CSV ↔ NDJSON

NDJSON, or JSON Lines, is the format with one JSON object per line. It became the standard input for stream processing tools, for bulk loading into document databases and for training sets, because it lets you read record by record without holding the whole file in memory. Spreadsheets and system exports, on the other hand, almost always come out as CSV.

The conversion runs both ways: pick the direction and paste. From CSV to NDJSON, each record becomes a JSON line with the header row naming the fields; on the way back, the keys of all objects form the header, in the order they first appeared. The CSV reader is complete enough where it counts: a quoted field containing a comma, doubled quotes representing a literal quote, and a line break inside a quoted field are all handled correctly — which is precisely where improvised converters fall over.

A note on types: values come out as text. A CSV carries no types at all — the number 007 might be a code that must keep its leading zeros, and converting it to a number would destroy it. If the destination needs real numbers or booleans, that type conversion is the consumer's decision, made with knowledge of what each column means.