跳到主要内容

TiDB 和 C# 的简单 CRUD 应用程序

作者:hey-hoho,不知名 CRUD 程序员,TiDB 社区技术布道师、版主,分布式技术爱好者。

本文演示如何使用 C# 语言实现对 TiDB 的基础增删改查操作,包含了 C# 中常用的几种数据库访问方式。

相关环境

  • Ubuntu 18.04
  • .NET 6.0
  • C# 10
  • Visual Studio Code 1.63.2
  • TiDB 6.0-DMR

创建 TiDB 测试集群

你可以使用以下方式快速搭建一个 TiDB 测试集群:

本文仅用于代码演示,在单机环境使用 TiUP Playground 搭建了一套最基础的测试集群:

[root@dbserver1 ~]# tiup playground v6.0.0 --db 1 --pd 1 --kv 3 --tag tidb
tiup is checking updates for component playground ...timeout!
Starting component `playground`: /root/.tiup/components/playground/v1.9.5/tiup-playground /root/.tiup/components/playground/v1.9.5/tiup-playground v6.0.0 --db 1 --pd 1 --kv 3 --tag tidb
Playground Bootstrapping...
Start pd instance:v6.0.0
Start tikv instance:v6.0.0
Start tikv instance:v6.0.0
Start tikv instance:v6.0.0
Start tidb instance:v6.0.0
Waiting for tidb instances ready
127.0.0.1:4000 ... Done
Start tiflash instance:v6.0.0
Waiting for tiflash instances ready
127.0.0.1:3930 ... Done
CLUSTER START SUCCESSFULLY, Enjoy it ^-^
To connect TiDB: mysql --comments --host 127.0.0.1 --port 4000 -u root -p (no password)
To view the dashboard: http://127.0.0.1:2379/dashboard
PD client endpoints: [127.0.0.1:2379]
To view the Prometheus: http://127.0.0.1:9090
To view the Grafana: http://127.0.0.1:3000

数据库启动成功之后,我们用官方提供的 Bookshop 示例应用作为测试数据,使用如下命令生成数据:

tiup demo bookshop prepare --users=1000 --books=5000 --authors=1000 --ratings=10000 --orders=2000 --drop-tables

最后看一下测试数据的生成情况:

+-----------------------+----------------+-----------+------------+--------+
| Table Name | Number of Rows | Data Size | Index Size | Total |
+-----------------------+----------------+-----------+------------+--------+
| bookshop.orders | 2000 | 0.08MB | 0.02MB | 0.09MB |
| bookshop.ratings | 10000 | 0.31MB | 0.31MB | 0.61MB |
| bookshop.book_authors | 5000 | 0.08MB | 0.08MB | 0.15MB |
| bookshop.authors | 1000 | 0.04MB | 0.00MB | 0.04MB |
| bookshop.users | 999 | 0.03MB | 0.01MB | 0.04MB |
| bookshop.books | 5000 | 0.28MB | 0.00MB | 0.28MB |
+-----------------------+----------------+-----------+------------+--------+
6 rows in set (0.01 sec)

创建 C# 应用程序

为了简化演示代码,这里构建一个最简单的控制台应用程序用于数据库访问。

dotnet new console --name tidb-example-csharp --framework net6.0

看一下这个控制台程序的项目结构:

dc@dc-virtual-machine:~/dotnet$ ll tidb-example-csharp/
total 20
drwxrwxr-x 3 dc dc 4096 May 17 16:43 ./
drwxrwxr-x 3 dc dc 4096 May 17 16:43 ../
drwxrwxr-x 2 dc dc 4096 May 17 16:43 obj/
-rw-rw-r-- 1 dc dc 105 May 17 16:43 Program.cs
-rw-rw-r-- 1 dc dc 305 May 17 16:43 tidb-example-csharp.csproj

tidb-example-csharp.csproj是项目工程文件,Program.cs是程序入口文件。

验证一下程序是否能正常运行:

dc@dc-virtual-machine:~/dotnet/tidb-example-csharp$ dotnet run
Hello, World!

