jq .[] | select and test to FilterThe more I learn about jq — and see its use in the field — the
more I realize it is by far the most important utility of the modern
cloud era. Everything is JSON these days (even though you might think
it is YAML). Even ProtoBuf ultimately ends up as JSON when being parsed
and queried from the command line. In short, jq is sed and awk
combined for anything with JSON. A lot of people know how to use basic
jq for regular things like looping through data, but only recently
I realize that an entire 20 line shell script I wrote some months back
to filter some resources with different labels could have been done in
one very powerful jq line. I’m never making that mistake again.
We have a label on all our namespaces at work called nstype which is
set to user or system depending on the type of namespace, pretty
typical stuff. Here’s the one line command to show all user labels:
k get ns -o json | jq -r '.items[].metadata | select (.labels.nstype == "user") | .name'
That’s all it takes to get a perfect, one-column list. To change to
system just change "user".
Direct matching is cool and all, but when you combine this with test()
regular expressions (PCRE, or course) you realize this is all that you
ever need. Say goodbye to awk and sed for this kind of stuff.
Here’s an example of everything that starts with the letter r:
k get ns -o json | jq -r '.items[].metadata | select (.labels.nstype == "user") | select(.name|test("^r") | .name'
Which you could have also done exactly the same way with just grep -P:
k get ns -o json | jq -r '.items[].metadata | select (.labels.nstype == "user") | .name' | grep -P ^r
It’s just nice to know you have everything you need in jq, which
really doesn’t violate the UNIX philosophy with all this. Its “one
thing” it does well is handle JSON, and it is by far the best at it.
#json #jq #tips #devops #coding #shell #bash