Jq: removal of nodes, by naming nodes to excise

[paul-hammant]

Input {"foo": 42, "bar": "less interesting data", "baz": "sdkjhsdf"}
Output {"foo": 42, "baz": "sdkjhsdf"}

The jq expression would cite bar, but not foo or baz. Possibly overloading the minus char somehow.

[stedolan]

Yes, this is definitely a missing feature. Options include:

1. . - "bar"
2. . - ["bar"]
3. remove("bar")
4. .bar = empty

I think 1 or 2 might be what you suggested. I'm not sure which I prefer yet, but I'm kind of partial towards 4 since it fits with the current meanings of assignment and empty, rather than adding a new construct.

[paul-hammant]

you're the domain expert, choose the most appropriate solution :)

[stedolan]

If you build from master, you can now do this:

del(.bar)

It's not documented yet, and the exact semantics of what happens when you replace .bar with something more complicated will probably change, but that should solve the problem.

[paul-hammant]

That's great dude. Is it recursive too ?

[stedolan]

Now I'm happier with what del(complicated_thing) does. You can del() all sorts of junk:

• del(.foo,.bar,.baz) deletes those three keys
• del(.foo[0].bar.baz) removes the baz key from the object at .foo[0].bar
• del(.foo[0,1,2]) deletes the first three items of the array at field "foo"
• del(.foo[] | select(.score < 20)) deletes items in the array at field "foo" whose "score" field is below 20.

Not quite sure what you meant by "recursive", though.

[paul-hammant]

By recursive, I mean that it's able to remove a node that repeats throughout a document structure. For example, Reddit's articles go fairly deep with repeating element/attr names:

• http://www.reddit.com/r/test/comments/vdbmn/test_123.json

[stedolan]

Ah. You can in fact do this. First you need a means to recurse down to all of the children of a node, which you can do by defining a recursive function:

def all_children: . , (.data.children | select(. != null) | .[] | all_children);

Putting a node through all_children returns the node itself and all child nodes. Then, to get all of the author fields from any node in your dataset, you'd need:

.[] | all_children | .author

You can then delete all of the author fields with:

del(.[] | all_children | .author)

Defining a new recursive function each time you want to do this is effort, so I've added a recurse function to the standard library, which is defined as:

def recurse(f): . , (f | select(. != null) | recurse(f));

The program to delete all author fields anywhere in the document now becomes:

del(.[] | recurse((.data.children // []) | .[]) | .author)

(foo // [] returns [] if foo is null, or foo otherwise.)