1
0
forked from bot/app

update: Liteyuki ORM

This commit is contained in:
远野千束 2024-03-02 02:43:18 +08:00
parent 8303514fa0
commit e7c29c1597
2 changed files with 248 additions and 51 deletions

63
db_test2.py Normal file
View File

@ -0,0 +1,63 @@
import json
import random
from src.api.data import LiteModel, SqliteORMAdapter
class Score(LiteModel):
subject: str
score: int
class Student(LiteModel):
name: str
age: str
sex: str
scores: list[Score] = []
class Teacher(LiteModel):
name: str
age: str
class Class(LiteModel):
name: str
students: list[Student]
test_json: dict[str, Student] = {'张三': Student}
class University(LiteModel):
name: str
address: str
rank: int = 0
classes: list[Class] = []
students: list[Student] = []
cqupt = University(name='重庆邮电大学', address='重庆', rank=100)
pku = University(name='北京大学', address='北京', rank=1)
cqu = University(name='重庆大学', address='重庆', rank=10)
student_name_list = ['张三', '李四', '王五', '赵六', '钱七', '孙八', '周九', '吴十']
student_list = [Student(name=name, age=19, sex=random.choice(['', ''])) for name in student_name_list]
# 构建复杂模型进行数据库测试
cqupt.students = student_list
cqupt.classes = [Class(name='软件工程', students=student_list, test_json={name: student for name, student in zip(student_name_list, student_list)})]
for student in student_list:
student.scores = [Score(subject='语文', score=random.randint(60, 100)), Score(subject='数学', score=random.randint(60, 100)),
Score(subject='英语', score=random.randint(60, 100))]
student.scores.append(Score(subject='物理', score=random.randint(60, 100)))
student.scores.append(Score(subject='化学', score=random.randint(60, 100)))
student.scores.append(Score(subject='生物', score=random.randint(60, 100)))
student.scores.append(Score(subject='历史', score=random.randint(60, 100)))
student.scores.append(Score(subject='地理', score=random.randint(60, 100)))
student.scores.append(Score(subject='政治', score=random.randint(60, 100))
)
print(json.dumps(cqupt.dict(), indent=4, ensure_ascii=False))
db = SqliteORMAdapter('test2.db')
db.auto_migrate(University, Class, Student, Score)
db.save(cqupt)
# 查询测试
db.first(University, 'name = ?', '重庆邮电大学')