驱动程序和 ORM

ADO.NET 提供了开发者在 .NET 平台上对各种数据源的一致性访问标准,类似于 Java 的 jdbc,或者 Golang 的 sql/database 。在此基础上,我们通过实现了 ADO.NET 接口的驱动程序就能访问和操作不同的数据库,包括 SQL Server、PostgreSQL、MySQL、Oracle、Sqlite、Access 等等。

ADO.NET 体系结构(图片来自微软官网):

TiDB 高度兼容 MySQL 5.7 协议,还实现了一部分 MySQL 8.0 特性,所以市面上 MySQL 能使用的驱动程序和 ORM 框架基本都能用在 TiDB 上面。

梳理一下可以发现,在 C# 中实现对 TiDB 的操作分为以下两种方式:

  • 基于 ADO.NET 的驱动程序,代表有Oracle Connector/NETMySqlConnector
  • ORM 数据访问框架,代表有Entity FrameworkDapper

下面分别演示如何对 TiDB 实现增删改查操作。

使用 Oracle Connector/NET

Connector/NET 是 MySQL 官方提供的符合标准 ADO.NET 体系的数据库访问驱动,由于 TiDB 高度兼容 MySQL 协议,所以市面上 MySQL 能使用的驱动基本都能用在 TiDB 上面。

如果要以 ADO.NET 接口方式访问 TiDB ,首先安装驱动程序包:

dotnet add package MySql.Data --version 8.0.29

测试是否能连接上 TiDB :

using MySql.Data.MySqlClient;

const string conectionStr = "Server=127.0.0.1;UserId=root;Password=;Port=4000;Database=bookshop";

public void TestConnection()
{
MySqlConnection conn = new MySqlConnection(conectionStr);
conn.Open();
var cmd = conn.CreateCommand();
cmd.CommandText = "select tidb_version()";
var result = cmd.ExecuteScalar();
Console.WriteLine(result);
conn.Close();
}
dc@dc-virtual-machine:~/dotnet/tidb-example-csharp$ dotnet run
Release Version: v6.0.0
Edition: Community
Git Commit Hash: 36a9810441ca0e496cbd22064af274b3be771081
Git Branch: heads/refs/tags/v6.0.0
UTC Build Time: 2022-03-31 10:33:28
GoVersion: go1.18
Race Enabled: false
TiKV Min Version: v3.0.0-60965b006877ca7234adaced7890d7b029ed1306
Check Table Before Drop: false

一个简单的查询示例:

public void TestRead()
{
MySqlConnection conn = new MySqlConnection(conectionStr);
conn.Open();
var cmd = conn.CreateCommand();
cmd.CommandText = "select * from books limit 5";
MySqlDataReader reader = cmd.ExecuteReader();
while (reader.Read())
{
Console.WriteLine($"id: {reader["id"]} title: {reader["title"]} type: {reader["type"]} published_at: {reader["published_at"]}");
}
conn.Close();
}
dc@dc-virtual-machine:~/dotnet/tidb-example-csharp$ dotnet run
id: 648872 title: Sam Mayert type: Life published_at: 2/26/1953 12:53:33 PM
id: 6583703 title: Aron Kilback type: Kids published_at: 11/23/1923 9:19:43 AM
id: 6810515 title: Chelsey Dickens type: Education & Reference published_at: 4/8/1985 9:23:37 PM
id: 7884508 title: Annetta Rodriguez type: Education & Reference published_at: 5/11/1962 9:54:58 PM
id: 8683541 title: The Documentary of hamster type: Magazine published_at: 10/3/1945 1:44:52 AM

一个简单的新增修改删除示例:

public void TestWrite()
{
MySqlConnection conn = new MySqlConnection(conectionStr);
conn.Open();

int bookId = 888888;

var cmd1 = new MySqlCommand("insert into books values (@id,@title,@type,@published_at,@stock,@price)", conn);
cmd1.Parameters.AddWithValue("@id", bookId);
cmd1.Parameters.AddWithValue("@title", "TiDB in action");
cmd1.Parameters.AddWithValue("@type", "Science & Technology");
cmd1.Parameters.AddWithValue("@published_at", DateTime.Now);
cmd1.Parameters.AddWithValue("@stock", 1000);
cmd1.Parameters.AddWithValue("@price", 66.66);
int insertCnt = cmd1.ExecuteNonQuery();
Console.WriteLine($"insert successed {insertCnt} books.");
TestQueryBook(conn, bookId);

