862 lines
28 KiB
Python
862 lines
28 KiB
Python
"""
|
||
PDF 用户上传接口（FastAPI）
|
||
- 处理文件名和其对应关系
|
||
"""
|
||
|
||
from typing import Dict, Tuple
|
||
import os
|
||
import re
|
||
from typing import Dict
|
||
from fastapi import FastAPI, HTTPException
|
||
import logging
|
||
import json
|
||
from modelsAPI.model_api import OpenaiAPI
|
||
from openai import OpenAI
|
||
from typing import Dict, List, Any
|
||
|
||
app = FastAPI(title="PDF Upload Service")
|
||
logging.basicConfig(level=logging.INFO, format="%(asctime)s | %(levelname)s | %(name)s | %(message)s")
|
||
logger = logging.getLogger("filename_neo4j_process")
|
||
|
||
|
||
# =========================
|
||
# 图方法
|
||
# =========================
|
||
|
||
|
||
def node_to_json(node):
|
||
props = dict(node)
|
||
# ks_raw = props.get("knowledge_source")
|
||
# knowledge_source = json.loads(ks_raw) if ks_raw else []
|
||
|
||
ks_raw = props.get("knowledge_source")
|
||
knowledge_source = []
|
||
|
||
if ks_raw and isinstance(ks_raw, str):
|
||
try:
|
||
ks_list = json.loads(ks_raw)
|
||
if isinstance(ks_list, list):
|
||
for ks in ks_list:
|
||
if isinstance(ks, dict):
|
||
ks.pop("info", None)
|
||
knowledge_source.append(ks)
|
||
except (json.JSONDecodeError, TypeError):
|
||
# 解析失败或类型错误时，保持 knowledge_source 为空列表
|
||
pass
|
||
|
||
name = props.get("名称")
|
||
props.pop("embedding", None)
|
||
props.pop("knowledge_source", None)
|
||
props.pop("name", None) # 现在安全地移除
|
||
props.pop("fulltext", None)
|
||
if "last_updated" in props:
|
||
props["最后更新时间"] = props.pop("last_updated")
|
||
|
||
if "created_at" in props:
|
||
props["创建时间"] = props.pop("created_at")
|
||
return {
|
||
"id": node.element_id,
|
||
"name": name, # 使用提前保存的值
|
||
"labels": list(node.labels),
|
||
"knowledge_source": knowledge_source,
|
||
"properties": props,
|
||
}
|
||
|
||
|
||
def rel_to_json(rel):
|
||
props = dict(rel)
|
||
props.pop("embedding", None)
|
||
props.pop("fact_timeline", None)
|
||
return {"id": rel.element_id, "label": rel.type, "source": rel.start_node.element_id, "target": rel.end_node.element_id, "properties": props}
|
||
|
||
|
||
def is_simple_path(path):
|
||
"""
|
||
判断是否为无环路径（simple path）
|
||
"""
|
||
seen = set()
|
||
for n in path.nodes:
|
||
nid = n.element_id
|
||
if nid in seen:
|
||
return False
|
||
seen.add(nid)
|
||
return True
|
||
|
||
def build_tree(nodes: List[Dict], links: List[Dict]) -> List[Dict]:
|
||
"""
|
||
按照固定层级构建树：舰船(根) -> 舰艇 -> 系统 -> 子系统
|
||
不再依赖 links 的连线，直接根据 labels 归类，彻底避免环状数据导致的死循环。
|
||
默认只展开第一艘舰艇的系统/子系统，其它舰艇 children 置空（前端可点击按需加载）。
|
||
"""
|
||
# 1. 辅助函数：判断标签是否包含关键字
|
||
def has_label(node, keyword):
|
||
labels = node.get("labels", [])
|
||
if isinstance(labels, list):
|
||
return any(keyword in lbl for lbl in labels)
|
||
return False
|
||
|
||