View File

@ -1,7 +1,10 @@
import json
import sqlite3
import types
import typing
from abc import ABC
from collections.abc import Iterable
from copy import deepcopy
from typing import Any
from pydantic import BaseModel
@ -10,7 +13,10 @@ BaseIterable = list | tuple | set | dict
class LiteModel(BaseModel):
pass
"""轻量级模型基类
类型注解统一使用Python3.9的PEP585标准如需使用泛型请使用typing模块的泛型类型
"""
id: Any = None
class BaseORMAdapter(ABC):
@ -67,6 +73,11 @@ class BaseORMAdapter(ABC):
class SqliteORMAdapter(BaseORMAdapter):
"""SQLiteORM适配器严禁使用FORIEGNID和JSON作为主键前缀严禁使用$ID:作为字符串值前缀
Attributes:
"""
type_map = {
# default: TEXT
str : 'TEXT',
@ -75,60 +86,79 @@ class SqliteORMAdapter(BaseORMAdapter):
bool : 'INTEGER',
list : 'TEXT'
}
FOREIGNID = 'FOREIGNID'
JSON = 'JSON'
ID = '$ID'
def __init__(self, db_name: str):
super().__init__()
self.conn = sqlite3.connect(db_name)
self.conn.row_factory = sqlite3.Row
self.cursor = self.conn.cursor()
def auto_migrate(self, *args: LiteModel):
def auto_migrate(self, *args: type(LiteModel)):
"""自动迁移,检测新模型字段和原有表字段的差异,如有差异自动增删新字段
Args:
*args:
*args: 模型类
Returns:
"""
for model in args:
# 检测并创建表
self.cursor.execute(f'CREATE TABLE IF NOT EXISTS {model.__name__}(id INTEGER PRIMARY KEY AUTOINCREMENT)')
model: type(LiteModel)
# 检测并创建表若模型未定义id字段则使用自增主键有定义的话使用id字段且id有可能为字符串
table_name = model.__name__
if 'id' in model.__annotations__ and model.__annotations__['id'] is not None:
# 如果模型定义了id字段那么使用模型的id字段
id_type = self.type_map.get(model.__annotations__['id'], 'TEXT')
self.cursor.execute(f'CREATE TABLE IF NOT EXISTS {table_name} (id {id_type} PRIMARY KEY)')
else:
# 如果模型未定义id字段那么使用自增主键
self.cursor.execute(f'CREATE TABLE IF NOT EXISTS {table_name} (id INTEGER PRIMARY KEY AUTOINCREMENT)')
# 获取表字段
self.cursor.execute(f'PRAGMA table_info({model.__name__})')
self.cursor.execute(f'PRAGMA table_info({table_name})')
table_fields = self.cursor.fetchall()
table_fields = [field[1] for field in table_fields]
# 获取模型字段
model_fields = model.__dict__.keys()
# 获取模型字段类型
model_types = [self.type_map.get(type(model.__dict__[field]), 'TEXT') for field in model_fields]
# 获取模型字段默认值
model_defaults = [model.__dict__[field] for field in model_fields]
raw_fields = model.__annotations__.keys()
# 获取模型字段若有模型则添加FOREIGNID前缀若为BaseIterable则添加JSON前缀用多行if判断
model_fields = []
model_types = []
for field in raw_fields:
if isinstance(model.__annotations__[field], type(LiteModel)):
model_fields.append(f'{self.FOREIGNID}{field}')
model_types.append('TEXT')
elif isinstance(model.__annotations__[field], types.GenericAlias):
model_fields.append(f'{self.JSON}{field}')
model_types.append('TEXT')
else:
model_fields.append(field)
model_types.append(self.type_map.get(model.__annotations__[field], 'TEXT'))
# 检测新字段
for field, type_, default in zip(model_fields, model_types, model_defaults):
for field, type_ in zip(model_fields, model_types):
if field not in table_fields:
self.cursor.execute(f'ALTER TABLE {model.__name__} ADD COLUMN {field} {type_} DEFAULT {default}')
print(f'ALTER TABLE {table_name} ADD COLUMN {field} {type_}')
self.cursor.execute(f'ALTER TABLE {table_name} ADD COLUMN {field} {type_}')
# 检测多余字段
# 检测多余字段除了id字段
for field in table_fields:
if field not in model_fields:
self.cursor.execute(f'ALTER TABLE {model.__name__} DROP COLUMN {field}')
if field not in model_fields and field != 'id':
self.cursor.execute(f'ALTER TABLE {table_name} DROP COLUMN {field}')
self.conn.commit()
def save(self, *models: LiteModel) -> int:
"""存储数据
def save(self, *models: LiteModel) -> int | tuple:
"""存储数据检查id字段如果有id字段则更新没有则插入
Args:
models: 数据
Returns:
id: 数据id多个数据返回最后一个数据id
id: 数据id如果有多个数据返回id元组
"""
_id = 0
ids = []
for model in models:
table_name = model.__class__.__name__
key_list = []
@ -136,41 +166,145 @@ class SqliteORMAdapter(BaseORMAdapter):
# 处理外键,添加前缀'$IDFieldName'
for field, value in model.__dict__.items():
if isinstance(value, LiteModel):
key_list.append(f'$id:{field}')
value_list.append(f'{value.__class__.__name__}:{self.save(value)}')
elif isinstance(value, list | tuple | set):
key_list.append(field)
value_list.append(json.dumps(value))
key_list.append(f'{self.FOREIGNID}{field}')
value_list.append(f'{self.ID}:{value.__class__.__name__}:{self.save(value)}')
