Convert an Excel Spreadsheet to PostgreSQL INSERT Statements for DB Seeding

Lookup table seed, ready for psql

A reference table with a boolean active flag. Note TRUE/FALSE (Postgres-native) and the BOOLEAN column type in the DDL.

Excel (Sheet1):
 id | code | active
 1  | EU   | TRUE
 2  | US   | FALSE

Generated SQL (table 'region'):
CREATE TABLE IF NOT EXISTS "region" (
  "id" INTEGER,
  "code" TEXT,
  "active" BOOLEAN
);

INSERT INTO "region" ("id", "code", "active")
VALUES
  (1, 'EU', TRUE),
  (2, 'US', FALSE);

Idempotent re-seed with ON CONFLICT

With the conflict toggle on, the INSERT ends with ON CONFLICT DO NOTHING, so re-running the seed against a populated table skips existing primary keys silently.

Options: PostgreSQL, ON CONFLICT DO NOTHING

INSERT INTO "region" ("id", "code", "active")
VALUES
  (1, 'EU', TRUE),
  (2, 'US', FALSE) ON CONFLICT DO NOTHING;

Reserved-word column survives quoting

A column literally named 'order' would be a syntax error unquoted. Postgres double-quoting keeps it valid.

Excel:
 id | order
 1  | 10
 2  | 20

Generated SQL:
INSERT INTO "data" ("id", "order")
VALUES
  (1, 10),
  (2, 20);

Dates: standardise, then CAST

Because Excel dates arrive as serials, generate them as ISO text first (via the date standardizer) so they land as TEXT, then alter the column to a real date type in Postgres.

Seed file value (after date standardizer → ISO text):
INSERT INTO "event" ("id", "starts")
VALUES (1, '2024-06-01');

Then in psql:
ALTER TABLE event
  ALTER COLUMN starts TYPE date USING starts::date;

Drop audit columns from the seed

Operational exports often carry imported_at / row_hash columns you do not want in version-controlled seed data. List them in Exclude columns.

Exclude columns: imported_at, row_hash

Result: those two headers are omitted from both
the CREATE TABLE and every INSERT, leaving only
the business columns in the seed file.

The Table name field is emitted as a single double-quoted identifier, so typing public.region produces "public.region" (one identifier with a dot) — not schema.table. For a schema target, edit the generated identifier to public."region" after copying.

Postgres won't accept a serial number into a date column. Standardise the column to ISO text with the date standardizer first; it then generates as quoted TEXT you can CAST/ALTER to date.

Decimals infer DOUBLE PRECISION, which is floating point. For currency, change the inferred type to NUMERIC(12,2) in the generated DDL before running, since DOUBLE PRECISION can introduce rounding.

ON CONFLICT DO NOTHING (without a target) relies on any unique or PK constraint existing on the table. If the table has none, nothing conflicts and every row inserts. Ensure the constraint exists, or this toggle is a no-op.

Only the first tab is read. Lookup tables spread across tabs must be generated one at a time, or combined first with the sheet joiner.

IDs beyond 2^53 lose precision during JS parsing. For bigint keys with very large values, format the Excel column as Text so they pass through verbatim as quoted strings, then cast in SQL.

The code-generation quota caps Free at 100 rows and 1 run/day; the seed is truncated past that. Pro raises it to 1,000 rows / 5 runs, Pro+Media to 10,000, Developer unlimited.

A blank numeric cell becomes NULL, not 0. If your Postgres column is NOT NULL DEFAULT 0, either fill the cell in Excel or let the default fire — the generated NULL would be rejected by a NOT NULL column without a default.

A column mixing numbers and text infers TEXT; numeric-looking values are quoted as strings. Clean or split the column first if Postgres needs it numeric.