Using Nx Core Without Plugins

The core of Nx is generic, simple, and unobtrusive. Nx plugins, although very useful for many projects, are completely optional. Most large Nx workspaces use plugins for some things and don't use plugins for others.

This guide will walk you through creating a simple Nx workspace with no plugins. It will help you see what capabilities of Nx are completely generic and can be used with any technology or tool.

Using Nx Core

Creating a New Workspace

Running yarn create nx-workspace --preset=npm creates an empty workspace.

This is what is generated:

packages/
nx.json
workspace.json
tsconfig.base.json
package.json

package.json contains Nx packages.

1{
2  "name": "myorg",
3  "version": "0.0.0",
4  "license": "MIT",
5  "scripts": {},
6  "private": true,
7  "devDependencies": {
8    "@nrwl/cli": "12.8.0",
9    "@nrwl/tao": "12.8.0",
10    "@nrwl/workspace": "12.8.0",
11    "@types/node": "14.14.33",
12    "typescript": "~4.3.5"
13  }
14}

nx.json contains the Nx CLI configuration.

1{
2  "extends": "@nrwl/workspace/presets/npm.json",
3  "npmScope": "myorg",
4  "tasksRunnerOptions": {
5    "default": {
6      "runner": "@nrwl/workspace/tasks-runners/default",
7      "options": {
8        "cacheableOperations": ["build", "lint", "test", "e2e"]
9      }
10    }
11  }
12}

Finally, workspace.json lists the workspace projects, and since we have none, it is empty.

Creating an NPM Package

Running nx g npm-package simple results in:

packages/
 simple/
   index.js
   package.json
nx.json
workspace.json
tsconfig.base.json
package.json

The generated simple/package.json:

1{
2  "name": "@myorg/simple",
3  "version": "1.0.0",
4  "scripts": {
5    "test": "node index.js"
6  }
7}

And workspace.json gets updated to:

1{
2  "version": 2,
3  "projects": {
4    "simple": {
5      "root": "packages/simple"
6    }
7  }
8}

With this you can invoke any script defined in simple/package.json via Nx. For instance, you can invoke the test script by running yarn nx test simple. And if you invoke this command a second time, the results are retrieved from cache.

In this example, we used a generator to create the package, but you could also have created it by hand or have copied it from another project.

The change in workspace.json is the only thing required to make Nx aware of the simple package. As long as you include the project into workspace.json, Nx will include that project source into its graph computation and source code analysis. It will, for instance, analyze the project's source code, and it will know when it can reuse the computation from the cache and when it has to recompute it from scratch.

Creating Second NPM Package and Enabling Yarn Workspaces

Running nx g npm-package complex results in:

packages/
 simple/
   index.js
   package.json
 complex/
   index.js
   package.json
nx.json
workspace.json
tsconfig.base.json
package.json

Now let's modify packages/complex/index.js to include require('@myorg/simple'). If you run yarn nx test complex, you will see an error saying that @myorg/simple cannot be resolved.

This is expected. Nx analyzes your source to enable computation caching, it knows what projects are affected by your PR, but it does not change how your npm scripts run. Whatever tools you use in your npm scripts will run exactly as they would without Nx. Nx Core doesn't replace your tools and doesn't change how they work.

To make it work add a dependency from complex to simple in package.json:

1{
2  "name": "@myorg/complex",
3  "version": "1.0.0",
4  "scripts": {
5    "test": "node index.js"
6  },
7  "dependencies": {
8    "@myorg/simple": "*"
9  }
10}

Then add the following to the root package.json (which enables Yarn Workspaces).

1{
2  "workspaces": ["packages/*"]
3}

Finally, run yarn.

yarn nx test complex works now.

Non-JS Projects

In this example both projects use JavaScript, but we could have created, say, a Rust project instead. The only thing we would have to do is to manually define targets.

This is an example of manually defining a target (read more about it):

1{
2  "version": 2,
3  "projects": {
4    "simple": {
5      "root": "packages/simple",
6      "targets": {
7        "test": {
8          "executor": "@nrwl/workspace:run-commands",
9          "options": {
10            "command": "npm run test"
11          }
12        }
13      }
14    }
15  }
16}

Using Yarn/PNPM/Lerna

The example uses Yarn to connect the two packages. Most of the time, however, there are better ways to do it. The React, Node, Angular plugins for Nx allow different projects in your workspace to import each other without having to maintain cumbersome package.json files. Instead, they use Webpack, Rollup and Jest plugins to enable this use case in a more elegant way. Read about the relationship between Nx and Yarn/Lerna/PNPM.

What Nx Core Provides

Nx Understands How Your Workspace is Structured

If you run yarn nx dep-graph you will see that complex has a dependency on simple. Any change to simple will invalidate the computation cache for complex, but changes to complex won't invalidate the cache for simple.

In contrast to more basic monorepo tools, Nx doesn't just analyze package.json files. It does much more. Nx also knows that adding a require() creates a dependency and that some dependencies cannot even be expressed in the source code.

Nx Orchestrates Tasks

Running yarn nx run-many --target=test --all --parallel will test all projects in parallel.

Running yarn nx run-many --target=build --projects=app1,app2 --parallel will build proj1 and proj2 and their dependencies in parallel. Note that if app1 depends on the output of its dependency (e.g., shared-components), Nx will build shared-components first and only then will build the app.

Nx Know What is Affected

Running yarn nx affected --target=test will test all the projects affected by the current PR.

Nx Caches and Distributes Tasks

Running yarn nx build app1 will cache the file artifacts and the terminal output, so if you run it again the command will execute instantly because the results will be retrieved from cache. If you use Nx Cloud the cache will be shared between you, your teammates, and the CI agents. Nx can also distribute tasks across multiple machines while preserving the developer experience of running it on a single machine.

This works because Nx's computation caching and distributed task execution work on the process level. It doesn't matter what build means. It can be an npm script, a custom Nx executor, a Gradle task. Nx will handle it in the same way.

Adding Plugins

As you can see, the core of Nx is generic, simple, and unobtrusive. Nx Plugins are completely optional, but they can really level up your developer experience. Watch this video to see the plugins in actions.