README

Extract information from Dutch addresses

I have been in this situation multiple times: You need to provide a housenumber and addition for some API. But the provided dataset only has the complete address. This library tries to solve this issue. For instance:

Kerkstraat 95A
- Street: Kerkstraat
- Housenumber: 95
- Addition: A

This is a lightweight repository, it has no runtime dependencies beyond the mbstring extension.

Upgrading from 1.x? See UPGRADE.md.

Requirements

PHP 8.1 or higher, with the mbstring extension. Version 1.x supports PHP 7.0 and up.

Installation

composer require michielgerritsen/extract-address-parts

Usage

use MichielGerritsen\ExtractAddressParts\CombinedAddressExtraction;
use MichielGerritsen\ExtractAddressParts\Exceptions\AddressExtractionError;
use MichielGerritsen\ExtractAddressParts\VO\AddressExtractionResult;

try {
    /** @var AddressExtractionResult $result */
    $result = CombinedAddressExtraction::create()->process('Kerkstraat 95A');
} catch (AddressExtractionError $exception) {
    die('Uh oh, this address seems to be invalid.');
}

$result->getStreet(); // Kerkstraat
$result->getHousenumber(); // 95
$result->getAddition(); // A

Pass an array when the address is split over multiple lines:

$result = CombinedAddressExtraction::create()->process(['Kerkstraat', '95A']);

Choosing a strategy

Strategy	How it works	Use when
`CombinedAddressExtraction`	Tries `PrecisionAddressExtraction` first and falls back to `AddressExtraction`.	You want the best result available. Start here.
`PrecisionAddressExtraction`	Matches against a list of all known Dutch streets.	You only want addresses whose street actually exists, and prefer an exception over a guess.
`AddressExtraction`	Matches the shape of an address with a regular expression.	The address may be outside the Netherlands, or you accept unverified input.

use MichielGerritsen\ExtractAddressParts\AddressExtraction;
use MichielGerritsen\ExtractAddressParts\PrecisionAddressExtraction;

$result = (new AddressExtraction())->process('Kerkstraat 95A');
$result = (new PrecisionAddressExtraction())->process('Kerkstraat 95A');

CombinedAddressExtraction::create() wires up the two default strategies. To use your own, pass any implementations of AddressExtractionInterface to the constructor. They are attempted in the order you give them:

new CombinedAddressExtraction(new PrecisionAddressExtraction(), new AddressExtraction());

Knowing how reliable a result is

PrecisionAddressExtraction verifies the street against a list of known Dutch streets, while AddressExtraction only checks that the input has the shape of an address. Because CombinedAddressExtraction falls back silently, a guess would otherwise be indistinguishable from a verified match. getExtractionMethod() tells you which of the two produced the result:

use MichielGerritsen\ExtractAddressParts\ExtractionMethod;

$result = CombinedAddressExtraction::create()->process('Qwertyuiopasdf 12');

$result->getExtractionMethod(); // ExtractionMethod::Pattern
$result->getExtractionMethod()->isVerified(); // false, the street was never confirmed to exist

That street does not exist, but it has the shape of an address, so the regex accepts it. Use isVerified() to decide what to trust, for example to accept verified addresses automatically and queue the rest for review.

Errors

An AddressExtractionError is thrown when:

no house number can be found, such as Kerkstraat on its own;
the street is not recognised (PrecisionAddressExtraction only);
the input is a postbox, such as Postbus 1234, which has no street or house number.

The street list

PrecisionAddressExtraction matches against the files in addresses/, one per starting character, holding every street name in the Netherlands in lowercase UTF-8. The data comes from the Kadaster BAG registry through the PDOK Locatieserver.

Regenerate it with:

composer update-addresses

The script walks all 342 gemeenten, because the API returns at most 100 rows per request and refuses an offset above 10000. It takes a few minutes, and refuses to write anything if it collects implausibly few streets, so a failed run cannot leave you with a truncated list.

License

The code is MIT licensed.

The street names in addresses/ are not. They come from the Basisregistratie Adressen en Gebouwen of the Dutch Kadaster, retrieved through PDOK, and are licensed CC BY 4.0. Keep that attribution if you redistribute the data.

Contributing

Pull the repository and run composer install.

composer test     # run the test suite
composer analyse  # run PHPStan

Both run in CI against every supported PHP version.

michielgerritsen / extract-address-parts

Maintainers

Package info

Statistics

Security