151 lines
7.3 KiB
Python
151 lines
7.3 KiB
Python
import json
|
||
import re
|
||
from flask_app.general.投标人须知正文提取指定内容 import process_nested_data, transform_json, get_requirements_with_gpt, \
|
||
extract_sections, concatenate_keys_values
|
||
|
||
|
||
# 对于每个target_value元素，如果有完美匹配json_data中的键，那就加入这个完美匹配的键名，否则，把全部模糊匹配到的键名都加入
|
||
def find_keys_by_value(target_value, json_data):
|
||
matched_keys = [k for k, v in json_data.items() if v == target_value] # 首先检查 JSON 中的每个键值对，如果值完全等于目标值，则将这些键收集起来。
|
||
if not matched_keys:
|
||
matched_keys = [k for k, v in json_data.items() if
|
||
isinstance(v, str) and v.startswith(target_value)] # 如果没有找到完全匹配的键，它会检查字符串类型的值是否以目标值开头，并收集这些键。
|
||
return matched_keys # eg:[3.1,3.1.1,3.1.2,3.2...]
|
||
|
||
|
||
# 定义查找以特定前缀开始的键的函数，eg:若match_keys中有3.1，那么以3.1为前缀的键都会被找出来，如3.1.1 3.1.2...
|
||
def find_keys_with_prefix(key_prefix, json_data):
|
||
subheadings = [k for k in json_data if k.startswith(key_prefix)]
|
||
return subheadings
|
||
|
||
#将 top_level_key 的值设为 target_value。
|
||
def extract_json(data, target_values):
|
||
results = {}
|
||
for target_value in target_values:
|
||
matched_keys = find_keys_by_value(target_value, data)
|
||
for key in matched_keys:
|
||
key_and_subheadings = find_keys_with_prefix(key, data)
|
||
for subkey in key_and_subheadings:
|
||
if "." in subkey:
|
||
parent_key = subkey.rsplit('.', 1)[0]
|
||
top_level_key = parent_key.split('.')[0] + '.'
|
||
# 特别处理定标相关的顶级键，确保不会重复添加其他键
|
||
if top_level_key not in results:
|
||
results[top_level_key] = target_value
|
||
# 添加或更新父级键
|
||
if parent_key not in results:
|
||
if parent_key in data:
|
||
results[parent_key] = data[parent_key]
|
||
# 添加当前键
|
||
results[subkey] = data[subkey]
|
||
return results
|
||
|
||
def extract_between_sections(data, target_values):
|
||
target_found = False
|
||
extracted_data = {}
|
||
current_section_title = ""
|
||
section_pattern = re.compile(r'^[一二三四五六七八九十]+$') # 匹配 "一", "二", "三" 等大标题
|
||
current_block = {}
|
||
|
||
# 遍历所有键值对
|
||
for key, value in data.items():
|
||
# 只匹配形如 "一": "竞争性磋商响应文件" 的章节标题
|
||
if section_pattern.match(key):
|
||
if target_found:
|
||
# 如果已经找到了符合的章节，并且遇到了另一个章节
|
||
# 保存当前块并重置
|
||
if current_block:
|
||
extracted_data[current_section_title] = current_block
|
||
current_block = {}
|
||
target_found = False
|
||
|
||
# 检查当前标题是否包含 target_values 中的任意关键词
|
||
if any(tv in value for tv in target_values):
|
||
target_found = True # 找到了目标章节，开始捕获后续内容
|
||
current_section_title = value # 保存章节标题内容
|
||
|
||
elif target_found: # 只捕获目标值之后的内容
|
||
current_block[key] = value
|
||
|
||
# 保存最后一个块（如果有的话）
|
||
if current_block:
|
||
extracted_data[current_section_title] = current_block
|
||
|
||
return extracted_data
|
||
|
||
def sort_clean_data_keys(data):
|
||
# 预处理：删除键名中的空格
|
||
def preprocess_key(key):
|
||
return re.sub(r'\s+', '', key)
|
||
|
||
# 将键转换成由整数构成的元组，作为排序依据
|
||
def key_func(key):
|
||
