pub.dev
runs on Google AppEngine, but we also support a local,
in-memory server that is able to run 95-99% of the site on a single
machine, without any need for AppEngine account or setup.
-
If you want to contribute to
pub.dev
, the current document will give you instructions how to work with the local, in-memory server (fake_server
). -
If you want to setup the project on Google AppEngine, follow this guide.
-
Development and runtime support is best on Linux, compatible system may work, Windows isn't supported yet (e.g. bash scripts are not compatible).
The main directories that one would be looking at:
app/
: contains the main server application.pkg/web_app
: contains the client-side part of the app.pkg/web_css
: contains the SCSS files.static
: contains static files and images deployed alongside the app.third_party
: contains 3rd-party assets that are deployed alongside the app.
The in-memory server starts with no data. Prepopulated data can be created with test-profiles, using the following steps:
A test-profile is a short description of packages, publishers, users, their flags and relations.
To create an initialize it with some minimal data, create the following YAML file:
defaultUser: '[email protected]'
packages:
- name: retry
- name: http
publishers:
- # Only `example.com`` and `verified.com`` will verify.
# See app/lib/fake/backend/fake_domain_verifier.dart
name: example.com
members:
- email: [email protected]
role: admin
Using the test-profile above, the following process will:
- create the publishers and users.
- fetch the latest versions and the archive file from pub.dev.
- publish the archive locally under the name of the user or the publisher in the test-profile description.
- analyze the packages and runs dartdoc on them
- stores the results and all the entities in a local file.
cd app/
dart bin/fake_server.dart init-data-file \
--test-profile=[the file you have created] \
--analysis=[ none | fake | local | worker ] \
--data-file=dev-data-file.jsonl
fake
analysis will use a deterministic, but random-looking process to create analysis results quickly.local
analysis will use the locally available SDKs fromPATH
.worker
analysis will runpana
anddartdoc
inside a docker container, similarly as the production server would run them, using the configured SDK versions.
After the data file has been created, you can start using it locally:
cd app/
dart bin/fake_server.dart run --data-file=dev-data-file.jsonl
The fake_server.dart script picks up changes and rebuilds your style sheets when initializing.
The web app and the API endpoints use a simple mechanism to map access tokens
to authenticated accounts: user-at-domain-dot-com
gets mapped to [email protected]
.
- On the web app, use
localhost:8080/[email protected]
to sign in. - On the API endpoints one should send the
Authorization
header withBearer $token
as value.
The application and various packages uses
builders to generate code based on source
annotations. Input files are usually listed in build.yaml
, and generated
files usually suffixed .g.dart
. To generate code use:
dart run build_runner build
The app uses mono_repo
to organize
multiple packages inside the repository.
To use mono_repo
, first activate it:
dart pub global activate mono_repo
Note: Run
pub global list
to make sure the version is at least2.0.0
.
mono_repo
has two kinds of configuration files:
mono_repo.yaml
(in the root directory)mono_pkg.yaml
(in each package directory)
dart pub global run mono_repo pub get
dart tool/update_mono_pkg_yaml.dart
dart pub global run mono_repo generate
-
Create
mono_pkg.yaml
for the package. (Use the existing ones as template.) -
Run
dart pub global run mono_repo generate
from the root.