Customize generated label names in time series visualization

Grafana Dashboards
1

Using Grafana OSS 11.5.1 I have this Postgres query:

SELECT 
    b.date_booked AS "time",
    s."name" as service_name,
    COUNT(*) AS metric
FROM 
    bookings b
JOIN 
    services s ON b.service_id = s.id
WHERE 
    -- $__timeFilter(b.date_booked) and
    b.status != 'X'
GROUP BY 
    b.date_booked, s."name"
ORDER BY 
    b.date_booked, metric DESC;

Using Transformations → Prepare Time Series → Multi-frame time series I get almost the output I want, with one metric per service displayed in the graph. The labels however look like this:

total_bookings Escalade Summer

… and I cannot figure out how to strip the "total_bookings " prefix??

Rename fields by regex apparently only works on raw query fields, not on the fields generated by the Prepare time series transformation.
Field override → Display Name supports regex bu apparently not back references to capture groups.

What is the right way to go about this?

Thanks!

  • What Grafana version and what operating system are you using?
    OSS 11.5.1

  • What are you trying to achieve?
    Display multiple time series on the Time Series visualization, with dynamic labels coming from the DB

  • How are you trying to achieve it?
    Using Prepare Time Series transformation to create a multi-frame time series, then a Rename field by regex to rename the new fields

  • What happened?
    The transformation is not applied to generated fields.

  • What did you expect to happen?
    The regex transform being applied to the fields generated by the Prepare time series step.

2

Hi @franck102

Rename fiends by regex transformation like this should work:
Match: ^total_bookings\s(.+)
Replace: $1

In action…

Data with no transformation:

image
image517×386 30.7 KB

and after applying this:

image
image358×116 5.27 KB

image
image517×386 30.9 KB

1 Like
3

Indeed, I swear I tried the transformation and somehow it didn’t affect my labels.

Thanks!

4

I believe you. Regex is really finicky and just the slightest omission or extra space will render nothing. Glad you are off and running now.

5

another way of doing it. You have the metric and value swapped. then no need for transformation if you choose Time Series Format.

image
image514×485 25.7 KB

SELECT 
    date_booked AS "time",
    "name" as metric,
    COUNT(*) AS value
FROM bookings b
1 Like
6

My keyboard is set up as US international, and the caret is a weird non-ascii character waiting to change into an ê or î until I hit space… that’s what tripped me :frowning:

7

Thanks for this, lots of things coming together thanks to your post, I had been really struggling with time series & Postgres (no timescaledb) so far.
The documentation didn’t help much, the page that describes time series (here) only has examples with multiple values on the same row.

Franck