To get started locally, follow these instructions:
- If you haven't done it already, make a fork of this repo.
- Clone to your local computer using
git
. - Make sure that you have Node
v14.20
,v16.13
orv18.10
installed. See instructions here. - Make sure that you have
yarn
installed; see instructions here. - Run
yarn
(no arguments) from the root of your clone of this project to install dependencies.
To make a local build:
yarn build --local
This generates a number of tarballs in the dist/
directory. To actually use
the locally built tools, switch to another repository reproducing the specific
issue you want to fix (or just generate a local repo with ng new
). Then
install the locally built packages:
cd "${EXAMPLE_ANGULAR_PROJECT_REPO}"
npm install -D ${CLI_REPO}/dist/*.tgz
Builds of this example project will use tooling created from the previous local build and include any local changes. When using the CLI, it will automatically check for a local install and use that if present. This means you can just run:
npm install -g @angular/cli
to get a global install of the latest CLI release. Then running any ng
command
in the example project will automatically find and use the local build of the
CLI.
Note: If you are testing ng update
, be aware that installing all the tarballs
will also update the framework (@angular/core
) to the latest version. In this
case, simply install the CLI alone with
npm install -D ${CLI_REPO}/dist/_angular_cli.tgz
, that way the rest of the
project remains to be upgraded with ng update
.
To debug an invocation of the CLI, build and install the CLI for an example
project, then run the desired ng
command
as:
node --inspect-brk node_modules/.bin/ng ...
This will trigger a breakpoint as the CLI starts up. You can connect to this
using the supported mechanisms for your IDE, but the simplest option is to open
Chrome to chrome://inspect and then click on the inspect
link for the node_modules/.bin/ng
Node target.
Unfortunately, the CLI dynamically require()
's other files mid-execution, so
the debugger is not aware of all the source code files before hand. As a result,
it is tough to put breakpoints on files before the CLI loads them. The easiest
workaround is to use the debugger;
statement to stop execution in the file you
are interested in, and then you should be able to step around and set breakpoints
as expected.
There are two different test suites which can be run locally:
- Run all tests:
yarn bazel test //packages/...
- Run a subset of the tests, use the full Bazel target example:
yarn bazel test //packages/schematics/angular:angular_test
- For a complete list of test targets use the following Bazel query:
yarn bazel query "tests(//packages/...)"
You can find more info about debugging tests with Bazel in the docs.
- Compile the packages being tested:
yarn build
- Run all tests:
node tests/legacy-cli/run_e2e.js
- Run a subset of the tests:
node tests/legacy-cli/run_e2e.js tests/legacy-cli/e2e/tests/i18n/ivy-localize-*
- Run on a custom set of npm packages (tar files):
node tests/legacy-cli/run_e2e.js --package _angular_cli.tgz _angular_create.tgz dist/*.tgz ...
When running the debug commands, Node will stop and wait for a debugger to attach. You can attach your IDE to the debugger to stop on breakpoints and step through the code. Also, see IDE Specific Usage for a simpler debug story.
When debugging a specific test, change describe()
or it()
to fdescribe()
and fit()
to focus execution to just that one test. This will keep the output clean and speed up execution by not running irrelevant tests.
Some additional tips for developing in specific IDEs.
To load the project in Intellij products, simply Open
the repository folder.
Do not Import Project
, because that will overwrite the existing
configuration.
Once opened, the editor should automatically detect run configurations in the
workspace. Use the drop down to choose which one to run and then click the Run
button to start it. When executing a debug target, make sure to click the
Debug
icon to automatically attach the debugger (if you click Run
, Node will
wait forever for a debugger to attach).
In order to debug some Angular CLI behaviour using Visual Studio Code, you can run npm run build
, and then use a launch configuration like the following:
{
"type": "node",
"request": "launch",
"name": "ng serve",
"cwd": "<path to an Angular project generated with Angular-CLI>",
"program": "${workspaceFolder}/dist/@angular/cli/bin/ng",
"args": [
"<ng command>",
...other arguments
],
"console": "integratedTerminal"
}
Then you can add breakpoints in dist/@angular
files.
For more informations about Node.js debugging in VS Code, see the related VS Code Documentation.
In order to investigate performance issues, CPU profiling is often useful.
Node.js 16+ users can use the Node.js command line argument --cpu-prof
to create a CPU profile.
node --cpu-prof node_modules/.bin/ng build
In addition to this one, another, more elaborated way to capture a CPU profile using the Chrome Devtools is detailed in angular#8259 (comment).
You can use the Chrome Devtools to process it. To do so:
- open
chrome://inspect
in Chrome - click on "Open dedicated DevTools for Node"
- go to the "profiler" tab
- click on the "Load" button and select the generated
.cpuprofile
file - on the left panel, select the associated file
Adding a package to this repository means running two separate commands:
schematics devkit:package PACKAGE_NAME
. This will update the.monorepo
file, and create the base files for the new package (package.json, src/index, etc).devkit-admin templates
. This will update the README and all other template files that might have changed when adding a new package.
For private packages, you will need to add a "private": true
key to your package.json manually.
This will require re-running the template admin script.