To import a large CSV file in Laravel 13 without running out of memory, read the file line by line with fgetcsv(), collect rows into batches of about 1,000, and write each batch with upsert(). Move the work into a queued job so the upload request finishes immediately.
What do you need before you start?
- Laravel 13 on PHP 8.3 or newer
- A
productstable withsku(unique),nameandpricecolumns - A CSV file with a header row:
sku,name,price
Why does loading the whole file fail?
Functions such as file() or Excel::toArray() load every row into memory. A file with a few hundred thousand rows can exceed memory_limit. Streaming reads keep memory use flat, because only one batch is in memory at a time.
How do you prepare the table?
Schema::create('products', function (Blueprint $table) {
$table->id();
$table->string('sku')->unique();
$table->string('name');
$table->decimal('price', 10, 2);
$table->timestamps();
});upsert() needs a unique index on the columns you pass as the "unique by" argument.
How do you write the importer?
<?php
namespace App\Services;
use App\Models\Product;
use RuntimeException;
class ProductCsvImporter
{
public function __construct(private int $batchSize = 1000)
{
}
public function import(string $path): int
{
$handle = fopen($path, 'r');
if ($handle === false) {
throw new RuntimeException("Cannot open {$path}");
}
$header = fgetcsv($handle);
$batch = [];
$total = 0;
while (($row = fgetcsv($handle)) !== false) {
if ($row === [null]) {
continue; // blank line
}
$data = array_combine($header, $row);
$batch[] = [
'sku' => trim($data['sku']),
'name' => trim($data['name']),
'price' => (float) $data['price'],
'created_at' => now(),
'updated_at' => now(),
];
if (count($batch) >= $this->batchSize) {
$total += $this->flush($batch);
$batch = [];
}
}
fclose($handle);
if ($batch !== []) {
$total += $this->flush($batch);
}
return $total;
}
private function flush(array $batch): int
{
Product::upsert($batch, ['sku'], ['name', 'price', 'updated_at']);
return count($batch);
}
}upsert() inserts new rows and updates existing ones with the same sku in a single query per batch. It does not fire Eloquent model events.
How do you validate rows?
Validate each row cheaply before it enters the batch and count the failures instead of stopping the import.
use Illuminate\Support\Facades\Validator;
$validator = Validator::make($data, [
'sku' => ['required', 'string', 'max:64'],
'name' => ['required', 'string', 'max:255'],
'price' => ['required', 'numeric', 'min:0'],
]);
if ($validator->fails()) {
$skipped++;
continue;
}How do you run it in a queue?
php artisan make:job ImportProductsCsv<?php
namespace App\Jobs;
use App\Services\ProductCsvImporter;
use Illuminate\Contracts\Queue\ShouldQueue;
use Illuminate\Foundation\Queue\Queueable;
class ImportProductsCsv implements ShouldQueue
{
use Queueable;
public int $timeout = 900;
public function __construct(public string $path)
{
}
public function handle(ProductCsvImporter $importer): void
{
$importer->import(storage_path('app/'.$this->path));
}
}$path = $request->file('csv')->store('imports');
ImportProductsCsv::dispatch($path);Run php artisan queue:work. Set the worker or job timeout longer than the expected import time.
How do you verify it works?
- Create a small CSV with a few rows and import it
- Run it again and confirm the row count does not grow, because
upsert()updates bysku - Import a large generated file and watch memory with
memory_get_peak_usage(true)
What are the common mistakes?
- Inserting one row at a time, which is slow for large files
- Batches that are too large and exceed the database packet or placeholder limit
- Forgetting the unique index required by
upsert() - Non-UTF-8 files that need conversion before reading
- Running the import in the web request and hitting the PHP time limit
FAQ
What batch size should I use?
Start with 500 to 1,000 rows and measure. The best size depends on column count and your database.
Can I use Laravel Excel instead?
Yes. Its chunk reading features can also handle large files, but the plain PHP approach has no extra dependency.
How do I report progress?
Store the total and processed counts in a table or cache key from the job and poll them from the UI.
