Reputation: 93
I have a Table Blob, which has a varbinary(max) as a column. Now I want to store data into the database using a Filestream. The data can be really big (in my case 1.5GB) so I dont want to load the whole data into a buffer.
What I tried:
using (FileStream fs = File.Open(@"BigData.iso", FileMode.Open))
{
using (SqlConnection conn = new SqlConnection())
{
conn.ConnectionString = @"...";
conn.Open();
SqlCommand command = new SqlCommand("INSERT INTO Blob Values (@0, @1)", conn);
command.Parameters.Add(new SqlParameter("0", Guid.NewGuid()));
var sqlb = new SqlBytes(fs);
command.Parameters.Add(new SqlParameter("1", SqlDbType.VarBinary, -1)).Value = sqlb;
command.ExecuteNonQuery();
}
}
But I got an OutOfMemoryException, because SqlBytes initialze its buffer to the whole size of the data.
I know there is a FILESTREAM feature from Microsoft, but I don't want to use it.
Is there a way to achieve this?
Upvotes: 4
Views: 1940
Reputation: 4036
You can read the file in small chunks and append them to the data column.
You will need an IDENTITY column or another column(s) that can be used as a key to execute UPDATE
statements. Here is an example using an IDENTITY
column:
CREATE TABLE [dbo].[table1](
[ID] [int] IDENTITY(1,1) PRIMARY KEY NOT NULL,
[Data] [varbinary](max) NULL,
)
private const string C_SqlConnectionString = @"Server=SERVERNAME;Database=DBNAME;Trusted_Connection=yes;";
private const int C_FileChunkSizeBytes = 1024 * 1024; // 1 MB
private static void storeFile(string filepath)
{
using (FileStream fs = File.Open(filepath, FileMode.Open))
{
using (SqlConnection conn = new SqlConnection())
{
conn.ConnectionString = C_SqlConnectionString;
conn.Open();
// Use a transaction to ensure that all parts of the file get stored to DB
SqlCommand command = new SqlCommand("BEGIN TRAN", conn);
command.ExecuteNonQuery();
var pos = 0;
byte[] fileBytes = null;
int sqlRowId = 0;
// Read the file in chunks
while (pos < fs.Length)
{
// Read file bytes
var bytesToRead = pos + C_FileChunkSizeBytes < fs.Length
? C_FileChunkSizeBytes
: (int)(fs.Length - pos);
fileBytes = new byte[bytesToRead];
fs.Read(fileBytes, 0, bytesToRead);
// Store bytes to a parameter
var varbinary = new SqlParameter("0", System.Data.SqlDbType.VarBinary, -1);
varbinary.Value = fileBytes;
if (pos == 0)
{
// If this is the first chunk, then we need to INSERT
// The HOLDLOCK hint will hold a lock on the table until transaction completes (or is rolled back)
command = new SqlCommand("INSERT INTO [dbo].[table1] WITH(HOLDLOCK) VALUES(@0)", conn);
command.Parameters.Add(varbinary);
command.ExecuteNonQuery();
// Get the row ID for the inserted row
command = new SqlCommand("SELECT @@IDENTITY", conn);
sqlRowId = Convert.ToInt32(command.ExecuteScalar());
}
else
{
// Update existing row and append data
command = new SqlCommand("UPDATE [dbo].[table1] SET [Data] = [Data] + @0 WHERE [ID] = @1", conn);
command.Parameters.Add(varbinary);
command.Parameters.Add(new SqlParameter("1", System.Data.SqlDbType.Int)).Value = sqlRowId;
command.ExecuteNonQuery();
}
// ** Good place for a breakpoint
pos += bytesToRead;
}
// Commit transaction
command = new SqlCommand("COMMIT TRAN", conn);
command.ExecuteNonQuery();
conn.Close();
}
}
}
Place a breakpoint in C# code at the bottom of the while
loop, eg at pos += bytesToRead;
.
Whilst executing the code, when code execution stops at the breakpoint, check the data in SQL:
SELECT *
,LEN([Data]) AS [Length]
FROM [dbo].[table1] WITH(NOLOCK)
The NOLOCK
hint will let us see data in uncommitted transactions. LEN([Data])
will show how field length grows after each iteration of the while
loop.
Upvotes: 1