無法將PostgreSQL中的文件作為大對象存儲並讀取

Question

我正在嘗試使用NpgSQL v3.0.4.0向PostgreSQL數據庫V9.4.x寫入和讀取大對象，因此我制定了一種將本地文件作為大對象存儲在數據庫中的方法，如下所示：

public static async Task<uint> InsertLargeObjectFileToDB(string theFilePath)
{
     // connecting to DB
     string connstring = MakeDatabaseConnectionString();
     // make a connection object
     NpgsqlConnection Conn = new NpgsqlConnection(connstring);
     try
     {
        await OpenDatabaseConnection(Conn); //open database connection
     }
     catch (Exception Ex)
     {
        throw (Ex);
     }

     uint oid; // to store object ID number
     try
     {
        // Reading and writing Large Objects requires the use of a transaction

        using (FileStream fs = new FileStream(theFilePath, FileMode.Open))
        {
           using (var transaction = Conn.BeginTransaction())
           {
              // Retrieve a Large Object Manager for this connection
              var manager = new NpgsqlLargeObjectManager(Conn);
              // Create a new empty file, returning the identifier to later access it
              oid = manager.Create();

              using (var DbStream = manager.OpenReadWrite(oid))
              {
                 long theFileSize = GetFileSizeInBytes(theFilePath);
                 StreamReader sr = new StreamReader(fs);
                 byte[] buffer = new byte[1024 * 1024];

                 while (sr.BaseStream.Position < theFileSize)
                 {
                    await fs.ReadAsync(buffer, 0, buffer.Length);
                    await DbStream.WriteAsync(buffer, 0, buffer.Length);
                 }
              }
              transaction.Commit();
              return oid;
           }
        }
     }
     catch // any error
     {
        // exception
        Exception ex = new Exception();
        ex.Data.Add(ex.Data.Count, "some error message");
        throw ex;
     }
}

然后，我制作了一種方法來讀取大對象並將其存儲在temp目錄中的隨機命名文件中，如下所示：

public static async Task<string> GetLargeObjectFileFromDB(uint oid)
{
     // connecting to DB
     string connstring = MakeDatabaseConnectionString();
     // make a connection object
     NpgsqlConnection Conn = new NpgsqlConnection(connstring);
     try
     {
        await OpenDatabaseConnection(Conn); //open database connection
     }
     catch (Exception Ex)
     {
        throw (Ex);
     }

     // getting a temorary file name from the system to use it to store the fetched file
     string TempFileName = GetRandomFileNameFromSystem();

     try
     {
        using (FileStream LocalStream = new FileStream(TempFileName, FileMode.Create))
        {
           using (var transaction = Conn.BeginTransaction())
           {
              // create a Large Object Manager for this connection
              var DbLargeObjectManager = new NpgsqlLargeObjectManager(Conn);

              using (var DbStream = await DbLargeObjectManager.OpenReadAsync(oid))
              {
                 byte[] buffer = new byte[1024 * 1024];
                 // get the length of the database object
                 long LengthOfDbObject = DbStream.Length;

                 while (DbStream.Position < LengthOfDbObject)
                 {
                    // read from the database to buffer
                    await DbStream.ReadAsync(buffer, 0, buffer.Length);
                    //write from buffer to local file
                    await LocalStream.WriteAsync(buffer, 0, buffer.Length);
                 }
              }
              transaction.Commit();
              return TempFileName;
           }
        }
     }
     catch // any error
     {
        // exception
        Exception ex = new Exception();
        ex.Data.Add(ex.Data.Count, "Error inserting object in database");
        throw ex;
     }
 }

如您所見，我一直在異步編寫。 問題是我對這2種方法進行了測試，並且此測試將6MB文件寫入數據庫，但是當我再次從數據庫中讀取該文件時，文件大約大了400 kb，（當然）MD5哈希不匹配。 不要忘了說，沒有例外發生過。 如果您關心的話，這是測試：

private async void button3_Click(object sender, EventArgs e)
  {
     listBox1.Items.Clear();

     // getting the MD5 hash of the source file
     string FirstMd5Hash = GetMd5OfFile(tbSourceFile.Text);

     // performance measurment ##########################################
     DateTime dt1 = new DateTime(DateTime.Now.Ticks);
     listBox1.Items.Add("Uploading file to database");
     //storing that file into database
     uint oid = await InsertLargeObjectFileToDB(tbSourceFile.Text);

     // performance measurment #########################################################
     DateTime dt2 = new DateTime(DateTime.Now.Ticks);
     TimeSpan ts = new TimeSpan(dt2.Ticks - dt1.Ticks);
     listBox1.Items.Add("Large object (oid = " + oid + ") inserted in " + ts.Seconds + "." + ts.Milliseconds + " seconds");

     // performance measurment ##########################################
     dt1 = new DateTime(DateTime.Now.Ticks);
     listBox1.Items.Add("reading file back from the database");
     // get that object back from the database into temporary file
     string ReturnedFileName = await PostgresqlLargeObject.GetLargeObjectFileFromDB(oid);
     // performance measurment #########################################################
     dt2 = new DateTime(DateTime.Now.Ticks);
     ts = new TimeSpan(dt2.Ticks - dt1.Ticks);
     listBox1.Items.Add("reading done in " + ts.Seconds + "." + ts.Milliseconds + " seconds");


     //calculate md5 of that file
     string SecondMd5Hash = GetMd5OfFile(ReturnedFileName);

     // compare the 2 hashes
     if (FirstMd5Hash == SecondMd5Hash)
     {
        listBox1.Items.Add("the hashes are match . MD5 = " + FirstMd5Hash);
     }
     else
     {
        listBox1.Items.Add("failed with oid = " + oid);
        tbFileBack.Text = ReturnedFileName;
     }
 }

怎么了？

Answer 1

好的，我已經解決了這個問題，事實證明（除了考慮Emil的答案），您還必須異步讀取然后同步寫入。 我不知道為什么。 此代碼的工作原理：

using (FileStream LocalStream = new FileStream(TempFileName, FileMode.Create))
            {
               using (var transaction = Conn.BeginTransaction())
               {
                  // create a Large Object Manager for this connection
                  var DbLargeObjectManager = new NpgsqlLargeObjectManager(Conn);

                  using (var DbStream = await DbLargeObjectManager.OpenReadAsync(oid))
                  {
                     byte[] buffer = new byte[262144]; //256KB
                     // query the database stream length
                     long DatabaseStreamLength = DbStream.Length;
                     while (DbStream.Position < DatabaseStreamLength)
                     {
                        // read from the database to buffer (async)
                        int bufferByteCount = await DbStream.ReadAsync(buffer, 0, buffer.Length);
                        //write from buffer to local file (sync)
                        LocalStream.Write(buffer, 0, bufferByteCount);
                     }
                  }
                  transaction.Commit();

Answer 2

當執行await <stream>.WriteAsync(buffer, 0, buffer.Length); 您應該寫入由上一個read方法實際讀取的字節數（將返回該值）。