bagisto-data-transfer

Data Transfer

packages/Webkul/DataTransfer moves bulk records into Bagisto from a file. The package ships three importers — products, customers and tax rates — and a queued pipeline that validates, imports, links and indexes in batches.

Import only. There is no exporter here; a DataGrid's own export handles outbound data — see the bagisto-datagrid-development skill.

Reference files

The registry

An importer is registered in Config/importers.php, merged into the top-level importers key (not data_transfer.importers):

'tax_rates' => [
    'title'        => 'data_transfer::app.importers.tax-rates.title',
    'importer'     => 'Webkul\DataTransfer\Helpers\Importers\TaxRate\Importer',

The admin create/edit screens iterate config('importers') directly, so a new entry appears in the type dropdown with no view change. Import resolves the class with config('importers.'.$type.'.importer') — the array key is the type stored on the import record, so renaming a key orphans existing imports.

Provide all four sample paths. The UI offers a sample download per format, and a missing file is a broken link rather than a graceful fallback.

The importer contract

Extend Helpers\Importers\AbstractImporter and implement exactly two methods:

Everything else is declared as properties — $validColumnNames, $masterAttributeCode, $permanentAttributes, $messages — or overridden as hooks. See importers.md.

Sources

Helpers\Sources\ supplies CSV, XLS, XLSX and XML, all extending AbstractSource, which is an iterator over rows plus generateErrorReport(array $errors). An importer never opens the file itself — it reads $this->source, so the same importer serves every format.

Non-negotiables

• Rows are validated before anything is written. validateRow() must be free of side effects: it runs over the whole file, and on stop-on-errors the import may never reach importBatch().
• Work in batches, never row-by-row over the whole file. AbstractImporter::BATCH_SIZE is 100 and the pipeline dispatches one job per batch. An importer that loads the file into memory defeats the design and fails on the file sizes this feature exists for.
• Go through repositories for writes, as everywhere else in Bagisto.
• Every message goes through trans() in the data_transfer:: namespace, in all 22 locales.
• Declare isLinkingRequired() / isIndexingRequired() honestly. Returning true adds a queued stage per batch; returning false when linking is needed leaves records half-related with no error.
• The queue must be running for anything past validation. With QUEUE_CONNECTION=sync the whole import runs inline in the request and will time out on a real file.

REQUIRED SUB-SKILL: Use bagisto-change-verification before calling any change done.`