据
+  - 找出原 through 表中 node_key 为空的数据（这些是重复或无效的）
+  - 显示前 100 条要删除的数据的 (asset_id, node_id)
+  - 分批删除（50k/batch）
+  - 输出删除的总数
+
+【阶段5】删除备份表 (或用户手动删除也可以)
+  - 验证 AssetNodeBackup 表中的数据（显示记录数）
+  - 清空备份表中的所有数据
+  - 删除 AssetNodeBackup 表
+  - 意义：迁移完成后，备份表已无用，清理数据库空间
+
+【数据一致性保证】
+  - 备份表：AssetNodeBackup 在阶段1中持久化所有原始数据，支持恢复
+  - 去重：阶段3 中使用 set 预检测避免重复
+  - 容错：阶段3 批插入失败时自动降级到单条插入
+  - 清理：阶段4 只删除 node_key 为空的无效数据
+  - 清理：阶段5 删除已完成使命的备份表
+
+【字段映射】
+  AssetNode.node_key 来自 Node.key
+"""
+
+import time
+from datetime import datetime
+from django.db import migrations, models, transaction
+from django.db.models import Count, Q
+import django.db.models.deletion
+import assets.models.asset.common
+
+
+# ============== 全局变量 ==============
+through_old_data = []
+migration_stats = {'backed_up': 0, 'restored': 0}
+
+
+def log(msg):
+    print(f"[{datetime.now().strftime('%H:%M:%S')}] {msg}")
+
+
+def load_data_from_backup(AssetNodeBackup):
+    """从 backup 表加载数据到内存"""
+    global through_old_data
+    total = AssetNodeBackup.objects.count()
+    
+    if total == 0:
+        log("⚠ backup 表为空，无数据可恢复")
+        return False
+    
+    log(f"从 backup 表加载 {total:,} 条数据...")
+    batch_size = 50000
+    start = time.time()
+    
+    for offset in range(0, total, batch_size):
+        batch = list(AssetNodeBackup.objects.all().values_list('asset_id', 'node_id')[offset:offset + batch_size])
+        through_old_data.extend(batch)
+        log(f"  已加载 {len(through_old_data):,}/{total:,}")
+    
+    log(f"✓ 从 backup 加载完成! 耗时 {time.time()-start:.1f}s")
+    return True
+
+
+def phase1_save_and_backup(apps, schema_editor):
+    """阶段1: 读取 through 所有数据，加载到全局变量，并保存到 backup 中"""
+    global through_old_data, migration_stats
+    Asset = apps.get_model('assets', 'Asset')
+    AssetNodeBackup = apps.get_model('assets', 'AssetNodeBackup')
+    asset_node_through = Asset.nodes.through
+    total = asset_node_through.objects.count()
+    
+    log(f"\n{'='*50}")
+    log("【阶段1】读取 through 数据并备份")
+    log(f"{'='*50}")
+    log(f"从 through 表读取 {total:,} 条数据...")
+    
+    batch_size = 50000
+    start = time.time()
+    backup_batch = []
+    
+    # 阶段1-1: 读取所有数据到内存
+    for offset in range(0, total, batch_size):
+        batch = list(asset_node_through.objects.all().values_list('asset_id', 'node_id')[offset:offset + batch_size])
+        through_old_data.extend(batch)
+        
+        # 准备备份数据
+        backup_objs = [AssetNodeBackup(asset_id=aid, node_id=nid) for aid, nid in batch]
+        backup_batch.extend(backup_objs)
+        
+        log(f"  已读取 {len(through_old_data):,}/{total:,} ({len(through_old_data)/total*100:.1f}%)")
+    
+    # 阶段1-2: 写入备份到数据库并立即提交
+    log(f"\n写入 {len(backup_batch):,} 条备份数据到数据库...")
+    backup_start = time.time()
+    
+    backup_batch_size = 50000
+    for i in range(0, len(backup_batch), backup_batch_size):
+        batch = backup_batch[i:i + backup_batch_size]
+        with transaction.atomic():
+            created = AssetNodeBackup.objects.bulk_create(batch, batch_size=backup_batch_size, ignore_conflicts=True)
+            migration_stats['backed_up'] += len(created)
+        log(f"  已备份 {min(i+backup_batch_size, len(backup_batch)):,}/{len(backup_batch):,}")
+    
+    log(f"✓ 阶段1完成! 读取耗时 {time.time()-start:.1f}s, 备份耗时 {time.time()-backup_start:.1f}s")
+    log(f"  内存缓存: {len(through_old_data):,} 条")
+    log(f"  数据库备份: {migration_stats['backed_up']:,} 条\n")
+
+
+def phase3_restore_data_and_set_node_key(apps, schema_editor):
+    """阶段3: 恢复数据时，先查看全局变量是否有，如果没有从backup中加载"""
+    global through_old_data, migration_stats
+    Node = apps.get_model('assets', 'Node')
+    AssetNode = apps.get_model('assets', 'AssetNode')
+    AssetNodeBackup = apps.get_model('assets', 'AssetNodeBackup')
+    
+    log(f"\n{'='*50}")
+    log("【阶段3】恢复数据并设置 node_key")
+    log(f"{'='*50}")
+    
+    # 检查内存是否有数据，如果没有则从 backup 加载
+    if not through_old_data:
+        log("内存缓存为空，从 backup 表加载数据...")
+        if not load_data_from_backup(AssetNodeBackup):
+            log("✗ 无法恢复数据：backup 表也为空")
+            return
+        log()
+    else:
+        log(f"使用内存缓存的 {len(through_old_data):,} 条数据\n")
+    
+    total = len(through_old_data)
+    log(f"开始恢复 {total:,} 条数据到 AssetNode 表...")
+    
+    # 预加载 node_key 映射
+    id_key_map = {str(item['id']): item['key'] for item in Node.objects.values('id', 'key')}
+    
+    # 预加载已存在的数据，避免重复
+    existing = set(AssetNode.objects.values_list('asset_id', 'node_id'))
+    log(f"数据库中已存在 {len(existing):,} 条记录\n")
+    
+    batch_size = 50000
+    start = time.time()
+    skipped = 0
+    
+    for i in range(0, total, batch_size):
+        batch = through_old_data[i:i + batch_size]
+        
+        # 去重：只保留不存在的记录
+        objs = []
+        for aid, nid in batch:
+            if (aid, nid) not in existing:
+                objs.append(AssetNode(asset_id=aid, node_id=nid, node_key=id_key_map.get(str(nid), '')))
+                existing.add((aid, nid))
+            else:
+                skipped += 1
+        
+        # 批量插入
+        if objs:
+            try:
+                AssetNode.objects.bulk_create(objs, batch_size=batch_size, ignore_conflicts=True)
+                migration_stats['restored'] += len(objs)
+            except Exception as e:
+                log(f"  ✗ 批插入失败: {str(e)}")
+                # 降级：逐条插入
+                for obj in objs:
+                    try:
+                        obj.save()
+                        migration_stats['restored'] += 1
+                    except Exception as save_err:
+                        log(f"    ✗ 跳过 asset_id={obj.asset_id}, node_id={obj.node_id}: {str(save_err)}")
+                        skipped += 1
+        
+        progress = min(i + batch_size, total)
+        log(f"  已恢复 {progress:,}/{total:,} (插入{migration_stats['restored']:,} 跳过{skipped:,})")
+    
+    log(f"✓ 阶段3完成! 耗时 {time.time()-start:.1f}s")
+    log(f"  插入: {migration_stats['restored']:,} 条")
+    log(f"  跳过: {skipped:,} 条\n")
+
+
+def phase4_cleanup_duplicates(apps, schema_editor):
+    """阶段4: 删除 through 表中 node_key 为空的数据"""
+    Asset = apps.get_model('assets', 'Asset')
+    asset_node_through = Asset.nodes.through
+    
+    log(f"\n{'='*50}")
+    log("【阶段4】清理 through 表中 node_key 为空的数据")
+    log(f"{'='*50}")
+    
+    # 找出 node_key 为空的记录
+    empty_node_key = asset_node_through.objects.filter(Q(node_key='') | Q(node_key__isnull=True))
+    total = empty_node_key.count()
+    
+    if total == 0:
+        log("✓ 没有 node_key 为空的数据，无需清理\n")
+        return
+    
+    log(f"发现 {total:,} 条 node_key 为空的数据")
+    
+    start = time.time()
+    batch_size = 50000
+    deleted = 0
+    
+    # 获取要删除的数据信息
+    to_delete_records = list(
+        empty_node_key.values_list('asset_id', 'node_id', 'id')
+    )
+    
+    log("删除详情:")
+    for aid, nid, record_id in to_delete_records[:100]:  # 显示前100条
+        log(f"  asset_id={aid}, node_id={nid}")
+    
+    if len(to_delete_records) > 100:
+        log(f"  ... 还有 {len(to_delete_records)-100:,} 条")
+    
+    # 分批删除
+    for offset in range(0, len(to_delete_records), batch_size):
+        batch_ids = [record_id for _, _, record_id in to_delete_records[offset:offset + batch_size]]
+        
+        if batch_ids:
+            delete_count, _ = asset_node_through.objects.filter(id__in=batch_ids).delete()
+            deleted += delete_count
+            log(f"  已删除 {deleted:,}/{total:,}")
+    
+    log(f"✓ 阶段4完成! 耗时 {time.time()-start:.1f}s")
+    log(f"  删除: {deleted:,} 条 node_key 为空的数据\n")
+
+
+def phase5_cleanup_backup_table(apps, schema_editor):
+    """阶段5: 删除备份表"""
+    log(f"\n{'='*50}")
+    log("【阶段5】删除 AssetNodeBackup 备份表")
+    log(f"{'='*50}")
+    
+    AssetNodeBackup = apps.get_model('assets', 'AssetNodeBackup')
+    total = AssetNodeBackup.objects.count()
+    
+    log(f"备份表中有 {total:,} 条数据")
+    
+    start = time.time()
+    
+    # 删除所有备份数据
+    delete_count, _ = AssetNodeBackup.objects.all().delete()
+    
+    log(f"✓ 删除 {delete_count:,} 条备份数据")
+    log(f"✓ 阶段5完成! 耗时 {time.time()-start:.1f}s\n")
+
+
+class Migration(migrations.Migration):
+
+    dependencies = [
+        ('assets', '0020_assetnodebackup'),
+    ]
+
+    operations = [
+        # 阶段1：保存和备份
+        migrations.RunPython(phase1_save_and_backup),
+        
+        # 阶段2：数据库表结构操作
+        migrations.RemoveField(
+            model_name='asset',
+            name='nodes',
+        ),
+        migrations.CreateModel(
+            name='AssetNode',
+            fields=[
+                ('id', models.AutoField(auto_created=True, primary_key=True, serialize=False, verbose_name='ID')),
+                ('node_key', models.CharField(db_index=True, default='', max_length=64, verbose_name='Node key')),
+                ('asset', models.ForeignKey(on_delete=django.db.models.deletion.CASCADE, to='assets.asset', verbose_name='Asset')),
+                ('node', models.ForeignKey(on_delete=django.db.models.deletion.CASCADE, to='assets.node', verbose_name='Node')),
+            ],
+            options={
+                'verbose_name': 'Asset Node',
+                'db_table': 'assets_asset_nodes',
+            },
+        ),
+        migrations.AddField(
+            model_name='asset',
+            name='nodes',
+            field=models.ManyToManyField(default=assets.models.asset.common.default_node, related_name='assets', through='assets.AssetNode', to='assets.node', verbose_name='Nodes'),
+        ),
+        migrations.AlterUniqueTogether(
+            name='assetnode',
+            unique_together={('asset', 'node_key'), ('asset', 'node')},
+        ),
+        migrations.AddIndex(
+            model_name='assetnode',
+            index=models.Index(fields=['node_key', 'asset_id'], name='idx_node_key_asset_id'),
+        ),
+        migrations.AddIndex(
+            model_name='assetnode',
+            index=models.Index(fields=['node_id', 'asset_id'], name='idx_node_id_asset_id'),
+        ),
+        migrations.AddIndex(
+            model_name='assetnode',
+            index=models.Index(fields=['asset_id', 'node_id', 'node_key'], name='idx_asset_id_node_id_key'),
+        ),
+        
+        # 阶段3：恢复数据
+        migrations.RunPython(phase3_restore_data_and_set_node_key),
+        
+        # 阶段4：清理重复数据
+        migrations.RunPython(phase4_cleanup_duplicates),
+        
+        # 阶段5：删除备份表 (或用户手动删除也可以)
+        # migrations.RunPython(phase5_cleanup_backup_table),
+        # migrations.DeleteModel(
+        #     name='AssetNodeBackup',
+        # ),
+    ]

apps/assets/models/__init__.py

@@ -9,3 +9,4 @@ from .node import *
 from .favorite_asset import *
 from .automations import *
 from .my_asset import *
+from .asset_node import *

