update: The issue was incorrectly described at first. I have completely rewritten the description, along with sharing a code snippet that may not be aesthetically pleasing but gets the job done to some extent.
Imagining an object:
const input = {
a: 1,
b: '2',
c: {
d: true,
e: '4'
},
f: [{
g: 5,
h: {
i: '6'
}
}, {
g: 7,
h: {
i: '8'
}
}]
}
My goal is to create a collection of all potential arrangements of nested arrays, where keys of the object are flattened and joined with ".", such as:
[{
a: 1,
b: '2',
'c.d': true,
'c.e': '4',
'f.g': 5,
'f.h.i': '6'
}, {
a: 1,
b: '2',
'c.d': true,
'c.e': '4',
'f.g': 7,
'f.h.i': '8'
}]
Note that non-primitive values for keys like 'f.h' pointing to an object are not included in this scenario.
To achieve this, I start by gathering all keys and adding a "#" sign to array item keys to signify "every index in that array":
function columns(data, prefix = '') {
if (_.isArray(data)) {
return columns(_.first(data), `${prefix}.#`);
} else if (_.isObject(data)) {
return _.filter(_.flatMap(_.keys(data), key => {
return _.concat(
!_.isObject(_.result(data, key)) ? `${prefix}.${key}` : null,
columns(data[key], `${prefix}.${key}`)
);
}));
} else {
return null;
}
}
console.log(columns(input)); // -> [".a", ".b", ".c.d", ".c.e", ".f.#.g", ".f.#.h.i"]
Using lodash, I flatten the object into a single-level object with unconventional keys:
function flattenKeys(original, keys) {
return _.mapValues(_.groupBy(_.map(keys, key => ({
key,
value: _.result(original, key)
})), 'key'), e => _.result(e, '0.value'));
}
console.log(flattenKeys(input, columns(input))) // -> {".a":1,".b":"2",".c.d":true,".c.e":"4"}
By traversing through every array-like property of the original object, I produce an array of objects setting keys like .f.#.h.i with values from .f.0.h.i for the first element, and so on:
function unfold(original, keys, iterables) {
if (!_.isArray(iterables)) {
return unfold(original, keys, _.uniq(_.map(_.filter(keys, key => /#/i.test(key)), key => _.replace(key, /\.\#.*/, '')));
} else if (_.isEmpty(iterables)) {
return [];
} else {
const first = _.first(iterables);
const rest = _.tail(iterables);
const values = _.result(original, first);
const flatKeys = _.mapKeys(_.filter(keys, key => _.includes(key, first)));
const updated = _.map(values, (v, i) => ({
...flattenKeys(original, keys),
..._.mapValues(flatKeys, k => _.result(original, _.replace(k, /\#/, i)))
}));
return _.concat(updated, unfold(original, keys, rest));
}
}
console.log(unfold(input, columns(input))) // -> [{".a":1,".b":"2",".c.d":true,".c.e":"4",".f.#.g":5,".f.#.h.i":"6"},{".a":1,".b":"2",".c.d":true,".c.e":"4",".f.#.g":7,".f.#.h.i":"8"}]
In conclusion, although the code may appear messy, cleaning up the keys isn't essential in this case.
The main concern remains how to adapt this solution for scenarios involving multiple array-like properties in original objects.
Upon reflection, it seems that this question would be better suited for CodeReview StackExchange. If someone decides to move it there, I am perfectly fine with that decision.