We're using a monorepo to manage our packages, as it becomes cumbersome to manage the overhead of multiple separate repos for each package. Using lerna to manage all of our packages in one repo simplifies the process.
There are many packages located in the packages directory. A few of the more important ones are:
- @deephaven/code-studio: Main web UI used with the deephaven-core backend. This package is the main front-end application, and depends on almost all other packages in the repository. It is often the easiest way to see the effect of your changes by opening this application. Follow the instructions in the code-studio README.md to get it started.
- @deephaven/components: Component library used within the web UI.
- @deephaven/grid: High-performance grid component used to display large tables of data.
- @deephaven/dashboard: Dashboards used in @deephaven/code-studio for displaying and organizing panels.
- @deephaven/golden-layout: Layout framework used in @deephaven/dashboard.
For details on how to contribute to this repository, please see the contributing guidelines.
We are using node 18.x and npm 8.x. If you are using nvm, there is an .nvmrc file, so just run nvm install
to get the latest 18.x/8.x versions (or set up your environment to automatically switch). Otherwise, download from the node homepage.
In order to use the UI, you must also be running a deephaven-core server on port 10000. The server provides APIs that the web-client-ui depends upon. An easy way to get started is to launch a Deephaven container from the quick start guide.
We recommend using Visual Studio Code and installing the recommended workspace extensions which VS Code will suggest when you open the repo or when you browse the extensions panel. There are a few workspace settings configured with the repo.
If using Linux, we recommend installing directly from the deb or rpm file from the VSCode website rather than through a package manager or store such as snap
or flatpak
. The other install methods (for example, apt on Ubuntu installs via snap) may end up in a sandboxed environment that cannot use the debug launch configs properly.
We use Chrome for development with the React and Redux extensions.
- React Developer Tools: Allows inspection/changing the props/state of react components.
- Redux DevTools: Inspect the redux store data.
npm install
: Install all dependencies and automatically bootstrap packages. Should be run before any of the other steps.npm start
: Start building all packages and watching them (when possible). Use when you're developing, and your changes will be picked up automatically. Servers will open up for code-studio and embed-widget. These will open on localhost on ports 4000 and 4010 respectively.npm test
: Start running tests in all packages and watching (when possible). Use when you're developing, and any breaking changes will be printed out automatically. See Unit tests for more details.npm run build
: Create a production build of all packages. Mainly used by CI when packaging up a production version of the app.npm run preview
: Runs the Vite preview server for the built code-studio and embed-widget. These will open on ports 4000 and 4010 respectively.npm run e2e
: Runs the Playwright end-to-end tests locally. See E2E Tests for more details.
If your DHC address is different from the default http://localhost:10000
, edit .env.local
in each package to point to your local DHC. This is needed for the session websocket and for things like notebooks to be proxied correctly by Vite.
VITE_PROXY_URL=http://<dhc-host>:<port>
The plugins repo supports serving plugins locally. DHC can be configured to proxy js-plugins
requests to the local dev server by setting VITE_JS_PLUGINS_DEV_PORT
in packages/code-studio/.env.development.local
.
e.g. To point to the default dev port:
VITE_JS_PLUGINS_DEV_PORT=4100
If you'd like to override the vite config for local dev, you can define a packages/code-studio/vite.config.local.ts
file that extends from vite.config.ts
. This file is excluded via .gitignore
which makes it easy to keep local overrides in tact.
The config can be used by running:
npm run start:app -- -- -- --config=vite.config.local.ts
For example, to proxy js-plugins
requests to a local server, you could use this vite.config.local.ts
:
export default defineConfig((config: ConfigEnv) => {
const baseConfig = (createBaseConfig as UserConfigFn)(config) as UserConfig;
return {
...baseConfig,
server: {
...baseConfig.server,
proxy: {
...baseConfig.server?.proxy,
'/js-plugins': {
target: 'http://localhost:5173',
changeOrigin: true,
rewrite: path => path.replace(/^\/js-plugins/, ''),
},
},
},
};
});
We have a pre-defined launch config that lets you set breakpoints directly in VSCode for debugging browser code. The Launch Deephaven
config will launch a new Chrome window that stores its data in your repo workspace. With this setup, you only need to install the React and Redux devtool extensions once. They will persist to future launches using the launch config.
