Rescale collection of timeseries by their first values

[gio]

Hello,

I’m basically trying to achieve what’s explained in this answer:

Normalized measurement starting at value 1

I want to normalize a measurement such that it starts by value 1. For a fixed starting point this is no problem using the code: start_value = 12930.0 from(bucket: "finance") |> range(start: v.timeRangeStart, stop: v.timeRangeStop) |> filter(fn: (r) => r["_measurement"] == "price") |> filter(fn: (r) => r["Stock_Name"] == "DAX 30") |> filter(fn: (r) => r["ISIN"] == "DE0008469008") |> map(fn: (r) => ({ r with _value: r._value / start_value})) But I don’t now how to code a dynamical rang…

with the only difference that I would like to rescale a collection of series labelled by tags, not just one single series. I tried throwing a group() in the pipeline but obviously it’s not that easy.

I also wonder if there’s a simple trick to rescale, maybe a multiplicative version of: data |> difference() |> cumulativeSum() used to shift the whole series by its first value. This would be something like: data |> 'ratio()' |> 'cumulativeProd()'.

Thanks!

[gio]

I ended up implementing the multiplicative version of “diff |> cumsum”. I’ll put it here in case anyone is interested. Essentially, it’s a good substitute for the method by @scott that I quoted in my question.

import "math"

logarithm = (tables=<-) =>
  tables
    |> map(fn: (r) => ({
        r with
        _value: math.log(x: r._value)
      })
    )

exponential = (tables=<-) =>
  tables
    |> map(fn: (r) => ({
        r with
        _value: math.exp(x: r._value)
      })
    )

data
  |> group(columns: ["mytag"])
  |> aggregateWindow(every: 1d, fn: mean) // to align the series in time
  |> logarithm()
  |> difference()
  |> cumulativeSum()
  |> exponential()