JSON to CSV Converter

How nested JSON maps to a flat CSV

CSV is a grid: one header row, then one row per record, with the same columns in every row. JSON is a tree. To bridge the two, every branch of the tree is turned into a single column whose name is the path to that leaf, joined with dots:

{ "name": "Alice", "address": { "city": "Berlin", "zip": "10115" } }

name , address.city , address.zip
Alice, Berlin       , 10115

The flattening rules this tool uses:

• Objects are walked recursively. a.b.c is the value at {"a":{"b":{"c": ...}}}.
• Arrays are handled per the Inner arrays option. Join writes the elements into one cell separated by ; (objects in the array are written as compact JSON). Indexed columns emits tags[0], tags[1], and keeps flattening object elements into items[0].sku.
• The header is the union of every leaf path seen in any row, in first-seen order. A key that appears in only some records still gets a column.
• A missing key is an empty cell, not the string null. An explicit JSON null is also written as an empty cell.
• Scalars are written as-is: numbers unquoted, true / false literally, strings quoted only when the content requires it.

The RFC 4180 quoting rules

RFC 4180 is the closest thing CSV has to a standard. A field must be wrapped in double quotes when it contains any of:

• the delimiter (a comma in a normal CSV),
• a double quote — which is then escaped by doubling it: She said "hi" becomes "She said ""hi""",
• a line break (CR or LF).

A field with none of those may be written bare. Quote every field mode wraps all of them regardless — slightly larger, but some brittle parsers are happier with it. Records are separated by CRLF in the spec; keep CRLF unless you know the consumer wants Unix line endings.

Arrays of arrays vs arrays of objects

Input shape	Example	What you get
Array of objects	[{"a":1,"b":2},{"a":3,"b":4}]	The normal case. Keys become the header; one row per object.
Array of arrays	[["a","b"],[1,2],[3,4]]	Positional columns col1, col2, ...; each inner array is a row. If the first row looks like headers, you may want a CSV that keeps it as data — check the output.
Array of scalars	["x","y","z"]	A single column named value, one value per row.
Single object	{"a":1,"b":2}	Treated as a one-row table.
A bare string, number, or boolean	42	Rejected with a message — there is no table to build.

Excel gotchas

• UTF-8 without a BOM shows as mojibake. Excel on Windows assumes the system code page unless the file starts with a UTF-8 byte-order mark. If names with accents or non-Latin text look wrong, re-export with Add UTF-8 BOM on. Google Sheets and most programming languages want it off.
• Leading zeros disappear. A ZIP code 01234 or a part number 007 is read as a number and the zeros are dropped. There is no fix inside the CSV; import via Data → From Text/CSV and mark the column as Text.
• Things become dates. 3-4, 1/2, and MAR-1 get silently converted. Same fix: typed import, or open in Sheets which prompts first.
• Long numbers lose precision. A 16+ digit ID becomes 1.23457E+17. Keep it as text on import.

Questions