Best practices for WPS

[cehbrecht]

For example:

• handling a set of output files (Create a prototype for pywps multiple outputs #13).
• describing more complex inputs (we have “dynamic" inputs in Copernicus with dependencies between the input parameters).

[huard]

Multiple output files

Problem description

A process operating on large datasets may want to offer users the possibility to split the output into multiple files (one per year, per region, per variable, etc). In many cases, user parameters will determine the number of output files. It seems that the WPS standard would suggest to define the process as having one ComplexOutput, whose content provides a description and links to the individual output files. If we are to realize the promise of interoperability, there should be some community agreement on the format and structure of this output file.

Possible solutions

The WPS standard references "nested output". It's not clear what this means however. Would it be possible to nest a list of ComplexOutputs inside the process ComplexOutput ? Is this supported by pywps ?

Another option would be to describe netCDF output using NcML: https://www.unidata.ucar.edu/software/thredds/v4.3/tds/tds4.2/tutorial/NcML.htm

Dynamic inputs

For some processes, inputs are interdependent. That is, specifying the value of an input conditions the range of potential values for another input. This is not supported by the WPS standard and can only be enforced once the process is launched. A mechanism to provide dynamic input validation before the process is launched would be useful in many situations.