SeaweedFS在.net core下的实践方案(续一)

前言

我们之前已经完成了SeaweedFS在.net core下的使用了，但是说实话，还是不够，于是，我的目光盯住了IApplicationBuilder的扩展方法UseStaticFiles

这个可是好东西啊，我们访问资源的静态文件映射，你懂我的意思吧，对这里下手~

前戏

开工之前，我们转到定义看看

StaticFileOptions，这个就是我们自定义乱嗨的前提

它有两个，我们DIY需要用到的参数

RequestPath、FileProvider

顾名思义，前者是访问路径的前置地址

RequestPath的值是wwwroot，那么我们访问

http://url/wwwroot/XX.后缀 才会触发这个，而且，一定要是带后缀的才触发

后者是，前置地址触发的基础上才调用的

我们给他安排一下

实现

因为核心参数FileProvider类型为IFileProvider，所以，我们写一个实现类吧

    public class SeaweedFSFileProvider : IFileProvider
    {
        public IDirectoryContents GetDirectoryContents(string subpath)
        {
            throw new NotImplementedException();
        }

        public IFileInfo GetFileInfo(string subpath)
        {
            throw new NotImplementedException();
        }

        public IChangeToken Watch(string filter)
        {
            throw new NotImplementedException();
        }
    }

我们现在需要用到的是GetFileInfo这个方法，另外两个，并不会触发（严谨一点说，GetDirectoryContents，我们访问无论是资源目录路径，还是完整的资源路径，都不会触发）

比如我们wwwroot这个文件夹是资源路径

无论是

http://url/wwwroot/ 还是 http://url/wwwroot，都不触发

这个也没触发~

emmmmm，可能研究太浅了，这两个接口方法是给其他实现类提供的定制化功能？比如FileServer?

GetFileInfo方法的返回值是IFileInfo

这个接口，触发文件返回的顺序是

Exists属性->Length属性->LastModified属性->Exists属性->PhysicalPath属性->CreateReadStream方法

我们写一个实现

    public class SeaweedFSFileInfo : IFileInfo
    {
        public bool Exists { get; set; }

        public long Length => new MemoryStream(Context).Length;

        public string PhysicalPath { get; set; }

        public string Name { get; set; }

        public DateTimeOffset LastModified { get; }

        public bool IsDirectory => false;

        private byte[] Context { get; }

        public SeaweedFSFileInfo()
        {
        }

        public SeaweedFSFileInfo(string physicalPath, byte[] context)
        {
            Context = context;
            PhysicalPath = physicalPath;
            Name = Path.GetFileName(PhysicalPath);
            LastModified = DateTimeOffset.Now;
            Exists = true;
        }

        public Stream CreateReadStream()
        {
            return new MemoryStream(Context);
        }
    }

我们修改一下SeaweedFSFileProvider，这里注入一个IFileService

因为我们希望整个SeaweedFSFileProvider他只依赖于IFileService，而不过多依赖SeaweedFS的实现，不会让代码简洁性受损

    public class SeaweedFSFileProvider : IFileProvider
    {
        private IFileService Service { get; }

        public SeaweedFSFileProvider(IFileService service)
        {
            Service = service;
        }

        public IDirectoryContents GetDirectoryContents(string subpath)
        {
            throw new NotImplementedException();
        }

        public IFileInfo GetFileInfo(string subpath)
        {
            throw new NotImplementedException();
        }

        public IChangeToken Watch(string filter)
        {
            throw new NotImplementedException();
        }
    }

我们转到接口IFileService，写一个接口

    public interface IFileService
    {
        Task GetUploadFileUrlAsync();

        Task UploadFileAsync(string url,byte[] context);

        IFileInfo GetFileInfo(string subpath);
    }

再转到实现类

增加这个实现

        public IFileInfo GetFileInfo(string subpath)
        {
            throw new NotImplementedException();
        }

我们去外层的SeaweedFSFileProvider修改一下

    public class SeaweedFSFileProvider : IFileProvider
    {
        private IFileService Service { get; }

        public SeaweedFSFileProvider(IFileService service)
        {
            Service = service;
        }

        public IDirectoryContents GetDirectoryContents(string subpath)
        {
            throw new NotImplementedException();
        }

        public IFileInfo GetFileInfo(string subpath)
        {
            return Service.GetFileInfo(subpath);
        }

        public IChangeToken Watch(string filter)
        {
            throw new NotImplementedException();
        }
    }

这样IFileService 里面变成什么样，都跟这层没关系了

我们定义一个IFileInfoFactory

    public interface IFileInfoFactory
    {
        bool Contains(string filepath);

        IFileInfo GetFileInfo(string filepath);

        IFileInfo AddFileInfo(string filepath, byte[] context);