# 2. 辅助函数：构建标准的树节点格式
|
||
def make_node(node_data):
|
||
return {
|
||
"id": node_data["id"],
|
||
"name": node_data.get("name", ""),
|
||
"labels": node_data.get("labels", []),
|
||
|
||
"properties": node_data.get("properties", {}),
|
||
"color": node_data.get("color"),
|
||
"size": node_data.get("size"),
|
||
"children": []
|
||
}
|
||
|
||
# 3. 将节点按层级分类存储到字典中，方便快速查找
|
||
fleet_roots = {} # 第零层：舰船/舰号（父根节点）
|
||
ships = {} # 第一层：舰艇
|
||
systems = {} # 第二层：系统（不含子系统）
|
||
sub_systems = {} # 第三层：子系统
|
||
|
||
for node in nodes:
|
||
# 注意：判断逻辑要互斥，防止"子系统"被误判为"系统"
|
||
if has_label(node, "舰船") or has_label(node, "舰号"):
|
||
fleet_roots[node["id"]] = make_node(node)
|
||
elif has_label(node, "舰艇"):
|
||
ships[node["id"]] = make_node(node)
|
||
elif has_label(node, "子系统"):
|
||
sub_systems[node["id"]] = make_node(node)
|
||
elif has_label(node, "系统"):
|
||
systems[node["id"]] = make_node(node)
|
||
else:
|
||
pass
|
||
|
||
# 4. 组装系统 -> 子系统、舰艇 -> 系统 关系（先把完整子树挂好）
|
||
for link in links:
|
||
if link.get("label") != "包含":
|
||
continue
|
||
|
||
src_id = link["source"]
|
||
tgt_id = link["target"]
|
||
|
||
if src_id in ships and tgt_id in systems:
|
||
ships[src_id]["children"].append(systems[tgt_id])
|
||
elif src_id in systems and tgt_id in sub_systems:
|
||
systems[src_id]["children"].append(sub_systems[tgt_id])
|
||
elif src_id in ships and tgt_id in sub_systems:
|
||
ships[src_id]["children"].append(sub_systems[tgt_id])
|
||
|
||
# 5. 所有舰艇 children 全部保留，仅用 collapsed 标记默认折叠状态。
|
||
# 前端读取 collapsed=False 的舰艇默认展开，其它折叠；
|
||
# 用户点击折叠的舰艇时，前端切换 collapsed 即可展示子树。
|
||
# 只保留名称包含 "122" 的舰艇（隐藏 JZ贱 等其它分支）
|
||
#sorted_ships = sorted(
|
||
# (s for s in ships.values() if "122" in s.get("name", "")),
|
||
# key=lambda s: s.get("name", ""),
|
||
#)
|
||
sorted_ships = sorted(ships.values(), key=lambda s: s.get("name", ""))
|
||
for idx, ship in enumerate(sorted_ships):
|
||
ship["collapsed"] = (idx != 0)
|
||
|
||
# 6. 挂载到舰船根节点；如果数据里没有舰船根，就创建一个虚拟根
|
||
if fleet_roots:
|
||
# 取第一个舰船根节点作为父（一般场景下只有一个）
|
||
root = next(iter(fleet_roots.values()))
|
||
root["children"] = sorted_ships
|
||
return [root]
|
||
else:
|
||
virtual_root = {
|
||
"id": "__fleet_root__",
|
||
"name": "舰号",
|
||
"labels": ["舰船"],
|
||
"properties": {},
|
||
"color": None,
|
||
"size": None,
|
||
"children": sorted_ships,
|
||
}
|
||
return [virtual_root]
|
||
def build_tree1(nodes: List[Dict], links: List[Dict]) -> List[Dict]:
|
||
"""
|
||
从 nodes + links 构建树结构。
|
||
- 根节点：label 包含 '舰艇' 的节点
|
||
- 父子关系：links 中 label == '包含' 的边 (source -> target)
|
||
- 支持多棵树（多个舰艇根节点）
|
||
"""
|
||
|
||
# 1. 建立 id -> node 的映射
|
||
node_map: Dict[str, Dict] = {n["id"]: n for n in nodes}
|
||
|
||
# 2. 建立 父id -> [子id列表] 的邻接表（只处理"包含"关系）
|
||
children_map: Dict[str, List[str]] = {n["id"]: [] for n in nodes}
