Previously, I showed how to combine two Prometheus queries into a single table in Grafana, and how to format the table with data links so that the entire table doesn’t turn into a giant HTML link by using a mix of Grafana’s data transformations and field overrides.
There’s actually a much cleaner way to do table transformations, if you’re using a recent enough version of Grafana with optional features enabled: SQL expressions. Grafana added these somewhere in the middle of Grafana 12 as an optional feature, and I think they’re enabled by default in Grafana 13.
With SQL Expressions, you can issue SQL queries against the results of other queries. For my Kubernetes CPU and RAM example, instead of 3 transforms we could just use a query like this:
SELECT A.namespace, A.pod, A.container, A.__value__ as CPU, B.__value__ as Memory FROM A, B WHERE A.namespace = B.namespace and A.pod = B.pod and A.container = B.container ORDER by CPU desc Obviously, you need to be familiar enough with SQL to be able to whip up queries like this, but if you are (and rather a lot of people are!), then it’s more flexible than using transforms and less surprising.

scottstuff.net

Grafana's SQL Expressions feature (available in Grafana 12 as optional, default in Grafana 13) lets you run SQL queries directly against the results of other queries in a panel. This replaces multiple data transformations with a single SQL JOIN, making it easier to combine Prometheus metrics like CPU and RAM into a unified table. You can also add extra columns for data links via the SELECT clause, eliminating the need for most transforms while retaining field overrides for formatting.

Grafana Table Joins With SQL Expressions