Using NestJS with SWC

This experimental project aims to explore the possibilities of using swc with NestJS. Tries to mimic a non-trivial project with TypeORM, nestjsx/crud, and dynamically loaded entities.

foo

Features

• dev-server with
  • experimental "hot-reloading"
  • transpiled by swc instead of ts-loader
  • does not write transpiled files to disk (no dist folder in dev)
  • supports source-maps (correct mappings of line numbers in errors to Typescript source files)
• alternative webpack setup (disclaimer: totally not happy with this)
  • uses swc-loader instead of ts-loader
  • removes ForkTsCheckerWebpackPlugin
  • uses a HMR config based on the docs
  • ⚠️ does not seem reliable (code changes are sometimes not reflected in responses). Unclear if swc-loader might cause or if the setup is wrong.
  • ⚠️ does not have proper source-map support
  • ⚠️ does not support generic glob references to entities
• testing (jest)
  • uses @swc/jest for transforms
  • supports coverage reports
  • supports watch mode
  • supports e2e-test. ⚠️ needed to change import * as request from "supertest"; to import request from "supertest"; to make it work

Benchmarks

production build

❯ time "dum build:nest-cli" "dum build:swc"
Benchmark #1: dum build:nest-cli
  Time (mean ± σ):      2.988 s ±  0.128 s    [User: 4.386 s, System: 0.365 s]
  Range (min … max):    2.904 s …  3.329 s    10 runs

Benchmark #2: dum build:swc
  Time (mean ± σ):     163.5 ms ±   7.3 ms    [User: 112.7 ms, System: 47.4 ms]
  Range (min … max):   156.9 ms … 186.8 ms    18 runs

Summary
  'dum build:swc' ran
   18.27 ± 1.13 times faster than 'dum build:nest-cli'

jest

unit tests

❯ time "dum test:ts-jest" "dum test:swc"
Benchmark #1: dum test:ts-jest
  Time (mean ± σ):      2.809 s ±  0.114 s    [User: 4.445 s, System: 0.489 s]
  Range (min … max):    2.723 s …  3.076 s    10 runs

Benchmark #2: dum test:swc
  Time (mean ± σ):      1.083 s ±  0.011 s    [User: 985.3 ms, System: 262.2 ms]
  Range (min … max):    1.072 s …  1.112 s    10 runs

Summary
  'dum test:swc' ran
    2.59 ± 0.11 times faster than 'dum test:ts-jest'

unit tests with coverage

❯ time "dum test:cov:ts-jest" "dum test:cov:swc"
Benchmark #1: dum test:cov:ts-jest
  Time (mean ± σ):     10.206 s ±  0.583 s    [User: 50.058 s, System: 4.228 s]
  Range (min … max):    9.651 s … 11.351 s    10 runs

Benchmark #2: dum test:cov:swc
  Time (mean ± σ):      2.044 s ±  0.170 s    [User: 5.328 s, System: 1.151 s]
  Range (min … max):    1.881 s …  2.302 s    10 runs

Summary
  'dum test:cov:swc' ran
    4.99 ± 0.50 times faster than 'dum test:cov:ts-jest'

e2e tests

Benchmark #1: dum test:e2e:ts-jest
  Time (mean ± σ):      3.817 s ±  0.190 s    [User: 5.280 s, System: 0.710 s]
  Range (min … max):    3.659 s …  4.204 s    10 runs

Benchmark #2: dum test:e2e:swc
  Time (mean ± σ):      2.138 s ±  0.114 s    [User: 1.934 s, System: 0.477 s]
  Range (min … max):    2.052 s …  2.436 s    10 runs

Summary
  'dum test:e2e:swc' ran
    1.79 ± 0.13 times faster than 'dum test:e2e:ts-jest'

Caveats

• Sometimes, TypeORM does not close the connection properly. In this case, the server needs to be restarted manually
• HMR is 100% experimental. Open to any ideas on how this could be implemented better.