Premise

Converting between different data types in JavaScript is easier than many languages because of the concise and declarative JSON. However, currently both imperative and functional approaches create very verbose code.

For example, given some data such as:

{
  "person": {
    "id": 7,
    "name": "bob",
    "tasks": [
      { "id": 11 },
      { "id": 23 },
      { "id": 99, "optional": [ "a", "b", "c" ] }
    ],
    "interests": [
      { "type": "sport", "name": "football" },
      { "type": "sport", "name": "rugby" },
      { "type": "sport", "name": "swimming" },
      { "type": "entertainment", "name": "tv" }
    ],
    "paymentHistory": [
      { "date": "today", "amount": "£10" },
      { "date": "yesterday", "amount": "£20" },
      { "date": "last week", "amount": "£30" }
    ]
  }
}

And desiring data in this format:

{
  "id": 7,
  "name": "bob",
  "lastPaymentAmount": "£10",
  "taskIds": [11, 23, 99],
  "taskIdsWithOptionalData": [99],
  "activities": {
    "sport": ["football", "rugby", "swimming"],
    "entertainment": ["tv"]
  }
}

You might write code like this:

var newData = {
  id: data.person.id,
  name: data.person.name,
  lastPaymentAmount: data.person.paymentHistory[0].amount,
  taskIds: [],
  taskIdsWithOptionalData: [],
  activities: {}
};
for (var i = 0; i < data.person.tasks.length; i++) {
  newData.taskIds.push(data.person.tasks[i].id);
  if (data.person.tasks[i].optional) {
    newData.taskIdsWithOptionalData.push(data.person.tasks[i].id);
  }
}
for (var it = 0; it < data.person.interests.length; it++) {
  if (newData.activities[data.person.interests[it].type]) {
    newData.activities[data.person.interests[it].type].push(data.person.interests[it].name);
  } else {
    newData.activities[data.person.interests[it].type] = [ data.person.interests[it].name ];
  }
}

Or this:

var newData = {
  id: data.person.id,
  name: data.person.name,
  lastPaymentAmount: data.person.paymentHistory[0].amount,
  taskIds: _.pluck(data.person.tasks, 'id'),
  taskIdsWithOptionalData: _.chain(data.person.tasks).filter(function (t) { return t.optional; }).pluck('id').value(),
  activities: _.chain(data.person.interests).groupBy(function (i) { return i.type; } ).map(function (interests, interestType) { return [interestType, _.pluck(interests, 'name')]; }).object().value()
};

Both of these approaches are time-consuming to write and there is no need for this to be the case.

I propose a simpler way of defining a transformation between one json structure and another in json itself.

For example, given:

{
  "id": "person/id",
  "name": "person/name",
  "lastPaymentAmount": "person/paymentHistory[0]/amount",
  "taskIds": "person/tasks[*]/id",
  "taskIdsWithOptionalData": j.sel("person/tasks[:hasOptionalData]/id", { hasOptionalData: has('optional') }),
  "activities": {
    "person/interests[*]/type": "person/interests[k]/name"
  }
}

Which would then output the desired data:

{
  "id": 7,
  "name": "bob",
  "lastPaymentAmount": "£10",
  "taskIds": [11, 23, 99],
  "taskIdsWithOptionalData": [99],
  "activities": {
    "sport": ["football", "rugby", "swimming"],
    "entertainment": ["tv"]
  }
}