_time column needed to get a result

1

I don’t understand the observed behavior with Influx 2.6.1.

This query also returns 5 rows with column “_value”:

from(bucket: "halfyear")
  |> range(start: -2m) // we know that data is flushed per-minute
  |> filter(fn: (r) => r["_measurement"] == "deviceStatus")
  |> keep(columns: ["_value"])
  |> yield(name: "network")

This query also returns 5 rows, with columns “name”, “_value”, and “_time”:

from(bucket: "halfyear")
  |> range(start: -2m) // we know that data is flushed per-minute
  |> filter(fn: (r) => r["_measurement"] == "deviceStatus")
  |> map(fn: (r) => ({ name: "Device", _value: if r._value == 0 then "fine" else if r._value == 1 then "warning" else "error", _time: r._time }))
  |> yield(name: "network")

This query returns no result!

from(bucket: "halfyear")
  |> range(start: -2m) // we know that data is flushed per-minute
  |> filter(fn: (r) => r["_measurement"] == "deviceStatus")
  |> map(fn: (r) => ({ name: "Device", _value: if r._value == 0 then "fine" else if r._value == 1 then "warning" else "error" }))
  |> yield(name: "network")

The only difference is that we omit _time column because it is not needed.
Is this a bug or expected behavior?

Thanks!

2

Hello @ypnos,
Hmmmm wtf that makes no sense.
Weird…can you please create an issue?

Out of curiosity does this work?

from(bucket: "halfyear")
  |> range(start: -2m) // we know that data is flushed per-minute
  |> filter(fn: (r) => r["_measurement"] == "deviceStatus")
  |> map(fn: (r) => ({ name: "Device"}))
  |> map(fn: (r) => ({ _value: if r._value == 0 then "fine" else if r._value == 1 then "warning" else "error"}))
  |> map(fn: (r) => ({ _time: r._time }))
  |> yield(name: "network")

Maybe try commenting out each column respectively to try and find the culprit?
Thank you

3

Only this version produces results:

from(bucket: "halfyear")
  |> range(start: -2m) // we know that data is flushed per-minute
  |> filter(fn: (r) => r["_measurement"] == "deviceStatus")
  //|> map(fn: (r) => ({ name: "Device"}))
  //|> map(fn: (r) => ({ _value: if r._value == 0 then "fine" else if r._value == 1 then "warning" else "error"}))
  |> map(fn: (r) => ({ _time: r._time }))
  |> yield(name: "network")

It appears like it really hates a row out of a map() without _time. While it is o.k. to get rid of _time via keep(). So this also works:

from(bucket: "halfyear")
  |> range(start: -2m) // we know that data is flushed per-minute
  |> filter(fn: (r) => r["_measurement"] == "deviceStatus")
  //|> map(fn: (r) => ({ name: "Device"}))
  //|> map(fn: (r) => ({ _value: if r._value == 0 then "fine" else if r._value == 1 then "warning" else "error"}))
  |> map(fn: (r) => ({ _time: r._time, name: "Device" }))
  |> keep(columns: ["name"])
  |> yield(name: "network")

I will file a bug report.

4

@ypnos, Where are you running the queries? In the Data Explorer?

5

Yes.

Sorry for late reply.