README

A small PHP 8.2+ library for the moment before an agency CSV enters your property catalogue. Map the provider's columns, validate records, review rejected rows, then export spreadsheet-safe results. No network requests, no CMS writes and no dependencies.

Published by NovAsia Estate, whose property research covers Cambodia, Thailand and Bali. This project handles the import boundary; it does not verify availability or the accuracy of an advertised price.

Why reject a plausible-looking value?

1,200 could be a grouping convention, a decimal convention or two CSV fields. $ does not identify one currency. A missing source date must not become today's date. This library accepts explicit decimal strings and an explicit supported currency instead of guessing.

Install and use

composer require novasia-estate/listing-import-guard

require 'vendor/autoload.php';
use NovAsia\ListingImport\ImportGuard;

$guard = new ImportGuard([
    'id' => 'Reference', 'title' => 'Headline',
    'price' => 'Amount', 'currency' => 'Currency',
    'source_url' => 'Source', 'observed_at' => 'Checked',
]);
$report = $guard->importCsv(file_get_contents('agency.csv'), '2026-10-29');
foreach ($report['rejected'] as $row) {
    echo $row['record'] . ': ' . implode(', ', $row['errors']) . PHP_EOL;
}
file_put_contents('reviewed.csv', ImportGuard::exportCsv($report['accepted']));

price remains a string, including decimal precision. It is never converted to a float or an exchange-rate equivalent. Accepted rows have all six canonical fields. Duplicate identifiers after the first accepted record are rejected. Rejection positions count CSV records (a quoted multiline field is still one record), not physical lines.

Contract

Field	Rule
id	Non-empty, up to 200 bytes, unique among accepted rows
title	Non-empty, up to 2,000 bytes
price	Non-negative plain decimal string, up to 40 bytes; explicit zero allowed
currency	USD, THB, IDR, KHR, EUR, GBP, AUD or SGD; uppercase and explicit
source_url	HTTP(S), no embedded credentials, no literal private/reserved IP
observed_at	Actual YYYY-MM-DD date, no later than explicit as-of date

CSV uses comma delimiters, a unique header, UTF-8 (optional BOM), at most 5 MB. Standard PHP fgetcsv quoting behavior applies. Extra unmapped columns are ignored. URL validation is syntactic; DNS aliases are not checked and this is not an SSRF guard for any later fetching code.

WordPress or framework integration

Run validation before calling wp_insert_post or writing an ORM model. Keep the rejection report for editorial review. Store price as DECIMAL or a string, not FLOAT. Escape titles for the final output context (esc_html in WordPress); this library does not turn text into trusted HTML. Do not automatically fetch a source URL just because its syntax passed.

Try the synthetic fixture

php examples/import.php
php tests/run.php

The first example row is accepted; the second is rejected for missing currency. The sample is invented and contains no real listing, customer record or market price.

Related workflow

Property Schema Validator checks a JSON representation before sharing it downstream.
CSV Quality Lab provides an interactive browser review when a PHP pipeline is not available.
NovAsia Estate supplies the publisher's property guides and research context; this library carries no live inventory.

MIT licensed. No telemetry, tracking pixels, API keys or external service required. Issues and small reproducible failing CSV examples are welcome; remove private contact details before opening an issue.

1.2: optional exact header contract

$guard = new ImportGuard($mapping, 'reject');
$report = $guard->importCsv($csv, $asOf);

The second constructor argument is ignore (default, unchanged) or reject. Strict mode requires the CSV header to contain exactly the six explicitly mapped input columns, in any order. It rejects every unmapped column before validating or accepting records and reports all offending one-based column positions, without echoing header text or row values. Custom mappings work the same way: it is the mapped source column names, not canonical output names, that define the contract. Header matching stays exact and case-sensitive; nothing is renamed or guessed. Empty and duplicate headers and missing required columns retain their original checks. An invalid policy fails in the constructor.

Use the default when intentionally importing a wider source table and ignoring unrelated columns. Use reject for a fixed integration contract where an added or mistyped optional column should stop the import. This mode does not make source facts trustworthy or change per-record validation. Run composer test for the original checks and strict-mode counterexamples.

novasia-estate / listing-import-guard

Package info

Security

Statistics