JSON parent/child CSV roundtrip guide
Split nested order JSON into parent and child CSVs, keep keys, child order, types and null, and change only item B quantity from 3 to 4.
The original orders array contains orders 001 and 002, with three items A, B and C. Separate parent and child tables can be restored without a Cartesian product when parent_id and zero-based ordinal stay attached to every child.
| Parent | Children | Relationship |
|---|---|---|
| 001 | 0:A / 1:B | two items |
| 002 | 0:C | one item |
parents.csv carries the order ID and note meaning. children.csv carries parent ID, order, SKU and quantity. If a source has no stable key, keep a documented in-file ID and do not mix it with another document. Store column, JSON path, type and null policy in schema/mapping.json.
parent_id=001ordinal=0 → Aordinal=1 → B001.note = null002.note = ""A blank is not null proofedited-children.csv changes only 001 / ordinal 1 / B from qty 3 to qty 4. Parent ID and note, items A and C, and every ordinal stay unchanged. Total quantity moves from 6 to 7.
| Row | Before | After | Changed path |
|---|---|---|---|
| 001 / 1 / B | qty 3 | qty 4 | /orders/0/items/1/qty |
| Everything else | unchanged | unchanged | none |
expected.json has two orders, three items and total quantity 7. Check ID and quantity types, child order, 001.note=null, 002.note="", and empty child arrays where applicable. Separate formatting or object-key order differences from value, type and path differences.
orders[0].items[1].qty: 3 → 4parents: 2 / children: 3 / total qty: 7orders[0].items stays A, Borphan.csv points to missing parent 999; keep it unresolved and do not invent a parent. duplicate-parent.csv repeats order ID 001; do not silently select the first row. Missing ordinals, untyped blanks and relationship drift should block output for review.
The original JSON and CSVs are fixtures for this guide. In real work, preserve the original and compare before and after independently.
expected.json by path, without sending input outside the browser.