How to create an efficient group by function without mutation?

Question

Is there a way to efficiently implement a group by function without mutation?

Naive implementation:

ReScript playground
JavaScript (see below)

var messages = [
  {insertedAt: "2021-01-10"},
  {insertedAt: "2021-01-12"},
  {insertedAt: "2021-01-13"},
  {insertedAt: "2021-01-13"},
  {insertedAt: "2021-01-13"},
  {insertedAt: "2021-01-14"},
  {insertedAt: "2021-01-15"},
  {insertedAt: "2021-01-15"},
  {insertedAt: "2021-01-16"},
  {insertedAt: "2021-01-17"},
  {insertedAt: "2021-01-17"},
  {insertedAt: "2021-01-17"},
  {insertedAt: "2021-01-18"},
  {insertedAt: "2021-01-18"},
]

var messagesGroupedByDate = messages.reduce(function (data, message) {
  if (
    data.some(function (point) {
      return point.date === message.insertedAt;
    })
  ) {
    return data.map(function (point) {
      if (point.date === message.insertedAt) {
        return {
          date: point.date,
          count: (point.count + 1) | 0,
        };
      } else {
        return point;
      }
    });
  } else {
    return data.concat([
      {
        date: message.insertedAt,
        count: 1,
      },
    ]);
  }
}, []);


console.log(messagesGroupedByDate);

For the sake of argument, there's no need to make this more generic. The problem I'm facing is that I'm looping three times:

once with Array.prototype.reduce which is necessary to loop over messages
once with Array.prototype.some to see if the date key already exists in the resulting array
in the case where a date key already exists, we loop again with Array.prototype.map to update a specific element of an array
otherwise, a new array is returned that contains the new element

If there's not really any good way to make this efficient in ReScript, then I can always use raw JavaScript for this function, but I'm curious if it's possible to do this efficiently without mutation.

user_9473764 · Answer

You can group more simply by building an object of counts by date, and then using Object.entries and Array.map to convert that to an array of objects as required:

var messages = [
  {insertedAt: "2021-01-10"},
  {insertedAt: "2021-01-12"},
  {insertedAt: "2021-01-13"},
  {insertedAt: "2021-01-13"},
  {insertedAt: "2021-01-13"},
  {insertedAt: "2021-01-14"},
  {insertedAt: "2021-01-15"},
  {insertedAt: "2021-01-15"},
  {insertedAt: "2021-01-16"},
  {insertedAt: "2021-01-17"},
  {insertedAt: "2021-01-17"},
  {insertedAt: "2021-01-17"},
  {insertedAt: "2021-01-18"},
  {insertedAt: "2021-01-18"},
];

var messagesGroupedByDate = Object.entries(
  messages.reduce((data, message) => {
    data[message.insertedAt] = data[message.insertedAt] || 0;
    data[message.insertedAt]++;
    return data;
  }, {})
).map(([date, count]) => ({ date, count }));

console.log(messagesGroupedByDate);

You can also create an object directly from the values in messages and then update the counts by looping over messages:

var messages = [
  {insertedAt: "2021-01-10"},
  {insertedAt: "2021-01-12"},
  {insertedAt: "2021-01-13"},
  {insertedAt: "2021-01-13"},
  {insertedAt: "2021-01-13"},
  {insertedAt: "2021-01-14"},
  {insertedAt: "2021-01-15"},
  {insertedAt: "2021-01-15"},
  {insertedAt: "2021-01-16"},
  {insertedAt: "2021-01-17"},
  {insertedAt: "2021-01-17"},
  {insertedAt: "2021-01-17"},
  {insertedAt: "2021-01-18"},
  {insertedAt: "2021-01-18"},
]

const data = Object.fromEntries(messages.map(({ insertedAt: date }) => [ date, 0 ]));

messages.forEach(({ insertedAt: date }) => data[date]++);

const messagesGroupedByDate = Object.entries(data).map(([date, count])=> ({date, count}));

console.log(messagesGroupedByDate);

user_7087502 · Answer

Just add data to a Map() and then convert to array, then to object. It doesn't mutate anything as per your request.

