Add aggregate_spatial_window

[clausmichele]

I created a new pull request from draft to draft, with the corrected json file, which now should be correctly formatted. Let's continue the discussion here.

[m-mohr]

Made a couple of detailed review comments. After working on the comments, have a look at the CI results (currently fails) and change the JSON file accordingly, please.

[m-mohr]

Does it have constraints? It seems like the input raster data cube must have only two dimensions? x and y...

[clausmichele]

In the implementation I am working on, it just needs to have x and y, if it is a timeseries then it will be applied on the full timeseries and return another timeseries. So the datacube dimensions could be (time,x,y) or (x,y) or (band,x,y) or (time,band,x,y) and so on. But we can also choose to limit it to 2d datacubes to ease the implementation.

[m-mohr]

Let's discuss later in the meeting.

[m-mohr]

This still needs clarification

[clausmichele]

what about:
"description": "A raster data cube with two spatial dimensions and arbitrary number of bands and times. If the datacube has multiple bands the process will be applied to all of them. Same for a timeseries."

Sorry I'm not good in writing this kind of descriptions.

[clausmichele]

I don't know why the build fails, it tells me that the word 'behavior' is misspelled

[m-mohr]

@clausmichele I guess it wants you to use 'behaviour'. Other than that it's failing because the JSON needs to be formatted with 4 spaces. Running the JSON through https://jsonformatter.org/ with 4 Tab Spaces should make it work.

[m-mohr]

This still needs clarification

[m-mohr]

I completely lost track in this PR of what still needs to be done. Too many comments and commits. Let's merge this into a new branch here and make a new PR for more detailed discussions that then merges into draft.