mongez-collection-recipes

name

mongez-collection-recipes

description

Idiomatic multi-method pipelines for `@mongez/collection` — top-N (`where` + `sortByDesc` + `take` + `pluck`), pagination (`skip` + `take`), bucket-then-aggregate (`groupBy` + `map` + `sum`), partition for dual pipelines, dedupe-and-project (`uniqueList` + `pluck`), `tap` for side-effects mid-chain, safe sort via `clone`, chunk-based batch processing, Map → collection conversion, lookup index via `reduce`, and intentional queue draining via `shift`. TRIGGER when: user asks "show me a worked example with @mongez/collection", "how do I do top-10 customers", "how to compute totals per group", "how to safely sort without mutating", "how to batch upload in chunks", "how to convert a Map into a collection", "how to drain a queue"; user wants a full chained pipeline rather than a single method. SKIP: single-method lookup — use the focused skill (`mongez-collection-where`, `mongez-collection-math`, `mongez-collection-sort-group`, `mongez-collection-pagination`, `mongez-collection-transforming`, `mongez-collection-mutation`, etc.); standalone array helpers — use `mongez-reinforcements-arrays` (no wrapper, no chain); React state — use `@mongez/atom`'s `atomCollection`.

Recipes

Cross-feature compositions and idiomatic chains. Each recipe is shaped around what you're trying to do; the chained methods follow from the goal.

Top 10 highest-spending customers

A dashboard widget showing the top customers by total spend, scoped to paid orders only.

collect(orders)
  .where("status", "paid")
  .sortByDesc("total")
  .take(10)
  .pluck("customerId")
  .all();

pluck is the lazy alternative to a .map(o => o.customerId) — same result, less ceremony.

Page through a long list

The list page is rendering page 3 of 20-per-page results. One small helper keeps the call sites clean.

function paginate<T>(c: ImmutableCollection<T>, page: number, perPage: number) {
  return c.skip((page - 1) * perPage).take(perPage);
}

paginate(collect(products), 3, 20);

There's no built-in paginate(page, perPage) that returns metadata — compose it from length + skip().take(). See mongez-collection-pagination for the slicing primitives.

Find active users seen in the last week

You're filtering for "currently engaged" users — both flagged active and seen recently — and showing the most recent first.

const recentActive = collect(users)
  .where("active", true)
  .where("lastSeenAt", ">", Date.now() - 7 * 86400_000)
  .sortByDesc("lastSeenAt");

Multiple where calls AND together. For OR semantics use a single filter with a function predicate.

Total spend per customer

You're producing a report: one row per customer, sorted by their total spend across all paid orders.

collect(orders)
  .groupBy("customer")
  .map(bucket => ({
    customer: bucket.customer,
    total: collect(bucket.items).sum("amount"),
  }))
  .sortByDesc("total");

groupBy returns one bucket per distinct key with the matching items inside. sum("amount") sums a numeric field across the bucket.

Split tasks into pending and completed, then route each elsewhere

One source list, two downstream pipelines — instead of running two filters with overlapping work, partition once.

const [pending, completed] = collect(tasks).partition(t => t.status === "open");

console.log(`Pending: ${pending.length}`);
console.log(`Done:    ${completed.length}`);

pending.where("priority", "high").forEach(notify);
completed.sortByDesc("completedAt").take(10).forEach(showInActivity);

partition walks the list exactly once. The result is a destructurable tuple of two collections.

Pick one entry per email, then keep just the IDs

A signup-registrations table where users may have registered multiple times under the same email. You want the first registration per email, projected down to just the IDs.

collect(registrations)
  .uniqueList("email")
  .pluck("id")
  .all();

uniqueList keeps the first item for each unique key, in source order. For "keep the most recent" semantics, sort by date descending before uniqueList.

Log an intermediate count without breaking the chain

You're debugging a long filter pipeline and want to see how many items survive a particular where. tap runs a side effect and passes the collection through unchanged.

collect(events)
  .where("type", "click")
  .tap(c => console.log(`Click events: ${c.length}`))
  .where("payload.target", "starts with", "/products/")
  .map(e => e.payload.target);

Useful inline-instrumentation that drops in and out without restructuring the pipeline.

Sort a list without mutating the original

All sort variants (sort, sortBy, sortByDesc) clone internally, so the source collection is never reordered. Safe to use even when something else holds a reference.

function sortedCopy<T>(c: ImmutableCollection<T>, comparator: (a: T, b: T) => number) {
  return c.sort(comparator);       // source `c` is untouched
}

function byAge<T>(c: ImmutableCollection<T>) {
  return c.sortBy("age");          // source preserved
}

Compared to Array.prototype.sort (which mutates), this lets you sort defensively without first cloning.

Upload a list in batches of 100

You're shipping thousands of records to an API that rate-limits per request. Split into batches, await each one in sequence.

async function uploadInBatches<T>(items: ImmutableCollection<T>, batchSize = 100) {
  const batches = items.chunk(batchSize, false);
  for (const batch of batches) {
    await uploadBatch(batch);
  }
}

For parallel batch uploads with bounded concurrency, see pMap from @mongez/reinforcements.

Rename a user without mutating the list

You have a list of users and want to update one of them by id — keeping the others untouched and the result a new immutable collection.

const idx = users.findIndex(u => u.id === id);
const updated = users.set(idx, { ...users.at(idx), name: "Renamed" });

Or the bulk-by-predicate form:

const updated = collect(users).map(u =>
  u.id === id ? { ...u, name: "Renamed" } : u,
);

Both produce a new collection; the original is untouched.

Chain over a Map's entries

You have a Map and want collection-style chaining over its entries.

import { collect } from "@mongez/collection";

const m = new Map([
  ["a", 1],
  ["b", 2],
]);

// As tuples
collect.fromIterator(m).all();          // [["a", 1], ["b", 2]]

// As objects
collect.fromIterator(m).map(([key, value]) => ({ key, value }));

fromIterator accepts any iterable — Maps, Sets, generator functions, async generators (in their sync-iterable form).

Build an id → user lookup table

Hot path needs O(1) lookups by id. Reduce the list into an index once and keep it around.

const byId = collect(users).reduce<Record<number, User>>(
  (acc, u) => ({ ...acc, [u.id]: u }),
  {},
);

byId[42];     // O(1) lookup

For frequent updates, switch to a Map<number, User> — the spread-into-accumulator pattern shown here is cheap to build once but expensive to mutate.

Drain a queue one item at a time

shift() and pop() on a collection return the first/last item without removing it (collections are immutable). A naive while (queue.length > 0) loop would spin forever — you need to advance the collection or drop the wrapper.

let queue = collect(jobs);
while (queue.length > 0) {
  const job = queue.shift();        // read the head…
  queue = queue.skip(1);            // …then advance the collection
  await process(job);
}

Simpler for true queue semantics — drop the wrapper:

const queue2 = [...collect(jobs)];
while (queue2.length > 0) {
  const job = queue2.shift();       // native Array.shift() mutates queue2 in place
  await process(job);
}

Native arrays are the better fit when you genuinely want destructive shift. Collections shine when the workload is functional — filter / sort / pluck / group — not mutate-and-drain.