For integration that requires synchronization in both sides, there is a possibility to declare export process on connector level. Your connector should implement Oro\Bundle\IntegrationBundle\Provider\TwoWaySyncConnectorInterface to expose the job name that will make export.
Export Job Definition
The definition of the export job is similar to import. It is an additional job for Oro\Bundle\BatchBundle that should be added to batch_job.yml. A job might be declared with multiple steps, but a good practice is to use one connector for one entity. In order to read a entity from the database, there is additional reader placed in OroIntegrationBundle oro_integration.reader.entity.by_id, it takes the EntityReaderById::ID_FILTER option from the context object(ContextInterface) for the matching entity to read.
For now only non-composite identifiers are supported.
#batch_job.yml example_export: title: Job title here type: export steps: export_entity_1: title: Step title here class: Oro\Bundle\BatchBundle\Step\ItemStep services: reader: oro_integration.reader.entity.by_id # read entity from database by identifier processor: YOUR_PROCESSOR # service which process each record. Could prepare changeset for writer. writer: YOUR_REVERSE_WRITER # service that are responsible for pushing data to remote instance parameters: ~ # .... another steps
Processor and writer could be initialized in your bundle in service.yaml.
YOUR_PROCESSOR: class: Acme\Bundle\AcmeBundle\Processor\YourProcessor YOUR_REVERSE_WRITER: class:Acme\Bundle\AcmeBundle\Writer\YourReverseWriter
Where YOUR_PROCESSOR.class - should implement Oro\Bundle\ImportExportBundle\Processor\ProcessorInterface and YOUR_REVERSE_WRITER.class - should implement Oro\Bundle\ImportExportBundle\Processor\WriterInterface
Implementation of those classes is platform-specific, so there is no abstraction layer.