return tuple(int(part) for part in re.split(r'\D+', key) if part)
|
||
|
||
# 创建一个新的字典，键名经过预处理
|
||
preprocessed_data = {preprocess_key(key): value for key, value in data.items()}
|
||
|
||
# 对预处理后的字典键进行排序
|
||
sorted_keys = sorted(preprocessed_data.keys(), key=key_func)
|
||
|
||
# 创建一个新的字典，按照排序后的键添加键值对
|
||
sorted_data = {key: preprocessed_data[key] for key in sorted_keys}
|
||
|
||
return sorted_data
|
||
|
||
"""
|
||
递归处理嵌套的数据结构（字典和列表）。
|
||
对最内层的字符串值应用 post_process 函数。
|
||
post_process 函数尝试将长字符串按特定模式分割成块，每块至少包含50个中英文字符。
|
||
如果字典中所有值都是 ""、"/" 或空列表，则返回'键'的列表。
|
||
"""
|
||
|
||
# 读取JSON数据，提取内容，转换结构，并打印结果
|
||
def extract_from_notice(merged_baseinfo_path,clause_path, type):
|
||
if type == 1:
|
||
target_values = ["投标","投标文件","响应文件"]
|
||
elif type == 2:
|
||
target_values = ["开标", "评标", "定标","磋商程序","中标"]
|
||
elif type == 3:
|
||
target_values = ["重新招标、不再招标和终止招标","重新招标","重新采购", "不再招标", "不再采购","终止招标","终止采购"]
|
||
elif type == 4:
|
||
target_values = ["评标"] # 测试
|
||
else:
|
||
raise ValueError(
|
||
"Invalid type specified. Use 1 for '投标文件, 投标' or 2 for '开标, 评标, 定标'or 3 for '重新招标'")
|
||
with open(clause_path, 'r', encoding='utf-8') as file:
|
||
data = json.load(file)
|
||
extracted_data = extract_between_sections(data, target_values) #先使用大章节'二、投标文件'这种筛选
|
||
if not extracted_data:
|
||
extracted_data = extract_json(data, target_values) # 若没有，再使用'3.投标文件' 筛选
|
||
if not extracted_data:
|
||
final_result = get_requirements_with_gpt(merged_baseinfo_path, type) # 万一都没，那就调用大模型
|
||
return final_result
|
||
print("老方法")
|
||
print(json.dumps(extracted_data,ensure_ascii=False,indent=4))
|
||
final_result=extract_sections(extracted_data,target_values) #后处理，生成键名
|
||
return final_result
|
||
else:
|
||
extracted_data_concatenated = {
|
||
section: concatenate_keys_values(content)
|
||
for section, content in extracted_data.items()
|
||
}
|
||
return extracted_data_concatenated
|
||
# print(json.dumps(res, ensure_ascii=False, indent=4))
|
||
# sorted_data = sort_clean_data_keys(extracted_data) # 对输入的字典 data 的键进行预处理和排序
|
||
# transformed_data = transform_json(sorted_data)
|
||
# print(json.dumps(transformed_data,ensure_ascii=False,indent=4))
|
||
# final_result = process_nested_data(transformed_data)
|
||
# return final_result
|
||
|
||
if __name__ == "__main__":
|
||
# file_path = 'C:\\Users\\Administrator\\Desktop\\fsdownload\\3bffaa84-2434-4bd0-a8ee-5c234ccd7fa0\\clause1.json'
|
||
merged_baseinfo_path=r"D:\flask_project\flask_app\static\output\output1\05339b83-50bf-4405-905c-38625928840e\merged_baseinfo_path_more.pdf"
|
||
clause_path=r"D:\flask_project\flask_app\static\output\output1\05339b83-50bf-4405-905c-38625928840e\clause1.json"
|
||
try:
|
||
res = extract_from_notice(merged_baseinfo_path,clause_path, 1) # 可以改变此处的 type 参数测试不同的场景
|
||
res2 = json.dumps(res, ensure_ascii=False, indent=4)
|
||
print(res2)
|
||
except ValueError as e:
|
||
print(e)
|