var cmd2 = new MySqlCommand("update books set stock=stock-1 where id=@id", conn);
cmd2.Parameters.AddWithValue("@id", bookId);
int updateCnt = cmd2.ExecuteNonQuery();
Console.WriteLine($"update successed {updateCnt} books.");
TestQueryBook(conn, bookId);

var cmd3 = new MySqlCommand("delete from books where id=@id", conn);
cmd3.Parameters.AddWithValue("@id", bookId);
int deleteCnt = cmd3.ExecuteNonQuery();
Console.WriteLine($"delete successed {updateCnt} books.");
TestQueryBook(conn, bookId);

conn.Close();
}
dc@dc-virtual-machine:~/dotnet/tidb-example-csharp$ dotnet run
insert successed 1 books.
id: 888888 title: TiDB in action type: Science & Technology published_at: 5/18/2022 3:22:02 PM stock: 1000 price: 66.66
update successed 1 books.
id: 888888 title: TiDB in action type: Science & Technology published_at: 5/18/2022 3:22:02 PM stock: 999 price: 66.66
delete successed 1 books.
book id 888888 not found.

更多 API 用法可以参考官方文档:

https://dev.mysql.com/doc/connector-net/en/connector-net-programming.html

https://dev.mysql.com/doc/connector-net/en/connector-net-tutorials.html`

注意:对于这种数据库 IO 类型请求,建议使用 API 的异步同名方法提高程序处理效率,例如ExecuteNonQueryAsync

这里梳理一下连接字符串中的核心参数。

参数名描述默认值
Server , Host , Data Source , DataSource数据库连接主机localhost
Port数据库连接端口3306
UserID , User Id , Username , Uid , User name , User数据库登录用户名null
Password , pwd 数据库登录用户密码
ConnectionTimeout , Connect Timeout , Connection Timeout数据量连接超时时间15s
DefaultCommandTimeout , Default Command TimeoutSQL 执行的超时时间30s
Pooling是否启用连接池true
ConnectionLifeTime , Connection Lifetime连接过期时间,超过这个时间的连接会被销毁重新创建0(不限制)
MaxPoolSize, Max Pool Size连接池最大连接数100
MinPoolSize, Min Pool Size连接池最小连接数0
TableCaching , Table Cache , TableCache是否开启客户端表数据缓存false
DefaultTableCacheAge , Default Table Cache Age客户端表数据缓存时间60s
AllowBatch , Allow Batch是否允许批量 SQL 执行true

关于连接池参数的最佳实践可以参考TiDB官网文档

使用 MySqlConnector

MySqlConnector 也是广泛使用的一种实现了 ADO.NET 接口的 MySQL 驱动,它提供了比MySql.Data更好的异步性能,很多 ORM 框架底层都是依赖于MySqlConnector 实现对 MySQL 的访问。

首先在项目中安装依赖包:

dotnet add package MySqlConnector --version 2.1.9

看一个数据库连接示例:

using MySqlConnector;

const string conectionStr = "Server=127.0.0.1;UserId=root;Password=;Port=4000;Database=bookshop";

public async Task TestConnection()
{
using (var conn = new MySqlConnection(conectionStr))
{
await conn.OpenAsync();
using (var cmd = new MySqlCommand("select tidb_version()", conn))
{
var result = await cmd.ExecuteScalarAsync();
Console.WriteLine(result);
}
}
}
dc@dc-virtual-machine:~/dotnet/tidb-example-csharp$ dotnet run
Release Version: v6.0.0
Edition: Community
Git Commit Hash: 36a9810441ca0e496cbd22064af274b3be771081
Git Branch: heads/refs/tags/v6.0.0
UTC Build Time: 2022-03-31 10:33:28
GoVersion: go1.18
Race Enabled: false
TiKV Min Version: v3.0.0-60965b006877ca7234adaced7890d7b029ed1306
Check Table Before Drop: false

可以看到使用方式和Connector/NET非常相似都是标准的 ADO.NET 风格,MySqlConnector的连接字符串参数绝大部分都兼容

Connector/NET,并且在此基础上提供了一些新的特性,比如负载均衡功能:

参数名说明默认值
Server, Host, Data Source, DataSource, Address, Addr, Network Address数据库请求地址,可以用逗号分隔填写多个地址实现负载均衡localhost
Load Balance, LoadBalance负载均衡策略,支持 RoundRobin、LeastConnections、Failover 三种模式,需要开启连接池RoundRobin

【注意】

有些参数在MySqlConnector已经禁用了,更多差异和新增功能参考官网文档:https://mysqlconnector.net/connection-options/

除了可以使用连接字符串,MySqlConnector还支持 Builder 对象模式,连接串里的参数都能在MySqlConnectionStringBuilder找到对应的字段,例如:

var builder = new MySqlConnectionStringBuilder
{
Server = "your-server",
UserID = "database-user",
Password = "P@ssw0rd!",
Database = "database-name",
};

// open a connection asynchronously
using var connection = new MySqlConnection(builder.ConnectionString);
await connection.OpenAsync();

一个简单的查询示例:

public async Task TestRead()
{
using (var conn = new MySqlConnection(conectionStr))
{
await conn.OpenAsync();

var cmd = conn.CreateCommand();
cmd.CommandText = "select * from users limit 5;";
MySqlDataReader reader =await cmd.ExecuteReaderAsync();
while (await reader.ReadAsync())
{
Console.WriteLine($"id: {reader.GetInt32("id")} balance: {reader.GetDecimal("balance")} nicknmame: {reader.GetString("nickname")} ");
}
}
}
dc@dc-virtual-machine:~/dotnet/tidb-example-csharp$ dotnet run
id: 525196 balance: 9490.89 nicknmame: Goodwin4601
id: 822804 balance: 1197.81 nicknmame: Treutel4269
id: 4147652 balance: 349.36 nicknmame: Pacocha6285
id: 9704562 balance: 2292.28 nicknmame: Grady8130
id: 17101775 balance: 5054.69 nicknmame: Macejkovic7559

一个简单的增删改示例:

public async Task TestWrite()
{
using (var conn = new MySqlConnection(conectionStr))
{
await conn.OpenAsync();

int userId = 888888;

var cmd1 = new MySqlCommand("insert into users values (@id,@balance,@nickname)", conn);
cmd1.Parameters.AddWithValue("@id", userId);
cmd1.Parameters.AddWithValue("@balance", 0.01);
cmd1.Parameters.AddWithValue("@nickname", "hey-hoho");
int insertCnt = await cmd1.ExecuteNonQueryAsync();
Console.WriteLine($"insert successed {insertCnt} users.");
TestQueryUser(conn, userId);

var cmd2 = new MySqlCommand("update users set balance=balance+99 where id=@id", conn);
cmd2.Parameters.AddWithValue("@id", userId);
int updateCnt =await cmd2.ExecuteNonQueryAsync();
Console.WriteLine($"update successed {updateCnt} users.");
TestQueryUser(conn, userId);

var cmd3 = new MySqlCommand("delete from users where id=@id", conn);
cmd3.Parameters.AddWithValue("@id", userId);
int deleteCnt = await cmd3.ExecuteNonQueryAsync();
Console.WriteLine($"delete successed {updateCnt} users.");
TestQueryUser(conn, userId);

}
}
dc@dc-virtual-machine:~/dotnet/tidb-example-csharp$ dotnet run
insert successed 1 users.
id: 888888 balance: 0.01 nicknmame: hey-hoho
update successed 1 users.
id: 888888 balance: 99.01 nicknmame: hey-hoho
delete successed 1 users.
user id 888888 not found.