apps/assets/models/asset/common.py

@@ -173,7 +173,12 @@ class Asset(NodesRelationMixin, LabeledMixin, AbsConnectivity, JSONFilterMixin,
         verbose_name=_("Zone"), on_delete=models.SET_NULL
     )
     nodes = models.ManyToManyField(
-        'assets.Node', default=default_node, related_name='assets', verbose_name=_("Nodes")
+        'assets.Node',
+        default=default_node,
+        related_name='assets',
+        verbose_name=_("Nodes"),
+        through='assets.AssetNode',  # 使用自定义 through 表
+        through_fields=('asset', 'node')
     )
     directory_services = models.ManyToManyField(
         'assets.DirectoryService', related_name='assets',

apps/assets/models/asset_node.py

@@ -0,0 +1,67 @@
+from django.db import models
+from django.db import models
+from django.utils.translation import gettext_lazy as _
+
+
+__all__ = ['AssetNode', 'AssetNodeBackup']
+
+
+class AssetNode(models.Model):
+    asset = models.ForeignKey(
+        'assets.Asset',
+        on_delete=models.CASCADE,
+        verbose_name=_('Asset'),
+        db_index=True
+    )
+    node = models.ForeignKey(
+        'assets.Node',
+        on_delete=models.CASCADE,
+        verbose_name=_('Node'),
+        db_index=True
+    )
+    node_key = models.CharField(
+        max_length=64,
+        verbose_name=_('Node key'),
+        db_index=True,
+        default=''
+    )
+    
+    class Meta:
+        db_table = 'assets_asset_nodes'
+        verbose_name = _('Asset Node')
+        
+        # unique_together: 定义哪些字段组合必须唯一
+        unique_together = [
+            ('asset', 'node'),  # asset_id + node_id 必须唯一
+            ('asset', 'node_key'), # 必须加上 node_key 保证唯一
+        ]
+        
+        # indexes: 定义查询优化的索引
+        indexes = [
+            # 索引 1: 按 node_key 查询所有资产（子孙节点）
+            # 查询: WHERE node_key LIKE '1.12:%'
+            models.Index(fields=['node_key', 'asset_id'], name='idx_node_key_asset_id'),
+            
+            # 索引 2: 按 node_id 查询所有直接资产
+            # 查询: WHERE node_id = ?
+            models.Index(fields=['node_id', 'asset_id'], name='idx_node_id_asset_id'),
+            
+            # 索引 3: 按 asset_id 查询 node_key 和 node_id
+            # 查询: WHERE asset_id = ? 获取 node_key 或 node_id
+            models.Index(fields=['asset_id', 'node_id', 'node_key'], name='idx_asset_id_node_id_key'),
+        ]
+
+
+class AssetNodeBackup(models.Model):
+    asset_id = models.CharField(
+        max_length=1024,
+        verbose_name=_('Asset ID'),
+    )
+    node_id = models.CharField(
+        max_length=1024,
+        verbose_name=_('Node ID'),
+    )
+    
+    class Meta:
+        db_table = 'assets_asset_nodes_backup'
+        verbose_name = _('Asset Node Backup')

apps/assets/models/node.py

@@ -65,7 +65,7 @@ class FamilyMixin:
     def get_nodes_children_key_pattern(cls, nodes, with_self=True):
         keys = [i.key for i in nodes]
         keys = cls.clean_children_keys(keys)
-        patterns = [cls.get_node_all_children_key_pattern(key) for key in keys]
+        patterns = [cls.get_node_all_children_key_pattern(key, with_self=with_self) for key in keys]
         patterns = '|'.join(patterns)
         return patterns
 

apps/assets/signal_handlers/__init__.py

@@ -1,3 +1,4 @@
 from .asset import *
 from .node_assets_amount import *
 from .node_assets_mapping import *
+from .node_asset import *

apps/assets/signal_handlers/node_asset.py

@@ -0,0 +1,24 @@
+from django.db.models.signals import pre_save, m2m_changed, post_save
+from django.dispatch import receiver
+
+from assets.models import AssetNode, Asset, Node
+
+
+@receiver(m2m_changed, sender=Asset.nodes.through)
+def fill_node_key_on_m2m_change(sender, instance, action, pk_set, **kwargs):
+    if action == 'post_add':
+        if isinstance(instance, Asset):
+            asset_ids = [str(instance.id)]
+            node_ids = [str(pk) for pk in pk_set]
+        elif isinstance(instance, Node):
+            asset_ids = [str(pk) for pk in pk_set]
+            node_ids = [str(instance.id)]
+        else:
+            return
+
+        id_key_pairs = Node.objects.filter(id__in=node_ids).values_list('id', 'key')
+        id_key_map = {str(id_): key for id_, key in id_key_pairs}   
+        rs = AssetNode.objects.filter(asset_id__in=asset_ids, node_id__in=node_ids)
+        for r in rs:
+            r.node_key = id_key_map.get(str(r.node_id), '')
+        AssetNode.objects.bulk_update(rs, ['node_key'], batch_size=5000)

