Histogram process

[jdries]

In some of our zonal statistics use cases, we compute histograms. I believe this is basically a type of reducer.
When the input data cube uses a type with a limited range, like byte, we compute a histogram with 256 buckets, and keep track of nodata. For float/double data types, we might need user definded buckets.

[m-mohr]

Could you come up with a concrete JSON-based proposal? If it's required for the four core use cases it should be in the 0.4 release.

[jdries]

I don't think it needs to be in 0.4!

[m-mohr]

@jdries Could you come up with a concrete JSON-based proposal for the next version?

[m-mohr]

Telco: Agreed to add. Not sure whether it's a reducer though, depends on the outcomes of #73, I guess. If there's a volunteer to help define this function, I'd appreciate it.

[bossie]

Histogram as a reducer process with discrete buckets might look like this: #97. No user-defined buckets are supported yet.

[m-mohr]

Seems to be relevant for UC3 (Platform), so giving it some priority. @jdries

[m-mohr]

Would this really go through reduce_dimension? I think it might need to go through apply_dimension (with a target_dimension?) instead as the histogram returns more than a single value, right?