        IFileInfo AddNotExists(string filepath);
    }

再写一个默认实现

    public class FileInfoFactory: IFileInfoFactory
    {
        private List FileInfo { get; } = new List();

        public bool Contains(string filepath)
        {
            return FileInfo.Any(file => file.PhysicalPath.Equals(filepath));
        }

        public IFileInfo GetFileInfo(string filepath)
        {
            return FileInfo.FirstOrDefault(file => file.PhysicalPath.Equals(filepath));
        }

        public IFileInfo AddFileInfo(string filepath,byte[] context)
        {
            var info = new SeaweedFSFileInfo(filepath, context);
            FileInfo.Add(info);

            return info;
        }

        public IFileInfo AddNotExists(string filepath)
        {
            var info = new SeaweedFSFileInfo();
            FileInfo.Add(info);

            return info;
        }
    }

我们修改一下

SeaweedFSService的默认实现，增加一个注入IFileInfoFactory

        private IFileInfoFactory FileInfoFactory { get; }

        public SeaweedFSService(IOptions options, IFileInfoFactory fileInfoFactory)
        {
            Configuration = options.Value;
            FileInfoFactory = fileInfoFactory;
        }

我们实现一下

GetFileInfo方法

        public IFileInfo GetFileInfo(string subpath)
        {
            using (var client = HttpClientFactory.Create())
            {
                var path = subpath.Replace(Path.GetExtension(subpath), "");
                var splits = path.Split("/".ToCharArray(), StringSplitOptions.RemoveEmptyEntries);

                if (splits.Length == 0)
                {
                    return FileInfoFactory.AddNotExists(subpath);
                }
                else
                {
                    var fid = $"{splits[0]},{splits[1]}";

                    var response = client.GetAsync($"http://{Configuration.BaseUrl}/{fid}")
                        .GetAwaiter()
                        .GetResult();

                    if (response.StatusCode == HttpStatusCode.NotFound)
                    {
                        return FileInfoFactory.AddNotExists(subpath);
                    }
                    else
                    {
                        var context = response.Content;
                        var bytes = context.ReadAsByteArrayAsync()
                            .GetAwaiter()
                            .GetResult();

                        if (FileInfoFactory.Contains(subpath))
                        {
                            return FileInfoFactory.GetFileInfo(subpath);
                        }
                        else
                        {
                            return FileInfoFactory.AddFileInfo(subpath, bytes);
                        }
                    }
                }
            }
        }

 

这个时候，我们测试一下

大功告成，撒花

优化

但是我们可能场景是这个文件上传了，就不再修改了，修改后文件，变成新路径，这样，文件就始终是静态的，那么这样反复http请求就没意义了

所以，我们修改一下

        private SeaweedFSServiceConfiguration Configuration { get; }
        private IFileInfoFactory FileInfoFactory { get; }
        private IDistributedCache Cache { get; }

        public SeaweedFSService(IOptions options, IFileInfoFactory fileInfoFactory, IDistributedCache cache)
        {
            Configuration = options.Value;
            FileInfoFactory = fileInfoFactory;
            Cache = cache;
        }

增加了一个分布式缓存

我们就找这个缓存，能不能找到，还能找到，就说明已经缓存了这个文件信息，就不再走http

修改一下GetFileInfo

        public IFileInfo GetFileInfo(string subpath)
        {
            var key = $"Distributed_Files_{subpath}";

            var contextBytes = Cache.Get(key);

            if (contextBytes != null && FileInfoFactory.Contains(subpath))
            {
                return FileInfoFactory.GetFileInfo(subpath);
            }
            else
            {
                using (var client = HttpClientFactory.Create())
                {
                    var path = subpath.Replace(Path.GetExtension(subpath), "");
                    var splits = path.Split("/".ToCharArray(), StringSplitOptions.RemoveEmptyEntries);

                    Cache.Set(key, new byte[] { });

                    if (splits.Length == 0)
                    {
                        return FileInfoFactory.AddNotExists(subpath);
                    }
                    else
                    {
                        var fid = $"{splits[0]},{splits[1]}";

                        var response = client.GetAsync($"http://{Configuration.BaseUrl}/{fid}")
                            .GetAwaiter()
                            .GetResult();

                        if (response.StatusCode == HttpStatusCode.NotFound)
                        {
                            return FileInfoFactory.AddNotExists(subpath);
                        }
                        else
                        {
                            var context = response.Content;
                            var bytes = context.ReadAsByteArrayAsync()
                                .GetAwaiter()
                                .GetResult();

                            if (FileInfoFactory.Contains(subpath))
                            {
                                return FileInfoFactory.GetFileInfo(subpath);
                            }
                            else
                            {
                                return FileInfoFactory.AddFileInfo(subpath, bytes);
                            }
                        }
                    }
                }
            }
        }

这样访问的地址，缓存没失效之前，并且在文件缓存里面，就不再走http请求了

 

附

我们附上入口的代码

ConfigureServices方法内增加

services.AddDistributedMemoryCache();

这样就启用了默认的分布式缓存接口，后期要替换的实现，只用更换这里的具体实现就好了，我们不依赖具体实现

Configure方法内增加代码

            using (var services = app.ApplicationServices.CreateScope())
            {
                var fileService = services.ServiceProvider.GetRequiredService();

                app.UseStaticFiles(
                    new StaticFileOptions
                    {
                        RequestPath = "/Resource",
                        FileProvider = new SeaweedFSFileProvider(fileService)
                    }
                );
            }

 

转载于:https://www.cnblogs.com/NCoreCoder/p/10266360.html