We may simplify this even more but it is 5:00 AM right now and my brain is asleep now.

var messages = [
  {insertedAt: "2021-01-10"},
  {insertedAt: "2021-01-12"},
  {insertedAt: "2021-01-13"},
  {insertedAt: "2021-01-13"},
  {insertedAt: "2021-01-13"},
  {insertedAt: "2021-01-14"},
  {insertedAt: "2021-01-15"},
  {insertedAt: "2021-01-15"},
  {insertedAt: "2021-01-16"},
  {insertedAt: "2021-01-17"},
  {insertedAt: "2021-01-17"},
  {insertedAt: "2021-01-17"},
  {insertedAt: "2021-01-18"},
  {insertedAt: "2021-01-18"},
];

const mapped = new Map();

messages.forEach(message => {
    // if date already seen before, increment the count
    if (mapped.has(message.insertedAt)) {
        const count = mapped.get(message.insertedAt);
        mapped.set(message.insertedAt, count+1);
    } else {
        // date never seen before, add to map with initial count
        mapped.set(message.insertedAt, 1);
    }
});

const msgArr = Array.from(mapped);

const final = msgArr.map(([date, count])=> ({date, count}));

console.log(final);

user_1244884 · Answer

You need a map as an intermediate data structure:

{"2021-01-18": 2, /*…*/}

Then you deconstruct it into pairs and remap the pairs into objects:

const count =
  xs =>
    Object.entries(
      xs.reduce((acc, {insertedAt: k}) =>
        (acc[k] = (acc[k] ?? 0) + 1, acc), {}))
          .map(([k, v]) => ({date: k, count: v}));

// similar to count above but 100% pure
const count_pure =
  xs =>
    Object.entries(
      xs.reduce((acc, {insertedAt: k}) =>
        ({...acc, [k]: (acc[k] ?? 0) + 1}), {}))
          .map(([k, v]) => ({date: k, count: v}));


console.log(count(messages));
console.log(count_pure(messages));

It is common when doing this to use the spread operator … but depending on how many items you need to process, this may very quickly turn out to be inefficient. See gist I wrote on the subject https://gist.github.com/customcommander/97eb4b3f1600773db59406d39f3f9cd7

user_11450873 · Answer

Though the question is little old, i thought i will share my code for future readers. The below code is written in rescript and completely immutable because I have used Immutable Map from rescript.

type message = {insertedAt: string}

let messages = [
  {insertedAt: "2021-01-10"},
  {insertedAt: "2021-01-12"},
  {insertedAt: "2021-01-13"},
  {insertedAt: "2021-01-13"},
  {insertedAt: "2021-01-13"},
  {insertedAt: "2021-01-14"},
  {insertedAt: "2021-01-15"},
  {insertedAt: "2021-01-15"},
  {insertedAt: "2021-01-16"},
  {insertedAt: "2021-01-17"},
  {insertedAt: "2021-01-17"},
  {insertedAt: "2021-01-17"},
  {insertedAt: "2021-01-18"},
  {insertedAt: "2021-01-18"},
]

// Map Values
// Reduce into an immutable map
// Convert to tuple Array
// Log it

messages
->Belt.Array.reduce(Belt.Map.String.empty, (m, v) =>
  // Here Every set is creating a new map
  m->Belt.Map.String.set(v.insertedAt, m->Belt.Map.String.getWithDefault(v.insertedAt, 0) + 1)
)
->Belt.Map.String.toArray
->Js.log

Run In Rescript Playground. More on Immutable Map in rescript here.

Output:

[ [ '2021-01-10', 1 ],
  [ '2021-01-12', 1 ],
  [ '2021-01-13', 3 ],
  [ '2021-01-14', 1 ],
  [ '2021-01-15', 2 ],
  [ '2021-01-16', 1 ],
  [ '2021-01-17', 3 ],
  [ '2021-01-18', 2 ] ]

How to create an efficient group by function without mutation?

4 Answers