mirror of
https://github.com/LiteyukiStudio/LiteyukiBot.git
synced 2024-11-26 14:05:03 +08:00
update: Liteyuki ORM
This commit is contained in:
parent
a3f63e383d
commit
8303514fa0
295
src/api/data.py
295
src/api/data.py
@ -1,241 +1,176 @@
|
||||
import json
|
||||
import sqlite3
|
||||
import uuid
|
||||
from abc import ABC
|
||||
from collections.abc import Iterable
|
||||
from typing import Any
|
||||
|
||||
from pymongo import MongoClient
|
||||
from pydantic import BaseModel
|
||||
|
||||
BaseIterable = list | tuple | set | dict
|
||||
|
||||
|
||||
class LiteModel(BaseModel):
|
||||
pass
|
||||
|
||||
|
||||
class BaseORM:
|
||||
|
||||
def __init__(self, *args, **kwargs):
|
||||
class BaseORMAdapter(ABC):
|
||||
def __init__(self):
|
||||
pass
|
||||
|
||||
def auto_migrate(self, *args, **kwargs):
|
||||
"""自动迁移数据库
|
||||
Args:
|
||||
*args:
|
||||
**kwargs:
|
||||
"""自动迁移
|
||||
|
||||
Returns:
|
||||
|
||||
"""
|
||||
raise NotImplementedError
|
||||
|
||||
def save(self, *args, **kwargs):
|
||||
"""创建数据
|
||||
Args:
|
||||
*args:
|
||||
**kwargs:
|
||||
"""存储数据
|
||||
|
||||
Returns:
|
||||
"""
|
||||
raise NotImplementedError
|
||||
|
||||
def update(self, *args, **kwargs):
|
||||
"""更新数据
|
||||
Args:
|
||||
*args:
|
||||
**kwargs:
|
||||
|
||||
Returns:
|
||||
"""
|
||||
raise NotImplementedError
|
||||
|
||||
def delete(self, *args, **kwargs):
|
||||
"""删除数据
|
||||
Args:
|
||||
*args:
|
||||
**kwargs:
|
||||
|
||||
Returns:
|
||||
"""
|
||||
raise NotImplementedError
|
||||
|
||||
def first(self, *args, **kwargs):
|
||||
"""查询第一条数据
|
||||
Args:
|
||||
*args:
|
||||
**kwargs:
|
||||
|
||||
Returns:
|
||||
"""
|
||||
raise NotImplementedError
|
||||
|
||||
def where(self, *args, **kwargs):
|
||||
"""查询数据
|
||||
Args:
|
||||
*args:
|
||||
**kwargs:
|
||||
|
||||
Returns:
|
||||
"""
|
||||
raise NotImplementedError
|
||||
|
||||
def all(self, *args, **kwargs):
|
||||
"""查询所有数据
|
||||
Args:
|
||||
*args:
|
||||
**kwargs:
|
||||
|
||||
Returns:
|
||||
|
||||
"""
|
||||
raise NotImplementedError
|
||||
|
||||
def delete(self, *args, **kwargs):
|
||||
"""删除数据
|
||||
|
||||
Returns:
|
||||
|
||||
"""
|
||||
raise NotImplementedError
|
||||
|
||||
def update(self, *args, **kwargs):
|
||||
"""更新数据
|
||||
|
||||
Returns:
|
||||
|
||||
"""
|
||||
raise NotImplementedError
|
||||
|
||||
|
||||
class SqliteORM(BaseORM):
|
||||
"""同步sqlite数据库操作"""
|
||||
class SqliteORMAdapter(BaseORMAdapter):
|
||||
type_map = {
|
||||
int: 'INTEGER',
|
||||
float: 'REAL',
|
||||
str: 'TEXT',
|
||||
bool: 'INTEGER',
|
||||
# default: TEXT
|
||||
str : 'TEXT',
|
||||
int : 'INTEGER',
|
||||
float: 'REAL',
|
||||
bool : 'INTEGER',
|
||||
list : 'TEXT'
|
||||
}
|
||||
|
||||
def __init__(self, db, *args, **kwargs):
|
||||
super().__init__(*args, **kwargs)
|
||||
self.db = sqlite3.connect(db)