apps/perms/utils/user_perm_tree_engine.py

@@ -0,0 +1,477 @@
+import re, time
+from typing import Optional
+from collections import defaultdict
+
+from django.db.models import F
+
+from users.models import User
+from assets.models import Asset, Node
+from perms.models import AssetPermission
+from common.utils import lazyproperty, timeit
+from orgs.utils import current_org
+from django.core.cache import cache
+from functools import wraps
+
+def cache_tree(ttl=30):
+    """权限树缓存装饰器"""
+    def decorator(func):
+        @wraps(func)
+        def wrapper(self, *args, **kwargs):
+            # 构建缓存 key
+            cache_key = f"perm_tree:{self._org_id}:{self._user_id}"
+            
+            # 尝试从缓存获取
+            result = cache.get(cache_key)
+            if result is not None:
+                return result
+            
+            # 执行原方法
+            result = func(self, *args, **kwargs)
+            
+            # 缓存结果
+            cache.set(cache_key, result, ttl)
+            return result
+        return wrapper
+    return decorator
+
+
+class TreeNode:
+
+    separator = ':'
+
+    class Type:
+        BRIDGE = 'bridge'
+        OWNER = 'owner'
+        DA = 'da'
+
+    def __init__(self, key, tp, assets=None):
+        self.key = key
+        self.type = tp
+        # 节点下的直接资产集合，不包含子孙节点的资产
+        self._assets = set() if assets is None else set(assets)
+        self._assets_amount = 0
+    
+    def add_assets(self, asset_ids):
+        self._assets.update(asset_ids)
+
+    @property
+    def assets(self):
+        return self._assets
+    
+    @property
+    def assets_amount(self):
+        return self._assets_amount
+
+    def assets_amount_increment(self, amount=1):
+        self._assets_amount += amount
+
+    def is_children(self, other: 'TreeNode'):
+        pattern = r'^{0}:[0-9]+$'.format(other.key)
+        return bool(re.match(pattern, self.key))
+    
+    def is_all_children(self, other: 'TreeNode'):
+        pattern = r'^{0}:'.format(other.key)
+        return bool(re.match(pattern, self.key))
+    
+    def can_be_overridden(self, other: 'TreeNode'):
+        """
+        # 不可以
+        owner owner
+        owner bridge
+        owner da
+
+        # 可以
+        da da
+        # 可以
+        da owner
+        # 不可以
+        da bridge
+
+        # 可以
+        bridge da
+        bridge owner
+        bridge bridge
+        """
+        if self.key != other.key:
+            return False
+        if self.type == self.Type.OWNER:
+            return False
+        if self.type == self.Type.DA and other.type == self.Type.BRIDGE:
+            return False
+        return True
+
+
+class Tree:
+
+    def __init__(self, nodes: Optional[list[TreeNode]] = None, org_id=None):
+        # {node_key: TreeNode}
+        self._nodes = defaultdict(TreeNode)
+        self._org_id = org_id
+        self.init(nodes)
+    
+    def init(self, nodes: Optional[list[TreeNode]]):
+        if nodes is None:
+            return
+        for node in nodes:
+            self.add_node(node)
+        self._reverse_generated()
+    
+    def _reverse_generated(self):
+        """ 逆向生成树 """
+        for key in list(self._nodes.keys()):
+            ancestor_keys = Node.get_node_ancestor_keys(key)
+            for ancestor_key in ancestor_keys:
+                # 自动生成的祖先节点默认标记为 bridge, 添加时会判断是否要覆盖已经存在的节点
+                tree_node = TreeNode(key=ancestor_key, tp=TreeNode.Type.BRIDGE)
+                self.add_node(tree_node)
+    
+    def merge(self, other: 'Tree') -> 'Tree':
+        merged_tree = Tree()
+        for node in self._nodes.values():
+            merged_tree.add_node(node)
+        for node in other._nodes.values():
+            merged_tree.add_node(node)
+        merged_tree._finalize()
+        return merged_tree
+    
+    def _finalize(self):
+        self._prune()
+        self._init_owner_nodes_children()
+        self._compute_assets_amount()
+        self._sorted()
+    
+    def _sorted(self):
+        self._nodes = defaultdict(
+            TreeNode, 
+            sorted(self._nodes.items(), key=lambda item: [int(i) for i in item[0].split(':')])
+        )
+    
+    @timeit
+    def _init_owner_nodes_children(self):
+        """ 初始化 Owner-Node 的所有子孙节点以及其下的直接资产 """
+        t1 = time.time()
+        owner_nodes = self._owner_nodes
+        if not owner_nodes:
+            return
+        nodes = Node.get_nodes_all_children(owner_nodes, with_self=True)
+        node_id_key_sets = nodes.annotate(char_id=F('id')).values_list('char_id', 'key')
+        node_id_key_mapper = dict(node_id_key_sets)
+
+        node_ids = node_id_key_mapper.keys()
+        t2 = time.time()
+        nid_aid_sets = Node.assets.through.objects.filter(node_id__in=node_ids).annotate(
+            char_nid=F('node_id'), char_aid=F('asset_id')).values_list('char_nid', 'char_aid')
+        
+        nid_aid_sets = list(nid_aid_sets)
+        t3 = time.time()
+        print('Fetch node-assets sets time by node_id__in: {:.1f}ms'.format((t3 - t2) * 1000))
+        t4 = time.time()
+        print('Fetch all node-assets sets time: {:.1f}ms'.format((t4 - t3) * 1000))
+        
+        for nid, aid in nid_aid_sets:
+            key = node_id_key_mapper.get(nid)
+            if not key:
+                continue
+            tree_node = self._nodes.get(key)
+            if tree_node:
+                tree_node.add_assets({aid})
+            else:
+                tree_node = self.wrap_as_tree_node(node_key=key, tp=TreeNode.Type.OWNER, assets={aid})
+                self.add_node(tree_node)
+    
+    def _compute_assets_amount(self):
+        """
+        生成数据结构:
+        {
+            "asset_id": set("node_key1", "node_key2" ...), # 资产所在的直接节点
+        }
+        迭代，对每个资产所在的节点的所有祖先节点取并集+去重, +1
+        """
+        aid_node_keys_mapper = defaultdict(set)
+        for node in self._nodes.values():
+            for aid in node.assets:
+                aid_node_keys_mapper[aid].add(node.key)
+        
+        for aid, node_keys in aid_node_keys_mapper.items():
+            ancestor_keys = set(self.get_ancestor_keys(node_keys)) # 必须去重
+            for ancestor_key in ancestor_keys:
+                tree_node = self._nodes.get(ancestor_key)
+                if not tree_node:
+                    continue
+                tree_node.assets_amount_increment()
+        
+    def get_ancestor_keys(self, keys, with_self=True):
+        ancestor_keys = set()
+        for k in keys:
+            _ancestor_keys = Node.get_node_ancestor_keys(k, with_self=with_self)
+            ancestor_keys.update(_ancestor_keys)
+        return ancestor_keys
+
+    def _prune(self):
+        self._prune_owner_nodes_branch()
+    
+    def _prune_owner_nodes_branch(self):
+        # 修剪所有 owner nodes 节点的分枝（保留每条 owner 节点分枝的最上一层，删除其所有子孙节点）
+        owner_node_keys = [n.key for n in self._owner_nodes]
+        for node in list(self._nodes.values()):
+            ancestor_keys = Node.get_node_ancestor_keys(node.key)
+            if set(ancestor_keys) & set(owner_node_keys):
+                self.remove_node(node)
+    
+    @property
+    def _owner_nodes(self):
+        return [node for node in self._nodes.values() if node.type == TreeNode.Type.OWNER]
+
+    def get_node_all_assets(self, node: TreeNode):
+        """ 获取节点下的所有资产 """
+        nodes = self.get_node_all_children(node)
+        assets = set()
+        for n in nodes:
+            assets.update(n.assets)
+        return assets
+    
+    def get_node_all_children(self, node: TreeNode):
+        """ 获取节点的所有子孙节点 """
+        children = [
+            n for n in self._nodes.values() if n.is_all_children(node)
+        ]
+        return children
+    
+    def get_node_children(self, node: TreeNode):
+        """ 获取节点的直接子节点 """
+        children = [
+            n for n in self._nodes.values() if n.is_children(node)
+        ]
+        return children
+
+    def get_node(self, key) -> Optional[TreeNode]:
+        return self._nodes.get(key)
+    
+    def add_node(self, node: TreeNode):
+        _node = self._nodes.get(node.key)
+        if _node is None:
+            self._nodes[node.key] = node
+            return
+        if _node.can_be_overridden(node):
+            self._nodes[node.key] = node
+            return
+    
+    def remove_node(self, node_or_key: 'TreeNode | str'):
+        if isinstance(node_or_key, TreeNode):
+            key = node_or_key.key
+        else:
+            key = node_or_key
+        self._nodes.pop(key, None)
+
+    @classmethod
+    def wrap_as_tree_node(cls, node_key, tp, assets=None):
+        return TreeNode(key=node_key, tp=tp, assets=assets)
+    
+    @classmethod
+    def wrap_as_tree_nodes(cls, node_keys, tp):
+        return [cls.wrap_as_tree_node(nk, tp) for nk in node_keys]
+
+    def print_nodes(self):
+        print('--- Tree Nodes ---')
+        for n in self._nodes.values():
+            print(f'{n.key}({n.assets_amount}) - {n.type}')
+
+
+class UserPermTreeEngine(object):
+    """
+        DA: Directly Permed Asset 
+        DN: Directly Permed Node
+
+        DA-Tree: 通过直接授权的资产生成的树
+        DN-Tree: 通过直接授权的节点生成的树
+
+        Perm-Tree: 最终的权限树，由 DA-Tree 和 DN-Tree 合并生成，bridge 和 da 节点全部保留，owner 节点只保留第一级
+
+        Tree-Node-Type:
+            bridge: 所有权桥梁节点，没有直接授权节点，也没有授权它下的资产
+            owner: 所有权节点，直接授权的节点
+            da: DA 节点，仅授权它下的资产
+    """
+
+    def __init__(self, user, org_id=None):
+        self.user = user
+        self._user_id = str(user.id)
+        self._org_id = org_id or current_org.id
+        self._tree = self.tree()
+
+    @cache_tree(ttl=5)
+    def tree(self):
+        da_tree = self._generate_da_tree()
+        dn_tree = self._generate_dn_tree()
+        tree = self._merge_trees(da_tree, dn_tree)
+        return tree
+
+    def _generate_da_tree(self):
+        node_assets_mapper = self._get_da_node_key_asset_ids_mapper()
+        tree_nodes = [
+            TreeNode(key=key, tp=TreeNode.Type.DA, assets=asset_ids) 
+            for key, asset_ids in node_assets_mapper.items()
+        ]
+        tree = Tree(nodes=tree_nodes)
+        return tree
+
+    def _get_da_node_key_asset_ids_mapper(self):
+        direct_asset_ids = AssetPermission.assets.through.objects \
+            .filter(assetpermission_id__in=self._perm_ids) \
+            .annotate(char_id=F('asset_id')).values_list('char_id', flat=True)
+        nid_aid_set = Asset.nodes.through.objects.filter(asset_id__in=direct_asset_ids) \
+            .annotate(char_nid=F('node_id'), char_aid=F('asset_id')).values_list('char_nid', 'char_aid')
+        nid_aid_mapper = dict(nid_aid_set)
+
+        node_ids = list(nid_aid_mapper.keys())
+        node_id_key_set = Node.objects.filter(id__in=node_ids).annotate(char_id=F('id')).values_list('id', 'key')
+        node_id_key_mapper = dict(node_id_key_set)
+
+        mapper = defaultdict(set)
+        for nid, aid in nid_aid_set:
+            key = node_id_key_mapper.get(nid)
+            if key:
+                mapper[key].add(aid)
+        return mapper
+
+    def _generate_dn_tree(self):
+        node_keys = self._get_dn_node_keys()
+        nodes = Tree.wrap_as_tree_nodes(node_keys, TreeNode.Type.OWNER)
+        tree = Tree(nodes=nodes)
+        return tree
+
+    def _get_dn_node_keys(self):
+        node_ids = AssetPermission.nodes.through.objects.filter(assetpermission_id__in=self._perm_ids) \
+            .annotate(char_id=F('node_id')).values_list('char_id', flat=True)
+        node_keys = Node.objects.filter(id__in=node_ids).values_list('key', flat=True)
+        return list(set(node_keys))
+    
+    def _merge_trees(self, da_tree: Tree, dn_tree: Tree) -> Tree:
+        tree = da_tree.merge(dn_tree)
+        return tree
+
+    @lazyproperty
+    def _perm_ids(self):
+        return self._get_permission_ids()
+
+    def _get_permission_ids(self):
+        user_perm_ids = AssetPermission.users.through.objects.filter(user_id=self._user_id).annotate(
+            char_id=F('assetpermission_id')).values_list('char_id', flat=True)
+        group_ids = User.groups.through.objects.filter(user_id=self._user_id).annotate(
+            char_id=F('usergroup_id')).values_list('char_id', flat=True)
+        group_perm_ids = AssetPermission.user_groups.through.objects.filter(usergroup_id__in=group_ids).annotate(
+            char_id=F('assetpermission_id')).values_list('char_id', flat=True)
+        perm_ids = set(user_perm_ids).union(set(group_perm_ids))
+        return perm_ids
+
+    def get_node_children(self, node_key, with_assets=False):
+        "Luna 页面会调用此方法"
+        tree_node = self.tree.get_node(node_key)
+        if not tree_node:
+            return None
+        children = self.tree.get_node_children(tree_node)
+        data = {"children": children}
+        if with_assets:
+            data.update({"assets": tree_node.assets})
+        return data
+
+    def get_node_all_assets(self, node_key):
+        " 用户详情页面会调用此方法 "
+        node = self.tree.get_node(node_key)
+        if not node:
+            return None
+        assets = self.tree.get_node_all_assets(node)
+        return {
+            "assets": assets
+        }
+
+    from common.utils import timeit
+    @timeit
+    def query_3_result(guessed_asset_amount=50000, view_old=False, sql2_raw=False, sql1_raw=False, sql3_raw=False):
+        from django.db import connection, connections
+        from django.conf import settings
+        from django.db.models import OuterRef, Subquery, Count
+        import time
+
+        settings.DEBUG = True
+        connections.close_all()
+
+        with connection.cursor() as cursor:
+            # 1 查询节点资产数量
+            # cursor.execute(sql1)
+            # node_key_asset_amount_tuple = cursor.fetchall()
+
+            t1 = time.time()
+            if sql1_raw:
+                # 不能用 raw 因为node下没有资产在表中不存在，node 获取不完整
+                sql1 = """
+                SELECT node_id, COUNT(*) AS assets_count
+                FROM assets_asset_nodes
+                GROUP BY node_id
+                """
+                cursor.execute(sql1)
+                node_id_asset_amount_rows = cursor.fetchall()
+            else:
+                count_sub = Node.assets.through.objects.filter(
+                    node_id=OuterRef("id")
+                ).values("node_id").annotate(c=Count("id")).values("c")
+
+                node_id_asset_amount_rows = Node.objects.annotate(
+                    assets_count=Subquery(count_sub)
+                ).values("id", "assets_count")
+            node_id_asset_amount_rows = list(node_id_asset_amount_rows)
+
+            t2 = time.time()
+            # 2 查询属于多个节点的资产ID
+            if sql2_raw:
+                sql2 = """
+                SELECT asset_id FROM assets_asset_nodes
+                GROUP BY asset_id HAVING COUNT(*) > 1
+                """
+                cursor.execute(sql2)
+                rows = cursor.fetchall()
+                asset_ids = [row[0] for row in rows]
+            else:
+                count_sub = Asset.nodes.through.objects.filter(
+                    asset_id=OuterRef("id")
+                ).values("asset_id").annotate(c=Count("id")).values("c")
+
+                asset_id_node_amount_row = Asset.objects.annotate(
+                    nodes_count=Subquery(count_sub)
+                ).values_list('id', 'nodes_count')
+                print(asset_id_node_amount_row[0])
+                asset_ids = [str(row[0]) for row in asset_id_node_amount_row if row[1] and row[1] > 1]
+            print('Assets belong to multiple nodes:', len(asset_ids))
+
+            t3 = time.time()
+            # 3 查询资产ID和节点ID的对应关系 (只查 2 的资产)
+            # 假设 asset_ids 不多，只查前 guessed_asset_amount 个
+            guessed_asset_ids = asset_ids[:guessed_asset_amount]
+            if sql3_raw:
+                print('Guessed asset ids count:', len(guessed_asset_ids))
+                sql3 = """
+                SELECT asset_id, node_id FROM assets_asset_nodes
+                WHERE asset_id IN ({})
+                """.format(','.join(['%s'] * len(guessed_asset_ids)))
+                cursor.execute(sql3, guessed_asset_ids)
+                aid_nid_set = cursor.fetchall()
+            else:
+                aid_nid_set = Node.assets.through.objects.filter(asset_id__in=guessed_asset_ids).values_list('asset_id', 'node_id')
+            aid_nid_set = list(aid_nid_set)
+            t4 = time.time()
+            # aid_nid_set 获取每个 aid 的 parent_ids, 获取两两 parent 的祖先节点的交集 -1
+        
+        print('Query times: sql1 {:.2f}s, sql2 {:.2f}s, sql3 {:.2f}s'.format(t2 - t1, t3 - t2, t4 - t3), 
+              len(node_id_asset_amount_rows), len(asset_ids), len(aid_nid_set))
+            
+
+        print('New ORM query time: {:.2f}s, total rows: {}'.format(t4 - t1, len(aid_nid_set)))
+
+        # old 
+
+        if view_old:
+            t1 = time.time()
+            old = list(Node.assets.through.objects.all())
+            t2 = time.time()
+            print('Old ORM query time: {:.2f}s, total rows: {}'.format(t2 - t1, len(old)))
+        return node_id_asset_amount_rows, asset_ids, aid_nid_set

