代码改变世界

解读ASP.NET 5 & MVC6系列(8):Session与Caching

2015-05-21 09:08  汤姆大叔  阅读(25176)  评论(19编辑  收藏  举报

在之前的版本中,Session存在于System.Web中,新版ASP.NET 5中由于不在依赖于System.Web.dll库了,所以相应的,Session也就成了ASP.NET 5中一个可配置的模块(middleware)了。

配置启用Session

ASP.NET 5中的Session模块存在于Microsoft.AspNet.Session类库中,要启用Session,首先需要在project.json中的dependencies节点中添加如下内容:

"Microsoft.AspNet.Session": "1.0.0-beta3"

然后在ConfigureServices中添加Session的引用(并进行配置):

services.AddCaching();  // 这两个必须同时添加,因为Session依赖于Caching
services.AddSession();
//services.ConfigureSession(null); 可以在这里配置,也可以再后面进行配置

最后在Configure方法中,开启使用Session的模式,如果在上面已经配置过了,则可以不再传入配置信息,否则还是要像上面的配置信息一样,传入Session的配置信息,代码如下:

app.UseInMemorySession(configure:s => { s.IdleTimeout = TimeSpan.FromMinutes(30); });
//app.UseSession(o => { o.IdleTimeout = TimeSpan.FromSeconds(30); });
//app.UseInMemorySession(null, null);   //开启内存Session
//app.UseDistributedSession(null, null);//开启分布式Session,也即持久化Session
//app.UseDistributedSession(new RedisCache(new RedisCacheOptions() { Configuration = "localhost" }));

对于UseInMemorySession方法,接收2个可选参数,分别是:IMemoryCache可用于修改Session数据的默认保存地址;Action<SessionOptions>委托则可以让你修改默认选项,比如Session cookie的路径、默认的过期时间等。本例中,我们修改默认过期时间为30分钟。

注意:该方法必须在app.UseMvc之前调用,否则在Mvc里获取不到Session,而且会出错。

获取和设置Session

获取和设置Session对象,一般是在Controller的action里通过this.Context.Session来获取的,其获取的是一个基于接口ISessionCollection的实例。该接口可以通过索引、Set、TryGetValue等方法进行Session值的获取和设置,但我们发现在获取和设置Session的时候,我们只能使用byte[]类型,而不能像之前版本的Session一样可以设置任意类型的数据。原因是因为,新版本的Session要支持在远程服务器上存储,就需要支持序列化,所以才强制要求保存为byte[]类型。所以我们在保存Session的时候,需要将其转换为byte[]才能进行保存,并且获取以后要再次将byte[]转换为自己的原有的类型才行。这种形式太麻烦了,好在微软在Microsoft.AspNet.Http命名空间(所属Microsoft.AspNet.Http.Extensions.dll中)下,为我们添加了几个扩展方法,分别用于设置和保存byte[]类型、int类型、以及string类型,代码如下:

public static byte[] Get(this ISessionCollection session, string key);
public static int? GetInt(this ISessionCollection session, string key);
public static string GetString(this ISessionCollection session, string key);
public static void Set(this ISessionCollection session, string key, byte[] value);
public static void SetInt(this ISessionCollection session, string key, int value);
public static void SetString(this ISessionCollection session, string key, string value);

所以,在Controller里引用Microsoft.AspNet.Http命名空间以后,我们就可以通过如下代码进行Session的设置和获取了:

Context.Session.SetString("Name", "Mike");
Context.Session.SetInt("Age", 21);

ViewBag.Name = Context.Session.GetString("Name");
ViewBag.Age = Context.Session.GetInt("Age");

自定义类型的Session设置和获取

前面我们说了,要保存自定义类型的Session,需要将其类型转换成byte[]数组才行,在本例中,我们对bool类型的Session数据进行设置和获取的代码,示例如下:

public static class SessionExtensions
{
    public static bool? GetBoolean(this ISessionCollection session, string key)
    {
        var data = session.Get(key);
        if (data == null)
        {
            return null;
        }
        return BitConverter.ToBoolean(data, 0);
    } 

    public static void SetBoolean(this ISessionCollection session, string key, bool value)
    {
        session.Set(key, BitConverter.GetBytes(value));
    }
}

定义bool类型的扩展方法以后,我们就可以像SetInt/GetInt那样进行使用了,示例如下:

Context.Session.SetBoolean("Liar", true);
ViewBag.Liar = Context.Session.GetBoolean("Liar");

另外,ISessionCollection接口上还提供了Remove(string key)和Clear()两个方法分别用于删除某个Session值和清空所有的Session值的功能。但同时也需要注意,该接口并没提供之前版本中的Abandon方法功能。

基于Redis的Session管理

使用分布式Session,其主要工作就是将Session保存的地方从原来的内存换到分布式存储上,本节,我们以Redis存储为例来讲解分布式Session的处理。

先查看使用分布式Session的扩展方法,示例如下,我们可以看到,其Session容器需要是一个支持IDistributedCache的接口示例。

public static IApplicationBuilder UseDistributedSession([NotNullAttribute]this IApplicationBuilder app, IDistributedCache cache, Action<SessionOptions> configure = null);

该接口是缓存Caching的通用接口,也就是说,只要我们实现了缓存接口,就可以将其用于Session的管理。进一步查看该接口发现,该接口中定义的Set方法还需要实现一个ICacheContext类型的缓存上下文(以便在调用的时候让其它程序进行委托调用),接口定义分别如下:

public interface IDistributedCache
{
    void Connect();
    void Refresh(string key);
    void Remove(string key);
    Stream Set(string key, object state, Action<ICacheContext> create);
    bool TryGetValue(string key, out Stream value);
}

public interface ICacheContext
{
    Stream Data { get; }
    string Key { get; }
    object State { get; }

    void SetAbsoluteExpiration(TimeSpan relative);
    void SetAbsoluteExpiration(DateTimeOffset absolute);
    void SetSlidingExpiration(TimeSpan offset);
}

接下来,我们基于Redis来实现上述功能,创建RedisCache类,并继承IDistributedCache,引用StackExchange.Redis程序集,然后实现IDistributedCache接口的所有方法和属性,代码如下:

using Microsoft.Framework.Cache.Distributed;
using Microsoft.Framework.OptionsModel;
using StackExchange.Redis;
using System;
using System.IO;

namespace Microsoft.Framework.Caching.Redis
{
    public class RedisCache : IDistributedCache
    {
        // KEYS[1] = = key
        // ARGV[1] = absolute-expiration - ticks as long (-1 for none)
        // ARGV[2] = sliding-expiration - ticks as long (-1 for none)
        // ARGV[3] = relative-expiration (long, in seconds, -1 for none) - Min(absolute-expiration - Now, sliding-expiration)
        // ARGV[4] = data - byte[]
        // this order should not change LUA script depends on it
        private const string SetScript = (@"
                redis.call('HMSET', KEYS[1], 'absexp', ARGV[1], 'sldexp', ARGV[2], 'data', ARGV[4])
                if ARGV[3] ~= '-1' then
                  redis.call('EXPIRE', KEYS[1], ARGV[3]) 
                end
                return 1");
        private const string AbsoluteExpirationKey = "absexp";
        private const string SlidingExpirationKey = "sldexp";
        private const string DataKey = "data";
        private const long NotPresent = -1;

        private ConnectionMultiplexer _connection;
        private IDatabase _cache;

        private readonly RedisCacheOptions _options;
        private readonly string _instance;

        public RedisCache(IOptions<RedisCacheOptions> optionsAccessor)
        {
            _options = optionsAccessor.Options;
            // This allows partitioning a single backend cache for use with multiple apps/services.
            _instance = _options.InstanceName ?? string.Empty;
        }

        public void Connect()
        {
            if (_connection == null)
            {
                _connection = ConnectionMultiplexer.Connect(_options.Configuration);
                _cache = _connection.GetDatabase();
            }
        }

        public Stream Set(string key, object state, Action<ICacheContext> create)
        {
            Connect();

            var context = new CacheContext(key) { State = state };
            create(context);
            var value = context.GetBytes();
            var result = _cache.ScriptEvaluate(SetScript, new RedisKey[] { _instance + key },
                new RedisValue[]
                {
                    context.AbsoluteExpiration?.Ticks ?? NotPresent,
                    context.SlidingExpiration?.Ticks ?? NotPresent,
                    context.GetExpirationInSeconds() ?? NotPresent,
                    value
                });
            // TODO: Error handling
            return new MemoryStream(value, writable: false);
        }

        public bool TryGetValue(string key, out Stream value)
        {
            value = GetAndRefresh(key, getData: true);
            return value != null;
        }

        public void Refresh(string key)
        {
            var ignored = GetAndRefresh(key, getData: false);
        }

        private Stream GetAndRefresh(string key, bool getData)
        {
            Connect();

