将 IDataReader 映射到 C# 对象的最有效方法,也适用于深度复制?

问题描述 投票:0回答:2

我们有一个场景,我们使用企业图书馆数据包来满足我们的数据访问需求。我们已经实现了一个Generic Data Reader来将我们的数据融入业务对象。然而,我们正在寻找一种更有效的方法来做到这一点,因为我们正在处理大量数据,这也将在不使用反射的情况下满足对象内子集合的深度复制。通用数据读取器在处理大量数据时也表现不佳。我知道如果我们使用 EF,它可以解决这个问题,但我们还没有达到可以替换整个数据访问层的阶段。是否有替代方案或行业标准来实现这一目标?

c# performance entity-framework data-access-layer idatareader
2个回答
5
投票

你为什么不手工编写映射器的代码?这是最快的方法,而且您在项目生命周期内只创建一次映射器,对吗?在活了五年十年的项目中,与你得到的相比,它所花费的时间可以忽略不计(容易纠正错误,不需要解释 OR/M 错误或它们的“简单性”)。

public interface IEntityMapper
{
   object Map(IDataRecord record);
}

public void UserMapper  : IEntityMapper
{
    public void Map(IDataRecord record)
    {
        var user = new User();
        user.FirstName = record["firstName"]
    }
}

如果您有大量记录,则需要使用

number
索引器来防止名称查找:

public void UserMapper  : IEntityMapper
{
    public void Map(IDataRecord record)
    {
        var user = new User();

        //requires that you specify columns in your SELECT query
        //to not break the mapper in future versions.
        user.FirstName = record[0]
    }
}

在您的存储库中,您可以这样做:

public class UserRepository
{
    private readonly IDbConnection _connection;
    private UserMapper _mapper = new UserMapper();

    public IReadOnlyList<User> GetAllUsers()
    {
        using (var cmd = _connection.CreateCommand())
        {
            cmd.CommandText = "SELECT Id, UserName FROM Users";
            using (var reader = cmd.ExecuteReader())
            {
                var users = new List<User>();
                while (reader.Read())
                {
                   var user = _mapper.Map(reader);
                   users.Add(user);
                }
                return users;
            }
        }
    }
}

您甚至可以将所有内容移动到扩展方法:

public static class CommandExtensions
{
    public static IReadOnlyCollection<T> ToList<T>(this IDbCommand cmd, IDataMapper mapper)
    {
        using (var reader = cmd.ExecuteReader())
        {
            var items = new List<T>();
            while (reader.Read())
            {
               var item= _mapper.Map(reader);
               items.Add(item);
            }
            return item;
        }
    }
}

为您提供更轻量级的存储库:

public class UserRepository
{
    private readonly IDbConnection _connection;
    private UserMapper _mapper = new UserMapper();

    public IReadOnlyList<User> GetAllUsers()
    {
        using (var cmd = _connection.CreateCommand())
        {
            cmd.CommandText = "SELECT Id, UserName FROM Users";
            return cmd.ToList<User>(_mapper);
        }
    }
}

深拷贝

关于深拷贝,我尽量避免使用复杂的实体(除非子实体实际上是值对象)。最好使用 ID 引用其他实体,否则很难在代码中的一个地方保持对实体的责任。

说到映射,我可能会引入一个约定,即使用映射接口中的替代方法映射子对象:

public interface IEntityMapper
{
   object Map(IDataRecord record);
   object Map(IDataRecord record, string prefix);
}

前缀可以用作约定,表示使用连接从多个表中获取信息。

public void UserMapper  : IEntityMapper
{
    private AddressMapper _childMapper = new AddressMapper();

    public void Map(IDataRecord record)
    {
        var user = new User();
        user.FirstName = record["FirstName"]
        user.Address = _childMapper.Map(record, "Address_");
    }
}

.. 这将使您的存储库方法看起来像这样:

public class UserRepository
{
    private readonly IDbConnection _connection;
    private UserMapper _mapper = new UserMapper();

    public IReadOnlyList<User> GetAllUsers()
    {
        using (var cmd = _connection.CreateCommand())
        {
            cmd.CommandText = @"SELECT Id, UserName, Address.City as Address_City 
                                FROM Users 
                                JOIN Address ON (Address.Id = Users.AddressId)";
            return cmd.ToList<User>(_mapper);
        }
    }
}

我写过关于 ADO.NET 和(我的)最佳实践的文章:http://blog.gauffin.org/2013/01/04/ado-net-the-right-way/