|
||
parent_map: Dict[str, str] = {} # 子id -> 父id，用于识别根节点
|
||
|
||
for link in links:
|
||
if link.get("label") == "包含":
|
||
src = link["source"]
|
||
tgt = link["target"]
|
||
if src in children_map:
|
||
children_map[src].append(tgt)
|
||
if tgt not in parent_map:
|
||
parent_map[tgt] = src
|
||
|
||
# 3. 递归构建子树
|
||
def build_subtree(node_id: str, ancestors: set = None) -> Dict[str, Any]:
|
||
if ancestors is None:
|
||
ancestors = set()
|
||
if node_id in ancestors:
|
||
return None
|
||
ancestors.add(node_id)
|
||
node = node_map[node_id]
|
||
subtree = {"id": node_id, "name": node.get("name", ""), "labels": node.get("labels", []), "properties": node.get("properties", {}), "color": node.get("color"), "size": node.get("size"), "children": []}
|
||
current_ancestors = ancestors | {node_id}
|
||
for child_id in children_map.get(node_id, []):
|
||
child = build_subtree(child_id, current_ancestors)
|
||
if child is not None:
|
||
subtree["children"].append(child)
|
||
return subtree
|
||
|
||
# def build_subtree(node_id: str) -> Dict[str, Any]:
|
||
# node = node_map[node_id]
|
||
# subtree = {
|
||
# "id": node_id,
|
||
# "name": node.get("name", ""),
|
||
# "labels": node.get("labels", []),
|
||
# "properties": node.get("properties", {}),
|
||
# "color": node.get("color"),
|
||
# "size": node.get("size"),
|
||
# "children": []
|
||
# }
|
||
# for child_id in children_map.get(node_id, []):
|
||
# subtree["children"].append(build_subtree(child_id))
|
||
# return subtree
|
||
|
||
# 4. 找出所有根节点：label 包含 '舰艇'，且没有父节点
|
||
roots = [n["id"] for n in nodes if "舰艇" in n.get("labels", []) and n["id"] not in parent_map]
|
||
|
||
# 5. 为每棵树构建结构，并打上树编号
|
||
forest = []
|
||
for i, root_id in enumerate(roots):
|
||
tree = build_subtree(root_id)
|
||
tree["tree_index"] = i # 区分不同树
|
||
forest.append(tree)
|
||
|
||
return forest
|
||
|
||
|
||
def _filter_searchable(graph: dict) -> dict:
|
||
"""剥掉节点的 Searchable 标签；若剥完后 labels 为空则整体移除该节点及其关联边"""
|
||
for n in graph["nodes"]:
|
||
n["labels"] = [l for l in n.get("labels", []) if l != "Searchable"]
|
||
valid_ids = {n["id"] for n in graph["nodes"] if n["labels"]}
|
||
return {
|
||
"nodes": [n for n in graph["nodes"] if n["id"] in valid_ids],
|
||
"links": [
|
||
l for l in graph["links"]
|
||
if l["source"] in valid_ids and l["target"] in valid_ids
|
||
],
|
||
}
|
||
|
||
|
||
def build_graph_from_paths(paths):
|
||
"""
|
||
核心公共方法
|
||
从 Neo4j 路径结果构建 graphData
|
||
"""
|
||
node_map: Dict[str, dict] = {}
|
||
link_map: Dict[str, dict] = {}
|
||
|
||
def add_edge(rel):
|
||
s = rel.start_node
|
||
t = rel.end_node
|
||
|
||
if s.element_id not in node_map:
|
||
node_map[s.element_id] = node_to_json(s)
|
||
|
||
if t.element_id not in node_map:
|
||
node_map[t.element_id] = node_to_json(t)
