↓Skip to main content

Printing Tables in the Terminal

bash
Table of Contents

With the column utility, you can write scripts that display outputs similar to kubectl. If you couple this with jq or yq, then you can translate ugly pages of JSON to pretty and readable (and highly customisable) tables.

Example: list pods with mounted PVCs #

Script #

kubectl get pods -A -ojson |
jq -r '
    ["NAMESPACE", "NAME", "PVC"],
    (
        .items[] |
        select(.spec.volumes[].persistentVolumeClaim) |
        [
            .metadata.namespace,
            .metadata.name,
            (.spec.volumes[].persistentVolumeClaim.claimName | select(.))
        ]
    ) |
    @tsv' |
column -t

Output #

NAMESPACE  NAME                       PVC
minio      minio-0                    minio-0
jenkins    jenkins-0                  jenkins
mytests    test-app-56bdb5cb59-t5f6m  test-app
vault      vault-0                    data-vault-0

Breakdown #

kubectl get pods -A -ojson gets all the cluster’s pods and displays them in JSON format.

jq -r receives that JSON stream from kubectl and runs it against the filters (the multiline, single-quoted string).

["NAMESPACE", "NAME", "PVC"], is where we define our table column headers. The trailing comma means were creating a new output “object”.

.items[] iterates over every pod listed by kubectl. When you list resources by type (kubectl get RESOURCE ...), it will return a list object, and every list object has an items field.

select(.spec.volumes[].persistentVolumeClaim) filters pods that have a PVC listed in its volume list. When JQ sees an empty volumes or a non-existent .persistentVolumeClaim, it generates a null value, which is falsy.

[.metadata.namespace, .metadata.name, ...] is where we have the equivalent of a JS or Haskell map function that transforms a pod object to an array—a row in our table.

(.spec.volumes[].persistentVolumeClaim.claimName | select(.)), similar to the earlier select(...) filter, we get all the PVCs as defined by the pod. The trailing | select(.) removes null objects. When combined with the array (the table row), this flattens out to multiple rows. So for example, if the pod has N PVCs, then this filter will generate N rows.

@tsv formats the arrays to TSV format. All of the left-hand side inputs need to be arrays—our column headers and every generated row entry.

column -t is the final stage of this pipeline. jq displays a TSV output, but is not aligned. The column program takes all of those outputs and reformats them by padding the columns as needed, producing the pretty table you saw earlier.

This is what the table looks like if we remove column -t in the pipeline:

NAMESPACE	NAME	PVC
minio	minio-0	minio-0
jenkins	jenkins-0	jenkins
mytests	test-app-56bdb5cb59-t5f6m	test-app
vault	vault-0	data-vault-0