Use processed string to refer as column name in Flux

Question

How to use refer column names using r.sourceProcessed and r.targetProcessed? The processed variables are exact same as the column name.

sourceProcessed = processStatusName(status: source)
targetProcessed = processStatusName(status: target)

from(bucket: bucket)
  |> range(start: startRange, stop: now())
  |> filter(fn: (r) =>  r._field == sourceProcessed or r._field == targetProcessed)
  |> pivot(rowKey:["_time"], columnKey: ["_field"], valueColumn: "_value")
  |> filter(fn: (r) => r.sourceProcessed != "" and r.targetProcessed != "")
  |> map(fn: (r) => {
      diffSeconds = float(v: int(v: r.targetProcessed) - int(v: r.sourceProcessed))
      time_diff = diffSeconds / 86400.0 
      return {_time: r._time, time_diff: time_diff}
    })
  |> group(columns: ["_start", "_stop"])
  |> mean(column: "time_diff")
  |> yield(name: "mean_time_diff")

This issue occurs only in bottom part. Following query works without throwing any errors.

from(bucket: bucket)
  |> range(start: startRange, stop: now())
  |> filter(fn: (r) =>  r._field == sourceProcessed or r._field == targetProcessed)
  |> pivot(rowKey:["_time"], columnKey: ["_field"], valueColumn: "_value")

Any ideas to fix this? TIA!