test_build_asset_tree_perm_tree.py

@@ -0,0 +1,926 @@
+#!/usr/bin/env python
+"""
+测试查询资产树和授权树下指定节点的资产总数
+"""
+
+import os, sys, django, json, time
+from datetime import datetime
+import copy
+
+os.environ.setdefault('DJANGO_SETTINGS_MODULE', 'jumpserver.settings')
+sys.path.insert(0, os.path.join(os.getcwd(), 'apps'))
+django.setup()
+
+from functools import reduce
+from operator import or_
+from operator import or_
+from django.db.models import Q
+from collections import defaultdict
+
+from orgs.models import Organization
+from users.models import User, UserGroup
+from rbac.models import OrgRoleBinding
+from assets.models import Node, Asset
+from perms.models import AssetPermission
+from django.core.cache import cache
+
+
+AssetNodeThrough = Asset.nodes.through
+
+BUILD_ASSET_TREE = True
+# BUILD_ASSET_TREE = False
+
+BUILD_USER_PERM_TREE = True
+# BUILD_USER_PERM_TREE = False
+TEST_USER_AMOUNT = 1
+
+# 输出文件路径
+OUTPUT_DIR = os.path.join(os.getcwd(), 'mapper_output')
+if not os.path.exists(OUTPUT_DIR):
+    os.makedirs(OUTPUT_DIR)
+
+#
+# =========================== 辅助函数 ===============================
+#
+
+def log(msg=''):
+    """打印带时间戳的日志"""
+    print(f"[{datetime.now().strftime('%H:%M:%S')}] {msg}")
+
+
+def remove_asset_belong_many_node_record_from_asset_node_through():
+    """删除 Asset.nodes.through 表中重复的记录：如果一个资产属于多个不同的 node_key，只保留一个"""
+    log("开始清理 AssetNodeThrough 表中一个资产属于多个节点的重复记录...")
+    
+    from django.db.models import Count
+    
+    # 获取所有属于多个 node_key 的 asset_id
+    duplicates = AssetNodeThrough.objects.values('asset_id').annotate(
+        node_key_count=Count('node_key', distinct=True)
+    ).filter(node_key_count__gt=1)
+    
+    total_deleted = 0
+    
+    for dup in duplicates:
+        asset_id = dup['asset_id']
+        
+        # 获取该资产关联的所有 node_key
+        records = AssetNodeThrough.objects.filter(asset_id=asset_id).order_by('id')
+        
+        # 保留第一条，删除其他的
+        records_to_delete = records[1:]
+        delete_count = records_to_delete.count()
+        
+        for record in records_to_delete:
+            record.delete()
+        
+        total_deleted += delete_count
+        
+        if delete_count > 0:
+            keep_record = records[0]
+            log(f"  资产 {asset_id} 从多个节点删除到只属于一个: {keep_record.node_key}, 删除数量: {delete_count}")
+    
+    log(f"✓ 清理完成，共删除 {total_deleted} 条重复记录")
+    return total_deleted
+
+
+def get_node_level(node_key):
+    """获取节点层级（根据冒号分割）"""
+    if not node_key:
+        return 0
+    return len(node_key.split(':'))
+
+
+def write_asset_mapper_to_file(org_id, mapper, node_times=None, asset_count=0, through_count=0, total_time=0):
+    """将 mapper 按树形结构写入文件，子节点放在父节点下，显示节点时间和下一级时间之和"""
+    output_file = os.path.join(OUTPUT_DIR, f'mapper_org_{org_id}.txt')
+    if node_times is None:
+        node_times = {}
+    
+    # 计算树的深度和节点数最大宽度
+    max_count = 0
+    level_mapper = defaultdict(list)
+    for key, count in mapper.items():
+        level = get_node_level(key)
+        level_mapper[level].append((key, count))
+        max_count = max(max_count, count)
+    
+    tree_depth = max(level_mapper.keys()) if level_mapper else 0
+    max_width = len(f"{max_count:,}")
+    
+    # 构建树形结构（字典树）
+    tree = {}
+    for key, count in mapper.items():
+        parts = key.split(':')
+        current = tree
+        for i, part in enumerate(parts):
+            node_key = ':'.join(parts[:i+1])
+            if node_key not in current:
+                current[node_key] = {'count': count if i == len(parts) - 1 else None, 'children': {}}
+            current = current[node_key]['children']
+    
+    # 递归写入树形结构
+    def write_tree(node_dict, f, prefix='', node_times_dict=None):
+        if node_times_dict is None:
+            node_times_dict = {}
+        
+        for i, (key, node_data) in enumerate(sorted(node_dict.items())):
+            is_last = (i == len(node_dict) - 1)
+            
+            # 获取当前节点的计算时间
+            current_time = node_times_dict.get(key, 0)
+            
+            # 计算下一级时间之和（直接子节点的时间）
+            next_level_time = 0
+            if node_data['children']:
+                for child_key in node_data['children'].keys():
+                    next_level_time += node_times_dict.get(child_key, 0)
+            
+            # 树形符号
+            current_prefix = prefix + ('└── ' if is_last else '├── ')
+            
+            # 获取直接下级节点数
+            child_count = len(node_data['children']) if node_data['children'] else 0
+            
+            # 写入节点信息：节点 (资产数) [时间信息]
+            if node_data['count'] is not None:
+                time_info = f"当前: {current_time:.3f}s, 下级: {next_level_time:.3f}s, 下级节点: {child_count}"
+                f.write(f"{current_prefix}{key} ({node_data['count']:,}) [{time_info}]\n")
+                # 空行但保持竖线连续
+                blank_prefix = prefix + ('    ' if is_last else '│   ')
+                f.write(f"{blank_prefix}\n")
+            
+            # 递归写入子节点
+            if node_data['children']:
+                next_prefix = prefix + ('    ' if is_last else '│   ')
+                write_tree(node_data['children'], f, next_prefix, node_times_dict)
+    
+    # 写入文件
+    with open(output_file, 'w') as f:
+        f.write(f"{'='*80}\n")
+        f.write(f"组织 ID: {org_id}\n")
+        f.write(f"生成时间: {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}\n")
+        f.write(f"总耗时: {total_time:.2f}s\n")
+        f.write(f"总节点数: {len(mapper):,}\n")
+        f.write(f"总资产数: {asset_count:,}\n")
+        f.write(f"总关系表数: {through_count:,}\n")
+        f.write(f"树深度: {tree_depth}\n")
+        f.write(f"节点数最大宽度: {max_width}\n")
+        f.write(f"{'='*80}\n\n")
+        
+        # 写入树形结构
+        write_tree(tree, f, node_times_dict=node_times)
+    
+    log(f"✓ mapper 已写入文件:")
+    log(f"  {output_file}")
+
+
+def write_user_perm_mapper_to_file(user_name, user_id, org_id, mapper, node_times, tree_nodes=None):
+    """将 mapper 按树形结构写入文件，显示竖线和横线，每个节点后显示计算时间和节点信息"""
+    if tree_nodes is None:
+        tree_nodes = {}
+    output_file = os.path.join(OUTPUT_DIR, f'mapper_tree_{user_name}_{user_id}_{org_id}.txt')
+    
+    # 构建树形结构 - 关键：每个节点只在其直接子节点中，不是所有节点都在顶层
+    tree = {}
+    for key, count in mapper.items():
+        parts = key.split(':')
+        current = tree
+        
+        # 从根节点开始逐级构建
+        for i, part in enumerate(parts):
+            node_key = ':'.join(parts[:i+1])
+            
+            # 如果这个节点还不存在，创建它
+            if node_key not in current:
+                # 只有最后一层才有真实的 count，中间层是 None
+                is_leaf = (i == len(parts) - 1)
+                current[node_key] = {
+                    'count': count if is_leaf else None,
+                    'children': {}
+                }
+            
+            # 移到这个节点的 children 字典，为下一层节点做准备
+            current = current[node_key]['children']
+    
+    # 计算子节点的时间信息
+    def get_children_time_info(node_dict):
+        """返回 (直接子节点时间和, 直接子节点数, 孙子及以后节点时间和)"""
+        direct_children_time = 0
+        direct_children_count = 0
+        deeper_descendants_time = 0
+        
+        for child_key, child_data in node_dict.items():
+            # 直接子节点时间
+            direct_children_time += node_times.get(child_key, 0)
+            direct_children_count += 1
+            
+            # 递归计算孙子及以后的节点时间
+            if child_data['children']:
+                deeper_direct_time, deeper_direct_count, deeper_deeper_time = get_children_time_info(child_data['children'])
+                deeper_descendants_time += deeper_direct_time + deeper_deeper_time
+        
+        return direct_children_time, direct_children_count, deeper_descendants_time
+    
+    # 递归输出树结构
+    def write_tree_recursive(node_dict, f, prefix=''):
+        items = sorted(node_dict.items())
+        
+        # 在每一层级的开头加一个空行
+        if items:
+            extension = '│   '
+            f.write(f"{prefix}{extension}\n")
+        
+        for i, (key, node_data) in enumerate(items):
+            is_last = (i == len(items) - 1)
+            
+            # 从 mapper 中获取真实的 count 值
+            count_value = mapper.get(key, None)
+            if count_value is not None:
+                count_str = f"({count_value})"
+            else:
+                count_str = "(None)"
+            
+            # 获取该节点的计算时间
+            node_time = node_times.get(key, 0)
+            
+            # 计算直接子节点和更深层节点的时间
+            direct_children_time, direct_children_count, deeper_descendants_time = get_children_time_info(node_data['children'])
+            
+            # 格式化时间信息
+            time_str = f" [当前: {node_time:.3f}s"
+            if direct_children_count > 0:
+                time_str += f", 直接子节点: {direct_children_time:.3f}s ({direct_children_count}个)"
+            time_str += "]"
+            
+            # 获取节点的 da_asset_ids 和 dn_node_keys 信息
+            tree_node = tree_nodes.get(key)
+            node_info_str = ""
+            if tree_node:
+                da_count = len(tree_node.da_asset_ids) if hasattr(tree_node, 'da_asset_ids') else 0
+                dn_count = len(tree_node.dn_node_keys) if hasattr(tree_node, 'dn_node_keys') else 0
+                if da_count > 0 or dn_count > 0:
+                    node_info_str = f" [da_assets: {da_count}, dn_nodes: {dn_count}]"
+            
+            # 当前行的连接符
+            connector = '└── ' if is_last else '├── '
+            line = f"{prefix}{connector}{key} {count_str}{time_str}{node_info_str}\n"
+            f.write(line)
+            
+            # 递归写入子节点
+            if node_data['children']:
+                # 为下一层计算新的前缀
+                extension = '    ' if is_last else '│   '
+                new_prefix = prefix + extension
+                write_tree_recursive(node_data['children'], f, new_prefix)
+            
+            # 如果不是最后一个节点，在后面加空行
+            if not is_last:
+                extension = '│   '
+                f.write(f"{prefix}{extension}\n")
+    
+    # 写入文件
+    with open(output_file, 'w') as f:
+        f.write(f"{'='*80}\n")
+        f.write(f"用户: {user_name}\n")
+        f.write(f"用户ID: {user_id}\n")
+        f.write(f"组织ID: {org_id}\n\n")
+        
+        # 输出统计信息（用 ===== 包起来）
+        f.write(f"{'='*80}\n")
+        f.write(f"【统计信息】\n")
+        f.write(f"{'='*80}\n\n")
+        
+        # 获取根节点 keys
+        root_keys = sorted(tree.keys())
+        
+        # 逐个输出根节点及其子树
+        for key in root_keys:
+            node_data = tree[key]
+            count_value = mapper.get(key, None)
+            if count_value is not None:
+                count_str = f"({count_value})"
+            else:
+                count_str = "(None)"
+            
+            # 获取根节点的计算时间
+            node_time = node_times.get(key, 0)
+            
+            # 计算直接子节点和更深层节点的时间
+            direct_children_time, direct_children_count, deeper_descendants_time = get_children_time_info(node_data['children'])
+            
+            # 格式化时间信息
+            time_str = f" [当前: {node_time:.3f}s"
+            if direct_children_count > 0:
+                time_str += f", 直接子节点: {direct_children_time:.3f}s ({direct_children_count}个)"
+            time_str += "]"
+            
+            # 获取节点的 da_asset_ids 和 dn_node_keys 信息
+            tree_node = tree_nodes.get(key)
+            node_info_str = ""
+            if tree_node:
+                da_count = len(tree_node.da_asset_ids) if hasattr(tree_node, 'da_asset_ids') else 0
+                dn_count = len(tree_node.dn_node_keys) if hasattr(tree_node, 'dn_node_keys') else 0
+                if da_count > 0 or dn_count > 0:
+                    node_info_str = f" [da_assets: {da_count}, dn_nodes: {dn_count}]"
+            
+            f.write(f"{key} {count_str}{time_str}{node_info_str}\n")
+            
+            # 递归输出子树
+            if node_data['children']:
+                write_tree_recursive(node_data['children'], f, prefix='')
+    
+    log(f"✓ mapper_tree 已写入文件:")
+    log(f"  {output_file}")
+
+
+def query_node_assets_count(node_key):
+    count = AssetNodeThrough.objects.filter(
+         Q(node_key=node_key) | Q(node_key__startswith=node_key + ':')
+    ).distinct('asset_id').count()
+    return count
+
+
+def build_org_asset_tree(org_id):
+    """构建单个组织的资产树"""
+    log(f"开始构建组织 {org_id} 的资产树...")
+    
+    # 获取该组织的统计信息
+    nodes = Node.objects.filter(org_id=org_id).order_by('key')
+    node_count = nodes.count()
+    
+    assets = Asset.objects.filter(org_id=org_id)
+    asset_count = assets.count()
+    
+    through_count = AssetNodeThrough.objects.filter(asset__org_id=org_id).count()
+    
+    log(f"  节点数: {node_count:,}")
+    log(f"  资产数: {asset_count:,}")
+    log(f"  关系表总数: {through_count:,}")
+    
+    node_times = {}  # 记录每个节点的计算时间
+    max_count = 0
+    level_mapper = defaultdict(list)
+    
+    # =======================核心代码 ==========================
+    step_start = datetime.now()
+    mapper = defaultdict(int)
+    for i, node in enumerate(nodes):
+        node_start = datetime.now()
+        
+        count = query_node_assets_count(node.key)
+        mapper[node.key] = count
+        max_count = max(max_count, count)
+        
+        level = get_node_level(node.key)
+        level_mapper[level].append(count)
+        
+        # 记录该节点的计算时间
+        node_times[node.key] = (datetime.now() - node_start).total_seconds()
+    # =======================核心代码 ==========================
+    
+        if (i + 1) % 1000 == 0:
+            log(f"  已处理 {i+1:,}/{node_count:,} 个节点")
+    step_time = (datetime.now() - step_start).total_seconds()
+    log(f"  耗时: {step_time:.2f}s")
+    
+    # 计算树深度和节点数最大宽度
+    tree_depth = max(level_mapper.keys()) if level_mapper else 0
+    max_width = len(f"{max_count:,}")
+    
+    log(f"  树深度: {tree_depth}")
+    log(f"  节点数最大宽度: {max_width}")
+    
+    
+    # 写入文件
+    write_asset_mapper_to_file(org_id, mapper, node_times, asset_count, through_count, step_time)
+    
+    return {
+        'org_id': org_id,
+        'node_count': node_count,
+        'asset_count': asset_count,
+        'through_count': through_count,
+        'time': step_time
+    }
+
+
+#
+# =========================== 构建资产树 ===============================
+#
+
+def build_orgs_asset_tree():
+    """构建所有组织的资产树"""
+    log(f"\n{'='*60}")
+    log("【开始构建所有组织的资产树】")
+    log(f"{'='*60}\n")
+    
+    total_start = datetime.now()
+    org_ids = list(Organization.objects.all().values_list('id', flat=True))[:1]
+    org_count = len(org_ids)
+    log(f"发现 {org_count:,} 个组织\n")
+    
+    # 收集每个组织的统计信息
+    org_stats = []
+    total_nodes = 0
+    total_assets = 0
+    total_through = 0
+    
+    # =======================核心代码 ==========================
+    for org_id in org_ids:
+        stats = build_org_asset_tree(str(org_id))
+    # =======================核心代码 ==========================
+        org_stats.append(stats)
+        total_nodes += stats['node_count']
+        total_assets += stats['asset_count']
+        total_through += stats['through_count']
+        log()  # 空行分隔
+    
+    total_time = (datetime.now() - total_start).total_seconds()
+    
+    # 输出总体统计
+    log(f"{'='*60}")
+    log("【组织统计详情】")
+    log(f"{'='*60}")
+    log(f"{'组织ID':<12} {'节点数':<12} {'资产数':<12} {'关系总数':<12} {'耗时(s)':<10}")
+    log(f"{'-'*60}")
+    
+    for stats in org_stats:
+        log(f"{str(stats['org_id']):<12} {stats['node_count']:<12,} {stats['asset_count']:<12,} {stats['through_count']:<12,} {stats['time']:<10.2f}")
+    
+    log(f"{'-'*60}")
+    log(f"{'合计':<12} {total_nodes:<12,} {total_assets:<12,} {total_through:<12,}")
+    
+    log(f"\n{'='*60}")
+    log("【全局统计】")
+    log(f"{'='*60}")
+    log(f"总组织数: {org_count:,}")
+    log(f"总节点数: {total_nodes:,}")
+    log(f"总资产数: {total_assets:,}")
+    log(f"关系表总数: {total_through:,}")
+    log(f"总耗时: {total_time:.2f}s")
+    if org_count > 0:
+        log(f"平均耗时/组织: {total_time/org_count:.2f}s")
+    log(f"输出目录: {OUTPUT_DIR}")
+    log(f"{'='*60}\n")
+
+
+#
+# =========================== 构建用户授权树 ===========================
+#
+
+
+class TreeNode:
+    class Type:
+        BRIDGE = 'bridge'
+        DA = 'da-node'
+        OWNER = 'owner-node'
+
+    def __init__(self, key, type=None):
+        self.key = key
+        self.type = type
+        self.assets_count = 0
+        self.da_asset_ids = set() # 当前节点下所有子孙的 da 资产 ID, 用于查询
+        self.dn_node_keys = set() # 当前节点下所有子孙的 dn 节点 key, 用于查询
+    
+    def overridden_da_from_dn(self, dn):
+        """da-node 被 dn 覆盖时，合并资产 ID"""
+        # bridge owner
+        # da owner
+        # bridge bridge
+        if dn.type == self.Type.OWNER:
+            return dn
+        if self.type in [self.Type.DA, self.Type.BRIDGE] and dn.type == self.Type.DA:
+            self.dn_node_keys.update(dn.dn_node_keys)
+        return self
+        
+
+    def get_ancestor_keys(self, with_self=False):
+        parent_keys = []
+        key_list = self.key.split(":")
+        if not with_self:
+            key_list.pop()
+        for i in range(len(key_list)):
+            parent_keys.append(":".join(key_list))
+            key_list.pop()
+        return parent_keys
+
+    def get_all_children_keys(self, with_self=False):
+        children_keys = []
+        key_prefix = self.key + ":"
+        if with_self:
+            children_keys.append(self.key)
+        for node_key in Node.objects.filter(key__startswith=key_prefix).values_list('key', flat=True):
+            children_keys.append(node_key)
+        return children_keys
+
+
+def reverse_build_da_perm_tree(da_nodes):
+    """逆向构建完整树节点，补全 bridge-node, 同时把 da_asset_ids 挂上到祖先节点中"""
+    complete_tree = defaultdict(TreeNode)
+    
+    # 创建所有缺失的 bridge 节点
+    for tn in da_nodes.values():
+        # 先把自己加进去
+        complete_tree[tn.key] = tn
+        ancestor_keys = tn.get_ancestor_keys(with_self=False)  # 不包含自己
+        for key in ancestor_keys:
+            an = complete_tree.get(key)
+            if not an:
+                an = TreeNode(key, type=TreeNode.Type.BRIDGE)
+            an.da_asset_ids.update(tn.da_asset_ids)
+            complete_tree[key] = an
+    
+    return complete_tree
+
+
+def prune_owner_nodes(tree):
+    # prune_owner_nodes 只保留最上一层的 owner-node
+    owner_nodes = {k: v for k, v in tree.items() if v.type == TreeNode.Type.OWNER}
+    for tn in owner_nodes.values():
+        ancestor_keys = tn.get_ancestor_keys(with_self=False)
+        if set(ancestor_keys) & set(owner_nodes.keys()):
+            tree.pop(tn.key)
+
+
+def reverse_build_dn_perm_tree(dn_nodes):
+    """逆向构建完整树节点，补全 bridge-node，同时把每一个 owner-node 挂上到祖先节点中"""
+    complete_tree = copy.deepcopy(dn_nodes)
+    
+    # 第一步：创建所有缺失的 bridge 节点
+    for tn in dn_nodes.values():
+        ancestor_keys = tn.get_ancestor_keys(with_self=False)  # 不包含自己
+        for key in ancestor_keys:
+            if key not in complete_tree:
+                complete_tree[key] = TreeNode(key, type=TreeNode.Type.BRIDGE)
+
+    # 第二步：给所有父节点挂上 owner-node
+    for tn in dn_nodes.values():
+        ancestor_keys = tn.get_ancestor_keys(with_self=False)  # 不包含自己
+        for key in ancestor_keys:
+            n = complete_tree[key]
+            n.dn_node_keys.add(tn.key)
+    
+    prune_owner_nodes(complete_tree)
+    return complete_tree
+
+
+def merge_trees(da_tree, dn_tree):
+    tree = {k: v for k, v in da_tree.items()}
+    # 合并 dn_tree 到 da_tree 上
+    for dn in dn_tree.values():
+        da = tree.get(dn.key)
+        if dn.type == TreeNode.Type.OWNER:
+            tree[dn.key] = dn
+            continue
+        if dn.type == TreeNode.Type.BRIDGE:
+            da.dn_node_keys.update(dn.dn_node_keys)
+            continue
+
+    # 最后要删除 owner node 下的所有节点
+    owner_nodes = {v for v in tree.values() if v.type == TreeNode.Type.OWNER}
+    for on in owner_nodes:
+        ac_keys = [k for k in tree.keys() if k.startswith(on.key + ':')]
+        for k in ac_keys:
+            tree.pop(k)
+
+    return tree
+
+
+def get_tree_node_all_da_asset_ids(tree, tree_node):
+    asset_ids = set()
+    all_da_nodes_children = [ 
+        n for n in tree.values()
+        if n.type == TreeNode.Type.DA and (n.key == tree_node.key or n.key.startswith(tree_node.key + ':')) 
+    ]
+    for node in all_da_nodes_children:
+        asset_ids.update(node.da_asset_ids)
+    return asset_ids
+
+
+def get_tree_node_all_dn_nodes(tree, tree_node):
+    all_dn_nodes_children = [ 
+        n for n in tree.values()
+        if n.type == TreeNode.Type.OWNER and (n.key == tree_node.key or n.key.startswith(tree_node.key + ':')) 
+    ]
+    return set(all_dn_nodes_children)
+
+
+def complete_all_dn_children_nodes(tree, org_node_keys):
+    """补全所有 dn 子节点"""
+    complete_tree = copy.deepcopy(tree)
+    for node in tree.values():
+        if node.type != TreeNode.Type.OWNER:
+            continue
+        _keys = [ k for k in org_node_keys if k.startswith(node.key + ':') ]
+        for key in _keys:
+            complete_tree[key] = TreeNode(key, type=TreeNode.Type.OWNER)
+    return complete_tree
+
+
+def query_user_perm_node_assets_count(da_ids, tree_node):
+    # 如果是 dn node, 查询方式和资产树节点的完全一样
+    # 构造当前节点下所有资产数量的 Q 对象
+    q_count = Q(node_key=tree_node.key) | Q(node_key__startswith=tree_node.key + ':')
+
+    if tree_node.type == TreeNode.Type.OWNER:
+        count = AssetNodeThrough.objects.filter(q_count).distinct('asset_id').count()
+        return count
+    # 如果是 da node 或 bridge node, 需要把 da_asset_ids 和 dn_node_keys 都加进去
+    # 构造查询用户当前节点下所有授权的全量资产的 Q 对象, 因为在 DB 中要基于这个子表查 (后面的逻辑和资产树节点的查询逻辑一样)
+    # 会在 sql 语句传输层面上让数据量小一些
+    q_assets = Q()
+
+    if tree_node.da_asset_ids:
+        _q = Q(asset_id__in=tree_node.da_asset_ids)
+        q_assets |= _q
+
+    if tree_node.dn_node_keys:
+        _q = [ 
+            Q(node_key=n_key) | Q(node_key__startswith=n_key + ':') 
+            for n_key in tree_node.dn_node_keys 
+        ]
+        q_assets |= reduce(or_, _q)
+    
+    # TODO: 待考虑 先 q_assets, 还是 q_count 效率更高?
+    count = AssetNodeThrough.objects.filter(q_assets).filter(q_count).distinct('asset_id').count()
+    return count
+
+
+def build_user_org_perm_tree(user, org_id, org=None):
+    """构建单个用户在组织下的授权树"""
