JSON Output

Output

Write the incoming row stream as a JSON array or as NDJSON (one JSON object per line). Optionally pretty-print the JSON array, and optionally reconstruct nested objects from dot-notation column names like address.city → { address: { city: ... } }.

Local processing

This transform runs entirely in your browser. No data is sent to external servers.

How it works

JSON Output is a sink node. It consumes the row stream and produces a Blob whose MIME type is application/json for arrays or application/x-ndjson for line-delimited output.

In json_array mode, every row is buffered before serialisation because JSON.stringify needs the full array. Pretty-printing adds two-space indentation. In ndjson mode each row is stringified and appended with a trailing newline as it arrives, so memory stays roughly constant.

When unflattenNested is true, dot-separated column names are split into nested object structure: address.city becomes { address: { city: ... } }. Reconstruction stops at five levels deep — keys with more dots than that are written as flat keys with literal dots in the name. Existing nested objects from the row are preserved as-is; only flat dotted keys are reshaped.

Input: A row stream from any upstream transform. Output: A Blob containing a JSON array or NDJSON.

Options

Option	Type	Default	Description
format	"json_array" | "ndjson"	"json_array"	Serialisation shape. json_array wraps everything in [...]. ndjson writes one object per line.
prettyPrint	boolean	false	Pretty-print with two-space indentation. Has no effect in NDJSON mode.
unflattenNested	boolean	false	Split dot-notation keys back into nested objects (e.g. address.city → { address: { city } }). Stops at five levels.

Examples

Default JSON array, compact

A small product catalogue exported as a flat JSON array.

Before:

sku	name	price
W-001	Widget	9.99
G-001	Gadget	24.50
D-001	Doohickey	4.25

Configuration: defaults — format: "json_array", prettyPrint: false, unflattenNested: false.

After (raw file content):

[
  { "sku": "W-001", "name": "Widget", "price": 9.99 },
  { "sku": "G-001", "name": "Gadget", "price": 24.5 },
  { "sku": "D-001", "name": "Doohickey", "price": 4.25 }
]

Pretty-printed array with unflattening

The upstream pipeline produced dot-notation columns for an address; downstream consumers expect nested objects.

Before:

customer	address.city	address.country	total
Acme Corp	Valencia	ES	4250
Beta Inc	Berlin	DE	1875

Configuration: format: "json_array", prettyPrint: true, unflattenNested: true.

After (raw file content):

[
  {
    "customer": "Acme Corp",
    "address": {
      "city": "Valencia",
      "country": "ES"
    },
    "total": 4250
  },
  {
    "customer": "Beta Inc",
    "address": {
      "city": "Berlin",
      "country": "DE"
    },
    "total": 1875
  }
]

NDJSON for log-style streaming

A consumer that ingests one record per line — typical for analytics pipelines or jq post-processing.

Before:

event	user	ts
signup	[email protected]	2026-02-10T08:30:00Z
login	[email protected]	2026-02-10T08:31:14Z
login	[email protected]	2026-02-10T09:02:55Z

Configuration: format: "ndjson", unflattenNested: false.

After (raw file content):

{"event":"signup","user":"[email protected]","ts":"2026-02-10T08:30:00Z"}
{"event":"login","user":"[email protected]","ts":"2026-02-10T08:31:14Z"}
{"event":"login","user":"[email protected]","ts":"2026-02-10T09:02:55Z"}

Tips and Edge Cases

• prettyPrint only applies to json_array. In NDJSON mode it is ignored — each object is single-line by definition. The UI hides the checkbox when NDJSON is selected, but if a config object sets both, the runtime silently ignores prettyPrint. See apps/web/src/transforms/output-json/logic.ts:101-117.
• Unflattening stops at depth 5. Keys with six or more dot segments are written as flat keys with literal dots. This protects against pathological inputs but means deeply nested column names round-trip differently than shallow ones. See apps/web/src/transforms/output-json/logic.ts:14, 36-42.
• NDJSON is streaming; JSON array is not. Choose ndjson when the consumer can read line-by-line and the dataset is large — memory stays roughly flat. json_array buffers every row before calling JSON.stringify. See apps/web/src/transforms/output-json/logic.ts:101-138.

Related Transforms

• CSV Output — write plain CSV when the consumer is a spreadsheet or shell tool.
• Excel Output — write .xlsx when the consumer is Excel or another office suite.
• JSON Input — round-trip JSON; pair with flattenNested and unflattenNested to flatten and rebuild nested shapes.