express-slonik
v3.2.0
Published
Slonik transaction middleware
Downloads
1,028
Maintainers
Readme
express-slonik
Slonik transaction middleware for Express.js with zero dependencies.
Table of Contents
Getting started
Install the middleware as a dependency in your Express.js project:
npm i -S slonik express-slonik
Usage
You can use the createMiddleware
function to create a request transaction context that contains
methods to wrap your route handlers in a PostgreSQL transaction.
Basic usage
Use the transaction.begin()
and transaction.end()
middleware to wrap your request handlers in a
transaction.
app.ts
:
import createMiddleware from "express-slonik";
import { sql } from "slonik";
import { z } from "zod";
const userSchema = z.object({
id: z.number().int(),
name: z.string(),
email: z.string().email(),
});
export const createServer = ({ app, pool }) => {
const transaction = createMiddleware(pool);
app.get(
"/user/:id",
// Starts the transaction.
transaction.begin(),
async (req, res, next) => {
try {
const user = await req.transaction.one(
sql.type(
userSchema
)`SELECT * FROM users WHERE users.id = ${req.params.id}`
);
res.json(user);
} catch (error) {
if (error instanceof NotFoundError) {
res.status(404).json({
name: error.name,
message: `User with given id (${req.params.id}) not found.`,
});
return;
}
next(error);
}
},
// Optional. If omitted, the transaction will automatically commit when the
// response is sent, or rollback if there are unhandled errors. Specifying
// transaction.end() is useful if you wish to have finer control over when
// the transaction commits in the middleware chain.
transaction.end()
);
const server = app.listen(8080);
// Cleanup when server closes or you might have something that keeps the process running.
server.on("close", async function () {
await pool.end();
});
return server;
};
server.ts
:
import { Server } from "http";
import express from "express";
import { createPool } from "slonik";
import { createServer, userSchema } from "./app";
/**
* Gracefully attempt to shut down the server.
*/
async function shutdownHandler(server: Server) {
// Handler that triggers a graceful shutdown. Server is responsible for cleaning up stragglers.
return async function () {
server.close();
};
}
(async function () {
const app = express();
const pool = await createPool(process.env.DATABASE_URL);
const server = createServer({ app, pool });
process
.on("SIGTERM", shutdownHandler(server))
.on("SIGINT", shutdownHandler(server));
})();
This is functionally equivalent to using pool.transaction
in your handler:
app.get("/user/:id", async (req, res, next) => {
try {
const user = await pool.transaction(async (transaction) => {
return await transaction.one(
sql.type(
userSchema
)`SELECT * FROM users WHERE users.id = ${req.params.id}`
);
});
res.json(user);
} catch (error) {
if (error instanceof NotFoundError) {
res.status(404).json({
name: error.name,
message: `User with given id (${req.params.id}) not found.`,
});
return;
}
next(error);
}
});
Sharing transaction with multiple route handlers or middleware
Suppose you had a middleware that returns the current user from the session or JWT. You can make sure the user edit handler are on the same database transaction as the current user middleware. This can prevent concurrent user updates from causing inconsistent the query result between the time the current user middleware and your user edit handler executes.