一个简单的事务使用示例:

public async Task TestTransaction()
{
// 用一个事务演示购书流程
using (var conn = new MySqlConnection(conectionStr))
{
await conn.OpenAsync();

int userId = 525196, bookId = 648872;
decimal price = 15;

var tnx = await conn.BeginTransactionAsync();

try
{
// 新增一个订单
var cmd1 = new MySqlCommand("insert into orders values(@id, @book_id, @user_id, @quality, @ordered_at)", conn, tnx);
cmd1.Parameters.AddWithValue("@id", 999999);
cmd1.Parameters.AddWithValue("@book_id", bookId);
cmd1.Parameters.AddWithValue("@user_id", userId);
cmd1.Parameters.AddWithValue("@quality", 1);
cmd1.Parameters.AddWithValue("@ordered_at", DateTime.Now);
await cmd1.ExecuteNonQueryAsync();

// 扣减账户余额
var cmd2 = new MySqlCommand("update users set balance=balance-@price where id =@id", conn, tnx);
cmd2.Parameters.AddWithValue("@id", userId);
cmd2.Parameters.AddWithValue("@price", price);
await cmd2.ExecuteNonQueryAsync();

// 更新商品库存
var cmd3 = new MySqlCommand("update books set stock=stock-1 where id =@id", conn, tnx);
cmd3.Parameters.AddWithValue("@id", bookId);
await cmd3.ExecuteNonQueryAsync();

// 提交事务
await tnx.CommitAsync();
}
catch (Exception ex)
{
Console.WriteLine(ex.ToString());
// 异常回滚事务
await tnx.RollbackAsync();
}
}
}

需要注意的是,如果有大量重复的 SQL 需要执行,建议使用 TiDB 的 Prepare Statement 特性,它能有效减少 SQL 编译解析的时间提升执行效率,使用示例:

cmd.CommandText = "insert into books values (@id, @title, @type, @published_at, @stock, @price)";
cmd.Parameters.Add("@id", MySqlDbType.Int32);
cmd.Parameters.Add("@title", MySqlDbType.String);
...
cmd.Prepare();

for (int i = 1; i <= 1000; i++)
{
cmd.Parameters["@id"].Value = i;
cmd.Parameters["@title"].Value = $"TiDB in action - {i}";
...
await cmd.ExecuteNonQueryAsync();
}

更多用法可以参考官网文档:https://mysqlconnector.net/

使用 Entity Framework

Entity Framework (EF) 是 .NET 领域最知名的跨平台数据库访问 ORM 框架,它最早在 2008 年作为 .NET Framework 的一部分发布,现在的最新版本是 EF Core 6.0,也已经开源

它支持丰富的数据访问驱动,基于这个特性我们可以使用一套统一的 API 接口访问各种类型的数据库,比如 Sqlite、SQL Server、MySQL、PostgreSQL、Spanner 等等。在 MySQL 协议上,广泛使用的驱动有两个:

在使用之前先安装 Entity Framework 的基础包:

dotnet add package Microsoft.EntityFrameworkCore --version 6.0.5
dotnet add package Microsoft.Extensions.Logging.Console //记录日志用,非必须

Pomelo.EntityFrameworkCore.MySql

Pomelo.EntityFrameworkCore.MySql是最流行的兼容 MySQL 协议的 EF Core 驱动,也是微软官方推荐使用的方式,它底层依赖于前面提到的MySqlConnector,所以在连接串配置上并无差别。

dotnet add package Pomelo.EntityFrameworkCore.MySql --version 6.0.1

先构造一个DbContext类型和对应的实体类:

using Microsoft.EntityFrameworkCore;
using Pomelo.EntityFrameworkCore.MySql.Infrastructure;
using Pomelo.EntityFrameworkCore.MySql;
using System.ComponentModel.DataAnnotations.Schema;
using Microsoft.Extensions.Logging;

public class BookShopContext : DbContext
{
const string conectionStr = "Server=127.0.0.1;UserId=root;Password=;Port=4000;Database=bookshop";

public DbSet<User> Users { get; set; }
public DbSet<Book> Books { get; set; }

protected override void OnConfiguring(DbContextOptionsBuilder options)
{
options.UseMySql(conectionStr, ServerVersion.Create(new Version("5.7.25"), ServerType.MySql));

options.UseLoggerFactory(LoggerFactory.Create(builder =>
{
builder.AddConsole();
}));
}
}

public class User
{
[Column("id")]
public long Id { get; set; }
[Column("balance")]
public decimal Balance { get; set; }
[Column("nickname")]
public string Nickname { get; set; }
}

public class Book
{
[Column("id")]
public long Id { get; set; }
[Column("title")]
public string Title { get; set; }
[Column("type")]
public string Type { get; set; }
[Column("published_at")]
public DateTime PublishedAt { get; set; }
[Column("stock")]
public int Stock { get; set; }
[Column("price")]
public decimal Price { get; set; }
}