|
||
|
||
if rel.element_id not in link_map:
|
||
link_map[rel.element_id] = rel_to_json(rel)
|
||
|
||
for record in paths:
|
||
p = record["p"]
|
||
|
||
if not is_simple_path(p):
|
||
continue
|
||
|
||
for r in p.relationships:
|
||
add_edge(r)
|
||
|
||
return _filter_searchable({"nodes": list(node_map.values()), "links": list(link_map.values())})
|
||
|
||
|
||
# def fetch_graph_sample(driver,limit: int = 200):
|
||
# """提取采样逻辑为可复用函数"""
|
||
# with driver.session() as session:
|
||
# result = session.run("""
|
||
# MATCH (n)
|
||
# WITH n
|
||
# ORDER BY elementId(n) ASC
|
||
# LIMIT $limit
|
||
# MATCH (n)-[rel]->(m)
|
||
# RETURN n, rel, m
|
||
# """, limit=limit)
|
||
|
||
# node_map, link_map = {}, {}
|
||
# for rec in result:
|
||
# n, r, m = rec["n"], rec["rel"], rec["m"]
|
||
# node_map[n.element_id] = node_to_json(n)
|
||
# node_map[m.element_id] = node_to_json(m)
|
||
# link_map[r.element_id] = rel_to_json(r)
|
||
|
||
# return {
|
||
# "nodes": list(node_map.values()),
|
||
# "links": list(link_map.values())
|
||
# }
|
||
|
||
# def fetch_graph_sample(driver, limit: int = 200):
|
||
# """
|
||
# 获取采样数据：
|
||
# 1. 选取前 $limit 个种子节点（Level 0 或 1，取决于业务定义，这里定义种子为 Level 0）。
|
||
# 2. 对每个节点向下探索最多 3 跳。
|
||
# 3. 每个节点对象中增加 level 字段，表示其距离种子节点的深度。
|
||
# """
|
||
# with driver.session() as session:
|
||
# # Cypher 逻辑：
|
||
# # - 选出种子节点 n
|
||
# # - 匹配 1-3 跳路径
|
||
# # - 返回种子节点 n，路径 path，以及该路径的长度（即层级）
|
||
# result = session.run("""
|
||
# MATCH (n)
|
||
# WITH n ORDER BY elementId(n) ASC LIMIT $limit
|
||
# OPTIONAL MATCH path = (n)-[*1..3]-(m)
|
||
# UNWIND nodes(path) AS x
|
||
# WITH n, path, m, COUNT(DISTINCT x) AS nodeCount
|
||
# WHERE path IS NULL OR nodeCount = LENGTH(path) + 1
|
||
# RETURN n, path, LENGTH(path) AS depth
|
||
# """, limit=limit)
|
||
|
||
# node_map, link_map = {}, {}
|
||
|
||
# for rec in result:
|
||
# seed_node = rec["n"]
|
||
# path = rec["path"]
|
||
# depth = rec["depth"] or 0 # 如果没有路径，深度为 0
|
||
|
||
# # 1. 处理种子节点 (Level 0)
|
||
# if seed_node.element_id not in node_map:
|
||
# node_data = node_to_json(seed_node)
|
||
# node_data["level"] = 0 # 种子节点设为 0 层
|
||
# node_map[seed_node.element_id] = node_data
|
||
|
||
# # 2. 如果存在路径，解析路径中的所有节点和关系
|
||
# if path:
|
||
# # 路径中的节点处理
|
||
# # path.nodes 包含了从起始到终点的所有节点
|
||
# for i, node in enumerate(path.nodes):
|
||
# if node.element_id not in node_map:
|
||
# node_data = node_to_json(node)
|
||
# # 层级即为该节点在当前路径中的索引
|
||
# node_data["level"] = i
|
||
# node_map[node.element_id] = node_data
|
||
# else:
|
||
# # 如果节点已存在，保留最小的 level (即最靠近种子的距离)
|
||
# node_map[node.element_id]["level"] = min(node_map[node.element_id].get("level", 3), i)
|
||
|
||
# # 路径中的关系处理
|
||
# for rel in path.relationships:
|
||
# if rel.element_id not in link_map:
|
||
