sanduteo95/npm-deps-tree

Computing the Dependency Tree for an NPM Package

This project exposes an API for computing a given NPM package's dependency tree. It was built using Node.js and Express and tested using Mocha.

The API was implemented with Node.js v12.

Note 1 The API only retrieves dependencies, so no devDependencies, peerDependencies, or optionalDepdendencies for now. Although it could be extended to include those very easily. Note 2 If provided with a version of latest, it never gets cached.

Project structure

.
+-- docs - documentation
+-- lib - business logic
+-- persistence - for now just the cache management logic
+-- routes - Express router and middleware
+-- test - unit tests
+-- utils - utility functions (error handling, logging, validation, and NPM registry management logic)
+-- app.js - Express app
+-- server.js

How to run

Run npm start and this will start up a server on port 3000 of localhost

API

Call the following APIs at http://localhost:3000:

• GET /package/<package>, where <package> is the name of a package published to NPM e.g.

curl -X GET http://localhost:3000/package/express

• POST /package/<package> with JSON body and a provided version field

curl -X POST http://localhost:3000/package/express -d '{ "version": "latest" }'

The response format is:

{
  "<package name>": {
    "version": "<package version>",
    "dependencies": [
      {
        "<dependency name>": {
          "version": "<dependency version>",
          "dependencies": {
              ...
          }
        }
      }
    ]
  }
}

How to test

Run npm t and this will lint the code, run Mocha tests and gather coverage statistics under /reports, and finally generate documentation under /docs.

Performance statistics

Using Artillery, I have gathered some statistics. These depend on external factors but show the benefits of caching.

To install Artillery, run:

npm i -g artillery

Without caching, we see that without caching 95% and 99% of the requests take 12s, whereas the minimum one can take is 4s.

artillery quick -c 10 -n 20 http://localhost:3000/package/express

Summary report @ 18:47:57(+0000) 2021-02-20
  Scenarios launched:  10
  Scenarios completed: 10
  Requests completed:  200
  Mean response/sec: 1.62
  Response time (msec):
    min: 1552.8
    max: 10919.2
    median: 5934.6
    p95: 8162.1
    p99: 9758.3
  Scenario counts:
    0: 10 (100%)
  Codes:
    200: 200

With caching, we see that 95% and 99% percent of the requests take 2s - 3s, whereas the minimum one can take is 0.6s:

artillery quick -c 10 -n 20 http://localhost:3000/package/express

Summary report @ 18:49:23(+0000) 2021-02-20
  Scenarios launched:  10
  Scenarios completed: 10
  Requests completed:  200
  Mean response/sec: 5.97
  Response time (msec):
    min: 343
    max: 4577.2
    median: 1385.3
    p95: 3296.6
    p99: 4186.5
  Scenario counts:
    0: 10 (100%)
  Codes:
    200: 200