The project is primarily composed of two areas:
-
UI -
src/explore-education-statistics-frontend
- Next.js React app
- Depends on:
- Content API
- Data API
- Notifier
-
Content API -
src/GovUk.Education.ExploreEducationStatistics.Content.Api
- Private API providing content for UI app
- Depends on:
- Publisher - to generate its cache
-
Data API -
src/GovUk.Education.ExploreEducationStatistics.Data.Api
- Private API providing statistics for UI app
- Depends on:
- SQLServer
statistics
database (known aspublic-statistics
in non-local environments)
- SQLServer
-
Notifier
- Azure function for adding users to GOV.UK Notify
-
Public Data API -
src/GovUk.Education.ExploreEducationStatistics.Public.Data.Api
- Public API providing statistics for public users
- Depends on:
- Postgres
public_data
database
- Postgres
-
UI -
src/explore-education-statistics-admin
- CRA React app
- Depends on:
- Admin API
-
Admin API -
src/GovUk.Education.ExploreEducationStatistics.Admin
- Depends on:
- SQLServer
content
database - SQLServer
statistics
database - Publisher
- Notifier
- Data Processor
- Public Data Processor
- SQLServer
- Depends on:
-
Publisher -
src/GovUk.Education.ExploreEducationStatistics.Publisher
- Azure function for publishing admin content to the public frontend
- Depends on:
- SQLServer
content
database - SQLServer
statistics
database
- SQLServer
-
Notifier -
src/GovUk.Education.ExploreEducationStatistics.Notifier
- Azure function for sending notifications
-
Data Processor -
src/GovUk.Education.ExploreEducationStatistics.Data.Processor
- Azure function for handling data set imports into the admin.
- Also referred to as the 'importer' or just 'processor'.
- Depends on:
- SQLServer
content
database - SQLServer
statistics
database
- SQLServer
-
Public Data Processor -
src/GovUk.Education.ExploreEducationStatistics.Public.Data.Processor
- Azure function for processing data sets so they can be used in the public API.
- Depends on:
- Postgres
public_data
database
- Postgres
You will need the following groups of dependencies to run the project successfully.
To run applications in this service you will require the following:
To run the databases:
To emulate Azure storage services (blobs, tables and queues) you will require one of the following options.
- Azurite for Docker and Docker Compose - recommended, see Setting up the storage emulator.
- You will also need to configure the storage emulator host
- Alternatively, if opting to not use a storage emulator at all, you could create your own Storage Account on Azure and amend your storage connection strings to point to this.
We use PNPM and PNPM workspaces to manage our dependencies. PNPM is a drop in replacement for NPM which has several advantages over its predecessor. You can read more about the benefits of PNPM here. This is installed & managed via corepack.
Corepack is a tool installed as part of your Node.js installation that allows you to install and
manage multiple package manager versions in your environment based on per-project configuration
(via the packageManager
field in package.json
).
We use corepack to ensure that everyone is using the same version of PNPM to avoid any issues when people are using different versions of PNPM.
In order to install Corepack and the right version of PNPM, run the following command:
corepack enable
If for some reason the above corepack command doesn't work, you can install PNPM manually by running:
PNPM_VERSION=$(node -e "console.log(require('./package.json').engines.pnpm)")
curl -fsSL https://get.pnpm.io/install.sh | env PNPM_VERSION=$PNPM_VERSION sh -
This will install the package manager version specified in the package.json
file. You can check
that this has been installed by running:
pnpm -v
Install the project's PNPM dependencies by simply running:
pnpm i
Add the following to your hosts
file:
127.0.0.1 db
127.0.0.1 data-storage
- On Unix this is located in
/etc/hosts
. - On Windows this is located in
C:\Windows\System32\drivers\etc\hosts
.
This step is only required if you are using Keycloak as the identity provider, or if you are using a custom identity provider and prefer to use a 'nicer' URL instead of http://localhost.
Add the following to your hosts
file:
127.0.0.1 ees.local
There are two options for setting up your database:
We regularly create new development databases that are uploaded to Google Drive. Ask a team member if you need access.
These are already bootstrapped with seed data to run tests and start the project. This is the recommended way of running the project.
The seed data file names are suffixed with a number to identify the latest. Download the most recent
ees-mssql-data-<number>.zip
and extract the ees-mssql
directory into the project's data
directory.
You must give all OS users appropriate access to this directory.