We prefer launching a new window instead of attaching to existing windows because it provides a cleaner debug environment (only development extensions). You would also need to launch Chrome with the remote debugging flag in order to attach to an existing instance.
If you are using Linux, you will likely need to use the direct install from the VSCode website (deb or rpm file) and not through a package manager. On Ubuntu, apt installs via snap
and some Linux flavors may use flatpak
. Both of these sandbox the VSCode instance in such a way that the launch debug configs will likely not work.
If you are not using Chrome (e.g. Chromium), you may need to do one of the following if the launch config is not working.
-
Alias
google-chrome-stable
to launchchromium-browser
. The launch config by default should try to launchgoogle-chrome-stable
, so if you can launch the browser withgoogle-chrome-stable
from a terminal, it should work. -
Add a new configuration to VSCode and copy the launch config from
settings.json
. Then add theruntimeExecutable
prop to point to your browser executable. VSCode unfortnuately does not merge workspacesettings.launch
with workspacelaunch.json
, so if we add more launch configs you would need to copy to your.vscode/launch.json
file to get the configs.
Depending on what your package is, there are a couple of different templates that may be appropriate.
A standalone application with it's own entry point. Recommend copying the embed-widget package, removing any dependencies and files not required.
A component/library package that can be imported into other packages. Recommend copying the components package, removing any dependencies and files not required.
We use Jest for running unit tests. We use an eslint runner in Jest to lint files as tests. We also use react-testing-library and react-hooks-testing-library for testing react components and hooks.
Run npm test
to start Jest in watch mode. By default, this will only run tests that have changed relative to your git origin/main. Keeping your origin/main up to date will help keep the number of tests run to a minimum.
While the tests are running, you can press p
to filter by filename or test name. You can also press Shift+P
to see a list of projects and disable certain projects like the eslint or stylelint.
Log messages from our @deephaven/log package are disabled by default in tests in jest.setup.ts. To change the log level, set the DH_LOG_LEVEL
env variable. For example, to enable all log messages run DH_LOG_LEVEL=4 npm test
.
Note that log messages from other sources such as react prop types will still be printed since they do not go through our logger.
If you want to collect coverage locally, run npm test -- --coverage
Unit tests can be debugged by running jest with the --inspect-brk
flag and attaching to the node process in vscode's debugger. There are 2 launch configs that make this easier:
-
Debug Jest Tests - This will prompt you for a test name or pattern and will then run tests in watch mode with an attached debugger.
-
Attach to Node Process - This will attempt to attach to an existing node process running with
--inspect-brk
. There is an npm scripttest:debug
that can start the process for you. By default, it will run with the same configuration asnpm test
, but you can supply additional parameters to narrow the scope of tests being run.e.g.
npm run test:debug ThemeUtils
would only run modules with "ThemeUtils" in the name.
We use Playwright for end-to-end tests. We test against Chrome, Firefox, and Webkit (Safari). Snapshots from E2E tests are only run against Linux so they can be validated in CI.
You should be able to pass arguments to these commands as if you were running Playwright via CLI directly. For example, to test only table.spec.ts
you could run npm run e2e -- ./tests/table.spec.ts
, or to test only table.spec.ts
in Firefox, you could run npm run e2e -- --project firefox ./tests/table.spec.ts
. See Playwright CLI for more details.
Snapshots are used by end-to-end tests to visually verify the output. Snapshots are both OS and browser specific. Sometimes changes are made requiring snapshots to be updated. Update snapshots locally by running npm run e2e:update-snapshots
.