验证是否能连接上 TiDB :

public async Task TestConnection()
{
using var context = new BookShopContext();
Console.WriteLine($"TiDB CanConnect: {await context.Database.CanConnectAsync()}");
}
dc@dc-virtual-machine:~/dotnet/tidb-example-csharp$ dotnet run
TiDB CanConnect: True

一个单表分页查询结果集的示例:

public async Task TestRead()
{
using var context = new BookShopContext();

var books = await context.Books.Where(b => b.Title.Contains("db")).OrderBy(b => b.Id).Skip(2).Take(5).ToListAsync();
foreach (var book in books)
{
Console.WriteLine($"id: {book.Id} title: {book.Title} type: {book.Type} published_at: {book.PublishedAt}");
}
}
dc@dc-virtual-machine:~/dotnet/tidb-example-csharp$ dotnet run
info: Microsoft.EntityFrameworkCore.Infrastructure[10403]
Entity Framework Core 6.0.5 initialized 'BookShopContext' using provider 'Pomelo.EntityFrameworkCore.MySql:6.0.1' with options: ServerVersion 5.7.25-mysql
info: Microsoft.EntityFrameworkCore.Database.Command[20101]
Executed DbCommand (47ms) [Parameters=[@__p_1='?' (DbType = Int32), @__p_0='?' (DbType = Int32)], CommandType='Text', CommandTimeout='30']
SELECT `b`.`id`, `b`.`price`, `b`.`published_at`, `b`.`stock`, `b`.`title`, `b`.`type`
FROM `Books` AS `b`
WHERE `b`.`title` LIKE '%db%'
ORDER BY `b`.`id`
LIMIT @__p_1 OFFSET @__p_0
id: 474329852 title: Geovany Padberg type: Science & Technology published_at: 2010/1/28 6:45:33
id: 1890134379 title: The Adventures of Hilda Padberg type: Comics published_at: 1943/7/17 16:28:02
id: 2181887016 title: Catalina Padberg type: Education & Reference published_at: 1961/3/28 18:18:37
id: 2193223665 title: Caroline Padberg type: Kids published_at: 1994/11/12 20:56:53
id: 2359817065 title: Darryl Padberg type: Science & Technology published_at: 1940/11/13 9:53:21

一个简单的增删改示例:

public async Task TestWrite()
{
using var context = new BookShopContext();

long userId = 888888;

var user = new User
{
Id = userId,
Balance = 0.01M,
Nickname = "hey-hoho"
};
context.Add(user);
int insertCnt = await context.SaveChangesAsync();
Console.WriteLine($"insert successed {insertCnt} users.");
TestQueryUser(context, userId);

user.Balance += 99;
int updateCnt = await context.SaveChangesAsync();
Console.WriteLine($"update successed {updateCnt} users.");
TestQueryUser(context, userId);

context.Remove(user);
int deleteCnt = await context.SaveChangesAsync();
Console.WriteLine($"delete successed {deleteCnt} users.");
TestQueryUser(context, userId);
}
dc@dc-virtual-machine:~/dotnet/tidb-example-csharp$ dotnet run
insert successed 1 users.
id: 888888 balance: 0.01 nicknmame: hey-hoho
update successed 1 users.
id: 888888 balance: 99.01 nicknmame: hey-hoho
delete successed 1 users.
user id 888888 not found.

【注意】

当只需要查询结果而不用对结果进行修改时,建议关闭 EF Core 的tracking功能,它能提高执行速度。比如:

// 单个查询关闭跟踪
var books = await context.Books.OrderBy(b => b.Id).Take(3).AsNoTracking().ToListAsync();
// 整个上下文关闭跟踪
options.UseQueryTrackingBehavior(QueryTrackingBehavior.NoTracking);

Pomelo.EntityFrameworkCore.MySql的文档地址:

https://github.com/PomeloFoundation/Pomelo.EntityFrameworkCore.MySql/wiki

MySql.EntityFrameworkCore

MySql.EntityFrameworkCore是 Oracle 官方发布的Entity Framework支持 MySQL 的数据库驱动,所以它的底层是依赖于Connector/NET

