Join datasets that are too large to fit in memory. Unlike regular join, joinEach fetches only the relevant rows from storage for each batch, making it suitable for large right-side datasets.
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\{data_frame, df, equal, from_array, int_schema, join_on, row, rows, schema, str_schema, to_output};
use Flow\ETL\{DataFrame, DataFrameFactory, Extractor, FlowContext, Rows};
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 $context): Generator
{
yield rows($this->schema, row(['id' => 1, 'sku' => 'PRODUCT01']), row(['id' => 2, 'sku' => 'PRODUCT02']));
yield rows($this->schema, row(['id' => 10_001, 'sku' => 'PRODUCT10_001']));
}
public function schema(): Schema
{
return $this->schema;
}
public function withSchema(Schema $schema): static
{
$clone = clone $this;
$clone->schema = $schema;
return $clone;
}
};
$dbDataFrameFactory = new class($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();