README

High-performance PHP adapter for DuckDB that executes SQL queries directly on massive datasets (CSV, JSON, Parquet) without preprocessing or imports. Features SQLite-compatible interface, FFI integration, and intelligent CSV parsing for zero-ETL analytics on files up to gigabytes in size.

Benchmark:

Tested with a large CSV dataset containing 1 million rows (150 MB) using PHP v8.3.

Note: No optimizations were applied. The benchmark includes database creation from compressed CSV and query execution.

CPU	Memory	Disk	OS	Timing
i7 (13K)	32 Go	NVMe	Windows 10 Pro x64	2.19s
i7 (8)	8 Go	NVMe	Windows 10 Pro x64	3.054s
Xeon (E22)	16 Go	SSD	Linux Debian 11	3.58s
i3 (3)	8 Go	SSD	Windows 10 Pro x64	30.23s

Requirements:

PHP exec function
PHP SQLite3 extension
PHP FFI extension (Used by Loader)

Install:

composer require jakiboy/pducky
vendor/bin/pducky install-binaries

Test:

git clone https://github.com/Jakiboy/pducky.git
cd pducky && composer dump-autoload
composer install-binaries
cd examples && bash generate.sh
bash test.sh

Examples:

Fetch single value:

$price = (new Pducky\Adapter('data.csv'))->import()->single(
	'SELECT `price` FROM `temp` WHERE `ean` = "4567890123456";'
); // 374.08$

Fetch rows:

$rows = (new Pducky\Adapter('data.csv'))->import()->query(
	'SELECT * FROM `temp` LIMIT 100;'
); // []

Create database:

Create database data with table product from a compressed CSV file data.csv.gz.

(new Pducky\Adapter('data.csv.gz'))->import('data', 'product');

Loader query (FFI):

$rows = (new Pducky\Loader())->connect('data.db')
	   ->importCsv('data.csv', 'product')
	   ->query('SELECT * FROM product LIMIT 100;'); // []

jakiboy / pducky

Maintainers

Details