0
投票

使用

System.Linq.Expressions.Expression
,可以动态构建接近硬编码效率的动态代码

动手做一个简单的ADO内核:

// Code by Flithor(Mr. Squirrel.Downy)
// License: MIT
// == Warning ==
// Does not work in every possible situation
// Use at your own risk

/// <summary>
/// Getter <see cref="MethodInfo"/> of each common type of <see cref="IDataRecord"/>
/// </summary>
private static readonly Dictionary<Type, MethodInfo> IDataRecordGetterMethods = typeof(IDataRecord).GetMethods()
    .Where(m => m.GetParameters().Length == 1 && m.Name == "Get" + m.ReturnType.Name)
    .ToDictionary(m => m.ReturnType);
/// <summary>
/// The <see cref="MethodInfo"/> of<see cref="IDataRecord.IsDBNull"/> of <see cref="IDataRecord"/>
/// </summary>
private static readonly MethodInfo isDbNullMethod = typeof(IDataRecord).GetMethod(nameof(IDataRecord.IsDBNull));

/// <summary>
/// Convert <see cref="IDataReader "/> to an entity sequence
/// </summary>
public static IEnumerable<T> CastTo<T>(this IDataReader reader) where T : new()
{
    Func<IDataRecord, T> instanceBuilder = null;
    while (reader.Read())
    {
        if (instanceBuilder == null)
        {
            var sw = Stopwatch.StartNew();
            // get orderd fields from first line
            var fields = Enumerable.Range(0, reader.FieldCount).Select(reader.GetName).ToArray();
            instanceBuilder = BuildAssginer<T>(fields);
            // build lambda just spent about 10 ms
            Console.WriteLine($"Build lambda cost:{sw.Elapsed}");
        }

        yield return instanceBuilder(reader);
    }
}

/// <summary>
/// Build Entity builder lambda
/// </summary>
/// <param name="fields">Orderd fields of result datas</param>
private static Func<IDataRecord, TResult> BuildAssginer<TResult>(string[] fields) where TResult : new()
{
    var createdType = typeof(TResult);

    // new Tresult()
    var ctor = Expression.New(createdType);
    var properties = typeof(TResult).GetProperties().ToDictionary(pinfo => pinfo.Name.ToLower());

    //Map indexs of fields on PropertyInfo, for efficiency
    var indexToPropMap = fields.Select((field, index) => new { field, index })
        .GroupJoin(properties, o => o.field.ToLower(), prop => prop.Key.ToLower(),
            (o, props) => new { o.index, prop = props.FirstOrDefault().Value })
        .Where(o => o.prop != null)
        .OrderBy(o => o.index)
        .ToDictionary(o => o.index, o => o.prop);
    //define lambda input parameter of IDataRecord type
    var recordParamExp = Expression.Parameter(typeof(IDataRecord));
    //get property 
    var propAssginExps = indexToPropMap.Select(kv =>
    {
        // get field type and the underlying type of nullable type(maybe)
        var propType = kv.Value.PropertyType;
        var underType = Nullable.GetUnderlyingType(propType);
        // build get value expression for most most appropriate case
        var getValuesExp = underType != null
            ? Expression.Convert(
                Expression.Call(recordParamExp, IDataRecordGetterMethods[underType],
                    Expression.Constant(kv.Key)), propType)
            : (Expression)Expression.Call(recordParamExp, IDataRecordGetterMethods[propType],
                Expression.Constant(kv.Key));
        // is not value type or is nullable
        var propCanNull = !propType.IsValueType || underType != null;
        var getFieldValueExp = propCanNull
            // may null field check is dbnull or not
            ? Expression.Condition(
                Expression.Call(recordParamExp, isDbNullMethod, Expression.Constant(kv.Key)),
                Expression.Default(propType),
                getValuesExp)
            // return not nullable value directly
            : getValuesExp;
        // set field
        return Expression.Bind(kv.Value, getFieldValueExp);
    });
    // entity initializer
    // sample:
    // new TResult
    // {
    //     No = record.GetInt32(0),
    //     Name = record.GetString(1),
    //     Num = record.IsNull(2) ? default(Int64) : record.GetInt64(2)
    // }
    var memberInit = Expression.MemberInit(ctor, propAssginExps);
    return Expression.Lambda<Func<IDataRecord, TResult>>(memberInit, recordParamExp).Compile();
}

知道怎么用

System.Linq.Expressions.Expression
,就可以自己写深拷贝了

© www.soinside.com 2019 - 2024. All rights reserved.