Once you are satisfied with the snapshots and everything is passing locally, you need to use the docker image to update snapshots for CI (unless you are running the same platform as CI (Ubuntu)). Run npm run e2e:update-ci-snapshots
to update the CI snapshots. The snapshots will be written to your local directories. The Linux snapshots should be committed to git (non-Linux snapshots should be automatically ignored by git).
Note that while both the GH actions and docker configuration use Ubuntu 22.04 images, their configurations are not identical. One known difference is the available system fonts. In some cases this can cause snapshots to differ when running locally vs in CI such as when rendering unicode characters. To mitigate this, some of our e2e tests have been configured to ensure a consistent unicode font fallback.
- The
DejaVu Sans
font gets installed via the Dockerfile. It already exists in the CI environment. - Font family stacks that involve unicode characters can explicitly fallback to
DejaVu Sans
if they impact snapshots. e.g We do so in Grids.tsx by setting the canvas font to12px Arial, "DejaVu Sans", sans-serif
These scripts are recommended while developing your tests as they will be quicker to iterate and offer headed mode for debugging. You will need to run npx playwright install
before using these scripts. You may also need to update the browsers with the same command any time the Playwright version is updated. See Playwright Browsers for more details.
You must have a web UI running on localhost:4000/ide/
(dev mode and build preview are both fine), and deephaven-core
running on port 10000 with anonymous authentication for E2E tests and application mode using the ./tests/docker-scripts/data/app.d folder. For example, the command you would run from the deephaven-core repo would be (replacing the path to point to the app.d directory on your machine):
START_OPTS="-DAuthHandlers=io.deephaven.auth.AnonymousAuthenticationHandler -Ddeephaven.application.dir=/path/to/web-client-ui/tests/docker-scripts/data/app.d" ./gradlew server-jetty-app:run
See this guide for more details on anonymous authentication.
npm run e2e
: Run end-to-end tests in headless mode.npm run e2e:headed
: Runs end-to-end tests in headed debug mode. Also ignores snapshots since a test suite will stop once 1 snapshot comparison fails. Useful if you need to debug why a particular test isn't working. For example, to debug thetable.spec.ts
test directly, you could runnpm run e2e:headed -- ./tests/table.spec.ts
.npm run e2e:codegen
: Runs Playwright in codegen mode which can help with creating tests. See Playwright Codegen for more details.npm run e2e:update-snapshots
: Updates the E2E snapshots for your local OS.
These scripts will create a production build from your current code and run the tests in a Docker image. There is no need to install any other dependencies or have any other services running as they are all contained within Docker. Since any source code changes will require a new build, these scripts are only recommended if having issues with CI or if you need to update the snapshots for CI.
npm run e2e:docker
: Runs the E2E tests in a docker image. This is useful for testing how tests should run in the CI environment locally.npm run e2e:update-ci-snapshots
: Updates the E2E snapshots for CI.
All new changes (bug fixes, feature requests) are merged to main
so they are always included in the latest release. We use semantic versioning for major/minor/patch releases.
We use 3 release types
-
stable
- Stable releases are created periodically off of themain
with the dist-taglatest
. These will include an appropriate version bump and release notes, detailing the changes that are in that version. -
nightly
- Nightly releases are published every night with the dist-tagnightly
to npm. You can reference the nightly release to always be on the latest by referencingnightly
as the version, though stability is not guaranteed, e.g.npm install --save @deephaven/grid@nightly
. -
hotfix
- For Long Term Support releases (versions we consume in the enterprise product), we create a new branch in Community matching the LTS version number (e.g. release/v0.6). Bug fixes/hotfixes are then either cherry-picked frommain
(if the fix has been merged to main), or directly merged into the hotfix branch (if code has changed inmain
and the fix only applies in the hotfix branch). -
alpha
/other
- For publishing WIP branches to NPM or testing CI/publishing changes. There is a GitHub Action calledPublish Alpha
which will prompt you for apreid
andref
to create the alpha version. These will publish tonpm
under thecanary
tag. The version number will bex.yy.z-alpha.0
or using thepreid
specified.