# link_map[rel.element_id] = rel_to_json(rel)
|
||
|
||
# return {
|
||
# "nodes": list(node_map.values()),
|
||
# "links": list(link_map.values())
|
||
# }
|
||
|
||
|
||
# def fetch_graph_sample(driver, limit=200, max_depth=3):
|
||
# with driver.session() as session:
|
||
# result = session.run("""
|
||
# MATCH (c)
|
||
# WITH c ORDER BY elementId(c) ASC LIMIT $limit
|
||
|
||
# MATCH path = (c)-[*0..3]-(n)
|
||
# WITH n, MIN(length(path)) AS dist
|
||
# OPTIONAL MATCH (n)-[r]-(m)
|
||
|
||
# RETURN n, dist, r, m
|
||
# """, limit=limit)
|
||
|
||
# node_map, link_map = {}, {}
|
||
|
||
# for rec in result:
|
||
# n = rec["n"]
|
||
# dist = rec["dist"]
|
||
# r = rec["r"]
|
||
# m = rec["m"]
|
||
|
||
# # ---- 节点 ----
|
||
# nid = n.element_id
|
||
# if nid not in node_map:
|
||
# node_data = node_to_json(n)
|
||
# node_data["level"] = min(dist, 3) if dist is not None else 0
|
||
# node_map[nid] = node_data
|
||
|
||
# # ---- 边 ----
|
||
# if r and m:
|
||
# mid = m.element_id
|
||
# if mid not in node_map:
|
||
# node_map[mid] = node_to_json(m)
|
||
|
||
# link_map.setdefault(r.element_id, rel_to_json(r))
|
||
|
||
# return {
|
||
# "nodes": list(node_map.values()),
|
||
# "links": list(link_map.values())
|
||
# }
|
||
|
||
def fetch_graph_sample(driver):
|
||
"""
|
||
查询所有舰艇及其关联的系统和子系统
|
||
路径结构：(舰艇) -> (系统) -> (子系统)
|
||
"""
|
||
|
||
with driver.session() as session:
|
||
# =========================
|
||
# 统一查询：舰艇 -> 系统 -> 子系统
|
||
# =========================
|
||
# 逻辑说明：
|
||
# 1. MATCH (ship:舰艇)：选中所有舰艇
|
||
# 2. -[*1..3]->(target)：查找深度为 1 到 3 的路径
|
||
# - 深度 1：舰艇 -> 系统
|
||
# - 深度 2：舰艇 -> 系统 -> 子系统
|
||
# 3. WHERE target:系统 OR target:子系统：确保终点是我们关心的节点类型
|
||
# (防止查询出其他无关的深层节点)
|
||
|
||
query = """
|
||
MATCH path = (ship:舰艇)-[*1..3]->(target)
|
||
WHERE target:系统 OR target:子系统
|
||
RETURN path
|
||
"""
|
||
|
||
result = session.run(query)
|
||
|
||
# =========================
|
||
# 解析 paths
|
||
# =========================
|
||
all_nodes = {}
|
||
all_links = {}
|
||
|
||
for record in result:
|
||
path = record["path"]
|
||
|
||
# 解析节点
|
||
for node in path.nodes:
|
||
# 兼容 ID 获取
|
||
node_key = node.element_id if hasattr(node, 'element_id') else node.id
|
||
if node_key not in all_nodes:
|
||
all_nodes[node_key] = node_to_json(node)
|
||
|
||
# 解析关系
|
||
for rel in path.relationships:
|
||
rel_key = rel.element_id if hasattr(rel, 'element_id') else rel.id
|
||
if rel_key not in all_links:
|
||
all_links[rel_key] = rel_to_json(rel)
|
||
|
||
return _filter_searchable({
|
||
"nodes": list(all_nodes.values()),
|
||
"links": list(all_links.values())
|
||
})
|
||
def fetch_graph_sample1(driver):
|
||
"""
|
||
查找路径：
|
||
1. 优先：舰艇 -> 子系统
|
||
2. fallback：舰艇 -> 系统
|
||
|
||
注意：根据数据库实际结构，使用 Label (:舰艇, :子系统) 进行匹配，
|
||
而不是 WHERE category = '...'