+    org_name = org.name if org else "没有组织"
+    log(f"  构建用户 {user.name}({user.id}) 在组织 {org_name}({org_id}) 的授权树")
+    step_start = datetime.now()
+    
+    node_times = {}
+    
+    # =======================核心代码 - 数据构造阶段 ==========================
+    build_start = datetime.now()
+    # 查询所有该用户在该组织下的资产权限 perm_ids
+    group_ids = User.groups.through.objects.filter(user_id=user.id).values_list('usergroup_id', flat=True)
+    group_perm_ids = AssetPermission.user_groups.through.objects.filter(usergroup_id__in=group_ids).values_list('assetpermission_id', flat=True)
+    user_perm_ids = AssetPermission.users.through.objects.filter(user_id=user.id).values_list('assetpermission_id', flat=True)
+    orgs_perm_ids = set(list(group_perm_ids) + list(user_perm_ids))
+    org_perm_ids = AssetPermission.objects.filter(id__in=orgs_perm_ids, org_id=org_id).values_list('id', flat=True)
+
+    # 构建直接授权资产的节点树
+    da_asset_ids = AssetPermission.assets.through.objects.filter(assetpermission_id__in=org_perm_ids).distinct('asset_id').values_list('asset_id', flat=True)
+    da_asset_parent_node_keys = AssetNodeThrough.objects.filter(asset_id__in=da_asset_ids).distinct('asset_id').values_list('asset_id', 'node_key')
+    # 挂上直接资产 da_asset_ids
+    da_tree_nodes = defaultdict(TreeNode)
+    for asset_id, node_key in da_asset_parent_node_keys: # 元组 (asset_id, node_key) 不同得 asset_id 可能对应相同 node_key
+        tn = da_tree_nodes.get(node_key)
+        if not tn:
+            tn = TreeNode(node_key, type=TreeNode.Type.DA)
+        tn.da_asset_ids.add(asset_id)
+        da_tree_nodes[node_key]  = tn
+    # 逆向生成 da-tree, 补全 bridge-node
+    da_tree = reverse_build_da_perm_tree(da_tree_nodes)
+
+    # 构建直接授权节点的节点树
+    dn_ids = AssetPermission.nodes.through.objects.filter(assetpermission_id__in=org_perm_ids).distinct().values_list('node_id', flat=True)
+    dn_node_keys = Node.objects.filter(id__in=dn_ids).values_list('key', flat=True)
+    dn_tree_nodes = {k: TreeNode(k, type=TreeNode.Type.OWNER) for k in dn_node_keys}
+    dn_tree = reverse_build_dn_perm_tree(dn_tree_nodes)
+    # 合并成一颗用户上半棵完整授权树
+    _tree = merge_trees(da_tree, dn_tree)
+    # 补全所有 dn 子节点
+    org_node_keys = set(Node.objects.filter(org_id=org_id).values_list('key', flat=True))
+    tree = complete_all_dn_children_nodes(_tree, org_node_keys)
+    
+    build_time = (datetime.now() - build_start).total_seconds()
+
+    calc_start = datetime.now()
+    node_details = {}  # 记录每个节点的详细信息
+
+    # 计算每个节点的资产数
+    # =======================核心代码 - Mapper计算阶段 ==========================
+    mapper = {}
+    for tn in tree.values():
+        node_start = datetime.now()
+        count = query_user_perm_node_assets_count(da_asset_ids, tn)
+        mapper[tn.key] = count
+    # =======================核心代码 =========================================
+        node_times[tn.key] = (datetime.now() - node_start).total_seconds()
+        # 记录节点的详细信息 - 用于查询时使用的 da_assets 和 dn_nodes
+        node_details[tn.key] = {
+            'query_da_assets': len(tn.da_asset_ids),
+            'query_dn_nodes': len(tn.dn_node_keys),
+            'node_type': tn.type
+        }
+    calc_time = (datetime.now() - calc_start).total_seconds()
+    
+    step_time = (datetime.now() - step_start).total_seconds()
+    log(f"    节点数: {len(mapper):,}")
+    log(f"    数据构造时间: {build_time:.2f}s")
+    log(f"    Mapper计算时间: {calc_time:.2f}s")
+    log(f"    总耗时: {step_time:.2f}s")
+    
+    # 写入文件
+    write_user_perm_mapper_to_file(user.name, user.id, org_id, mapper, node_times, tree_nodes=tree)
+    
+    return {
+        'user_id': user.id,
+        'org_id': org_id,
+        'node_count': len(mapper),
+        'time': step_time
+    }
+
+
+def build_users_perm_tree():
+    """构建所有用户的授权树"""
+    orgs_mapper = { str(org.id): org for org in Organization.objects.all() }
+    log(f"\n{'='*60}")
+    log("【开始构建所有用户的授权树】")
+    log(f"{'='*60}\n")
+    
+    total_start = datetime.now()
+    users = User.objects.all()[:TEST_USER_AMOUNT]
+    user_count = users.count()
+    log(f"发现 {user_count:,} 个用户\n")
+
+    
+    user_index = 1
+    # =======================核心代码 ==========================
+    for user in users:
+        org_ids = OrgRoleBinding.objects.filter(user=user).distinct('org_id').values_list('org_id', flat=True)
+        log(f"用户 {user_index}: {user.name}({user.id}) - {len(org_ids)} 个组织")
+        for org_id in org_ids:
+            org = orgs_mapper.get(str(org_id))
+            build_user_org_perm_tree(user, str(org_id), org)
+    # =======================核心代码 ==========================
+        log()  # 用户之间的空行
+        user_index += 1
+    
+    total_time = (datetime.now() - total_start).total_seconds()
+    log(f"\n{'='*60}")
+    log("【用户授权树构建完成】")
+    log(f"总耗时: {total_time:.2f}s")
+    log(f"{'='*60}\n")
+
+
+def get_ancestor_keys(key):
+    parent_keys = []
+    key_list = key.split(":")
+    for i in range(len(key_list)):
+        parent_keys.append(":".join(key_list))
+        key_list.pop()
+    return parent_keys
+
+
+#
+# ============== X 方案 ==================
+#
+
+
+def x_build_user_org_perm_tree(user, org_id, use_cache=False):
+    print('构建用户授权树....')
+    t1 = time.time()
+    user_perm_tree = cache.get("user_perm_tree") if use_cache else None
+    if not user_perm_tree:
+        group_ids = User.groups.through.objects.filter(user_id=user.id).values_list('usergroup_id', flat=True)
+        group_perm_ids = AssetPermission.user_groups.through.objects.filter(usergroup_id__in=group_ids).values_list('assetpermission_id', flat=True)
+        user_perm_ids = AssetPermission.users.through.objects.filter(user_id=user.id).values_list('assetpermission_id', flat=True)
+        orgs_perm_ids = set(list(group_perm_ids) + list(user_perm_ids))
+        org_perm_ids = AssetPermission.objects.filter(id__in=orgs_perm_ids, org_id=org_id).values_list('id', flat=True)
+
+        da_ids = AssetPermission.assets.through.objects.filter(assetpermission_id__in=org_perm_ids).distinct('asset_id').values_list('asset_id', flat=True)
+
+        dn_ids = AssetPermission.nodes.through.objects.filter(assetpermission_id__in=org_perm_ids).distinct().values_list('node_id', flat=True)
+        dn_keys = Node.objects.filter(id__in=dn_ids).values_list('key', flat=True)
+
+        print("未命中缓存，查询所有资产节点关系对...")
+
+        # q 就是查询用户所有授权的资产
+
+        # 这是冗余 node_key 方案: 最最根本的原因
+        # 这一步是关键，根据直接授权节点，找到所有子孙节点
+        # 找出所有子孙节点 的 id
+        t11 = time.time()
+
+        # q = [Q(node_key=key) | Q(node_key__startswith=key + ':') for key in dn_keys] + [Q(asset__id__in=da_ids)]
+        # aid_nk_pairs = list(AssetNodeThrough.objects.filter(reduce(or_, q)).values_list('asset_id', 'node_key'))
+
+        t12 = time.time()
+        print(f"冗余 key: 查询资产节点关系对耗时: {t12 - t11:.2f}s, 关系对数量: {len(aid_nk_pairs):,}")
+
+        # 不冗余 node_key 方案: 直接计算所有 dn_ids 下的所有子孙节点
+        node_ids = set()
+        node_id_key_pairs = dict(Node.objects.filter(org_id=org_id).values_list('id', 'key'))
+        for nid in dn_ids:
+            node_ids.add(nid)
+            nk = node_id_key_pairs[nid]
+            children_ids = [ _id for _id, key in node_id_key_pairs.items() if key.startswith(nk + ':') ]
+            node_ids.update(children_ids)
+        
+        aid_nid_pairs = list(AssetNodeThrough.objects.filter( Q(asset__id__in=da_ids) | Q(node_id__in=node_ids) ).values_list('asset_id', 'node_id'))
+        aid_nk_pairs = [ (aid, node_id_key_pairs[nid]) for aid, nid in aid_nid_pairs ]
+
+        t13 = time.time()
+        print(f"非冗余 key: 查询资产节点关系对耗时: {t13 - t12:.2f}s, 关系对数量: {len(aid_nk_pairs):,}")
+
+
+        mapper = defaultdict(set)
+        for aid, nk in aid_nk_pairs:
+            mapper[nk].add(aid)
+            an_ks = get_ancestor_keys(nk)
+            for ak in an_ks:
+                mapper[ak].add(aid)
+        user_perm_tree = {k: len(v) for k, v in mapper.items()}
+        cache.set("user_perm_tree", user_perm_tree, 3600)
+    else:
+        print("命中缓存，直接使用缓存的资产节点关系对...")
+
+
+    t2 = time.time()
+
+
+    # 格式化打印前10个 mapper key
+    # mapper_keys_sorted = sorted(mapper.keys())
+    # print("\n【前10个 mapper key (已排序)】")
+    # for i, key in enumerate(mapper_keys_sorted[:10], 1):
+    #     print(f"  {i}. {key} ({len(mapper[key]):,} 资产)")
+
+    print(
+        f"""
+        ===============用户授权树构建 - 基本信息=============================
+        用户: {user.name}({user.id})
+        组织: {org_id}
+        资产总数: {Asset.objects.filter(org_id=org_id).count():,}
+        节点总数: {Node.objects.filter(org_id=org_id).count():,}
+        总耗时: {t2 - t1:.2f}s
+
+        测试数据: root-node - {user_perm_tree['1']}
+        ===================================================================
+        """
+    )
+
+
+def x_build_org_asset_tree(org_id, use_cache=False):
+    print('构建资产树....')
+    t1 = time.time()
+    asset_tree = cache.get("asset_tree") if use_cache else None
+
+    if not asset_tree:
+        # node_keys = Node.objects.filter(org_id=org_id).values_list('key', flat=True)
+        print("未命中缓存，查询所有资产节点关系对...")
+        node_id_key_pairs = dict(Node.objects.filter(org_id=org_id).values_list('id', 'key'))
+        node_ids = list(node_id_key_pairs.keys())
+        aid_nid_pairs = list(AssetNodeThrough.objects.filter(node_id__in=node_ids).values_list('asset_id', 'node_id'))
+        mapper = defaultdict(set)
+        for aid, nid in aid_nid_pairs:
+            nk = node_id_key_pairs[nid]
+            mapper[nk].add(aid)
+            an_ks = get_ancestor_keys(nk)
+            for ak in an_ks:
+                mapper[ak].add(aid)
+        asset_tree = {k: len(v) for k, v in mapper.items()}
+        cache.set("asset_tree", asset_tree, 3600)
+    else:
+        print("命中缓存，直接使用缓存的资产树...")
+    t2 = time.time()
+    
+    print('..........', asset_tree['1'])
+    print(
+        f"""
+        ===============构建信息=============================
+        组织: {org_id}
+        资产总数: {Asset.objects.filter(org_id=org_id).count():,}
+        节点总数: {Node.objects.filter(org_id=org_id).count():,}
+        总耗时: {t2 - t1:.2f}s
+
+        测试数据: root-node - {asset_tree['1']}
+        ===================================================================
+        """
+    )
+
+
+
+def main():
+    u = User.objects.get(username='admin')
+    org_id = Organization.default().org_id
+    try:
+        # =======================核心代码 ==========================
+        # if BUILD_ASSET_TREE:
+            # build_orgs_asset_tree()
+        # if BUILD_USER_PERM_TREE:
+            # build_users_perm_tree()
+        t1 = time.time()
+        x_build_user_org_perm_tree(u, org_id)
+        # =======================核心代码 ==========================
+        t2 = time.time()
+        x_build_org_asset_tree(org_id)
+        log(f"\n总耗时: {t2 - t1:.2f}s")
+        
+        log(f"{'='*60}")
+        log("✓ 所有操作完成！")
+        log(f"{'='*60}\n")
+        
+    except Exception as e:
+        log(f"\n✗ 发生错误: {str(e)}")
+        import traceback
+        traceback.print_exc()
+
+if __name__ == '__main__':
+    # remove_asset_belong_many_node_record_from_asset_node_through()
+    main()

