Feature/upsert

[puzzledpolymath]

This pull request adds support for performing upsert queries. Ticket #50 is what sparked interest for this feature.

🔍 What was changed

• Each driver (MySQL, Postgres, SQLite and SQLServer) compiler supports an UpsertQuery.
• Differences between SQL dialects have been accounted for.

🤔 Why?

1. Upsert allows developers to combine insert and update operations into a single command, reducing the need for explicit existence checks and multiple database calls.
2. Modern databases support atomic upsert operations, which help maintain data integrity even in concurrent environments.
3. Upsert reduces the number of database round-trips, eliminating separate existence checks followed by insert or update, likely leading to better performance, especially in high-volume scenarios.

📝 Checklist

• Closes 💡 Add support for MySQL: ON DUPLICATE KEY UPDATE, PostgreSQL: ON CONFLICT #50
• How was this tested:
  • Unit tests added

📃 Documentation

Note that specifying conflict index column names is redundant and not required for MySQL.

PHP

$upsert = $this->database->upsert('users')
    ->conflicts('email')
    ->columns('email', 'name')
    ->values('adam@email.com', 'Adam')
    ->values('bill@email.com', 'Bill');

$upsert = $this->database->upsert('users')
    ->conflicts(['email'])
    ->values([
        ['email' => 'adam@email.com', 'name' => 'Adam'],
        ['email' => 'bill@email.com', 'name' => 'Bill'],
    ]);

MyQL

INSERT INTO `users` (`email`, `name`)
VALUES
    ('adam@email.com', 'Adam'),
    ('bill@email.com', 'Bill')
ON DUPLICATE KEY UPDATE
    `email` = VALUES(`email`),
    `name` = VALUES(`name`);

Postgres / SQLite

INSERT INTO `users` (`email`, `name`)
VALUES
    ('adam@email.com', 'Adam'),
    ('bill@email.com', 'Bill')
ON CONFLICT (`email`) DO UPDATE SET
    `email` = EXCLUDED.`email`,
    `name` = EXCLUDED.`name`

MSSQL

MERGE INTO [users] WITH (holdlock) AS [target] 
USING ( VALUES 
    ('adam@email.com', 'Adam'),
    ('bill@email.com', 'Bill') 
) AS [source] ([email], [name]) 
ON [target].[email] = [source].[email] 
WHEN MATCHED THEN 
    UPDATE SET 
        [target].[email] = [source].[email], 
        [target].[name] = [source].[name] 
WHEN NOT MATCHED THEN 
    INSERT ([email], [name]) 
    VALUES ([source].[email], [source].[name])

[puzzledpolymath]

One thing I'm uncertain about or dislike at the moment, is the need to specify conflicting index column names for non MySQL drivers. Having said that, most database layers operate in this way, making the user specify conflicting columns.

While likely not an appropriate location for such logic, perhaps something like below could work. E.g. gather info from the schema to determine unique indexes and their columns.

class UpsertQuery extends ActiveQuery
{
    public function getTokens(): array
    {
        $conflicts = $this->conflicts;

        if ($conflicts === [] && $this->driver) {
            $schema = $this->driver->getSchemaHandler()->getSchema($this->table);
            foreach ($schema->getIndexes() as $index) {
                if ($index->isUnique()) {
                    $conflicts = $conflicts + $index->getColumns();
                }
            }
            $conflicts = array_unique($conflicts);
        }

        return [
            'table'     => $this->table,
            'columns'   => $this->columns,
            'values'    => $this->values,
            'conflicts' => $conflicts,
        ];
    }
}

[puzzledpolymath]

@roxblnfk is there anything needed from me to get past that labeler error?

Warning: The action requires write permission to add labels to pull requests. For more information please refer to the action documentation: https://github.com/actions/labeler#permissions
Error: Resource not accessible by integration

[roxblnfk]

@roxblnfk is there anything needed from me to get past that labeler error?

No, don't worry about it.

[puzzledpolymath]

Latest changes have been merged, I've bumped UPSERT_QUERY since the sub query feature was committed first.

    public const SUBQUERY = 9;
    public const UPSERT_QUERY = 10;

[roxblnfk]

I haven't had time to look at this PR yet, but it seems you've done a great job. Why did you decide to close it?

[puzzledpolymath]

I deleted the forked repositories forgetting this PR was still open, I’ll sort it out first thing in the morning

[puzzledpolymath]

@roxblnfk #231