In Linux:
- The ees-mssql folder needs to be present in an unencrypted folder / partition. The
ees-mssql
folder in the unencrypted location can then be symlinked in to thedata
folder usingln -s /path/to/unencrypted/ees-mssql /path/to/ees/data/ees-mssql
. - The Docker container user needs ownership fo the ees-mssql folder. Run
sudo chown -R 10001 /path/to/ees-mssql
to give this Docker user (with id 10001) appropriate permissions.
All the data in the data/ees-mssql
directory will be mounted and loaded automatically when the SQL Server Docker container starts.
The service can be started against a set of non-existent database. If no pre-existing content
or
statistics
databases yet exist on the target SQL Server instance:
-
Start the SQL Server Docker container:
docker compose up -d db
-
Create empty
content
andstatistics
databases. -
Perform a one-off creation of database logins and users. Using Azure Data Studio or similar, connect to these new databases and run:
-- Against the `master` database CREATE LOGIN [adminapp] WITH PASSWORD = 'Your_Password123'; CREATE LOGIN [importer] WITH PASSWORD = 'Your_Password123'; CREATE LOGIN [publisher] WITH PASSWORD = 'Your_Password123'; CREATE LOGIN [notifier] WITH PASSWORD = 'Your_Password123'; CREATE LOGIN [content] WITH PASSWORD = 'Your_Password123'; CREATE LOGIN [data] WITH PASSWORD = 'Your_Password123'; CREATE LOGIN [public_data_processor] WITH PASSWORD = 'Your_Password123'; -- Against the `content` database CREATE USER [adminapp] FROM LOGIN [adminapp]; ALTER ROLE [db_ddladmin] ADD MEMBER [adminapp]; ALTER ROLE [db_datareader] ADD MEMBER [adminapp]; ALTER ROLE [db_datawriter] ADD MEMBER [adminapp]; ALTER ROLE [db_securityadmin] add member [adminapp]; GRANT ALTER ANY USER TO [adminapp]; -- Against the `statistics` database CREATE USER [adminapp] FROM LOGIN [adminapp]; ALTER ROLE [db_ddladmin] ADD MEMBER [adminapp]; ALTER ROLE [db_datareader] ADD MEMBER [adminapp]; ALTER ROLE [db_datawriter] ADD MEMBER [adminapp]; ALTER ROLE [db_securityadmin] add member [adminapp]; GRANT ALTER ANY USER TO [adminapp]; GRANT EXECUTE ON TYPE::IdListGuidType TO [adminapp]; GRANT EXECUTE ON OBJECT::FilteredFootnotes TO [adminapp]; GRANT SELECT ON OBJECT::geojson TO [adminapp];
This will create contained users for the
content
andstatistics
databases as well as allowing theadminapp
user to manage the permissions of the contained users. -
Start the Admin project and this will configure the contained users' permissions via database migrations. The other projects will then be able to be started, using their own contained users to connect to the databases.
The Azurite Docker container can be started by one of the following methods:
-
Indirectly by starting the admin via the start script
pnpm start admin
-
Directly via the start script using:
pnpm start dataStorage
-
Directly via Docker Compose
docker compose up data-storage
The project uses an OpenID Connect Identity Provider (IdP) to allow login to the Admin service.
An out-of-the-box IdP is provided for ease of setup which runs Keycloak in a Docker container and contains a number of users for different roles. Alternatively, you can follow the steps to use a custom Identity Provider.
The Keycloak Docker container can be started by one of the following methods:
-
Indirectly by starting the admin via the
start
script:pnpm start admin
This will start the Keycloak container before the admin starts if you haven't created a custom
appsettings.Idp.json
file (see earlier). -
Directly via the
start
script:pnpm start idp
-
Directly via Docker Compose:
src docker compose up idp
All the standard seed data users can be supported with Keycloak, and use their standard email addresses and the
password password
to log in.
The standard accounts used day to day are:
- bau1 - username
bau1
and passwordpassword
- bau2 - username
bau2
and passwordpassword
- analyst - username
analyst
and passwordpassword
- analyst2 - username
analyst2
and passwordpassword
The Keycloak Admin login is available with username admin
and password
admin
. From here, users and OpenID Connect settings can be administered.
Additional seed data users can be added to Keycloak by manually adding new entries to the "users" array in
keycloak-ees-realm.json, ensuring to supply unique GUIDs to the user
and
credentials
Ids. If copying and pasting from an existing user record in the array, the new user password will be
"password" also.
After this, existing Keycloak Docker containers will need to be rebuilt in order to pick up the new user list.
