Data sources
============

A processing workflow is a data processing pipeline. How does the data reach the data processing pipeline? 

There are several approaches to provide data to the data processing pipeline:

* A local file containing the inputs as references to data or values (one per line)
* A list of comma-separated values containing the inputs as references to data or values
* A reference to a catalogue 

Local file 
**********

This approach is very useful during the early stages of the application development where you download a few files to your sandbox and use these data to run the first tests.

.. note:: 
  We do not recommend using this method beyond the early stages of the application development.

Comma-separated values 
**********************

This approach is very straight-forward to test the application against a small number of values. 
On the other hand, during the exploitation phase with an application exposed as an OGC WPS service, the comma-separated values list is very well managed and allows clients to provide several values to the application.

Reference to a catalogue
************************

This approach allows tapping on large repositories of Earth Observation data exposing an OpenSearch catalogue (see :ref:`da-catalog`)
It is the preferred option to process large datasets and to expose catalogue queriable in the application Web Processing Service interface.


.. tip:: 
  For details on how to implement this, read :doc:`Application descriptor reference </reference/application/index>`