Note: Only repo admins can do this. These steps apply to main
and any release/*
branches.
We use lerna and Conventional Commits to automatically handle incrementing the version, generate the changelog, and create the release.
-
Generate a GitHub Personal access token:
-
Under
Repository Access
, selectOnly select repositories
and adddeephaven/web-client-ui
. -
Under
Repository Permissions
, setAccess: Read and write
forContents
. This will be necessary to push your version bump and create the release. -
Copy the token created and replace
<token>
with it in the next step.
-
-
Checkout the branch you want to release.
main
orrelease/*
-
Run this npm script to bump the version, update the changelog, and create a release on GitHub. If your local remote name for the upstream repo is not origin, then set the proper
git-remote
(e.g.upstream
).GH_TOKEN=<token> npm run version-bump -- --git-remote origin
-
IMPORTANT: If releasing a
release/*
branch, you need to reset thelatest
tag on GitHub. To do that, go to the releases page and find the release that is actually thelatest
based on the versionmain
is on.-
Click edit on the actual
latest
release -
Check the box for
Set as the latest release
-
Update release
This ensures that an older version doesn't get marked as
latest
on GitHub. It will not effect npm. -
After the release is created, you can go to the actions page to see the publish action being kicked off.
Note: Only repo admins can do this since release/*
branches are protected.
For our enterprise product, we occasinally need to make hotfixes (patch releases) to older versions that the enterprise client depends on. Since it would be unwise to update an old version of enterprise to the newest community UI components, we keep hotfix release/*
branches. There are a few steps to create one.
-
Checkout the tag for the release where the branch should fork.
-
Create a new branch from this point called
release/vX.YY
. If forking fromv0.31.1
then the branch isrelease/v0.31
. -
Change the
distTag
inlerna.json
tovX.YY
which corresponds to the version from the branch name. This will ensure packages are published to npm under thedistTag
instead of the default oflatest
. -
Commit and push the branch
Once the branch is pushed to origin, new commits will require PRs into the branch. To create a patch release, refer to the Releasing a New Version section.
Periodically dependencies should be updated such that we're using the latest and greatest.
- Security updates: Run
npm audit fix
weekly to ensure any known security vulnerabilities are updated. - Dependency updates: At the beginning of a release cycle, run
npm update
to update thepackage-lock.json
with the latest version of dependencies. Afterwards, runnpm outdated
to see if there are any dependencies with major version changes that can be updated. There are two ways you can upgrade a dependency:- Run a tool like lerna-update-wizard by running
npx lerna-update-wizard
to go through steps to automatically update all child packages, OR - Manually update the
package.json
of all packages with that dependency to the latest version.
- Run a tool like lerna-update-wizard by running
- When updating the major version of dependencies, be sure to check the release notes for any breaking changes/migration notes. After updating dependencies, run an
npm install
andnpm test
to make sure all tests pass.
When adding new features, it can be useful to analyze the final package size to see what's in the package. Use source-map-explorer to see what's taking up the most size in the package. First run npm run build
to build a production bundle, then run npx source-map-explorer 'packages/<package-name>/build/static/js/*.js'
, e.g.:
npm run build
npx source-map-explorer 'packages/code-studio/build/static/js/*.js'
Support is best for Google Chrome and Chromium based browsers (such as Microsoft Edge based on Chromium). We also maintain compatibility with Mozilla Firefox and Apple Safari.
Officially, the following browsers and corresponding configurations are supported:
- Chrome: Versions from the last 0.5 years
- Edge: Versions from the last 0.5 years
- Opera: Versions from the last 0.5 years
- Firefox: Versions from the last 0.5 years, including Firefox ESR
- Safari: Versions from the last 1 year
If you encounter an issue specific to a browser, check that your browser is up to date, then check issues labeled with firefox or safari for a list of known browser compatibility issues before reporting the issue.