Convert CSV to JSON for MongoDB Import

Default settings produce a typed array. age is a number and active a boolean, so a Mongoose schema with those paths casts cleanly on insert.

Input (users.csv):
name,age,active
Ada,36,true
Linus,54,false

Output (array, inference on):
[
  { "name": "Ada", "age": 36, "active": true },
  { "name": "Linus", "age": 54, "active": false }
]

mongoimport --db app --collection users \
  --file users.json --jsonArray

Switch to NDJSON when you want the line-delimited form mongoimport reads with no flag — easier to stream for a large collection.

Input:
name,age
Ada,36
Linus,54

Output (NDJSON):
{"name":"Ada","age":36}
{"name":"Linus","age":54}

mongoimport --db app --collection users --file users.ndjson

An external account ID 00042 must stay verbatim. Turn inference off so every value, including the id, is a string.

Input:
_id,name
00042,Ada
00043,Linus

Output (inference OFF):
[
  { "_id": "00042", "name": "Ada" },
  { "_id": "00043", "name": "Linus" }
]

Quote-aware parsing keeps the comma inside the quoted bio cell, so the document has one field, not two — the failure mode of importing the CSV directly.

Input (note the quoted field):
name,bio
Ada,"Mathematician, writer, and analyst"

Output:
[
  { "name": "Ada", "bio": "Mathematician, writer, and analyst" }
]

Use the downloaded array directly in an insertMany seed. Typed values mean no manual casting.

// seed.js
const data = require('./users.json'); // the array from this tool
await User.insertMany(data);
// data[0].age === 36 (number), data[0].active === true (boolean)

With inference on, 00042 becomes 42. For an _id or external reference that must round-trip exactly, turn off Infer numbers, booleans, null so every field, including the id, is a string. There is no per-column toggle — if only the id column is affected and you need the others typed, convert with inference off and cast the numeric fields in your Mongoose schema.

An ID beyond JavaScript's safe-integer range is left as a string by inference (it checks Number.isSafeInteger). This is the safe outcome — Mongo would otherwise have to store a rounded Double. If you genuinely need a BSON Long, import it as a string and cast in an aggregation, or use --columnsHaveTypes in mongoimport on the CSV path instead.

2026-06-12 and other date formats do not match the number/boolean/null patterns, so they stay strings. Mongo stores them as strings, not BSON Date. Cast on import with $dateFromString in an aggregation, or set the Mongoose path to Date and let casting convert the ISO string on insert.

A header address.city becomes a top-level field literally named address.city, not a nested { address: { city } }. mongoimport's CSV path interprets dotted headers as nesting; this JSON converter does not. To get subdocuments, convert here then run the array through json-unflattener to expand dotted keys into nested objects.

Two columns named email map to one JSON key; the later one wins and the earlier value is lost. BSON documents cannot have duplicate field names, so this is unavoidable in the JSON. Rename one column before converting so both values land in distinct fields.

By default a blank cell becomes "", so the field exists with an empty value — different from the field being absent. If you want optional fields to be omitted (so Mongo treats them as unset), turn on Skip empty cells; then blank cells drop the field from that document.

Free tier caps at 2 MB / 500 rows. A real collection export usually needs Pro (100 MB / 100,000 rows) or Developer (5 GB, no row cap). For a one-off seed, split the CSV into chunks under the cap, convert each to NDJSON, and cat them together before mongoimport.

Grouped mode produces { "key": [ … ] }, which mongoimport will not read as a set of documents. Use it only when you actually want one keyed object (e.g. to build a lookup map). For seeding documents, use Array or NDJSON.