什么是插入/更新数据到数据库的最佳方法

Question

Given that we have the following table:

CREATE TABLE TestData 
(
    ID INT NOT NULL PRIMARY KEY,
    Value INT
)

When writing data to the table, I will try to UPDATE the table if the ID already exists or else INSERT a new row.

There are several options that I've written or come across in production code:

1. Perform an UPDATE query first, if no row is updated, perform an INSERT query

2. Perform an Insert query first first, if it fails, performs an INSERT query

3. Perform a SELECT query with new ID first, then if the row exists do UPDATE else INSERT query

4. Perform one big query to do all select update or insert

My question is what is the best practice or most efficient or preferred method amongst the above options (or your own method)?

Example code for above options:

(1) Perform an UPDATE query first, if no row is updated, perform an INSERT query

int affectedRows = 0;
using (IDbCommand updateCmd = new SqlCommand("UPDATE TestData SET Value = @Value WHERE ID = @ID", myConn))
{
    try
    {
        // ...add params
        affectedRows = updateCmd.ExecuteNonQuery();
    }
    catch (SqlException) { /*...*/ }
}

if (affectedRows == 0)
{
    using (IDbCommand insertCmd = new SqlCommand("INSERT INTO TestData VALUES (@ID, @Value)", myConn))
    {
        try
        {
            // ...add params
            affectedRows = insertCmd.ExecuteNonQuery();
        }
        catch (SqlException) { /*...*/ }
    }
}  

(2) Perform an Insert query first first, if it fails, performs an INSERT query

int affectedRows = 0;
using (IDbCommand insertCmd = new SqlCommand("INSERT INTO TestData VALUES (@ID, @Value)", myConn))
{
    try
    {
        // ... add params
        affectedRows = insertCmd.ExecuteNonQuery();
    }
    catch (SqlException) { /*...*/ }
}

if (affectedRows == 0)
{
    using (IDbCommand updateCmd = new SqlCommand("UPDATE TestData SET Value = @Value WHERE ID = @ID", myConn))
    {
        try
        {
            // ...add params
            affectedRows = updateCmd.ExecuteNonQuery();
        }
        catch (SqlException) { /*...*/ }
    }
}

(3) Perform a SELECT query with new ID first, then if the row exists do UPDATE else INSERT query

bool dataExist = false;
// ... 
using (IDbCommand selectCmd = new SqlCommand("SELECT ID FROM TestDate WHERE ID=@ID", myConn))
{
    try
    {
        // ... add param
        using (IDataReader reader = selectCmd.ExecuteReader())
            dataExist = reader.Read();
    }
    catch (SqlException) { /*...*/ }
}
if (dataExist) { /* update query, similar to above one*/ }
else { /* insert, similar to above one */ }

(4) Perform one big query to do all select update or insert

string query = "IF EXISTS (SELECT ID FROM TestData WHERE ID=@ID) " +
               "UPDATE TestData SET Value = @Value WHERE ID = @ID " +
               "ELSE INSERT INTO TestData VALUES (@ID, @Value)";

using (IDbCommand bigQueryCmd = new SqlCommand(query, myConn))
{
    try
    {
        // ... add param
        bigQueryCmd.ExecuteNonQuery();
    }
    catch (SqlException) { /*...*/ }
}

Answer 1

You could use MERGE .

Performs insert, update, or delete operations on a target table based on the results of a join with a source table.

For concurrency reasons it is much better to use MERGE than IF EXISTS/UPDATE/INSERT and it is only one statement so less code to maintain:

Something like:

SqlFiddleDemo

CREATE TABLE TestData (
   ID INT NOT NULL PRIMARY KEY,
   Value INT);

INSERT INTO TestData
VALUES (1,3);


DECLARE @id INT = 1, @value INT = 10;    -- will update
--DECLARE @id INT = 2, @value INT = 10;  -- will insert

;MERGE TestData AS TGT
USING (SELECT * FROM (SELECT @id AS ID, @value AS VALUE) AS t) AS SRC
   ON TGT.ID = SRC.ID
WHEN MATCHED THEN
   UPDATE SET Value = SRC.[VALUE]
WHEN NOT MATCHED THEN
   INSERT (ID, [VALUE])
   VALUES (SRC.ID, SRC.[VALUE]);

SELECT *
FROM TestData;