merge_cubes and overlap resolver

[clausmichele]

Parsing a process graph containing a merge_cubes process with an overlap resolver doesn't work properly, or at least I can't understand why it is giving the current output. For example, parsing the following process graph, for the subtract process (overlap resolver of merge_cubes):
{'process_id': 'subtract', 'arguments': {'x': [{'from_node': '3_2'}, {'from_node': '4_4'}], 'y': [{'from_node': '3_2'}, {'from_node': '4_4'}]}, 'result': True}
which is confusing, since I can't understand which parameter is x and which is y. If I revert the parameter in the code:

              "process_id": "subtract",
              "arguments": {
                "x": {
                  "from_parameter": "y"
                },
                "y": {
                  "from_parameter": "x"
                }
              },

I get the same output.
Could you please check this?

{
  "process_graph": {
    "1": {
      "process_id": "load_collection",
      "arguments": {
        "id": "WETSNOW2",
        "spatial_extent": {
          "east": 11.416511535644531,
          "south": 46.406617041056165,
          "north": 46.550361051816196,
          "west": 11.257896423339844
        },
        "temporal_extent": [
          "2015-11-06T00:00:00.000Z",
          "2016-09-25T00:00:00.000Z"
        ]
      }
    },
    "2": {
      "process_id": "filter_bands",
      "arguments": {
        "data": {
          "from_node": "1"
        },
        "bands": [
          "VH"
        ],
        "common_names": [],
        "wavelengths": []
      }
    },
    "3": {
      "process_id": "reduce_dimension",
      "arguments": {
        "data": {
          "from_node": "2"
        },
        "dimension": "t",
        "reducer": {
          "process_graph": {
            "max1": {
              "process_id": "max",
              "arguments": {
                "data": {
                  "from_parameter": "data"
                }
              },
              "result": true
            }
          },
          "parameters": [
            {
              "name": "data",
              "description": "",
              "schema": {}
            }
          ]
        }
      }
    },
    "4": {
      "process_id": "filter_bands",
      "arguments": {
        "data": {
          "from_node": "1"
        },
        "bands": [
          "VH"
        ],
        "common_names": [],
        "wavelengths": []
      }
    },
    "6": {
      "process_id": "save_result",
      "arguments": {
        "data": {
          "from_node": "7"
        },
        "format": "NETCDF",
        "options": {}
      },
      "result": true
    },
    "7": {
      "process_id": "merge_cubes",
      "arguments": {
        "cube1": {
          "from_node": "3"
        },
        "cube2": {
          "from_node": "4"
        },
        "overlap_resolver": {
          "process_graph": {
            "1": {
              "process_id": "subtract",
              "arguments": {
                "x": {
                  "from_parameter": "x"
                },
                "y": {
                  "from_parameter": "y"
                }
              },
              "result": true
            }
          }
        }
      }
    }
  }
}

[clausmichele]

In my opinion, the merge_cubes definition in the API https://processes.openeo.org/#merge_cubes is not clear enough, since it does not clarify what the x and y of the overlap resolver are, if they are cube1 and cube2 or what else.
I encountered a confusing behavior also in the GEE backend Open-EO/openeo-earthengine-driver#47

[claxn]

Thanks for reporting this, the general behaviour for such processes is still not clear to me. I have implemented the following solution for the time being:
Open-EO/openeo-processes#184 (comment)
But I hope we can clarify this and make it more transparent. Please try it again with the branch bug_fixes.

[claxn]

@clausmichele : maybe it is better to revert this behaviour again. Please have a look at: Open-EO/openeo-processes#184 (comment)

[claxn]

@clausmichele The current output from bug_fixes for this process graph is:

Node ID: 1_0 
Node Name: 1 
{'arguments': {'id': 'WETSNOW2',
               'spatial_extent': {'east': 11.416511535644531,
                                  'north': 46.550361051816196,
                                  'south': 46.406617041056165,
                                  'west': 11.257896423339844},
               'temporal_extent': ['2015-11-06T00:00:00.000Z',
                                   '2016-09-25T00:00:00.000Z']},
 'process_id': 'load_collection'}

Node ID: 2_1 
Node Name: 2 
{'arguments': {'bands': ['VH'],
               'common_names': [],
               'data': {'from_node': '1_0'},
               'wavelengths': []},
 'process_id': 'filter_bands'}

Node ID: 3_2 
Node Name: 3 
{'arguments': {'data': {'from_node': '2_1'},
               'dimension': 't',
               'reducer': {'from_node': 'max1_3'}},
 'process_id': 'reduce_dimension'}

Node ID: max1_3 
Node Name: max1 
{'arguments': {'data': {'from_parameter': 'data'}},
 'process_id': 'max',
 'result': True}

Node ID: 4_4 
Node Name: 4 
{'arguments': {'bands': ['VH'],
               'common_names': [],
               'data': {'from_node': '1_0'},
               'wavelengths': []},
 'process_id': 'filter_bands'}

Node ID: 6_5 
Node Name: 6 
{'arguments': {'data': {'from_node': '7_6'}, 'format': 'NETCDF', 'options': {}},
 'process_id': 'save_result',
 'result': True}

Node ID: 7_6 
Node Name: 7 
{'arguments': {'cube1': {'from_node': '3_2'},
               'cube2': {'from_node': '4_4'},
               'overlap_resolver': {'from_node': '1_7'}},
 'process_id': 'merge_cubes'}

Node ID: 1_7 
Node Name: 1 
{'arguments': {'x': {'from_parameter': 'x'}, 'y': {'from_parameter': 'y'}},
 'process_id': 'subtract',
 'result': True}

So all 'from_parameter' statements need now to be resolved by the back-end. But I have added a few more functions, which could be useful to resolve the relationships in a fluent manner (updated docs will follow soon, please track #29). Please close this issue if you are happy with the result, thanks!