Writing my own Provider Class in ASP.NET - asp.net

Question

Note: I DON't want to write custom membership provider.
I want to write my own Provider class so I can define it in web.config and access it like Membership class.
Here is a sample of my class (it has many other static methods):
public static class MySqlHelper
{
private static string constring = ConfigurationManager.ConnectionStrings["MyConnString"].ConnectionString;
public static int ExecuteNonQuery(string mysqlquery)
{
SqlConnection conn = new SqlConnection(connString);
SqlCommand cmd = new SqlCommand(mysqlquery, conn);
int result;
try
{
conn.Open();
result= cmd.ExecuteNonQuery();
}
finally
{
conn.Close();
}
return result;
}
}
Usage: MySqlHelper.ExecuteNonQuery("select * from customers");
Now as you see I have hard-coded the name of connectionstring i.e. "MyConnString". I am planning to make it dynamic.
So I was wondering if I can make it like static built-in Membership class, where I can define the connectionStringName in web.config. This way the class can be made re-usable without always naming my connectionstring in web.config to "MyConnString".
1: I DON'T want to pass connectionstring in every static method as a parameter.
2: I must be able to access the methods similar to Membership.CreateUser i.e. static.
I am looking over the web in parallel but any inputs/guidance will help.
Edited: I have updated my code sample, to clear some confusion about issues using static class. Here is a new question I posted to clarify that. Sorry about confusion.

Answer 1

the only thing i can think of that meets the qualifications you laid out is to use dependency injection, a static constructor, and inject in an something like an IConnectionStringProvider. this seems like about the most convoluted thing i can think of, so you might like it. :)
edit
after reading your comment, it seems like you just want to be able to reference any connection string, but only one connection string per application. i'd say just add an element to appSettings named MySqlProviderConnection with the value being the name of the connection string you want to use.
then in your helper, check for the existence of the appsetting, get its value, and pass it in to your ConfigurationManager.ConnectionStrings call. that way your provider could use any connection you want, without changing any code.

Answer 2

I typically discourage sharing one SqlConnection instance across several requests. Even if you enable MARS, you can run into performance issues. I think when your connection receives a non-read command, the connection buffer will pause all current reads until the write finishes. The only thing you're really saving is the time it takes to establish a connection.
SqlConnections are pooled so you can configure the provider to have a min / max number of instances available to soliciting clients. Keep in mind this is also controlled by whatever database you're connecting to; assuming you're connecting to a SQL Server instance, SQL Server has its own maximum connections allowed setting.
Instead of allowing clients to determine when to open/close a shared SqlConnection instance, I suggest having your public members take in either a command string or command parameters. Then, similar to what your sample has suggested, open a connection from the pool and execute the command.
public IEnumerable<SqlResults> ExecuteStoredProcedure(string procedure, params SqlParameter[] parameters) {
using(SqlConnection connection = new SqlConnection(MyConnectionStringProperty)) {
try {
connection.Open();
using(SqlCommand command = new SqlCommand(procedure, connection)) {
command.CommandType = CommandType.StoredProcedure;
if(parameters != null) {
command.Parameters.AddRange(parameters);
}
// yield return to handle whatever results from proc execution
// can also consider expanding to support reader.NextResult()
using(SqlDataReader reader = command.ExecuteReader()) {
yield return new SqlResults {
Reader = reader;
};
}
}
}
finally {
if(connection.State != ConnectionState.Closed) {
connection.Close();
}
}
}
}
The sample code above is just that - a sample of a concept I use at work. The sample does now have maximized error handling but is very flexible in how results are returned and handled. The SqlResults class simply contains a SqlDataReader property and can be expanded to include errors.
As far as making any of this static, it should be fine as long as you enable a way to make a singleton instance of the provider class and continue to not have any mutable properties be shared (potentially across various requests/threads). You may want to consider some sort of IoC or Dependency Injection approach for providing the connection string given your request.
EDIT
Yield allows the caller to use the returned object before the execution context returns to the method yielding the return for continued execution. So in the sample above, a caller can do something like this:
// Since it's an IEnumerable we can handle multiple result sets
foreach(SqlResults results in MySqlHelper.ExecuteStoredProcedure(myProcedureName, new SqlParameter("myParamName", myParamValue)) {
// handle results
}
without the connection closing while we handle the results. If you notice in the sample, we have using statements for our SqlClient objects. This approach allows result set handling to be decoupled from MySqlHelper as the provider class will take care of the would-be-duplicate SQL provision code, delegate result handling to the caller, then continue with what it has to do (i.e. close the connection).
As for IoC/DI, I personally use Castle Windsor. You can inject dependency objects as properties or construction parameters. Registering an Inversion of Control container as your dependency resource manager will allow you to (among other things) return the same object when a type of resource is requested. Basically, for every caller class that needs to use MySqlHelper, you can inject the same instance when the caller class is instantiated or when the caller class references its public MySqlHelper property. I, personally, prefer constructor injection whenever possible. Also, when I say inject, I mean you don't have to worry about setting the property value as your IoC/DI does it for you (if configured properly). See here for a more in depth explanation.
As another note, the IoC/DI approach would really only come into play if your class is non-static such that each application can have its own singleton instance. If MySqlHelper is static, then you could only support one connection string unless you pass it in, which in your original question, you'd prefer not to do so. IoC/DI will allow you to use your MySqlHelper property member as if it were static though since the registered container would ensure that the property has a proper instance.