|
||
"""
|
||
|
||
# 定义采样数量
|
||
SAMPLE_LIMIT = 400
|
||
|
||
with driver.session() as session:
|
||
# =========================
|
||
# 第一阶段：舰艇 -> 子系统
|
||
# =========================
|
||
# 修改点：直接使用 (:舰艇) 和 (:子系统) 标签匹配，去掉 WHERE 子句
|
||
result = session.run(
|
||
"""
|
||
MATCH (ship:舰艇)
|
||
WITH ship LIMIT $limit
|
||
|
||
MATCH path = (ship)-[*1..3]->(sub:子系统)
|
||
RETURN path
|
||
""",
|
||
limit=SAMPLE_LIMIT,
|
||
)
|
||
|
||
paths = [record["path"] for record in result]
|
||
|
||
# =========================
|
||
# fallback：舰艇 -> 系统
|
||
# =========================
|
||
if not paths:
|
||
# 修改点：直接使用 (:舰艇) 和 (:系统) 标签匹配
|
||
result = session.run(
|
||
"""
|
||
MATCH (ship:舰艇)
|
||
WITH ship LIMIT $limit
|
||
|
||
MATCH path = (ship)-[*1..2]->(sys:系统)
|
||
RETURN path
|
||
""",
|
||
limit=SAMPLE_LIMIT,
|
||
)
|
||
|
||
paths = [record["path"] for record in result]
|
||
|
||
# =========================
|
||
# 解析 paths
|
||
# =========================
|
||
all_nodes = {}
|
||
all_links = {}
|
||
|
||
for path in paths:
|
||
# 节点
|
||
for node in path.nodes:
|
||
# 兼容 Neo4j 5.x (element_id) 和 4.x (id)
|
||
node_key = node.element_id if hasattr(node, "element_id") else node.id
|
||
if node_key not in all_nodes:
|
||
all_nodes[node_key] = node_to_json(node)
|
||
|
||
# 关系
|
||
for rel in path.relationships:
|
||
rel_key = rel.element_id if hasattr(rel, "element_id") else rel.id
|
||
if rel_key not in all_links:
|
||
all_links[rel_key] = rel_to_json(rel)
|
||
|
||
return {"nodes": list(all_nodes.values()), "links": list(all_links.values())}
|
||
|
||
|
||
# =============================
|
||
# 文件名处理方法
|
||
# =============================
|
||
|
||
BASE_DIR = os.path.dirname(os.path.abspath(__file__))
|
||
TREE_JSON_PATH = os.path.join(BASE_DIR, "tree_data.json")
|
||
|
||
|
||
def _load_tree_data() -> Dict:
|
||
# 注意：实际运行时请确保 result.json 存在，此处仅为代码结构展示
|
||
if not os.path.exists(TREE_JSON_PATH):
|
||
return {}
|
||
with open(TREE_JSON_PATH, "r", encoding="utf-8") as f:
|
||
return json.load(f)
|
||
|
||
|
||
def _resolve_ship_root(data: Dict, ship_code: str) -> Dict:
|
||
raw = data.get(ship_code)
|
||
if not isinstance(raw, dict):
|
||
return {}
|
||
|
||
# 结构 1：直接是节点 {"code","name","children"}
|
||
if {"code", "name", "children"}.issubset(raw.keys()):
|
||
return raw
|
||
|
||
# 结构 2：包装层 {"101": {...root node...}}
|
||
if ship_code in raw and isinstance(raw[ship_code], dict):
|
||
node = raw[ship_code]
|
||
if {"code", "name", "children"}.issubset(node.keys()):
|
||
return node
|
||
|
||
# 兜底：取第一个 node-like 值
|
||
for value in raw.values():
|
||
if isinstance(value, dict) and {"code", "name", "children"}.issubset(value.keys()):
|
||
return value
|
||
|
||
return {}
|
||
|
||
|
||
def _longest_common_substring_len(a: str, b: str) -> int:
|
||
if not a or not b:
|
||
return 0
|
||
# 优化：确保 a 是较短的字符串以节省空间（可选），原逻辑保持不动也没问题
|
||
dp = [0] * (len(b) + 1)
|
||
best = 0
|
||
for i in range(1, len(a) + 1):
|
||
prev = 0
|
||
for j in range(1, len(b) + 1):
|
||
temp = dp[j]
|
||
if a[i - 1] == b[j - 1]:
|
||
dp[j] = prev + 1
|
||
if dp[j] > best:
|
||
best = dp[j]
|
||
else:
|
||
dp[j] = 0
|
||
prev = temp
|
||
return best
|
||
|
||
|
||
def _match_level4_tool(filename_text: str, device_node: Dict) -> Tuple[str, str]:
|
||
"""
|
||
匹配 Level 4 零件。
|
||
规则：如果文件名与 Level 3 (设备名) 的最大共同字符串长度 >= 与任何 Level 4 (零件名) 的长度，
|
||
则只匹配到 Level 3，返回空。
|
||
只有当 Level 4 的匹配度严格高于 Level 3 时，才返回 Level 4 信息。
|
||
"""
|
||
children = device_node.get("children") or {}
|
||
|
||
# 1. 计算文件名与当前设备 (Level 3) 名称的匹配度
|
||
device_name = str(device_node.get("name") or "")
|
||
level3_score = _longest_common_substring_len(filename_text, device_name)
|
||
|
||
best_name = ""
|
||
best_code = ""
|
||
best_score = 0 # 记录 Level 4 中的最佳得分
|
||
|
||
# 2. 遍历 Level 4 子节点
|
||
for child in children.values():
|
||
tool_name = str(child.get("name") or "")
|
||
tool_code = str(child.get("code") or "")
|
||
|
||
score = _longest_common_substring_len(filename_text, tool_name)
|
||
|
||
# 关键修改：只有当 Level 4 的得分严格大于 Level 3 的得分时，才视为有效匹配
|
||
# 并且要比当前找到的其他 Level 4 更好
|
||
if score > level3_score and score > best_score:
|
||
best_score = score
|
||
best_name = tool_name
|
||
best_code = tool_code
|
||
|
||
# 如果没有找到比 Level 3 匹配度更高的 Level 4，则返回空
|
||
if best_score <= level3_score:
|
||
return "", ""
|
||
|
||
return best_name, best_code
|
||
|
||
|
||
def process_filename(filename: str) -> Dict[str, str]:
|
||
name, _ = os.path.splitext(os.path.basename(filename))
|
||
parts = re.split(r"[_-]", name)
|
||
|
||
xian_number = ""
|
||
level_1_system_name = ""
|
||
level_2_system_name = ""
|
||
device_name = ""
|
||
tool_name = ""
|
||
tool_code = ""
|
||
|
||
if not parts:
|
||
return {}
|
||
|
||
xian_number_code = parts[0].strip()
|
||
level1_code = ""
|
||
level2_code = ""
|
||
device_code = ""
|
||
|
||
if len(parts) > 1:
|
||
part = parts[1].strip()
|
||
# 防止索引越界
|
||
if len(part) < 2:
|
||
return {}
|
||
|
||
level1_code = part[:2]
|
||
level2_code = part[2:4] if len(part) > 2 else ""
|
||
device_code = part[4:] if len(part) > 4 else ""
|
||
# 若 parts[2] 存在，则用 parts[2] 作为 device_code（支持 122-06A0014-B01001_发动机 格式）
|
||
if len(parts) > 2:
|
||
device_code_alt = parts[2].strip()
|
||
if device_code_alt:
|
||
device_code = device_code_alt
|
||
|
||
data = _load_tree_data()
|
||
root = _resolve_ship_root(data, xian_number_code)
|
||
if not root:
|
||
return {}
|
||
|
||
xian_number = root.get("name", "")
|
||
|
||
# 一级
|
||
s1 = root.get("children", {}).get(level1_code)
|
||
if s1:
|
||
level_1_system_name = s1.get("name", "")
|
||
|
||
# 二级
|
||
s2 = s1.get("children", {}).get(level2_code)
|
||
if s2:
|
||
level_2_system_name = s2.get("name", "")
|
||
|
||
# 三级设备
|
||
device = s2.get("children", {}).get(device_code)
|
||
if device:
|
||
device_name = device.get("name", "")
|
||
# 若存在 level4，匹配与文件名最大相同字符串的零件
|
||
# 内部已包含逻辑：如果设备名匹配度更高，则不返回零件
|
||
tool_name, tool_code = _match_level4_tool(name, device)
|
||
|
||
return {
|
||
"Xian_Number": xian_number,
|
||
"Xian_Number_code": xian_number_code,
|
||
"Level_1_System_Name": level_1_system_name,
|
||
"level1_code": level1_code,
|
||
"Level_2_System_Name": level_2_system_name,
|
||
"level2_code": level2_code,
|
||
"Device_Name": device_name,
|
||
"device_code": device_code,
|
||
"tool_name": tool_name,
|
||
"tool_code": tool_code,
|
||
}
|
||
|
||
|
||
def get_entity(filename):
|
||
result = process_filename(filename)
|
||
entities = []
|
||
relationships = []
|
||
low_level = ""
|