|
||||
def __init__(self, db_name: str):
|
||||
super().__init__()
|
||||
self.conn = sqlite3.connect(db_name)
|
||||
self.cursor = self.conn.cursor()
|
||||
|
||||
@staticmethod
|
||||
def get_model_table_name(model: type(LiteModel) | LiteModel | str) -> str:
|
||||
"""获取模型对应的表名"""
|
||||
if isinstance(model, str):
|
||||
return model
|
||||
elif isinstance(model, LiteModel):
|
||||
return model.__class__.__name__
|
||||
elif isinstance(model, type(LiteModel)):
|
||||
return model.__name__
|
||||
def auto_migrate(self, *args: 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
|
||||
|
||||
def auto_migrate(self, *args: type(LiteModel) | LiteModel | str, **kwargs):
|
||||
"""自动迁移数据库
|
||||
Args:
|
||||
*args: BaseModel
|
||||
**kwargs:
|
||||
delete_old_columns: bool = False # 是否删除旧字段
|
||||
add_new_columns: bool = True # 添加新字段
|
||||
*args:
|
||||
|
||||
Returns:
|
||||
|
||||
"""
|
||||
for model in args:
|
||||
# 获取模型对应的表名
|
||||
table_name = self.get_model_table_name(model)
|
||||
# 检测并创建表
|
||||
self.cursor.execute(f'CREATE TABLE IF NOT EXISTS {model.__name__}(id INTEGER PRIMARY KEY AUTOINCREMENT)')
|
||||
# 获取表字段
|
||||
self.cursor.execute(f'PRAGMA table_info({model.__name__})')
|
||||
table_fields = self.cursor.fetchall()
|
||||
table_fields = [field[1] for field in table_fields]
|
||||
|
||||
# 获取表中已有的字段
|
||||
existing_columns = set()
|
||||
cursor = self.db.execute(f"PRAGMA table_info({table_name})")
|
||||
for column_info in cursor.fetchall():
|
||||
existing_columns.add(column_info[1])
|
||||
# 获取模型字段
|
||||
model_fields = model.__dict__.keys()
|
||||
|
||||
# 创建表,如果不存在的话
|
||||
self.db.execute(f'CREATE TABLE IF NOT EXISTS {table_name} (id INTEGER PRIMARY KEY AUTOINCREMENT)')
|
||||
# 获取模型字段类型
|
||||
model_types = [self.type_map.get(type(model.__dict__[field]), 'TEXT') for field in model_fields]
|
||||
|
||||
# 检测模型中的字段并添加新字段,按照类型添加
|
||||
for field_name, field_type in model.__annotations__.items():
|
||||
if field_name not in existing_columns:
|
||||
self.db.execute(f'ALTER TABLE {table_name} ADD COLUMN {field_name} {self.type_map.get(field_type, "TEXT")}')
|
||||
# 获取模型字段默认值
|
||||
model_defaults = [model.__dict__[field] for field in model_fields]
|
||||
|
||||
# 提交事务
|
||||
self.db.commit()
|
||||
# 检测新字段
|
||||
for field, type_, default in zip(model_fields, model_types, model_defaults):
|
||||
if field not in table_fields:
|
||||
self.cursor.execute(f'ALTER TABLE {model.__name__} ADD COLUMN {field} {type_} DEFAULT {default}')
|
||||
|
||||
# 检测多余字段
|
||||
for field in table_fields:
|
||||
if field not in model_fields:
|
||||
self.cursor.execute(f'ALTER TABLE {model.__name__} DROP COLUMN {field}')
|
||||
|
||||
self.conn.commit()
|
||||
|
||||
def save(self, *models: LiteModel) -> int:
|
||||
"""存储数据
|
||||
|
||||
def save(self, model: LiteModel) -> int:
|
||||
"""保存或创建数据,对嵌套模型扁平化处理,加特殊前缀表示为模型,实际储存模型id,嵌套对象单独储存
|
||||
Args:
|
||||
model: BaseModel
|
||||
Returns: id主键
|
||||
"""
|
||||
# 先检测表是否存在,不存在则创建
|
||||
table_name = self.get_model_table_name(model)
|
||||
self.db.execute(f'CREATE TABLE IF NOT EXISTS {table_name} (id INTEGER PRIMARY KEY AUTOINCREMENT)')
|
||||
models: 数据
|
||||
|
||||
# 构建插入语句
|
||||
column_list = []
|
||||
value_list = []
|
||||
for key, value in model.dict().items():
|
||||
if isinstance(value, LiteModel):
|
||||
# 如果是嵌套模型,先保存嵌套模型
|
||||
nested_model_id = self.save(value)
|
||||
# 保存嵌套模型的id, 以特殊前缀表示为模型
|
||||
column_list.append(f'$id_{key}')
|
||||
value_list.append(f'{value.__class__.__name__}_{nested_model_id}')
|
||||
elif isinstance(value, list):
|
||||
# 如果是列表,先保存列表中的所有嵌套模型,有可能有多种类型的嵌套模型
|
||||
# 列表内存'ModelType_ModelId',以特殊前缀表示为模型类型
|
||||
nested_model_ids = []
|
||||
for nested_model in value:
|
||||
nested_model_id = self.save(nested_model)
|
||||
nested_model_ids.append(f'{nested_model.__class__.__name__}_{nested_model_id}')
|
||||
column_list.append(f'$ids_{key}')
|
||||
value_list.append(nested_model_ids)
|
||||
|
||||
columns = ', '.join(column_list)
|
||||
placeholders = ', '.join(['?' for _ in value_list])
|
||||
values = tuple(value_list)
|
||||
print(model.dict())
|
||||
print(table_name, columns, placeholders, values)
|
||||
|
||||
# 插入数据
|
||||
self.db.execute(f'INSERT INTO {table_name} ({columns}) VALUES ({placeholders})', values)
|
||||
self.db.commit()
|
||||
return self.db.execute(f'SELECT last_insert_rowid()').fetchone()[0]
|
||||
|
||||
def where(self, model_type: type(LiteModel) | str, conditions: str, *args, objectify: bool = True) -> list[LiteModel]:
|
||||
"""查询数据
|
||||
Args:
|
||||
objectify: bool: 是否将查询结果转换为模型
|
||||
model_type: BaseModel
|
||||
conditions: str
|
||||
*args:
|
||||
Returns:
|
||||
"""
|
||||
table_name = self.get_model_table_name(model_type)
|
||||
self.db.execute(f'CREATE TABLE IF NOT EXISTS {table_name} (id INTEGER PRIMARY KEY AUTOINCREMENT)')
|
||||
return [self._convert_to_model(model_type, item) for item in self.db.execute(f'SELECT * FROM {table_name} WHERE {conditions}', args).fetchall()]
|
||||
|
||||
def first(self, model_type: type(LiteModel) | str, conditions: str, *args, objectify: bool = True):
|
||||
"""查询第一条数据"""
|
||||
table_name = self.get_model_table_name(model_type)
|
||||
self.db.execute(f'CREATE TABLE IF NOT EXISTS {table_name} (id INTEGER PRIMARY KEY AUTOINCREMENT)')
|
||||
return self._convert_to_model(model_type, self.db.execute(f'SELECT * FROM {table_name} WHERE {conditions}', args).fetchone())
|
||||
|
||||
def all(self, model_type: type(LiteModel) | str, objectify: bool = True):
|
||||
"""查询所有数据"""
|
||||
table_name = self.get_model_table_name(model_type)
|
||||
self.db.execute(f'CREATE TABLE IF NOT EXISTS {table_name} (id INTEGER PRIMARY KEY AUTOINCREMENT)')