            // This also resets the LRU status as desired.
            // TODO: Can this be done in one operation on the server side? Probably, the trick would just be the DateTimeOffset math.
            RedisValue[] results;
            if (getData)
            {
                results = _cache.HashMemberGet(_instance + key, AbsoluteExpirationKey, SlidingExpirationKey, DataKey);
            }
            else
            {
                results = _cache.HashMemberGet(_instance + key, AbsoluteExpirationKey, SlidingExpirationKey);
            }
            // TODO: Error handling
            if (results.Length >= 2)
            {
                // Note we always get back two results, even if they are all null.
                // These operations will no-op in the null scenario.
                DateTimeOffset? absExpr;
                TimeSpan? sldExpr;
                MapMetadata(results, out absExpr, out sldExpr);
                Refresh(key, absExpr, sldExpr);
            }
            if (results.Length >= 3 && results[2].HasValue)
            {
                return new MemoryStream(results[2], writable: false);
            }
            return null;
        }

        private void MapMetadata(RedisValue[] results, out DateTimeOffset? absoluteExpiration, out TimeSpan? slidingExpiration)
        {
            absoluteExpiration = null;
            slidingExpiration = null;
            var absoluteExpirationTicks = (long?)results[0];
            if (absoluteExpirationTicks.HasValue && absoluteExpirationTicks.Value != NotPresent)
            {
                absoluteExpiration = new DateTimeOffset(absoluteExpirationTicks.Value, TimeSpan.Zero);
            }
            var slidingExpirationTicks = (long?)results[1];
            if (slidingExpirationTicks.HasValue && slidingExpirationTicks.Value != NotPresent)
            {
                slidingExpiration = new TimeSpan(slidingExpirationTicks.Value);
            }
        }

        private void Refresh(string key, DateTimeOffset? absExpr, TimeSpan? sldExpr)
        {
            // Note Refresh has no effect if there is just an absolute expiration (or neither).
            TimeSpan? expr = null;
            if (sldExpr.HasValue)
            {
                if (absExpr.HasValue)
                {
                    var relExpr = absExpr.Value - DateTimeOffset.Now;
                    expr = relExpr <= sldExpr.Value ? relExpr : sldExpr;
                }
                else
                {
                    expr = sldExpr;
                }
                _cache.KeyExpire(_instance + key, expr);
                // TODO: Error handling
            }
        }

        public void Remove(string key)
        {
            Connect();

            _cache.KeyDelete(_instance + key);
            // TODO: Error handling
        }
    }
}

在上述代码中,我们使用了自定义类RedisCacheOptions作为Redis的配置信息类,为了实现基于POCO的配置定义,我们还继承了IOptions接口,该类的定义如下:

public class RedisCacheOptions : IOptions<RedisCacheOptions>
{
    public string Configuration { get; set; }

    public string InstanceName { get; set; }

    RedisCacheOptions IOptions<RedisCacheOptions>.Options
    {
        get { return this; }
    }

    RedisCacheOptions IOptions<RedisCacheOptions>.GetNamedOptions(string name)
    {
        return this;
    }
}

第三部,定义委托调用时使用的缓存上下文类CacheContext,具体代码如下:

using Microsoft.Framework.Cache.Distributed;
using System;
using System.IO;

namespace Microsoft.Framework.Caching.Redis
{
    internal class CacheContext : ICacheContext
    {
        private readonly MemoryStream _data = new MemoryStream();

        internal CacheContext(string key)
        {
            Key = key;
            CreationTime = DateTimeOffset.UtcNow;
        }

        /// <summary>
        /// The key identifying this entry.
        /// </summary>
        public string Key { get; internal set; }