test_build_asset_tree_perm_tree_x.py

@@ -0,0 +1,243 @@
+#!/usr/bin/env python
+"""
+测试查询资产树和授权树下指定节点的资产总数
+"""
+
+import os, sys, django, json, time
+from datetime import datetime
+import copy
+
+os.environ.setdefault('DJANGO_SETTINGS_MODULE', 'jumpserver.settings')
+sys.path.insert(0, os.path.join(os.getcwd(), 'apps'))
+django.setup()
+
+from functools import reduce
+from operator import or_
+from operator import or_
+from django.db.models import Q
+from collections import defaultdict
+
+from orgs.models import Organization
+from users.models import User, UserGroup
+from rbac.models import OrgRoleBinding
+from assets.models import Node, Asset
+from perms.models import AssetPermission
+from django.core.cache import cache
+
+
+#
+# ============== X 方案 ==================
+#
+
+AssetNodeThrough = Node.assets.through
+
+
+def get_ancestor_keys(key):
+    parent_keys = []
+    key_list = key.split(":")
+    for i in range(len(key_list)):
+        parent_keys.append(":".join(key_list))
+        key_list.pop()
+    return parent_keys
+
+
+def x_build_user_org_perm_tree(user, org_id, use_cache=False):
+    print(f'构建用户授权树 ({user.username})....')
+    t1 = time.time()
+    user_perm_tree = cache.get("user_perm_tree") if use_cache else None
+    if not user_perm_tree:
+        group_ids = User.groups.through.objects.filter(user_id=user.id).values_list('usergroup_id', flat=True)
+        group_perm_ids = AssetPermission.user_groups.through.objects.filter(usergroup_id__in=group_ids).values_list('assetpermission_id', flat=True)
+        user_perm_ids = AssetPermission.users.through.objects.filter(user_id=user.id).values_list('assetpermission_id', flat=True)
+        orgs_perm_ids = set(list(group_perm_ids) + list(user_perm_ids))
+        org_perm_ids = AssetPermission.objects.filter(id__in=orgs_perm_ids, org_id=org_id).values_list('id', flat=True)
+
+        # 使用 set() 在 Python 中去重，而不是使用 distinct('asset_id')（MySQL 不支持）
+        # da_ids_all = AssetPermission.assets.through.objects.filter(assetpermission_id__in=org_perm_ids).distinct('asset_id').values_list('asset_id', flat=True)
+        da_ids_all = AssetPermission.assets.through.objects.filter(assetpermission_id__in=org_perm_ids).values_list('asset_id', flat=True)
+        da_ids = set(da_ids_all)
+
+        # 获取所有直接授权的节点 ID
+        dn_ids_all = AssetPermission.nodes.through.objects.filter(assetpermission_id__in=org_perm_ids).values_list('node_id', flat=True)
+        dn_ids = set(dn_ids_all)
+
+        print("未命中缓存，查询所有资产节点关系对...")
+
+
+        # 这是冗余 node_key 方案: 最最根本的原因
+        # 这一步是关键，根据直接授权节点，找到所有子孙节点
+        # 找出所有子孙节点 的 id
+        t11 = time.time()
+        # q 就是查询用户所有授权的资产
+        # dn_keys = Node.objects.filter(id__in=dn_ids).values_list('key', flat=True)
+        # q = [Q(node_key=key) | Q(node_key__startswith=key + ':') for key in dn_keys] + [Q(asset__id__in=da_ids)]
+        # aid_nk_pairs = list(AssetNodeThrough.objects.filter(reduce(or_, q)).values_list('asset_id', 'node_key'))
+
+        # t12 = time.time()
+        # print(f"冗余 key: 查询资产节点关系对耗时: {t12 - t11:.2f}s, 关系对数量: {len(aid_nk_pairs):,}")
+
+        # 不冗余 node_key 方案: 直接计算所有 dn_ids 下的所有子孙节点
+        node_ids = set()
+        node_id_key_pairs = dict(Node.objects.filter(org_id=org_id).values_list('id', 'key'))
+        for nid in dn_ids:
+            node_ids.add(nid)
+            nk = node_id_key_pairs[nid]
+            children_ids = [ _id for _id, key in node_id_key_pairs.items() if key.startswith(nk + ':') ]
+            node_ids.update(children_ids)
+        
+        aid_nid_pairs = list(AssetNodeThrough.objects.filter( Q(asset__id__in=da_ids) | Q(node_id__in=node_ids) ).values_list('asset_id', 'node_id'))
+        aid_nk_pairs = [ (aid, node_id_key_pairs[nid]) for aid, nid in aid_nid_pairs ]
+
+        t13 = time.time()
+        print(f"非冗余 key: 查询资产节点关系对耗时: {t13 - t11:.2f}s, 关系对数量: {len(aid_nk_pairs):,}")
+
+
+        mapper = defaultdict(set)
+        for aid, nk in aid_nk_pairs:
+            mapper[nk].add(aid)
+            an_ks = get_ancestor_keys(nk)
+            for ak in an_ks:
+                mapper[ak].add(aid)
+        user_perm_tree = {k: len(v) for k, v in mapper.items()}
+        cache.set("user_perm_tree", user_perm_tree, 3600)
+    else:
+        print("命中缓存，直接使用缓存的资产节点关系对...")
+
+
+    t2 = time.time()
+
+
+    # 格式化打印前10个 mapper key
+    # mapper_keys_sorted = sorted(mapper.keys())
+    # print("\n【前10个 mapper key (已排序)】")
+    # for i, key in enumerate(mapper_keys_sorted[:10], 1):
+    #     print(f"  {i}. {key} ({len(mapper[key]):,} 资产)")
+
+    ROOT_KEY = [k for k in user_perm_tree.keys() if k.isdigit() ]
+    if ROOT_KEY:
+        ROOT_KEY = ROOT_KEY[0]
+    else:
+        print("用户授权树中没有根节点，可能用户没有任何授权资产。")
+        return
+
+    print(
+        f"""
+        ===============用户授权树构建 - 基本信息=============================
+        用户: {user.name}({user.id})
+        组织: {org_id}
+        资产总数: {Asset.objects.filter(org_id=org_id).count():,}
+        节点总数: {Node.objects.filter(org_id=org_id).count():,}
+        资产节点关系对总数: {AssetNodeThrough.objects.filter(node__org_id=org_id).count():,}
+
+        查出来的资产节点关系对总数: {len(aid_nk_pairs):,}
+        直接授权资产数(用于查询): {len(da_ids):,}
+        直接授权节点数: {len(dn_ids):,}
+        用于查询的节点数: {len(node_ids):,}
+        总耗时: {t2 - t1:.2f}s
+
+        ROOT_KEY: {ROOT_KEY},
+        用于验证数据: ROOT Node count- {user_perm_tree[ROOT_KEY]} 通过rebuild_user_org_perm_tree 函数构建的授权树资产数一致就对了。
+        ===================================================================
+        """
+    )
+    from perms.utils.user_perm_tree import UserPermTreeRefreshUtil
+    from perms.models import UserAssetGrantedTreeNodeRelation
+    UserPermTreeRefreshUtil(user=user)._rebuild_user_perm_tree_for_org(org_id)
+    count = UserAssetGrantedTreeNodeRelation.objects.filter(user=user).get(node__key=ROOT_KEY).node_assets_amount
+    print(f'''
+            使用原始方法获取授权树 ROOT Node count:
+            通过 UserAssetGrantedTreeNodeRelation 刷新得到的授权 ROOT Node count - {count} 
+          ''')
+
+
+
+def x_build_org_asset_tree(org_id, use_cache=False):
+    print('构建资产树....')
+    t1 = time.time()
+    asset_tree = cache.get("asset_tree") if use_cache else None
+
+    if not asset_tree:
+        # node_keys = Node.objects.filter(org_id=org_id).values_list('key', flat=True)
+        print("未命中缓存，查询所有资产节点关系对...")
+        node_id_key_pairs = dict(Node.objects.filter(org_id=org_id).values_list('id', 'key'))
+        node_ids = list(node_id_key_pairs.keys())
+        aid_nid_pairs = list(AssetNodeThrough.objects.filter(node_id__in=node_ids).values_list('asset_id', 'node_id'))
+        mapper = defaultdict(set)
+        for aid, nid in aid_nid_pairs:
+            nk = node_id_key_pairs[nid]
+            mapper[nk].add(aid)
+            an_ks = get_ancestor_keys(nk)
+            for ak in an_ks:
+                mapper[ak].add(aid)
+        asset_tree = {k: len(v) for k, v in mapper.items()}
+        cache.set("asset_tree", asset_tree, 3600)
+    else:
+        print("命中缓存，直接使用缓存的资产树...")
+    t2 = time.time()
+    
+    ROOT_KEY = [k for k in asset_tree.keys() if k.isdigit() ][0]
+    r_aids = list(dict(aid_nid_pairs).keys())
+    a_ids = list(Asset.objects.filter(org_id=org_id).values_list('id', flat=True))
+    _aids = Asset.objects.filter(nodes__isnull=True).count()
+    print(
+        f"""
+        ===============构建信息=============================
+        组织: {org_id}
+        组织内资产总数: {Asset.objects.filter(org_id=org_id).count():,}
+        组织内节点总数: {Node.objects.filter(org_id=org_id).count():,}
+        资产节点关系对总数: {AssetNodeThrough.objects.count():,}
+        总耗时: {t2 - t1:.2f}s
+
+        用于验证数据: Root node - {asset_tree[ROOT_KEY]} 等于组织内资产总数就对了。
+        ===================================================================
+
+        找出资产在关系表中 和 资产表中 不一致的数据:
+        关系表中资产数: {len(r_aids):,}
+        资产表中资产数: {len(a_ids):,}
+        在关系表中有，但资产表中没有的数量: {len(set(r_aids) - set(a_ids)):,} 说明资产被删除了，但是资产和节点的关系还在
+        在资产表中有，但关系表中没有的数量: {len(set(a_ids) - set(r_aids)):,}, 说明资产没有挂载节点上
+        游离资产总数: { _aids }
+        """
+    )
+
+
+
+def main():
+    users = User.objects.filter(username='admin')
+    # users = User.objects.all()
+    org = Organization.objects.filter(name__contains='性能').first()
+    # org = Organization.objects.filter().first()
+    x_build_org_asset_tree(org.id)
+    for u in users:
+        x_build_user_org_perm_tree(u, org.id)
+
+    # print(f"\n总耗时: {t2 - t1:.2f}s")
+    
+    print(f"{'='*60}")
+    print("✓ 所有操作完成！")
+    print(f"{'='*60}\n")
+        
+    
+def setup_test_data():
+    print("设置测试数据...")
+
+    org = Organization.objects.filter(name__contains='性能').first()
+    if not org:
+        print("未找到测试组织，请先创建一个名称包含 '性能' 的组织。")
+        return
+    user = User.objects.filter(username='admin').first()
+    a_ids = Asset.objects.filter(org_id=org.id).values_list('id', flat=True)[:30000]
+    n_ids = Node.objects.filter(org_id=org.id).values_list('id', flat=True)[:1000]
+    p1, p2 = AssetPermission.objects.filter(org_id=org.id)[:2]
+    p1.users.add(user)
+    p1.assets.add(*a_ids)
+
+    p2.users.add(user)
+    p2.nodes.add(*n_ids)
+    print(f'给用户 {user.username} 分配两个权限..., 一个直接授权 {len(a_ids)} 资产，一个直接授权 {len(n_ids)} 节点')
+
+if __name__ == '__main__':
+    setup_test_data()
+    main()
+
+    # 