To do this, you can run one of the following:
# Using start script
pnpm start idp --rebuild-docker
# Using Docker
docker compose up --build --force-recreate idp
If you have your own custom OpenID Connect identity provider (IdP), you can provide its configuration by creating
an appsettings.Idp.json
file in the src/GovUk.Education.ExploreEducationStatistics.Admin
project directory, which is
excluded from Git.
You can use appsettings.Keycloak.json as a template for how the configuration should be structured.
Note that it must have Implicit Flow enabled and be using the OpenID Connect protocol. It must be set to issue ID Tokens.
If you wish, you can explicitly choose which config to load using the IdpConfig
environment variable:
IdpConfig=Keycloak # To use default Keycloak
IdpConfig=Idp # To use custom IdP
This might be useful if you want to toggle between Keycloak and your custom IdP config, but otherwise,
just remove the appsettings.Idp.json
to default back to Keycloak.
If you are wanting to use Keycloak but with a fresh database, set the following environment variable:
BootstrapUsers=Keycloak
This environment variable tells the Admin application to generate a set of BAU users on startup that are specified in the appsettings.KeycloakBootstrapUsers.json file.
This allows immediate use of the service with Keycloak against an empty database, as corresponding users will now be in both Keycloak and in the SQL Server database.
Shared test credentials are available to team members for use during development and testing. This step is only required for creating additional users, or for non-team members within the open source community to create initial users.
If you are using your own IdP config (via appsettings.Idp.json
), you can bootstrap users who you want to
have access to the system straight away by creating a appsettings.IdpBootstrapUsers.json
, which is excluded from Git.
This should contain a set of user emails in a format similar to appsettings.KeycloakBootstrapUsers.json.
Then set the following environment variable before starting the Admin:
BootstrapUsers=Idp
A good way of running applications/functions is directly through an IDE like Rider.
Alternatively, you can use our start
script. This is a simple wrapper around the various CLI
commands you need to start the applications.
This can be most easily accessed via from the root package.json as pnpm start
.
You will need to ensure you have all the project dependencies as specified in Requirements.
The script provides additional information about its usage by adding the --help
flag:
pnpm start --help
Examples:
-
To start the public frontend backend APIs:
pnpm start data content
-
To start the public frontend:
pnpm start frontend
-
To start the admin (front and backend):
pnpm start admin
-
To start admin dependency services:
pnpm start publisher processor publicProcessor
-
To start the public data API:
pnpm start publicData
If running the
start
script for the first time, and using the seed data downloaded for the recommended database setup, you may encounter an SQL error implying a login failure due to the server being in "script upgrade mode", this is due to the data still being processed by SQL Server. Depending on the size of the data, this may take a minute or two to complete, after which thestart
script can be re-ran.
The frontend applications can be accessed via:
http://localhost:3000
for the public frontendhttps://localhost:5021
for the admin frontend
A nice and convenient way to access the start script from anywhere on your machine is to create a
custom function in your .bashrc
, .zshrc
or similar.
This function would look like the following (change to your liking):
function ees()
{
(cd your-ees-directory && pnpm start $*)
}
You would then be able to use this like:
ees content data
Out of the box, each sub-project should come with a default set of environment variables (in .env
)
that should work for development.
If you need to change any environment variables only for your local environment, you can create a
corresponding .env.local
file which will be loaded in preference to .env
. .env.local
is
ignored by Git.
Various .env.{env}
files are checked into Git for use in our deployment environments, but will not
generally be used for local development.
Required environment variables should be supplied to both the specific .env.{env}
file and the
.env.example
file (example versions of variables should be placed here).
The .env.example
file is used to validate that the .env.{env}
file in use is not missing any
required variables and consequently needs to be in sync with any changes.
No secrets/keys etc. should be added to these environment variables.
The project currently uses PNPM and PNPM workspaces to handle dependencies as we have adopted a monorepo project structure and have dependencies between sub-projects. These dependencies are established using symlinks that PNPM creates.
-
explore-education-statistics-admin
- Contains the admin frontend application.
- Single page application based on Create React App.
-
explore-education-statistics-frontend
- Contains the public frontend application.
- This is a server side rendered Next application.
-
explore-education-statistics-common
- Contains common code between the other sub-projects for re-use.
-
explore-education-statistics-ckeditor
-
Contains the customised CKEditor build for the admin application.
When adding new NPM dependencies, be aware that we need to be careful about where we add them in the
package.json
file. We very deliberately add our dependencies to either devDependencies
or
dependencies
depending on the subproject.
