🧭 Pro Tips & Gotchas

What the converters do

• Array-of-objects in, one of two things out: CSV (first row becomes the header) or a SQL INSERT with keys as columns and values single-quoted.
• Non-string values are stringified rather than dropped, so nested objects arrive as [object Object] — a loud, if ugly, signal that the shape is wrong.
• Strings are escaped by doubling single quotes, which is correct for PostgreSQL/SQLite/MySQL-in-default-mode but not for a database running with NO_BACKSLASH_ESCAPES disabled expectations.

Gotchas

• Ragged objects lose data: the header comes from the first record, so any key that only appears later is silently omitted from every row. Merge keys before converting.
• null, undefined, and missing keys are not the same thing in JSON but often collapse to the same empty CSV cell, and an empty cell is not NULL on import — it is the empty string. Quote or NULL-cast explicitly.
• The SQL output has no transactions, no ON CONFLICT, and no batch limits. One giant INSERT with 200k rows will hit max_allowed_packet on MySQL and lock the table on Postgres; chunk it.
• Booleans print as true/false, which Postgres accepts and MySQL does not (it wants 1/0). Dates print as ISO strings, which is fine for timestamptz and a timezone trap for timestamp.
• For CSV: this is a convenience converter, not an RFC 4180 writer. Values containing commas, quotes, or newlines need quoting that a naive join does not provide — use Python csv, Node csv-stringify, or COPY ... TO CSV instead.
• Never paste generated SQL from untrusted JSON without review: an escaping bug plus a value containing '); DROP TABLE is the whole exploit chain. Parameterised inserts exist for a reason.