Lesson 30 / 35
Common Pitfalls and How to Avoid Them
Avoid the mistakes that bite real ORM projects: destructive sync, missing awaits, BigInt serialisation, timezones and hidden soft deletes.
By Rahul Agarwal · 15 years in web development
Bugs that every team meets once
Some ORM mistakes are so common they are worth memorising. Destructive sync: sequelize.sync({ force: true }) drops and recreates tables, and alter: true or TypeORM synchronize: true can drop columns; keep them out of production code paths. Missing await: forgetting it means the query may run after the response is sent, errors become unhandled rejections, and you get a pending Promise instead of data. BigInt: databases return 64-bit integers that may not fit a JavaScript number, so Prisma maps BigInt columns (and some raw COUNT results) to the JavaScript bigint type, which JSON.stringify refuses to serialise. Dates and timezones: store timestamps in UTC and convert only when displaying, otherwise IST users see times shifted by five and a half hours. Soft deletes: Sequelize paranoid: true and TypeORM @DeleteDateColumn hide deleted rows from normal queries, which surprises anyone writing raw SQL or reports. Knowing these saves hours of debugging.
The BigInt surprise
This plain Node.js snippet shows the error you get when an API returns a raw query result containing a bigint, and one way to convert it.
const row = { authorId: 7, n: 12n }; // what a raw COUNT(*) can come back as
try {
console.log(JSON.stringify(row));
} catch (e) {
console.log(e.name + ': ' + e.message);
}
console.log(JSON.stringify(row, (k, v) => (typeof v === 'bigint' ? Number(v) : v)));
Output:
TypeError: Do not know how to serialize a BigInt
{"authorId":7,"n":12}Number() is only safe below 2^53
Converting a bigint with Number() silently loses precision above Number.MAX_SAFE_INTEGER. For counts that is fine; for IDs or money stored as large integers, send them to the client as strings instead.
Quick check: A Sequelize model uses paranoid: true. A user is 'deleted' with destroy(). What happens to the row?
- It is permanently removed from the table
- It is moved to an archive table automatically
- The whole table is truncated
- It stays in the table with deletedAt set, and normal finders exclude it
Answer
It stays in the table with deletedAt set, and normal finders exclude it — Paranoid models perform soft deletes by setting deletedAt. Finders skip such rows unless you pass paranoid: false, and destroy({ force: true }) removes them for real.