@openimport/json

JSON and JSON Lines format plugins for OpenImport.

import { createImporter } from '@openimport/core'
import { json, jsonl } from '@openimport/json' // or the combined default export `jsonFormats()`

const importer = createImporter({ schema, plugins: [json(), jsonl()] })

JSON: automatic record discovery

Every array of objects in the document becomes a dataset, addressed with a JSONPath-like id:

{ "data": { "customers": [{ "name": "Ann", "address": { "city": "Köln" }, "orders": [{ "sku": "A" }] }] } }

| Dataset | Records | | -------------------------------- | ------------------------ | | $.data.customers[*] | every customer | | $.data.customers[*].orders[*] | every order of every customer |

• Top-level arrays come first, then shallower and larger collections
• Arrays of primitives are values, not datasets
• An object of objects ({ "u1": {…}, "u2": {…} }) is a keyed collection; the key is exposed as _key
• A single object is imported as one record

Nested objects are flattened by the core into dot paths (address.city) so they can be mapped like columns.

JSON Lines / NDJSON

.jsonl / .ndjson files are streamed line by line. Blank lines are skipped, primitives are wrapped as { value }, and parse errors report the line number.

Options

| Option | Default | Description | | ----------- | ------- | ---------------------------------------- | | chunkSize | 1000 | Records per yielded chunk | | maxDepth | 8 | How deep to search for record arrays (JSON) |