test_generate_asset_node_through_data.py

@@ -0,0 +1,350 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+生成随机的 Asset-Node 关系数据
+用法: python generate_asset_node_through_data.py
+"""
+
+import os
+import sys
+import django
+import random
+from datetime import datetime
+
+# 配置 Django 设置
+os.environ.setdefault('DJANGO_SETTINGS_MODULE', 'jumpserver.settings')
+sys.path.insert(0, '/Users/bryan/JumpServer/jumpserver/apps')
+django.setup()
+
+from django.db import connections, transaction
+from assets.models import Asset, Node, AssetNode
+from orgs.models import Organization
+
+
+# ============== 配置 ==============
+ORG_ID = str(Organization.default().id)  # 指定组织 ID，可根据需要修改
+# ORG_ID = None
+TARGET_COUNT = 1000  # 目标数据量：50万条
+BATCH_SIZE = 50000  # 批处理大小
+INCLUDE_NODE_KEY = True # 是否生成 node_key 字段（如果 through 表有该字段）
+
+
+def log(msg):
+    """打印带时间戳的日志"""
+    print(f"[{datetime.now().strftime('%H:%M:%S')}] {msg}")
+
+
+def get_asset_and_node_ids(org_id):
+    """获取指定组织的 Asset 和 Node ID"""
+    log(f"获取组织 {org_id} 的 Asset 和 Node 数据...")
+    
+    # 获取 Asset ID 列表
+    assets = Asset.objects.filter(org_id=org_id).values_list('id', flat=True)
+    asset_ids = list(assets)
+    
+    # 获取 Node ID 列表
+    nodes = Node.objects.filter(org_id=org_id).values_list('id', flat=True)
+    node_ids = list(nodes)
+    
+    if not asset_ids:
+        log(f"✗ 错误：组织 {org_id} 中没有 Asset 数据")
+        return None, None
+    
+    if not node_ids:
+        log(f"✗ 错误：组织 {org_id} 中没有 Node 数据")
+        return None, None
+    
+    log(f"✓ 获取到 {len(asset_ids):,} 个 Asset")
+    log(f"✓ 获取到 {len(node_ids):,} 个 Node\n")
+    
+    return asset_ids, node_ids
+
+
+def generate_and_insert_data_v2(asset_ids, node_ids, batch_size, include_node_key=False):
+    """生成数据：每个 asset 随机关联 1-3 个节点
+    
+    Args:
+        asset_ids: Asset ID 列表
+        node_ids: Node ID 列表
+        batch_size: 批处理大小
+        include_node_key: 是否生成 node_key 字段
+    """
+    log(f"开始为所有 Asset 生成关联节点数据...")
+    log(f"每个 asset 随机关联 1-3 个节点")
+    log(f"总 asset 数: {len(asset_ids):,}")
+    log(f"总 node 数: {len(node_ids):,}")
+    log(f"批处理大小: {batch_size:,}")
+    log(f"是否生成 node_key: {'是' if include_node_key else '否'}\n")
+    
+    start_time = datetime.now()
+    inserted = 0
+    
+    # 获取 AssetNode through 表
+    asset_node_through = Asset.nodes.through
+    
+    # 如果需要生成 node_key，预加载 Node.key 映射
+    node_key_map = {}
+    if include_node_key:
+        log("加载 Node key 映射...")
+        node_key_map = {nid: key for nid, key in Node.objects.filter(id__in=node_ids).values_list('id', 'key')}
+        log(f"✓ 加载了 {len(node_key_map):,} 个 node_key\n")
+    
+    # 检查 through 表是否有 node_key 字段
+    has_node_key_field = 'node_key' in [f.name for f in asset_node_through._meta.get_fields()]
+    if include_node_key and not has_node_key_field:
+        log("⚠ 警告：配置要求生成 node_key，但 through 表没有该字段，将跳过 node_key 生成\n")
+        include_node_key = False
+    
+    # 清空现有数据（如果需要）
+    current_count = asset_node_through.objects.count()
+    if current_count > 0:
+        log(f"⚠ 清空已有数据: {current_count:,} 条")
+        asset_node_through.objects.all().delete()
+        log(f"✓ 清空完成\n")
+    
+    try:
+        # 为每个 asset 生成 1-3 个 node 关联
+        objs = []
+        
+        for asset_idx, asset_id in enumerate(asset_ids):
+            # 为这个 asset 随机选择 1-3 个节点
+            num_nodes = random.randint(1, 3)
+            selected_nodes = random.sample(node_ids, min(num_nodes, len(node_ids)))
+            
+            # 为每个选中的节点创建关联
+            for node_id in selected_nodes:
+                if include_node_key and has_node_key_field:
+                    node_key = node_key_map.get(node_id, '')
+                    objs.append(
+                        asset_node_through(asset_id=asset_id, node_id=node_id, node_key=node_key)
+                    )
+                else:
+                    objs.append(
+                        asset_node_through(asset_id=asset_id, node_id=node_id)
+                    )
+            
+            # 定期批量插入
+            if len(objs) >= batch_size:
+                with transaction.atomic():
+                    asset_node_through.objects.bulk_create(objs, batch_size=batch_size, ignore_conflicts=True)
+                    inserted += len(objs)
+                
+                # 显示进度
+                progress_pct = (asset_idx + 1) / len(asset_ids) * 100
+                elapsed = (datetime.now() - start_time).total_seconds()
+                log(f"  已处理 {asset_idx + 1:,}/{len(asset_ids):,} ({progress_pct:.1f}%) asset - 已插入 {inserted:,} 条数据 - 耗时 {elapsed:.1f}s")
+                objs = []
+        
+        # 插入剩余的数据
+        if objs:
+            with transaction.atomic():
+                asset_node_through.objects.bulk_create(objs, batch_size=batch_size, ignore_conflicts=True)
+                inserted += len(objs)
+        
+        # 统计最终结果
+        total_now = asset_node_through.objects.count()
+        elapsed = (datetime.now() - start_time).total_seconds()
+        avg_nodes_per_asset = total_now / len(asset_ids) if asset_ids else 0
+        
+        log(f"\n✓ 数据生成完成!")
+        log(f"  总 asset 数: {len(asset_ids):,}")
+        log(f"  生成关系数: {inserted:,} 条")
+        log(f"  平均每个 asset 关联节点数: {avg_nodes_per_asset:.2f}")
+        log(f"  表总数: {total_now:,} 条")
+        log(f"  耗时: {elapsed:.1f}s")
+        if elapsed > 0:
+            log(f"  平均速率: {inserted/elapsed:.0f} 条/秒\n")
+        
+        return True
+        
+    except Exception as e:
+        log(f"\n✗ 插入数据时出错: {str(e)}")
+        import traceback
+        traceback.print_exc()
+        return False
+
+
+def generate_and_insert_data(asset_ids, node_ids, target_count, batch_size, include_node_key=False):
+    """生成随机数据并插入数据库
+    
+    Args:
+        asset_ids: Asset ID 列表
+        node_ids: Node ID 列表
+        target_count: 目标生成数量
+        batch_size: 批处理大小
+        include_node_key: 是否生成 node_key 字段
+    """
+    log(f"开始生成 {target_count:,} 条 Asset-Node 关系数据...")
+    log(f"批处理大小: {batch_size:,}")
+    log(f"是否生成 node_key: {'是' if include_node_key else '否'}\n")
+    
+    start_time = datetime.now()
+    inserted = 0
+    
+    # 获取 AssetNode through 表
+    asset_node_through = Asset.nodes.through
+    
+    # 如果需要生成 node_key，预加载 Node.key 映射
+    node_key_map = {}
+    if include_node_key:
+        log("加载 Node key 映射...")
+        node_key_map = {nid: key for nid, key in Node.objects.filter(id__in=node_ids).values_list('id', 'key')}
+        log(f"✓ 加载了 {len(node_key_map):,} 个 node_key\n")
+    
+    # 检查 through 表是否有 node_key 字段
+    has_node_key_field = 'node_key' in [f.name for f in asset_node_through._meta.get_fields()]
+    if include_node_key and not has_node_key_field:
+        log("⚠ 警告：配置要求生成 node_key，但 through 表没有该字段，将跳过 node_key 生成\n")
+        include_node_key = False
+    
+    # 清空现有数据（如果需要）
+    current_count = asset_node_through.objects.count()
+    if current_count > 0:
+        log(f"⚠ through 表中已有 {current_count:,} 条数据，继续追加...\n")
+    
+    try:
+        # 分批生成和插入数据
+        # 为了确保每个 asset 只属于一个 node，我们需要追踪已分配的 asset
+        asset_to_node_map = {}
+        
+        for batch_start in range(0, target_count, batch_size):
+            batch_end = min(batch_start + batch_size, target_count)
+            batch_size_actual = batch_end - batch_start
+            
+            # 生成随机数据
+            objs = []
+            for _ in range(batch_size_actual):
+                # 如果所有 asset 都已分配，重新开始循环
+                if len(asset_to_node_map) >= len(asset_ids):
+                    asset_to_node_map = {}
+                
+                # 找一个还没分配过的 asset
+                attempts = 0
+                while attempts < 100:
+                    asset_id = random.choice(asset_ids)
+                    if asset_id not in asset_to_node_map:
+                        break
+                    attempts += 1
+                
+                # 如果找不到未分配的 asset，就跳过这个
+                if asset_id in asset_to_node_map:
+                    continue
+                
+                # 为这个 asset 随机选择一个 node
+                node_id = random.choice(node_ids)
+                asset_to_node_map[asset_id] = node_id
+                
+                # 构建对象
+                if include_node_key and has_node_key_field:
+                    node_key = node_key_map.get(node_id, '')
+                    objs.append(
+                        asset_node_through(asset_id=asset_id, node_id=node_id, node_key=node_key)
+                    )
+                else:
+                    objs.append(
+                        asset_node_through(asset_id=asset_id, node_id=node_id)
+                    )
+            
+            # 批量插入
+            if objs:
+                with transaction.atomic():
+                    asset_node_through.objects.bulk_create(objs, batch_size=batch_size, ignore_conflicts=True)
+                    inserted += len(objs)
+            
+            # 显示进度
+            progress_pct = (batch_end / target_count) * 100 if target_count > 0 else 0
+            elapsed = (datetime.now() - start_time).total_seconds()
+            log(f"  已插入 {inserted:,}/{target_count:,} ({progress_pct:.1f}%) - 耗时 {elapsed:.1f}s")
+        
+        # 统计最终结果
+        total_now = asset_node_through.objects.count()
+        elapsed = (datetime.now() - start_time).total_seconds()
+        
+        log(f"\n✓ 数据生成完成!")
+        log(f"  插入: {inserted:,} 条新数据")
+        log(f"  表总数: {total_now:,} 条")
+        log(f"  耗时: {elapsed:.1f}s")
+        log(f"  平均速率: {inserted/elapsed:.0f} 条/秒\n")
+        
+        return True
+        
+    except Exception as e:
+        log(f"\n✗ 插入数据时出错: {str(e)}")
+        import traceback
+        traceback.print_exc()
+        return False
+
+
+def main():
+    """主函数"""
+    log(f"\n{'='*60}")
+    log("【Asset-Node 关系数据生成工具 (v2 - 每个asset关联1-3个节点)】")
+    log(f"{'='*60}\n")
+    
+    log(f"配置:")
+    log(f"  组织 ID: {ORG_ID}")
+    log(f"  批处理大小: {BATCH_SIZE:,}")
+    log(f"  包含 node_key: {INCLUDE_NODE_KEY}\n")
+
+    if ORG_ID is None:
+        log("✗ 错误：请在脚本中设置 ORG_ID 变量")
+        return False
+    
+    # 获取 Asset 和 Node ID
+    asset_ids, node_ids = get_asset_and_node_ids(ORG_ID)
+    if not asset_ids or not node_ids:
+        return False
+    
+    # 生成并插入数据 (v2: 每个 asset 关联 1-3 个节点)
+    success = generate_and_insert_data_v2(asset_ids, node_ids, BATCH_SIZE, include_node_key=INCLUDE_NODE_KEY)
+    
+    if success:
+        log("="*60)
+        log("✓ 所有操作完成！")
+        log("="*60)
+        return True
+    else:
+        log("="*60)
+        log("✗ 操作失败")
+        log("="*60)
+        return False
+
+def check_node_key_is_empty():
+    if not INCLUDE_NODE_KEY:
+        return
+    # 校验是否有 node_key 为空的数据
+    has_node_key_field = 'node_key' in [f.name for f in asset_node_through._meta.get_fields()]
+    if has_node_key_field:
+        from django.db.models import Q
+        empty_node_key = asset_node_through.objects.filter(Q(node_key='') | Q(node_key__isnull=True)).count()
+        log(f"  node_key 为空的数据: {empty_node_key:,}")
+        if empty_node_key > 0:
+            log(f"  ⚠ 警告：存在 {empty_node_key:,} 条 node_key 为空的数据")
+        else:
+            log(f"  ✓ 所有数据的 node_key 都已填充")
+    else:
+        log(f"  ⚠ through 表没有 node_key 字段")
+
+if __name__ == '__main__':
+    try:
+        success = main()
+        
+        # 输出最终统计
+        asset_node_through = Asset.nodes.through
+        total_count = asset_node_through.objects.count()
+        log(f"{'='*60}")
+        log(f"最终统计：")
+        log(f"  AssetNode through 表总数据量: {total_count:,}")
+        check_node_key_is_empty()
+        
+        log(f"{'='*60}")
+        
+        sys.exit(0 if success else 1)
+    except KeyboardInterrupt:
+        log("\n✗ 用户中断操作")
+        sys.exit(1)
+    except Exception as e:
+        log(f"\n✗ 发生错误: {str(e)}")
+        import traceback
+        traceback.print_exc()
+        sys.exit(1)

test_query_node_assets.py

@@ -0,0 +1,113 @@
+#!/usr/bin/env python
+"""
+测试查询资产树和授权树下指定节点的资产总数
+迭代查询，一个节点一个SQL
+"""
+
+import os, sys, django
+
+os.environ.setdefault('DJANGO_SETTINGS_MODULE', 'jumpserver.settings')
+sys.path.insert(0, '/Users/bryan/JumpServer/jumpserver/apps')
+django.setup()
+
+def test():
+    from django.db.models import Count, Q
+    from assets.models import Node
+
+
+def main():
+    # 示例：查询这些节点的资产数
+    import os, time, json, random
+    from functools import reduce
+    from operator import or_
+    from django.db.models import Q
+    from collections import defaultdict
+    from assets.models import Node, Asset
+
+    # 资产树
+    nodes = Node.objects.all().order_by('key')[:1000]
+    node_keys = [n.key for n in nodes]
+    t1 = time.time()
+    t_mapper = defaultdict(int)
+    for key in node_keys[:1]:
+        count = Node.assets.through.objects.filter(
+             Q(node_key=key) | Q(node_key__startswith=key + ':')
+        ).distinct('asset_id').count()
+        t_mapper[key] = count
+    t2 = time.time()
+    t_root = t_mapper.get('1', 0)
+    count_rs = Node.assets.through.objects.all().count()
+    count_asset = Asset.objects.all().count()
+    count_node = Node.objects.all().count()
+    # print('=' * 50, '查询资产树', '=' * 50)
+    # print(json.dumps(t_mapper, indent=4))
+    # print('总资产数:', count_asset, '总节点数:', count_node, '关联表总数:', count_rs)
+    # print(f"查询 {len(node_keys)} 个节点资产数耗时: {t2 - t1:.4f} 秒")
+
+    # 授权树
+    guessed_perm_das_count = 10000
+    guessed_perm_dn_count = 100
+
+    # guessed_perm_das_count = 5000
+    # guessed_perm_dn_count = 50
+
+    # guessed_perm_das_count = 1000
+    # guessed_perm_dn_count = 20
+
+    # guessed_perm_das_count = 100
+    # guessed_perm_dn_count = 10
+
+    # guessed_perm_das_count = 0
+    # guessed_perm_dn_count = 20
+
+    # guessed_perm_das_count = 1000
+    # guessed_perm_dn_count = 0
+
+    # guessed_perm_das_count = 200000
+    # guessed_perm_dn_count = 0
+
+    # guessed_perm_das_count = 0
+    # guessed_perm_dn_count = 1000
+
+    guessed_perm_das_count = 200000
+    guessed_perm_dn_count = 1000
+
+    perm_dns = Node.objects.all()[:guessed_perm_dn_count]
+    perm_dn_keys = [str(n.key) for n in perm_dns]
+    perm_das = Asset.objects.all()[:guessed_perm_das_count]
+    perm_da_ids= [str(a.id) for a in perm_das]
+    # qn_keys = Node.objects.all().values_list('key', flat=True)[:100]
+    qn_keys = ['1']
+    t3 = time.time()
+    p_mapper = defaultdict(int)
+    # 新方案： 比如查询 A 节点，
+    # 如果是 A 是 owner-node 那就跟资产一样 ms 级直接返回
+    # 如果是 A bridge 或 da-node:
+    #   只需要将A节点下的所有直接资产id和A节点下的所有 owner key 作第一个 filter 的过滤条件即可（base 资产表）
+    #   (in=A-da-ids | =A.key | startswith=A.key + ':')
+    q_user_assets = [
+        Q(node_key=key) | Q(node_key__startswith=key + ':') for key in perm_dn_keys
+    ] + [ Q(asset_id__in=perm_da_ids) ]
+    qs_base = Node.assets.through.objects.filter(reduce(or_, q_user_assets))
+    for qn_key in qn_keys:
+        count = qs_base.filter(
+            Q(node_key=qn_key) | Q(node_key__startswith=qn_key + ':')
+        ).distinct('asset_id').count()
+        p_mapper[qn_key] = count
+    t4 = time.time()
+    p_root = p_mapper.get('1', 0)
+    print(json.dumps(t_mapper, indent=4))
+    print(json.dumps(p_mapper, indent=4))
+
+    print('=' * 50, '查询资产树', '=' * 50)
+    print('总资产数:', count_asset, '总节点数:', count_node, '关联表总数:', count_rs)
+    print(f"查询 {len(node_keys)} 个节点资产数耗时: {t2 - t1:.4f} 秒")
+
+    print('='*50, '查询用户权限下节点资产数', '=' * 50)
+    print('假设直接授权节点数: ', len(perm_dn_keys), '假设直接授权资产数:', len(perm_da_ids))
+    print(f"查询 {len(qn_keys)} 个节点资产数耗时: {t4 - t3:.4f} 秒")
+    print(f"资产树根节点资产数: {t_root}, 授权树根节点资产数: {p_root}")
+
+
+if __name__ == '__main__':
+    main()