        /// <summary>
        /// The state passed into Set. This can be used to avoid closures.
        /// </summary>
        public object State { get; internal set; }

        public Stream Data { get { return _data; } }

        internal DateTimeOffset CreationTime { get; set; } // 可以让委托设置创建时间

        internal DateTimeOffset? AbsoluteExpiration { get; private set; }

        internal TimeSpan? SlidingExpiration { get; private set; }

        public void SetAbsoluteExpiration(TimeSpan relative) // 可以让委托设置相对过期时间
        {
            if (relative <= TimeSpan.Zero)
            {
                throw new ArgumentOutOfRangeException("relative", relative, "The relative expiration value must be positive.");
            }
            AbsoluteExpiration = CreationTime + relative;
        }

        public void SetAbsoluteExpiration(DateTimeOffset absolute) // 可以让委托设置绝对过期时间
        {
            if (absolute <= CreationTime)
            {
                throw new ArgumentOutOfRangeException("absolute", absolute, "The absolute expiration value must be in the future.");
            }
            AbsoluteExpiration = absolute.ToUniversalTime();
        }

        public void SetSlidingExpiration(TimeSpan offset) // 可以让委托设置offset过期时间
        {
            if (offset <= TimeSpan.Zero)
            {
                throw new ArgumentOutOfRangeException("offset", offset, "The sliding expiration value must be positive.");
            }
            SlidingExpiration = offset;
        }

        internal long? GetExpirationInSeconds()
        {
            if (AbsoluteExpiration.HasValue && SlidingExpiration.HasValue)
            {
                return (long)Math.Min((AbsoluteExpiration.Value - CreationTime).TotalSeconds, SlidingExpiration.Value.TotalSeconds);
            }
            else if (AbsoluteExpiration.HasValue)
            {
                return (long)(AbsoluteExpiration.Value - CreationTime).TotalSeconds;
            }
            else if (SlidingExpiration.HasValue)
            {
                return (long)SlidingExpiration.Value.TotalSeconds;
            }
            return null;
        }

        internal byte[] GetBytes()
        {
            return _data.ToArray();
        }
    }
}

最后一步定义,RedisCache中需要的根据key键获取缓存值的快捷方法,代码如下:

using StackExchange.Redis;
using System;

namespace Microsoft.Framework.Caching.Redis
{
    internal static class RedisExtensions
    {
        private const string HmGetScript = (@"return redis.call('HMGET', KEYS[1], unpack(ARGV))");

        internal static RedisValue[] HashMemberGet(this IDatabase cache, string key, params string[] members)
        {
            var redisMembers = new RedisValue[members.Length];
            for (int i = 0; i < members.Length; i++)
            {
                redisMembers[i] = (RedisValue)members[i];
            }
            var result = cache.ScriptEvaluate(HmGetScript, new RedisKey[] { key }, redisMembers);
            // TODO: Error checking?
            return (RedisValue[])result;
        }
    }
}

至此,所有的工作就完成了,将该缓存实现注册为Session的provider的代码方法如下:

app.UseDistributedSession(new RedisCache(new RedisCacheOptions()
{
    Configuration = "此处填写 redis的地址",
    InstanceName = "此处填写自定义实例名"
}), options =>
{
    options.CookieHttpOnly = true;
});

参考:http://www.mikesdotnetting.com/article/270/sessions-in-asp-net-5

关于Caching

默认情况下,本地缓存使用的是IMemoryCache接口的示例,可以通过获取该接口的示例来对本地缓存进行操作,示例代码如下:

var cache = app.ApplicationServices.GetRequiredService<IMemoryCache>();
var obj1 = cache.Get("key1");
bool obj2 = cache.Get<bool>("key2");

对于,分布式缓存,由于AddCaching,默认将IMemoryCache实例作为分布式缓存的provider了,代码如下:

public static class CachingServicesExtensions
{
    public static IServiceCollection AddCaching(this IServiceCollection collection)
    {
        collection.AddOptions();
        return collection.AddTransient<IDistributedCache, LocalCache>()
            .AddSingleton<IMemoryCache, MemoryCache>();
    }
}

所以,要使用新的分布式Caching实现,我们需要注册自己的实现,代码如下:

services.AddTransient<IDistributedCache, RedisCache>();
services.Configure<RedisCacheOptions>(opt =>
{
    opt.Configuration = "此处填写 redis的地址";
    opt.InstanceName = "此处填写自定义实例名";
});

基本的使用方法如下:

var cache = app.ApplicationServices.GetRequiredService<IDistributedCache>();
cache.Connect();
var obj1 = cache.Get("key1"); //该对象是流,需要将其转换为强类型,或自己再编写扩展方法
var bytes = obj1.ReadAllBytes();

同步与推荐

本文已同步至目录索引:解读ASP.NET 5 & MVC6系列