-
explore-education-statistics-frontend
dependencies are in eitherdependencies
ordevDependencies
to avoid including build dependencies in thenode_modules
that are deployed to environments.This is beneficial for cutting down build times, but in the past, we've also experienced weird issues with being unable to deploy to the Azure App Service when there are too many
node_modules
(related to Windows). -
explore-education-statistics-commmon
dependencies are independencies
as these must all be included in the final build (admin or public). -
explore-education-statistics-admin
dependencies are independencies
simply for consistency and simplicity. We need all dependencies to create the build, so it doesn't make sense to split out separatedevDependencies
.
To install new dependencies, you will need to use PNPM to do this, with the following steps:
- Directly add dependencies to any required
package.json
file(s). - Run the following:
pnpm i
During development, you might end up in an inconsistent state where your sub-project node_modules
are broken for whatever reason. Consequently, it is advisable to clean down your sub-project
node_modules
by running the following from the project root.
pnpm clean
These scripts can generally be run from most package.json
files across the project.
-
pnpm test
- Run all tests. -
pnpm tsc
- Run Typescript compiler to check types are correct. Does not build anything. -
pnpm lint
- Lint projects using Stylelint and ESLint.pnpm lint:js
- Run ESLint only.pnpm lint:style
- Run Stylelint only.
-
pnpm fix
- Fix any lint that can be automatically fixed by the linters.pnpm fix:js
- Fix only ESLint lints.pnpm fix:style
- Fix only Stylelint lints.
-
pnpm format
- Format codebase using Prettier.
These can only be run from the project root package.json
.
pnpm clean
- Remove anynode_modules
directories across any sub-projects.
These can only be run from a sub-project package.json
.
pnpm start
- Start a sub-project dev server.
We enforce the project code style via ESLint and Stylelint. Both are configured with the following rule sets:
- Airbnb style guide for TypeScript/JavaScript.
- typescript-eslint recommended rules for TypeScript.
- stylelint-config-sass-guidelines for SCSS.
We also combine this with Prettier to format our code to avoid disagreements on formatting.
To enforce these code styles, we run linting and formatting tasks upon save, commit and build.
Typically, we should aim to stick as close as possible to the rule set defaults. However, we do allow exceptions for rules that:
- Are incompatible with existing/legacy code
- Make the developer experience considerably worse
- Are buggy and don't work correctly
- The team generally disagrees with
If required, you can disable linting upon save by adding one (or both) of the following to your
.env.local
:
ESLINT_DISABLE=true
STYLELINT_DISABLE=true
Of course, we don't recommend that you do this as it's usually a good idea to get immediate feedback that something is wrong.
Note that linting will still run upon commit. Ideally, every commit should have the project in a buildable state (rather than be completely broken).
.NET code style and formatting rules are imposed by dotnet format because it is IDE and platform agnostic, and natively included in the .NET SDK. Rule are dictated by an .editorconfig file.
The warnings and errors about violations of these rules are surfaced at build time by enabling Enforce Code Style In Build () in each .csproj
file.
To that end - regardless of your IDE of choice - if your solution builds without warnings then your changes should not have caused any style violations.
NOTE We currently have a large number of warnings, many of which we're present before introducing an .editorconfig
file, and many more of which have been created by imposing style rules we previously weren't. Once the work has been done to resolve these warnings, we can turn on TreatWarningsAsErrors
to make the build itself fail if there are style violations.
Because the .editorconfig
file is attached to the solution, both Rider and Visual Studio should automatically use its rules in place of their own defaults.
- on Visual Studio, you may need to visit
Analyze -> Code Cleanup -> Configure Code Cleanup
, then change your default profile to include only theFix all warnings and errors set in EditorConfig
step. - on Rider, you may need to visit
Settings -> Editor -> Inspection Settings
then ensureRead settings from editorconfig, project settings and rule sets
is ticked.
The easiest way is probably through an IDE, configured as per the step above. Most IDEs can format single files if you right-click it in the explorer.
Failing that, the faffier harder way is to pass an --include <PATH>
argument to dotnet format
on the command line, providing it the files you want it to format.
You can perform the same check as the pre-commit hooks / GitHubs actions do and receive a report by running the following command from the repository root:
dotnet format src/GovUk.Education.ExploreEducationStatistics.sln --verify-no-changes --report dotnet-format-report.json
This will create a report named dotnet-format-report.json
in the repository root. This has been added to the .gitignore
file so should not cause a tracked change.
The --verify-no-changes
argument tells format
to make no changes. If you want it to automatically apply fixes, simply remove this argument:
dotnet format src/GovUk.Education.ExploreEducationStatistics.sln
If you want to run it against one or more specific directories in the solution (or indeed exclude one or more), these can be specified through the --include <PATH>
and --exclude <PATH>
arguments.
