tabloid is a weekend project. The goal is to be able to parse inputs from several command line applications like kubectl and docker that use a tabwriter to format their output: this is, they write column-based outputs where the first line is the column title -- often uppercased -- and the values come below, and they're often perfectly aligned.
Here's an example: more often than not though, you want one field from that output instead of a tens-of-lines long output. So your first attempt is to resort to grep:
$ kubectl get pods --all-namespaces | grep frontend
team-a-apps frontend-5c6c94684f-5kzbk 1/1 Running 0 8d
team-a-apps frontend-5c6c94684f-k2d7d 1/1 Running 0 8d
team-a-apps frontend-5c6c94684f-ppgkx 1/1 Running 0 8dYou have a couple of issues here:
- The first column disappeared, which holds the titles. I'm often forgetful and won't remember what each column is supposed to be. Maybe for some outputs, but not all (looking at you,
kubectl api-resources!) - There's some awkward space between the columns now, since the columns keep the original formatting.
We could fix the first issue by using awk instead of grep:
$ kubectl get pods --all-namespaces | awk 'NR == 1 || /frontend/'
NAMESPACE NAME READY STATUS RESTARTS AGE
team-a-apps frontend-5c6c94684f-5kzbk 1/1 Running 0 8d
team-a-apps frontend-5c6c94684f-k2d7d 1/1 Running 0 8d
team-a-apps frontend-5c6c94684f-ppgkx 1/1 Running 0 8dMuch better! Now if this works for you, you can stop reading here. Chances are, you won't need tabloid. But if you want:
- Some more human-readable filters than
awk - The ability to customize the columns' order
- The ability to filter with
ANDandORrules - Or filter using regular expressions
Then tabloid is the right tool for you. Here's an example:
# show only pods whose name starts with `frontend` or `redis`
$ kubectl get pods --all-namespaces | tabloid --expr 'name =~ "^frontend" || name =~ "^redis"'
NAMESPACE NAME READY STATUS RESTARTS AGE
team-a-apps frontend-5c6c94684f-5kzbk 1/1 Running 0 8d
team-a-apps frontend-5c6c94684f-k2d7d 1/1 Running 0 8d
team-a-apps frontend-5c6c94684f-ppgkx 1/1 Running 0 8d
team-a-apps redis-follower-dddfbdcc9-9xd8l 1/1 Running 0 8d
team-a-apps redis-follower-dddfbdcc9-l9ngl 1/1 Running 0 8d
team-a-apps redis-leader-fb76b4755-6t5bk 1/1 Running 0 8dOr better even...
# show only pods whose name starts with `frontend` or `redis`
# and only display the columns `namespace` and `name`
$ kubectl get pods --all-namespaces | tabloid \
> --expr '(name =~ "^frontend" || name =~ "^redis") && namespace == "team-a-apps"' \
> --column namespace,name
NAMESPACE NAME
team-a-apps frontend-5c6c94684f-5kzbk
team-a-apps frontend-5c6c94684f-k2d7d
team-a-apps frontend-5c6c94684f-ppgkx
team-a-apps redis-follower-dddfbdcc9-9xd8l
team-a-apps redis-follower-dddfbdcc9-l9ngl
team-a-apps redis-leader-fb76b4755-6t5bkOr we can also reorder the output:
# show only pods whose name starts with `frontend` or `redis`
# and only display the columns `namespace` and `name`, but reverse
$ kubectl get pods --all-namespaces | tabloid \
> --expr '(name =~ "^frontend" || name =~ "^redis") && namespace == "team-a-apps"' \
> --column name, namespace
NAME NAMESPACE
frontend-5c6c94684f-5kzbk team-a-apps
frontend-5c6c94684f-k2d7d team-a-apps
frontend-5c6c94684f-ppgkx team-a-apps
redis-follower-dddfbdcc9-9xd8l team-a-apps
redis-follower-dddfbdcc9-l9ngl team-a-apps
redis-leader-fb76b4755-6t5bk team-a-appsThe following features are available:
- Column titles are always on by default and their titles are normalized for querying with the expression language. Additionally, columns can be reordered.
- There's a powerful expression filtering with several additional built-in functions to handle specific filtering (like
kubectldurations or pod restart count). - Extra whitespaces (like the one that
awkorgrepcould produce) is removed automatically, and space count is recalculated.
The answer is "maybe". In short, I wanted to create a tool that serves my own purpose, with a quick and easy to use interface where I don't have to remember either cryptic languages or need to hack my way through to get the outputs I want.
While it's possible for kubectl, for example, to output JSON or YAML and have that parsed instead, I want this tool to be a one-size-fits-most in terms of column parsing. I build my own tools around the same premise of the 3+ tab padding and using Go's amazing tabwriter, so why not make this tool work with future versions of my own apps and potentially other 3rd-party apps?
Absolutely! This was a weekend project and really doesn't have much testing. Parsing columns might sound like a simple task, but you see, given the following input to the best tool out there to parse columns, awk, you'll see how quickly it goes wrong:
NAMESPACE NAME (PROVIDED) READY STATUS RESTARTS AGE
argocd argocd-application-controller-0 1/1 Running 0 8d
argocd argocd-dex-server-6dcf645b6b-nf2xb 1/1 Running 0 8d
argocd argocd-redis-5b6967fdfc-48z9d 1/1 Running 0 8d
argocd argocd-repo-server-7598bf5999-jfqlt 1/1 Running 0 8d
argocd argocd-server-79f9bc9b44-5fdsp 1/1 Running 0 8d
$ awk '{ print $2 }' pods-wrong-title.txt
NAME
argocd-application-controller-0
argocd-dex-server-6dcf645b6b-nf2xb
argocd-redis-5b6967fdfc-48z9d
argocd-repo-server-7598bf5999-jfqlt
argocd-server-79f9bc9b44-5fdsp
The name of the 2nd column is NAME (PROVIDED), yet awk parsed it as just NAME. awk is suitable for more generic approaches, while this tool works in harmony with tabwriter outputs, and as such, we can totally parse the column well:
$ cat pods-wrong-title.txt | tabloid --column name_provided
# or --column "NAME (PROVIDED)"
# or --column "name (provided)"
NAME (PROVIDED)
argocd-application-controller-0
argocd-dex-server-6dcf645b6b-nf2xb
argocd-redis-5b6967fdfc-48z9d
argocd-repo-server-7598bf5999-jfqlt
argocd-server-79f9bc9b44-5fdspBack to the point at hand though... Absolutely! Feel free to send any PRs you might want to see fixed/improved.