joinEach() asks a DataFrameFactory for the right side once per left batch, so the right side is
never loaded whole - here, the rows the database does not know about yet (Join::left_anti). Use
it when the right side is too large for join().
Examples
Joins
Description
Documentation
A join matches rows from two frames on a condition.
Code
<?php
declare(strict_types=1);
use function Flow\ETL\DSL\{array_to_rows, data_frame, df, equal, from_array, int_schema, join_on, schema, str_schema, to_output};
use Flow\ETL\{DataFrame, DataFrameFactory, Extractor, FlowContext, Rows};
use Flow\ETL\Extractor\Statistics;
use Flow\ETL\Join\Join;
use Flow\ETL\Schema;
require __DIR__ . '/vendor/autoload.php';
$apiSchema = schema(int_schema('id'), str_schema('sku'));
$apiExtractor = new class($apiSchema) implements Extractor {
public function __construct(private Schema $schema)
{
}
public function extract(FlowContext $context, ?int $limit = null): Generator
{
yield array_to_rows([['id' => 1, 'sku' => 'PRODUCT01'], ['id' => 2, 'sku' => 'PRODUCT02']], $this->schema);
yield array_to_rows([['id' => 10_001, 'sku' => 'PRODUCT10_001']], $this->schema);
}
public function schema(): Schema
{
return $this->schema;
}
public function statistics(): Statistics
{
return new Statistics();
}
public function withSchema(Schema $schema): static
{
return new static($schema);
}
};
$dbDataFrameFactory = new class($apiSchema) implements DataFrameFactory {
public function __construct(private Schema $schema)
{
}
public function from(Rows $rows): DataFrame
{
// in real life this is one SQL query for $rows->reduceToArray('id')
$known = [
['id' => 1, 'sku' => 'PRODUCT01'],
['id' => 2, 'sku' => 'PRODUCT02'],
];
return df()->read(from_array($known, $this->schema));
}
};
data_frame()
->extract($apiExtractor)
// left_anti keeps the rows the database does not know about
->joinEach($dbDataFrameFactory, join_on(equal('id', 'id')), Join::left_anti)
->write(to_output(truncate: false))
->run();