|
||||
return [self._convert_to_model(model_type, item) for item in self.db.execute(f'SELECT * FROM {table_name}').fetchall()]
|
||||
|
||||
def update(self, model_type: type(LiteModel) | str, operation: str, conditions: str, *args):
|
||||
"""更新数据
|
||||
Args:
|
||||
model_type: BaseModel
|
||||
operation: str: 更新操作
|
||||
conditions: str: 查询条件
|
||||
*args:
|
||||
Returns:
|
||||
id: 数据id,多个数据返回最后一个数据id
|
||||
"""
|
||||
table_name = self.get_model_table_name(model_type)
|
||||
self.db.execute(f'CREATE TABLE IF NOT EXISTS {table_name} (id INTEGER PRIMARY KEY AUTOINCREMENT)')
|
||||
self.db.execute(f'UPDATE {table_name} SET {operation} WHERE {conditions}', args)
|
||||
self.db.commit()
|
||||
_id = 0
|
||||
for model in models:
|
||||
table_name = model.__class__.__name__
|
||||
key_list = []
|
||||
value_list = []
|
||||
# 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))
|
||||
else:
|
||||
key_list.append(field)
|
||||
value_list.append(value)
|
||||
|
||||
def _convert_to_model(self, model_type: type(LiteModel), item: tuple) -> LiteModel:
|
||||
"""将查询结果转换为模型,处理嵌套模型"""
|
||||
# 获取表中已有的字段,再用字段值构建字典
|
||||
table_name = self.get_model_table_name(model_type)
|
||||
cursor = self.db.execute(f"PRAGMA table_info({table_name})")
|
||||
columns = [column_info[1] for column_info in cursor.fetchall()]
|
||||
item_dict = dict(zip(columns, item))
|
||||
# 遍历字典,处理嵌套模型
|
||||
new_item_dict = {}
|
||||
for key, value in item_dict.items():
|
||||
if key.startswith('$id_'):
|
||||
# 处理单个嵌套模型类型时从model_type中获取键
|
||||
new_item_dict[key.replace('$id_', '')] = self.first(model_type.__annotations__[key.replace('$id_', '')], 'id = ?', value.split('_')[-1])
|
||||
elif key.startswith('$ids_'):
|
||||
# 处理多个嵌套模型类型使用eval获取数据库对应索引的键
|
||||
new_item_dict[key.replace('$ids_', '')] = [self.first(eval(type_id.split('_')[0]), 'id = ?', type_id.split('_')[-1]) for type_id in value]
|
||||
else:
|
||||
new_item_dict[key] = value
|
||||
def flat(self, data: Iterable) -> Any:
|
||||
if isinstance(data, dict):
|
||||
for k, v in data.items():
|
||||
if isinstance(v, dict | list | tuple | set):
|
||||
self.flat(v)
|
||||
else:
|
||||
print(k, v)
|
||||
|
||||
return model_type(**new_item_dict)
|
||||
def first(self, model: type(LiteModel), *args, **kwargs):
|
||||
self.cursor.execute(f'SELECT * FROM {model.__name__} WHERE {args[0]}', args[1])
|
||||
return self.convert2dict(model, self.cursor.fetchone())
|
||||
|
||||
def convert2dict(self, data: dict) -> dict | list:
|
||||
"""将模型转换为dict
|
||||
|
||||
Args:
|
||||
data: 数据库查询结果
|
||||
|
||||
Returns:
|
||||
模型对象
|
||||
"""
|
||||
for field, value in data.items():
|
||||
if field.startswith('$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))
|
||||
|
Loading…
Reference in New Issue
Block a user