||
|
||
if result:
|
||
xian_number = result.get("Xian_Number")
|
||
xian_number_code = result.get("Xian_Number_code")
|
||
level_1_system_name = result.get("Level_1_System_Name")
|
||
level1_code = result.get("level1_code")
|
||
level_2_system_name = result.get("Level_2_System_Name")
|
||
level2_code = result.get("level2_code")
|
||
device_name = result.get("Device_Name")
|
||
device_code = result.get("device_code")
|
||
tool_name = result.get("tool_name")
|
||
tool_code = result.get("tool_code")
|
||
|
||
if xian_number:
|
||
entities.append(
|
||
{
|
||
"type": "舰艇",
|
||
"properties": {
|
||
"名称": xian_number,
|
||
"舷号": xian_number_code,
|
||
},
|
||
}
|
||
)
|
||
low_level = xian_number
|
||
|
||
if level_1_system_name:
|
||
entities.append(
|
||
{
|
||
"type": "系统",
|
||
"properties": {
|
||
"名称": level_1_system_name,
|
||
"系统编码": level1_code,
|
||
},
|
||
}
|
||
)
|
||
low_level = level_1_system_name
|
||
|
||
if xian_number and level_1_system_name:
|
||
relationships.append(
|
||
{
|
||
"type": "包含",
|
||
"from_entity": xian_number,
|
||
"to_entity": level_1_system_name,
|
||
}
|
||
)
|
||
|
||
if level_2_system_name:
|
||
entities.append(
|
||
{
|
||
"type": "子系统",
|
||
"properties": {
|
||
"名称": level_2_system_name,
|
||
"子系统编码": level2_code,
|
||
},
|
||
}
|
||
)
|
||
low_level = level_2_system_name
|
||
|
||
if level_1_system_name and level_2_system_name:
|
||
relationships.append(
|
||
{
|
||
"type": "包含",
|
||
"from_entity": level_1_system_name,
|
||
"to_entity": level_2_system_name,
|
||
}
|
||
)
|
||
|
||
if device_name:
|
||
entities.append(
|
||
{
|
||
"type": "设备",
|
||
"properties": {
|
||
"名称": device_name,
|
||
"设备编码": device_code,
|
||
},
|
||
}
|
||
)
|
||
low_level = device_name
|
||
|
||
if level_2_system_name and device_name:
|
||
relationships.append(
|
||
{
|
||
"type": "包含",
|
||
"from_entity": level_2_system_name,
|
||
"to_entity": device_name,
|
||
}
|
||
)
|
||
|
||
if tool_name:
|
||
entities.append(
|
||
{
|
||
"type": "零件",
|
||
"properties": {
|
||
"名称": tool_name,
|
||
"零件编码": tool_code,
|
||
},
|
||
}
|
||
)
|
||
low_level = tool_name
|
||
|
||
if device_name and tool_name:
|
||
relationships.append(
|
||
{
|
||
"type": "包含",
|
||
"from_entity": device_name,
|
||
"to_entity": tool_name,
|
||
}
|
||
)
|
||
|
||
final_result = {
|
||
"entities": entities,
|
||
"relationships": relationships,
|
||
"low_level": low_level,
|
||
}
|
||
|
||
return final_result
|
||
|
||
|
||
# 示例测试逻辑（非必须，仅供验证）
|
||
if __name__ == "__main__":
|
||
# 阀控蓄电池脉冲式快速充电装置
|
||
test_list = [
|
||
"777-11A1014-B01001_发动机-维修手册.pdf",
|
||
|
||
]
|
||
for i in test_list:
|
||
result = get_entity(i)
|
||
lower_entity = result.get("low_level")
|
||
print(result)
|
||
print(lower_entity)
|
||
#from neo4j import GraphDatabase
|
||
|
||
#NEO4J_URI = "bolt://192.168.0.111:7687"
|
||
#NEO4J_USER = os.getenv("NEO4J_USER", "neo4j")
|
||
#NEO4J_PASSWORD = os.getenv("NEO4J_PASSWORD","zdht123@") # 不设默认值，强制要求提供
|
||
#NEO4J_DATABASE = os.getenv("NEO4J_DATABASE", "neo4j")
|
||
#driver = GraphDatabase.driver(
|
||
# NEO4J_URI,
|
||
# auth=(NEO4J_USER, NEO4J_PASSWORD),
|
||
# database=NEO4J_DATABASE
|
||
#)
|
||
#res = fetch_graph_sample(driver)
|
||
#print(res)
|
||
|
||
|
||
|
||
|
||
|
||
|