elif isinstance(value, BaseIterable):
key_list.append(f'{self.JSON}{field}')
value_list.append(self.flat(value))
else:
key_list.append(field)
value_list.append(value)
# 更新或插入数据,用?占位
self.cursor.execute(f'INSERT OR REPLACE INTO {table_name} ({",".join(key_list)}) VALUES ({",".join(["?" for _ in key_list])})', value_list)
def flat(self, data: Iterable) -> Any:
if isinstance(data, dict):
for k, v in data.items():
if isinstance(v, dict | list | tuple | set):
self.flat(v)
else:
print(k, v)
ids.append(self.cursor.lastrowid)
self.conn.commit()
return ids[0] if len(ids) == 1 else tuple(ids)
def first(self, model: type(LiteModel), *args, **kwargs):
self.cursor.execute(f'SELECT * FROM {model.__name__} WHERE {args[0]}', args[1])
return self.convert2dict(model, self.cursor.fetchone())
def convert2dict(self, data: dict) -> dict | list:
"""将模型转换为dict
def flat(self, data: Iterable) -> str:
"""扁平化数据,返回扁平化对象
Args:
data: 数据库查询结果
data: 数据可迭代对象
Returns: json字符串
"""
if isinstance(data, dict):
return_data = {}
for k, v in data.items():
if isinstance(v, LiteModel):
return_data[f'{self.FOREIGNID}{k}'] = f'{self.ID}:{v.__class__.__name__}:{self.save(v)}'
elif isinstance(v, BaseIterable):
return_data[f'{self.JSON}{k}'] = self.flat(v)
else:
return_data[k] = v
elif isinstance(data, list | tuple | set):
return_data = []
for v in data:
if isinstance(v, LiteModel):
return_data.append(f'{self.ID}:{v.__class__.__name__}:{self.save(v)}')
elif isinstance(v, BaseIterable):
return_data.append(self.flat(v))
else:
return_data.append(v)
else:
raise ValueError('数据类型错误')
return json.dumps(return_data)
def first(self, model: type(LiteModel), conditions, *args, default: Any = None) -> LiteModel | None:
"""查询第一条数据
Args:
model: 模型
conditions: 查询条件
*args: 参数化查询条件参数
default: 未查询到结果默认返回值
Returns: 数据
"""
table_name = model.__name__
self.cursor.execute(f'SELECT * FROM {table_name} WHERE {conditions}', args)
data = dict(self.cursor.fetchone())
return model(**self.convert_to_dict(data)) if data else default
def all(self, model: type(LiteModel), conditions, *args, default: Any = None) -> list[LiteModel] | None:
"""查询所有数据
Args:
model: 模型
conditions: 查询条件
*args: 参数化查询条件参数
default: 未查询到结果默认返回值
Returns: 数据
"""
table_name = model.__name__
self.cursor.execute(f'SELECT * FROM {table_name} WHERE {conditions}', args)
data = self.cursor.fetchall()
return [model(**self.convert_to_dict(d)) for d in data] if data else default
def delete(self, model: type(LiteModel), conditions, *args):
"""删除数据
Args:
model: 模型
conditions: 查询条件
*args: 参数化查询条件参数
Returns:
模型对象
"""
for field, value in data.items():
if field.startswith('$id'):
# 外键处理
table_name = value[1:].split('_')[0]
id_ = value.split('_')[1]
key_tuple = self.cursor.execute(f'PRAGMA table_info({table_name})').fetchall()
value_tuple = self.cursor.execute(f'SELECT * FROM {table_name} WHERE id = ?', id_).fetchone()
field_dict = dict(zip([field[1] for field in key_tuple], value_tuple))
table_name = model.__name__
self.cursor.execute(f'DELETE FROM {table_name} WHERE {conditions}', args)
self.conn.commit()
def update(self, model: type(LiteModel), conditions: str, *args, operation: str):
"""更新数据
Args:
model: 模型
conditions: 查询条件
*args: 参数化查询条件参数
operation: 更新操作
Returns:
"""
table_name = model.__name__
self.cursor.execute(f'UPDATE {table_name} SET {operation} WHERE {conditions}', args)
self.conn.commit()
def convert_to_dict(self, data: dict) -> dict:
"""将json字符串转换为字典
Args:
data: json字符串
Returns: 字典
"""
def load(d: BaseIterable) -> BaseIterable:
"""递归加载数据,去除前缀"""
if isinstance(d, dict):
new_d = {}
for k, v in d.items():
if k.startswith(self.FOREIGNID):
new_d[k.replace(self.FOREIGNID, '')] = load(dict(self.cursor.execute(f'SELECT * FROM {v.split(":")[1]} WHERE id = ?', (v.split(":")[2],)).fetchone()))
elif k.startswith(self.JSON):
new_d[k.replace(self.JSON, '')] = load(json.loads(v))
else:
new_d[k] = v
elif isinstance(d, list | tuple | set):
new_d = []
for i, v in enumerate(d):
if isinstance(v, str) and v.startswith(self.ID):
new_d.append(load(dict(self.cursor.execute(f'SELECT * FROM {v.split(":")[1]} WHERE id = ?', (v.split(":")[2],)).fetchone())))
elif isinstance(v, BaseIterable):
new_d.append(load(v))
else:
new_d = d
return new_d
return load(data)