Preprocessor feature request: Common mask for multiple datasets

[BenMGeo]

For multiple datasets it is necessary to produce a common mask and apply it to all of them if you want to compare respective results.

As far as I know, there is currently no preprocessor for producing and applying such a common mask.

Is there a work-around? Is this a needed/wanted feature? If I'm the only one, I can focus on doing this in the diagnostic.

[BenMGeo]

I had an offline discussion with @axel-lauer and he pointed me to the threshold functionality.

mask_fillvalues:
        threshold_fraction: 0.0

This does the trick, but seems rather counter-intuitive, in my eyes.

Do we want to have a wrapper here?

[bouweandela]

I would prefer better documentation over a wrapper.

[BenMGeo]

Besides, is there any need for a dataset based threshold functionality? One might think about splitting the current combined version into 2 seperate steps?

[valeriupredoi]

I just finished writing the documentation for the preprocessor (including mask_fillvalues) have a look at #113

you can throttle the threshold_fraction indeed and that will combine the masks into one cumulative one, cheers for raising the issue since I can include a ProTip to the documentation.

From a functional point of view the fillvalues mask was one of the first preprocessor we implemented, and is rather important, combining masks without filling is of course a special case of it and I don't think there is a need for yet another masking operation since this one does it a-ok 🍺

[valeriupredoi]

also, the threshold is important for filling the missing values, mask_fillvalues is something that @mattiarighi bought me a beer for implementing, so it's solid 😁

[valeriupredoi]

just added a note in #113 to the preprocessor documentation documenting this, cheers @BenMGeo 🍺

[valeriupredoi]

can we close this then?

[BenMGeo]

If it's just me, who thinks it's an odd way to do/call it, I'm fine with leaving it and closing. I'm having a look at #113 right now.

Thanks @valeriupredoi for implementing it.

[BenMGeo]

Solved via #113.