Feature request: allow custom join implementations

[andig]

My use case is comparing daily PV data with averages data across the previous week.

One way to do this writing one query per day, adjusting timestamps and joining the tables. However the result- new table with duplicate field names with suffix per series is hard to work with, especially if not all series are always populated (null values/ empty windows).

Instead, it would seem plausible and straight forward to allow a custom join function based on an array of input rows, somewhat like this:

join(
    tables: {d1:d1, d2:d2, d3:d3, ...}, 
    on: ["_time"],
    fn: (tables, r) => (
        sum = 0
        count = 0
        for (i=0; i<tables.length, i++) {
            sum += tables[i]._value
            count++
        }
        if (count > 0) {
            r.value = sum/count
        }
    )
)

Implementation detail would be if time and window columns should already be pre-populated inside the function or left to the custom implementation. This would be orthogonal to #84.

Would this be helpful? Is the use case better suited for using other functions or would it rather make sense to combine timeshift and window(aggregate) into a new higher-order function?

[github-actions]

This issue has had no recent activity and will be closed soon.