Answer 3

Here is the complete code of a SqlHelper that I'd used on some small projects.
But carefull with static for this kind of class. If you will use it for Web project, remember that the connection will be shared at the same instance for all users, which can cause bad problems...
using System.Data;
using System.Data.SqlClient;
using System.Web.Configuration;
public class SqlHelper
{
private SqlConnection connection;
public SqlHelper()
{
connection = new SqlConnection();
}
public void OpenConnection()
{
// Updated code getting the ConnectionString without hard naming it.
// Yes, if you have more than 1 you'll have problems... But, how many times it happens?
if (WebConfigurationManager.ConnectionStrings.Length == 0)
throw new ArgumentNullException("You need to configure the ConnectionString on your Web.config.");
else
{
connection.ConnectionString = WebConfigurationManager.ConnectionStrings[0].ConnectionString;
connection.Open();
}
}
public void CloseConnection()
{
if (connection != null && connection.State != ConnectionState.Closed)
connection.Close();
}
public DataTable ExecuteToDataTable(string sql)
{
DataTable data;
SqlCommand command = null;
SqlDataAdapter adapter = null;
try
{
if (connection.State != ConnectionState.Open)
OpenConnection();
command = new SqlCommand(sql, connection);
adapter = new SqlDataAdapter(command);
retorno = new DataTable();
adapter.Fill(data);
}
finally
{
if (command != null)
command.Dispose();
if (adapter != null)
adapter.Dispose();
CloseConnection();
}
return data;
}
public int ExecuteNonQuery(string sql)
{
SqlCommand command = null;
try
{
if (connection.State != ConnectionState.Open)
OpenConnection();
command = new SqlCommand(sql, connection);
return command.ExecuteNonQuery();
}
finally
{
if (command != null)
command.Dispose();
CloseConnection();
}
}
public object ExecuteScalar(string sql)
{
SqlCommand command = null;
try
{
if (connection.State != ConnectionState.Open)
OpenConnection();
command = new SqlCommand(sql, connection);
return command.ExecuteScalar();
}
finally
{
if (command != null)
command.Dispose();
CloseConnection();
}
}
}
Sample usage:
SqlHelper sql = new SqlHelper();
DataTable data = sql.ExecuteToDataTable("SELECT * FROM Customers");
int affected = sql.ExecuteNonQuery("INSERT Customers VALUES ('Test')");
But if you really want static (if you is on a single user enviroment), just put static on all methods.