schemas/user.ts
:
import { z } from "zod";
export const userSchema = z.object({
id: z.number().int(),
name: z.string(),
email: z.string().email(),
});
middleware/current-user.ts
:
import { userSchema } from "schemas/user";
export default function currentUser() {
return async function (req, res, next) {
try {
req.currentUser = await req.transaction.one(
sql.type(
userSchema
)`SELECT * FROM users WHERE users.id = ${req.session.userId}`
);
next();
} catch (error) {
next(error);
}
};
}
app.ts
:
import express, { json } from "express";
import createMiddleware from "express-slonik";
import { createPool, sql } from "slonik";
import currentUser from "./middleware/current-user";
import { userSchema } from "./schemas/user";
const pool = createPool("postgres://localhost:5432/example_db");
const transaction = createMiddleware(pool);
const app = express();
app
.use(json())
.put(
"/user/:id",
transaction.begin(),
currentUser(),
async (req, res, next) => {
try {
// Same transaction as currentUser middleware
await req.transaction.query(
sql.type(
z.object({})
)`UPDATE users SET email = ${req.body.email} WHERE users.id = ${req.params.id}`
);
const updatedUser = await req.transaction.one(
sql.type(
userSchema
)`SELECT * FROM users WHERE users.id = ${req.params.id}`
);
res.json(updatedUser);
} catch (error) {
next(error);
}
},
transaction.end()
)
.use((error, req, res, next) => {
res.status(401).end();
});
app.listen(8080);
This behavior is especially helpful when you are using a custom validator or sanitizor in libraries like express-validator:
import express, { json } from "express";
import createMiddleware from "express-slonik";
import { body, validationResult } from "express-validator";
import { createPool, sql } from "slonik";
import currentUser from "./middleware/current-user";
import { teamSchema } from "./schemas/team";
import { userSchema } from "./schemas/user";
const pool = createPool("postgres://localhost:5432/example_db");
const transaction = createMiddleware(pool);
const app = express();
app
.use(json())
.put(
"/user/:id",
transaction.begin(),
body("email").isEmail().normalizeEmail(),
body("team_id")
.toInt()
.custom(async (value, { req }) => {
// Fail validation if client is attempting to add user to a non-existant team
const isValidTeam = await req.transaction.exists(
sql.type(
teamSchema
)`SELECT * FROM teams WHERE teams.id = ${req.body.team_id}`
);
if (!isValidTeam) {
throw new Error("Invalid team");
}
}),
async (req, res, next) => {
const errors = validationResult(req);
if (!errors.isEmpty()) {
res.status(422).json(errors.array());
return;
}
// We can assume the request body is valid and sanitized by the time we reach this point.
try {
await req.transaction.query(sql.unsafe`
UPDATE users SET
email = ${req.body.email},
team_id = ${req.body.team_id}
WHERE
users.id = ${req.params.id}
`);
const user = await req.transaction.one(
sql.type(
userSchema
)`SELECT * FROM users WHERE users.id = ${req.params.id}`
);
res.status(200).json(user);
} catch (error) {
next(error);
}
},
transaction.end()
)
.use((error, req, res, next) => {
res.status(401).end();
});
app.listen(8080);
Setting isolation levels
The transaction.begin
can take an optional argument to specify transaction isolation levels. It
defaults to READ COMMITTED isolation level is left empty.
There are three isolation levels: READ COMMITTED
, REPEATABLE READ
, and SERIALIZABLE
.
import createMiddleware, { IsolationLevels } from "express-slonik";
import { createPool } from "slonik";
const transaction = createMiddleware(
createPool("postgres://localhost:5432/example_db")
);
app.get(
"/posts/:postId/comments",
transaction.begin(IsolationLevels.SERIALIZABLE),
// ...
transaction.end()
);
For more information on the differences between transaction isolation levels, please refer to: 13.2. Transaction Isolation — PostgreSQL documentation.
Version compatibility
express-slonik follows Semantic Versioning specification. Each major version breaks backwards compatibility with Slonik and Express.js versions (although Express v5 has been extremely slow to come out of beta).
Refer to the compatibility chart below for picking the express-slonik version that works with Slonik versions in your project.
| express-slonik | slonik | | -------------: | --------------------------------: | | ^3.0.0 | ^33.0.0 || ^34.0.0 | | ^2.0.0 | ^30.0.0 || ^31.0.0 || ^32.0.0 | | ^1.1.0 | ^28.0.0 || ^29.0.0 | | ≥1.0.0 <1.1.0 | ^28.0.0 |
Minor version will always add support for Slonik versions that doesn't introduce major backwards incompatibility that breaks interoperability with this library. For instance, the breaking changes introduced between Slonik v28 and v29 are fairly minor and can be used with express-slonik without any major refactor to how you use this middleware. However, the difference between v29 and v30 introduces a major change to the API surface (which also affects other Slonik utility packages). In this case, the major version of express-slonik will be bumped up to indicate that there will be backwards breaking changes.
Other projects
Need to isolate database calls in your tests? Check out mocha-slonik!