If you want to see only errors, or include suggestions, pass a new value to the severity argument (accepted values are error
, warn
, and info
):
dotnet format src/GovUk.Education.ExploreEducationStatistics.sln --severity <SEVERITY>
See the docs for more on this.
The backend c# projects use code first migrations to generate the application's database schema. The migration tool is installed by running:
dotnet tool restore
To generate a migration for the content db:
cd src/GovUk.Education.ExploreEducationStatistics.Admin
dotnet ef migrations add EES1234_MigrationNameHere --context ContentDbContext --output-dir Migrations/ContentMigrations -v
To generate a migration for the statistics db:
cd src/GovUk.Education.ExploreEducationStatistics.Data.Api
dotnet ef migrations add EES1234_MigrationNameHere --context StatisticsDbContext --project ../GovUk.Education.ExploreEducationStatistics.Data.Model -v
To generate a migration for the UsersAndRolesDbContext:
cd src/GovUk.Education.ExploreEducationStatistics.Admin
dotnet ef migrations add EES1234_MigrationNameGoesHere --context UsersAndRolesDbContext --output-dir Migrations/UsersAndRolesMigrations -v
To generate a migration for the public data API db:
cd src/GovUk.Education.ExploreEducationStatistics.Public.Data.Api
dotnet ef migrations add EES1234_MigrationNameHere --context PublicDataDbContext --project ../GovUk.Education.ExploreEducationStatistics.Public.Data.Model -v
The backend c# projects have a number of unit and integration tests. From the project root, run:
cd src
dotnet clean
dotnet test
Note that the clean
is necessary due to an issue with AspectInjector
whereby compilation of code over already-compiled code will add AOP execution code on top of existing AOP execution
code, leading to AOP code being invoked multiple times rather than just once. This would result in test failures, as we
assert that AOP code is executed only once.
Due to the resource requirements of the integration tests, Linux users need to ensure that the system running the tests
is capable of doing so. The max_user_watches
and max_user_instances
settings must be set to a high enough limit,
for example by running:
echo fs.inotify.max_user_watches=524288 | sudo tee -a /etc/sysctl.conf
echo fs.inotify.max_user_instances=524288 | sudo tee -a /etc/sysctl.conf
sudo sysctl -p
During development you might want to reset your Azurite instance to clear out all data from blobs, queues and tables. This is typically done at the same time as resetting the databases.
To delete all data in Azurite simply delete the Azurite docker container, remove the Azurite volume and recreate it:
docker compose up data-storage
If wanting to add more users to the standard set of users we use and are using Keycloak as the Identity Provider, the users will firstly need to be added to Keycloak in the EES realm and then the realm exported. To export the realm you can run:
docker exec -it ees-idp /opt/jboss/keycloak/bin/standalone.sh -Djboss.socket.binding.port-offset=100 -Dkeycloak.migration.action=export \
-Dkeycloak.migration.provider=singleFile -Dkeycloak.migration.realmName=ees-realm -Dkeycloak.migration.usersExportStrategy=REALM_FILE -Dkeycloak.migration.file=/tmp/new-ees-realm.json
Wait for the above process to complete by waiting for the console output Admin console listening on http://127.0.0.1:10090
, then shut it down.
Then simply copy the file from the /tmp/new-ees-realm.json
file in the ees-idp
container to src/keycloak-ees-realm.json
in order for future restarts
of the IdP to use this new realm configuration. From the project root, run:
docker cp ees-idp:/tmp/new-ees-realm.json docker/keycloak/keycloak-ees-realm.json
During manual or automated testing, it is handy to have a way to schedule releases for publishing but to trigger that process to occur on demand, rather than having to wait for a lengthly period before the scheduled Publisher Functions run. For this, we provide 2 Functions that can be triggered by HTTP requests; one stages scheduled Releases, whilst the other completes the publishing process for any staged Releases and makes them live.
See the Publisher Functions README for more information.
In some cases, it may be useful to change the version of the magic.mgc
file that is used for
file validation in a project.
You can create an appsettings.Local.json
file in the relevant project e.g.
touch src/GovUk.Education.ExploreEducationStatistics.Data.Processor/appsettings.Local.json
Then ensure it has the following:
{
"MagicFilePath": "/usr/lib/file/magic.mgc"
}
The above example uses the default magic.mgc
used by Ubuntu, but you can change this path to
whatever you want on your filesystem.
Aside from unit tests for each project, we maintain suites of Robot Framework tests that can be found in tests
.
See the Robot Framework tests README for more information.
This application is licensed under the MIT License.