UPGRADING.md
sqlite3 is maintained again, we switched back to it. If you are using @vscode/sqlite3 driver dependency, please replace it with sqlite3 in your package.json;sqlite3 driver dependency, please replace it with @vscode/sqlite3 in your package.json;RETURNING operations now always return an object with column names;import Knex from 'knex'; used to import the knex instantiation function, the namespace and the interface for the knex instantiation function/object, there is now a clear distinction between them:import { knex } from 'knex'; // this is a function that you call to instantiate knex
import { Knex } from 'knex'; // this is a namespace, and a type of a knex object
import KnexTimeoutError = Knex.KnexTimeoutError; // this is a class from the Knex namespace
const config: Knex.Config = {}; // this is a type from the Knex namespace
const knexInstance: Knex = knex(config);
If your code looked like this:
import knex from 'knex';
const config: knex.Config = {}; // this is a type from the Knex namespace
const knexInstance = knex(config);
Change it to
import { knex, Knex } from 'knex';
const config: Knex.Config = {}; // this is a type from the Knex namespace
const knexInstance = knex(config);
Config or QueryBuilder directly, use Knex namespace instead.So change this:
import { QueryBuilder } from 'knex';
const qb: QueryBuilder = knex('table').select('*');
to this:
import { Knex } from 'knex';
const qb: Knex.QueryBuilder = knex('table').select('*');
const knex = require('knex').knex({
//connection parameters
});
It also works when using ESM imports:
import { knex } from 'knex';
const kn = knex({
//connection parameters
});
For usage as param it can be addressed like this:
/**
* @param {import("knex").Knex} db
*/
function up(db) {
// Your code
}
declare module 'knex' {
interface QueryBuilder {
paginate<TResult = any[]>(
params: IPaginateParams
): KnexQB<any, IWithPagination<TResult>>;
}
}
This should be changed into this:
declare module 'knex' {
namespace Knex {
interface QueryBuilder {
paginate<TResult = any[]>(
params: IPaginateParams
): KnexQB<any, IWithPagination<TResult>>;
}
}
}
TypeScript version 4.1+ is needed when using knex types now.
MSSQL driver was completely reworked in order to address the multitude of connection pool, error handling and performance issues. Since the new implementation uses tedious library directly instead of mssql, please replace mssql with tedious in your dependencies if you are using a MSSQL database.
Transaction rollback does not trigger a promise rejection for transactions with specified handler. If you want to preserve previous behavior, pass config object with doNotRejectOnRollback: false:
await knex.transaction(
async (trx) => {
const ids = await trx('catalogues').insert({ name: 'Old Books' }, 'id');
},
{ doNotRejectOnRollback: false }
);
Connection url parsing changed from legacy url.parse to WHATWG URL. If you have symbols, unusual for a URL (not A-z, not digits, not dot, not dash) - check Node.js docs for details
Global static Knex.raw support dropped, use instance knex.raw instead. (require('knex').raw() won't work anymore)
v8 flags are no longer supported in cli. To pass these flags use NODE_OPTIONS environment variable.
For example NODE_OPTIONS="--max-old-space-size=1536" npm run knex
Clients are now classes instead of new-able functions. Please migrate your custom clients to classes.
const Client = require('knex');
const { inherits } = require('util');
// old
function CustomClient(config) {
Client.call(this, config);
// construction logic
}
inherits(CustomClient, Client);
CustomClient.prototype.methodOverride = function () {
// logic
};
// new
class CustomClient extends Client {
// node 12+
driverName = 'abcd';
constructor(config) {
super(config);
this.driverName = 'abcd'; // bad way, will not work
// construction logic
}
methodOverride() {
// logic
}
}
// alternative to declare driverName
CustomClient.prototype.driverName = 'abcd';
There was a major internal restructuring and renaming effort. Most dialect-specific compilers/builder have dialect name as a prefix now. Also some files were moved. Make sure to make adjustments accordingly if you were referencing specific knex library files directly from your code.
"first" and "pluck" can no longer be both chained on the same operation. Previously only the last one chained was used, now this would throw an error.
Trying to execute an operation resulting in an empty query such as inserting an empty array, will now throw an error on all database drivers.
beforeDestroy pool configuration option was removed. You should use tarn.js event handlers if you still need similar functionality..catch() and then() not being recognized.table.datetime and table.timestamp methods instead of argument options. See documentation for these methods for more details;Node.js older than 6 is no longer supported, make sure to update your environment;
MSSQL: Creating a unique index on the table targeted by stored procedures that were created with QUOTED_IDENTIFIER = OFF fails.
You can use this query to identify all affected stored procedures:
SELECT name = OBJECT_NAME([object_id]), uses_quoted_identifier
FROM sys.sql_modules
WHERE uses_quoted_identifier = 0;
The only known solution is to recreate all stored procedures with QUOTED_IDENTIFIER = OFF
mariadb dialect is no longer supported;Instead, use "mysql" or "mysql2" dialects.
await knex.migrate.latest({
directory: 'src/services/orders/database/migrations',
tableName: 'orders.orders_migrations',
});
Instead, starting from 0.14.5 you should use new parameter schemaName:
await knex.migrate.latest({
directory: 'src/services/orders/database/migrations',
tableName: 'orders_migrations',
schemaName: 'orders',
});