如果你想使用Connector/NET 8.0.28 及以后的版本,那么安装这个包:

dotnet add package MySql.EntityFrameworkCore --version 6.0.1 //版本对应.NET的版本

如果你使用的是Connector/NET早期版本,那么安装这个包:

dotnet add package MySql.Data.EntityFrameworkCore --version 8.0.22

在使用方式上只有配置连接串上的一点点差别:

protected override void OnConfiguring(DbContextOptionsBuilder options)
{
options.UseMySQL("server=localhost;database=library;user=user;password=password");
}

其他数据库操作都是用标准的 EF API。

MySql.EntityFrameworkCore的文档地址:

https://dev.mysql.com/doc/connector-net/en/connector-net-entity-framework.html

总体来说,使用Entity Framework操作 TiDB 并没有什么特殊的地方,有使用经验的开发者几乎不用任何学习成本就能快速上手。

Entity Framework更多用法可以参考官网文档:https://docs.microsoft.com/en-us/ef/core/

使用Dapper

Dapper是 StackExchange 开源的一款轻量级 ORM 框架,它以高性能著称,在 .NET 领域使用非常广泛。它扩展了 ADO.NET 的IDbConnection接口,底层同样依赖于MySqlConnector或者Connector/NET,使用方式介于原生 ADO.NET 和 Entity Framework 之间。

dotnet add package Dapper.StrongName --version 2.0.123

测试数据库连接:

using Dapper;
using MySqlConnector;

const string conectionStr = "Server=127.0.0.1;UserId=root;Password=;Port=4000;Database=bookshop";

public async Task TestConnection()
{
MySqlConnection conn = new MySqlConnection(conectionStr);
var version =await conn.ExecuteScalarAsync<string>("select tidb_version()");
Console.WriteLine(version);
}
dc@dc-virtual-machine:~/dotnet/tidb-example-csharp$ dotnet run
Release Version: v6.0.0
Edition: Community
Git Commit Hash: 36a9810441ca0e496cbd22064af274b3be771081
Git Branch: heads/refs/tags/v6.0.0
UTC Build Time: 2022-03-31 10:33:28
GoVersion: go1.18
Race Enabled: false
TiKV Min Version: v3.0.0-60965b006877ca7234adaced7890d7b029ed1306
Check Table Before Drop: false

使用强类型的查询示例:

public class Book
{
public long Id { get; set; }
public string Title { get; set; }
public string Type { get; set; }
public DateTime Published_At { get; set; }
public int Stock { get; set; }
public decimal Price { get; set; }
}

public async Task TestQuery()
{
MySqlConnection conn = new MySqlConnection(conectionStr);

var book = await conn.QueryFirstAsync<Book>("select * from books where id = @id", new { id = 8683541 });
Console.WriteLine($"id: {book.Id} title: {book.Title} type: {book.Type} published_at: {book.Published_At} stock: {book.Stock} price: {book.Price}");

var books = await conn.QueryAsync<Book>("select * from books where type = @type and stock < @stock order by published_at limit 3", new { type = "Sports" stock = 100 });
books.ToList().ForEach(b =>
Console.WriteLine($"id: {b.Id} title: {b.Title} type: {b.Type} published_at: {b.Published_At} stock: {b.Stock} price: {b.Price}")
);
}
dc@dc-virtual-machine:~/dotnet/tidb-example-csharp$ dotnet run
id: 8683541 title: The Documentary of hamster type: Magazine published_at: 1945/10/3 1:44:52 stock: 190 price: 74.38

id: 1201887882 title: Nayeli Luettgen type: Sports published_at: 1901/11/20 19:37:37 stock: 53 price: 141.50
id: 3705800883 title: Arianna Considine type: Sports published_at: 1905/8/31 5:25:53 stock: 72 price: 312.93
id: 1811359739 title: Dawson Hackett type: Sports published_at: 1910/12/1 21:22:47 stock: 72 price: 415.72

