Any idea how to find easily values of (deeply) nested keys, when the keys not always exist?

[halloleo]

I have a JSON file like this:

{"details": {
  "sections": [
    {"fields": [
      { "x": "don't", "v": "Yes!" },
      { "x": "don't", "w": "don't" }
    ]},
    {"fields": [
      { "x": "don't", "v": "Indeed!" },
    ]}],
  "other-stuff": "don't"}
}

I need to extract the values to the "v" keys, i.e "Yes!" and "Indeed!" Because the "v" key is not always present, I do something like this:

cat data.json | jello -c '\
res = []
for s in _.details.sections:
    for f in s.fields:
        try:
            res.append(f.v)
        except KeyError:
            pass
res'

This works, but feels pretty cumbersome. Is there a way to gather the values easier then with my approach? Also it would be nice if I didn't need to know how deep my "v" keys are located.

[kellyjonbrazil]

Your solution works, but I agree it might be nicer to do it without a try/except block. One thing you can do when you don't know if a key exists is to use the in operator on the dictionary. It would look something like this:

cat data.json | jello -c '\
res = []
for s in _.details.sections:
    for f in s.fields:
        if "v" in f:
            res.append(f.v)
res'

The output should be:

["Yes!","Indeed!"]

To find the keys no matter how deep they are, you may need to use a recursive method. Fortunately, the -s option in jello happens to be a recursive function, so you can sort of hack a solution using -s with grep:

$ cat data.json | jello -s | grep ".v ="
.details.sections[0].fields[0].v = "Yes!";
.details.sections[1].fields[0].v = "Indeed!";

Then you would use normal text utilities (sed/awk/etc.) to grab the data. Or you could use that notation back in jello to get those values back into a list:

$ cat data.json | jello -c '\
res = []
res.append(_.details.sections[0].fields[0].v)
res.append(_.details.sections[1].fields[0].v)
res'

["Yes!","Indeed!"]

Happy JSON filtering!

[halloleo]

Thanks for your reply, @kellyjonbrazil. I like in particular your idea to utilise -s!

And I agree, this question is better placed here in the Discussions. Sorry!

[kellyjonbrazil]

No worries - any time! :)

[halloleo]

BTW, just wanted to mention: I really like the concept of jello!

For a while I have used jq on the command line but felt always unfamiliar with it; then I have used plain Python for mangling JSON, but getting the data from and to the command line is slightly cumbersome, particular because piping is difficult. - jello combines the two worlds beautifully!