README.md

jCue

jCue is a horrible way to use CUE as if it were jq. Well, for (relatively) statically patched structured data, at least.

jCue exposes a single definition: #Patch. This definition has a couple of fields:

• #Input: put your source data here
• #Output: get your patched data back from here

A "patch" is simply a struct containing data that you'd like to take precedence over the source data, in exactly the same shape as the source data. It can contain a subset of the source data structure, and only those fields present in the patch are overridden - the rest of the source data is left unchanged.

To apply a patch, simply unify it with the #Output field:

output: patch & {jCue.#Patch & {#Input: input}}.#Output

Example

See demo.cue:

package demo

import "jonathanmatthews.com/cuetils/cue/jCue"

input: {
	x: 1
	y: "string in input"
	z: {
		a: 56
	}
	b: [{
		name: "foo"
		val:  10
	}, {
		name: "string in input"
		val:  20
	},
	]
}

patch: {
	x: 2
	y: "string in patch"
	b: [_,
		{name: "string in patch"},
	]
}

output: patch & {jCue.#Patch & {#Input: input}}.#Output

Run this demo as follows:

$ cue export .:demo -e output
{
    "x": 2,
    "y": "string in patch",
    "z": {
        "a": 56
    },
    "b": [
        {
            "name": "foo",
            "val": 10
        },
        {
            "name": "string in patch",
            "val": 20
        }
    ]
}

Caveats

1. This isn't /exactly/ a joke, but don't use it in production.
   Or do; I'm a README, not a cop.
2. Your source data and patch can't contain any CUE defaults
3. Performance is ok for small source data, but it will rapidly approach "very very poor" as source data size grows

Bugs

Yes.