带事务的增删改示例:

public class Order
{
public long Id { get; set; }
public long Book_Id { get; set; }
public long User_Id { get; set; }
public int Quality { get; set; }
public DateTime Ordered_At { get; set; }
}
// 用一个事务演示购书流程
public async Task TestTransaction()
{
MySqlConnection conn = new MySqlConnection(conectionStr);

int userId = 525196, bookId = 648872;
decimal price = 15;

await conn.OpenAsync();
var tnx = await conn.BeginTransactionAsync();

try
{
// 新增一个订单
var order = new Order
{
Id = 666666,
Book_Id = bookId,
User_Id = userId,
Quality = 1,
Ordered_At = DateTime.Now
};
await conn.ExecuteAsync("insert into orders values(@id, @book_id, @user_id, @quality, @ordered_at)", order, tnx);

// 更新账户余额
await conn.ExecuteAsync("update users set balance=balance-@price where id =@id", new { id = userId, price = price }, tnx);

// 更新商品库存
await conn.ExecuteAsync("update books set stock=stock-1 where id =@id", new { id = bookId }, tnx);

// 提交事务
await tnx.CommitAsync();
}
catch (Exception ex)
{
Console.WriteLine(ex.ToString());
// 异常回滚事务
await tnx.RollbackAsync();
}
}

Dapper一般情况下会自动管理连接的开启关闭状态,使用起来比原生ADO.NET更加方便。但是要注意的是,在使用显式事务的时候要提前手动打开连接,否则会报异常System.InvalidOperationException: Connection is not open.

Dapper会缓存每一次查询语句,因此推荐的做法是使用参数化方式进行传参,一方面能提高 SQL 执行效率,另一方面可以减少内存占用。

更多用法可以参考官方文档:https://github.com/DapperLib/Dapper

最佳实践

原生 ADO.NET 能够带来非常优秀的性能,但是缺点就是需要大量的手写 SQL 和类型转换,对于应用开发不是很友好。而 ORM 框架虽然解决了代码复杂度的问题,但也带来了新的问题,就是无法精准控制 SQL 的行为不够灵活,以及大幅性能损耗。

所以在实际项目中,推荐使用Entity Framework+Dapper的组合,底层数据驱动选择MySqlConnector,简单读写场景交给Entity Framework去处理,想高度控制 SQL 的场景交给Dapper去处理,兼顾了性能和开发效率两方面。Entity Framework的上下文对象提供了一个访问原始 ADO.NET Connection 的入口,这就使得把两者结合起来非常方便。

需要导入的程序集:

  • Microsoft.EntityFrameworkCore
  • Pomelo.EntityFrameworkCore.MySql
  • Dapper

使用示例:

using Microsoft.EntityFrameworkCore;
using Microsoft.EntityFrameworkCore.Storage;
using Pomelo.EntityFrameworkCore.MySql.Infrastructure;
using Dapper;

public async Task TestWithDapper()
{
using var context = new BookShopContext();
var conn = context.Database.GetDbConnection();

// 使用 dapper 查询
string version = conn.ExecuteScalar<string>("select tidb_version()");
Console.WriteLine(version);

// 使用 EF 查询
var book = context.Books.Where(u => u.Id == 1).FirstOrDefault();

// 共享事务,类似的还可以使用 TransactionScope 或 EF 的 UseTransaction
using IDbContextTransaction tnx = context.Database.BeginTransaction();
conn.ExecuteScalar<dynamic>("select /*+ read_from_storage(tiflash[ratings]) */ book_id,count(*),avg(score) from ratings group by book_id", transaction: tnx.GetDbTransaction());
context.Books.Add(new Book { });
await tnx.CommitAsync();
}

【注意】

using会自动处理IDbContextTransaction对象的rollbackdispose,如果没有使用using语法,需要手动处理事务异常情况。

还有其他一些 ORM 框架例如 NHibernateFreeSql 等使用方式大同小异,底层都是依赖前面提到的数据库驱动,参考各自 API 文档即可。

总结

以上演示了 C# 在多种数据库访问方式下实现对 TiDB 的简单 CRUD 功能,整体来说和操作 MySQL 区别不大,但是在这个过程中我们要善于利用 TiDB 的特性提升应用程序的处理能力。

关于更多的复杂场景,如何在 TiDB 上做 SQL 开发和性能调优,推荐阅读以下文档: