This repo is an archival copy of the code that runs civilrights.justice.gov.
In order to be more responsive to the public’s changing communication needs and the increased reporting volume,the Civil Rights Division, in close collaboration with 18F, launched a user-friendly online submission experience at civilrights.justice.gov. The project has transformed the way the Civil Rights Division collects, sorts, and responds to public reports of civil rights violations.
See the project in it's new home: https://github.com/usdoj-crt/crt-portal
Here is a blog post about the launch of this project with more details.
Install Docker
https://www.docker.com/get-started
Create ssh key and add it to GitHub.
Clone the project locally:
git clone [email protected]:usdoj-crt/crt-portal.git
In the top level directory create a .env file in the top of your directory and set SECRET_KEY to a long, random string.
SECRET_KEY=''
To build the project You will need to build the project for the first time and when there are package updates to apply.
docker-compose up -d --build
To run the project This is a quicker way to start the project as long as you don't have new packages to install.
docker-compose up
Visit the site locally at [http://0.0.0.0:8000/report] 🎉
Create a superuser for local admin access
docker-compose run web python /code/crt_portal/manage.py createsuperuser
To add some test data with the form http://0.0.0.0:8000/report and then you can check it out in the backend view http://0.0.0.0:8000/form/view and the admin view at http://0.0.0.0:8000/admin.
Generate the SASS for the front end with gulp:
If you are doing front end work, you will want to have gulp compile the css so you can instantly see changes.
To ensure we are all using the same versions of our front-end dependencies, we use nvm to peg a version of node to this project.
Check that nvm is installed with nvm --version.
If not, run the following command to install it:
curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.35.1/install.sh | bash
source ~/.bash_profile
If you get an error, and don't have a bash_profile file, create one first with touch ~/.bash_profile, then run the command above again.
Then, if this is your first time installing the project or nvm, run nvm install.
Finally, nvm use && npm install
Now to compile the sass files into css, run:
npm run sass:watch
Also note, that the staticfiles folder is the destination of all static assets when you or a script runs manage.py collectstatic so don't make your changes there, or they will be overwritten.
In Django, when you update the data models you need to create migrations and then apply those migrations, you can do that with:
docker-compose run web python /code/crt_portal/manage.py makemigrations
docker-compose run web python /code/crt_portal/manage.py migrate
To install a new Python package, run:
docker-compose run web pipenv install name-of-package
To ssh into your local Docker web container run:
docker exec -it crt-portal_web_1 /bin/bash
To log into your local Docker database for debugging purposes, first run:
docker exec -it crt-portal_db_1 /bin/bash
Then from, within the container, you can run:
psql -U postgres
As a logged-in local Postgres user, you can run queries directly against the database, for example: select * from cts_forms_report; to see report data in your local database.
In production, we use django-auth-adfs and new endpoints are behind authentication by default. To create a public page, you must update LOGIN_EXEMPT_URLS In settings.py to include the endpoint(s) which are to be available without requiring authentication.
We also explicitly add login required to views and functions that need authentication. If you are making a new path that requires authentication, add a test the login required test class.
Important commands to use during internationalization (i18n):
When you run makemessages, Django will search through .py, .txt, and .html files to find strings marked for translation. Django finds these strings through the gettext function or its lazy-loading equivalent (in Python) or the trans function (in HTML). This adds the marked strings to .po files where translators will do their work.
docker-compose run -w /code/crt_portal/cts_forms web django-admin makemessages -l es
After the strings translated, the translation can be compiled back to Django-readable .mo files using run compilemessages:
docker-compose run -w /code/crt_portal/cts_forms web django-admin compilemessages
To add support for a new language, first generate a .po file for the new language using the makemessages command described above. Once the file is generated, open it and set the Language metadata item near the top of the file. This value should be either the language name (Spanish, Chinese Traditional, Korean, etc.) or the language code (es, zh, ko, etc.). In order to properly translate plurals, you will also need to set the Plural-Forms metadata item. Translators should provide and/or verify the appropriate value for this field.
Next, use po-to-xls to convert the .po file to an Excel spreadsheet. This spreadsheet can be generated in a tmp or working directory, and should not be commited to source control. This spreadsheet can be distributed to the translation staff to fill in.
po-to-xls path/to/new/django.po -o /tmp/new-language-translations.xlsx
Once the translation process is complete and the spreadsheet contains the translations, it is time to convert it back into a .po file. This .po file is transient and should be generated in a tmp or working directory. This can be done using the same tool as above:
xls-to-po <language> /tmp/new-language-translations.xlsx /tmp/updated.po
* The language argument in the xls-to-po command should match the value that was entered in the Language metadata field in the original .po file.
At this point, we have updated.po, which contains the translated strings, and we have django.po which contains some useful comments (e.g. source files and line numbers for each string) that we'd like to preserve. So, the last step is to merge these two files using msgmerge:
msgmerge /tmp/updated.po path/to/django.po -o path/to/django.po
django.po now contains the comments generated by makemessages, along with the translation strings from updated.po.
Now that the translations are available, we need to update the LANGUAGES variable in settings.py to include our new language. Lastly, run compilemessages as described above, or rebuild the application with:
docker-compose up --build
If your local database is in a wonky state, you might want to try tearing it all down and rebuilding from scratch.
First, shut down any containers you have running locally. Then run:
docker system prune --volumes
The volumes are the data elements in Docker. Note that you will need to re-create any local user roles after running this command, and the database will be in an empty state, with no complaint records.
Tests run automatically with repos that are integrated with Circle CI. You can run those tests locally with the following instructions.
Run unit test on Windows:
- Ensure docker is running
- Start a powershell as admin (git bash has issue running ssh console in docker)
- Find the id for the web container
docker container ls - Identify the id for the crt-portal_web_1
- SSH to web container in docker:
docker exec -it [id for the crt-portal_web goes here] /bin/bash (see below) docker exec -it 0a6039095e34 /bin/bash - Once you are in the SSH ./code run the test command below:
python crt_portal/manage.py test cts_forms - If you're lucky your tests will all pass otherwise, inspect the output to find the failing tests and errors to work on!
Run unit test on MAC:
You can also run project tests using docker with:
docker-compose run web python /code/crt_portal/manage.py test cts_forms --parallel
This will run all of the tests located in the tests folder. where the business logic tests live.
The test suite includes test_all_section_assignments.py, a script that generates a csv in the /data folder which has the relevant permutations of form fields and runs the out put from the section assignment function. If you are updating section assignments, you will want to edit the csv in data/section_assignment_expected.csv to what the new assignment should be. Additionally, if there are new factors to consider for routing, you will want to add those factors to the test so that they are accounted for.
You can also run a subset of tests by specifying a path to a specific test class or module. For example:
docker-compose run web python /code/crt_portal/manage.py test cts_forms.tests.test_forms.ComplaintActionTests --parallel
We use the unit tests for calculating code coverage. Tests will fail if code coverage is below 89%. You can run code coverage locally with:
docker-compose run web coverage run --source='.' /code/crt_portal/manage.py test cts_forms --parallel
docker-compose run web coverage report --fail-under=89 -m
The -m will give you the line numbers in code that that are not tested by any unit tests. You can use that information to add test coverage.
Please keep in mind that the quality of tests is more important than the quantity. Be on the look out for key functionality and logic that can be documented and confirmed with good tests.
🏃♀️🏃♂️ To speed up your tests, after you have migrated your database at least once, you can run:
docker-compose run web python /code/crt_portal/manage.py test --settings=crt_portal.test_settings cts_forms
That will create a file locally that tells your machine to not load zip codes on migration. These are not needed for tests and take a while.
For accessibility testing with Pa11y, we generally want a test on each unique view. You can run Pa11y locally, if you have npm installed locally:
- If you are doing this for the first time, log into your local admin and make an account for testing at localhost:8000/admin/auth/user/add/ The user name is
pa11y_testerpassword isimposing40Karl5monomial. Never make this account in the dev or staging environments. Circle testing happens in a disposable database and these are not run against live sites. (Prod doesn't use password accounts but no reason to make it there either.) - Run
export PA11Y_PASSWORD=imposing40Karl5monomialin the command line. - run
npm run test:a11y
This will run all the tests, look in package.json for a listing of tests, if you want to run them individually. See full accessibility testing guidelines in our A11y plan.
You can check for Python style issues by running flake8:
docker-compose run web flake8
If you have a a reason why a line of code shouldn't apply flake8 you can add # noqa, but try to use that sparingly.
You can check for JS style issues by running Prettier:
npm run lint:check
Prettier can automatically fix JS style issues for you:
npm run lint:write
You can scan the code for potential python security flaws using bandit. Run bandit manually:
docker-compose run web bandit -r crt_portal/
If there is a false positive you can add # nosec at the end of the line that is triggering the error. Please also add a comment that explains why that line is a false positive.
We use OWASP ZAP for security scans. Here is an intro to OWASP ZAP that explains the tool. You can also look at the scan configuration documentation.
You can run and pull down the container to use locally:
docker pull owasp/zap2docker-weekly
Run OWASP ZAP security scans with docker using the GUI:
docker run -u zap -p 8080:8080 -p 8090:8090 -i owasp/zap2docker-weekly zap-webswing.sh
you can see the GUI at http://localhost:8080/zap/
Do use caution when using any "attack" tests, we generally run those in local or sandboxed environments.
To stop the container, find the container id with:
docker container ls
Then you can stop the container with:
docker stop <container_id>
Run OWASP ZAP security scans with docker using the command line. Here is an example of running the baseline, passive scan locally targeting the development site:
docker run -v $(pwd):/zap/wrk/:rw -t owasp/zap2docker-weekly zap-baseline.py \
-t https://crt-portal-django-dev.app.cloud.gov/report/ -c .circleci/zap.conf -z "-config rules.cookie.ignorelist=django_language"
That will produce a report locally that you can view in your browser. It will give you a list of things that you should check. Sometimes there are things at the low or informational level that are false positives or are not worth the trade-offs to implement. The report will take a minute or two to generate.
We're using Playwright for automated end-to-end testing. Currently this test suite can be run locally by developers with the following command.
pipenv run pytest crt_portal/cts_forms/tests/integration/*.py --base-url=http://localhost:8000
These tests are automatically executed for every pull request against a development instance in CircleCI.
As we build out our end-to-end test suite it will be further automated and integrated into our CI/CD pipeline.
For the general public facing pages, we regularly test against Microsoft Edge, Internet Explorer 11, and Google Chrome, and occasionally test against Safari and Firefox.
For the staff-only pages, we only support the Microsoft Edge and Google Chrome browsers. Internet Explorer 11 will not work.
You only need to get the services stood up and configure the S3 bucket once.
For working with cloud.gov directly, you will need to install the cloud foundry cli. That will allow you to run the cf commands in a terminal.
First, login to cloud.gov at https://login.fr.cloud.gov/login and then, get a passcode https://login.fr.cloud.gov/passcode.
Log on with cf login -a api.fr.cloud.gov --sso-passcode <put_passcode_here>
First, log into the desired space.
Create postgres DB and S3 with development settings:
cf create-service aws-rds shared-psql crt-db
cf create-service s3 basic-public crt-s3
cf create-service s3 basic sso-creds
Or, for prod use the following production settings:
cf create-service aws-rds medium-psql-redundant crt-db
cf create-service s3 basic-public crt-s3
cf create-service s3 basic sso-creds
The medium-psql-redundant instance will provide a more resilient database approved for production and sso-creds is a private bucket used for authentication- see more details on that in the single sign on docs.
Store environment variables
cf cups VCAP_SERVICES -p SECRET_KEY
when prompted give it the secret key
You will needed to enable CORS via awscli, for each bucket instructions are here: https://cloud.gov/docs/services/s3/#allowing-client-side-web-access-from-external-applications
Create a service account for deployment for each space you are setting up. (Replace "SPACE" with the name of the space you are setting up.)
cf create-service cloud-gov-service-account space-deployer crt-service-account-SPACE
cf create-service-key crt-service-account-SPACE crt-portal-SPACE-key
cf service-key crt-service-account-SPACE crt-portal-SPACE-key
Those credentials will need to be added to CircleCI as environment variables: CRT_USERNAME_SPACE CRT_PASSWORD_SPACE (replace "SPACE" with the relevant space).
Right now, the route is set for the production space, we will want to pass in different routes for different spaces but that can be handled when we add the automation.
To deploy manually, make sure you are logged in, run the push command and pass it the name of the manifest for the space you want to deploy to:
cf push -f manifest_space.yaml
That will push to cloud.gov according to the instructions in the manifest and Profile.
A network policy needs to be configured to allow communication between our web application and the ClamAV REST API.
Direct traffic from the portal to the ClamAV REST API:
cf add-network-policy crt-portal-django --destination-app clamav-rest --protocol tcp --port 9000
As of October 2019, we have two user roles in the system:
-
Staff. Logged-in staff can view the table of complaints at
/form/view. -
Admin (superusers). Logged-in admins can add and remove users, adjust form settings such as the list of protected classes, and view the table of complaints at
/form/view.
Please update the Accounts Spreadsheet if you create or modify any user accounts.
As we build out the product, we expect to add more granular user roles and permissions.
For production, we use DOJ's Single Sign on for authentication. For the ADFS authentication, you need to make sure that public urls are listed in settings.py in the AUTH_ADFS section.
For the dev and staging site, you will want to make sure the @login_required decorator are on all views that need authentication.
Setting up single sign on is documented in docs/maintenance_or_infrequent_tasks.
Need to ssh to create superuser (would like to do this automatically in another PR):
cf ssh crt-portal-django
Once in, activate local env:
/tmp/lifecycle/shell
Then, you can create a superuser:
python crt_portal/manage.py createsuperuser
Or change a user's password:
python crt_portal/manage.py changepassword {{username}}
We deploy from CircleCI. The circle config contains rules that will deploy the site to different environments using a set of rules.
We leverage the CircleCI dependency caching feature to accelerate the build process. There are times, such as when upgrading to a new runtime version, when it becomes neccesary to clear the cached dependencies manually. This can be done by incrementing the value of the CACHE_VERSION environment variable from the CircleCI project settings.
GitFlow is a tool that can make it easier to handle branching. On a Mac with homebrew it can be installed with:
brew install git-flow
Then, in the top level folder of the project, you will want to run:
git flow init
We are using the defaults, so you can press enter for all the set up options. You may need to type 'master' production releases if you haven't been on the master branch yet.
-
The app will deploy to dev when the tests pass and a PR is merged into
develop. You should do this in GitHub. -
The app will deploy to stage when the tests pass and when we make or update a branch that starts with
release/.- Make sure the develop branch is approved for deploy by the product owner
- Look at ZenHub and see if there is anything in "Dev done" flag that for approval so those issues are in "Ready for UAT" when you make the release
- Check out the develop branch and do a
git pull origin develop - You can create a release with the command
git flow release start <date-of-planned-relase> - Finally, push the release branch
git push origin release/<date-of-planned-release> - Create a PR from github repository for the newly pushed branch
- Ensure base is master while creating the PR
-
The app will re-deploy to stage if a new PR is merged into a
release/branch. We do not do this often, but sometimes we want to get fixes in staging before the next release. You should do this in GitHub.- Make sure we understand why the PR is updating stage instead of going through the normal process -- there should be a good reason as to why.
- Verify that the PR has a
release/<date-of-planned-release>branch as a base branch - If all looks good, merge the PR via Github "merge" button
- Verify that tests pass and that the deploy to stage succeeded on CircleCI.
- Note: the developer will be responsible for cherry picking or otherwise merging these staging changes back to dev.
-
The app will deploy to prod when the tests pass and a PR is merged into
master. You can also do this in GitHub once you confirm approval with the product owner. If there are any merge conflicts, you will want to resolve them on the staging branch first.- If have not been any PRs directly to the release branch you can merge in GitHub
- If there are PRs to the release branch merge with gitflow:
- Approve the relese on GitHub
- Check out the release branch and do a
git pull origin release/<name-of-release> - Check out the master branch and do a
git pull origin master - Check out the develop branch and do a
git pull origin develop - Run
git flow release finish - Check out the develop branch and do a
git push origin develop - Once that succeeds, check out the master branch do a
git push origin master
When CircleCI tries to deploy two PRs back-to-back, one of them can fail. In this case, you can restart the failed deploy process by clicking the "Rerun Workflow" button.
Hot fixes will be needed when we find urgent bugs or problems with production. This is where git-flow becomes very useful. To make the fix:
- Check out the master branch and do a
git pull origin master - Create a branch for your work with
git flow hotfix start <name-your-hotfix>- This will make a new branch based on the master branch with the naming convention
hotfix/name-of-hotfix
- This will make a new branch based on the master branch with the naming convention
- Commit and push your branch, make a PR that targets the master branch for review
To deploy the fix:
- Make sure the product owner is in the loop with any errors and fixes.
- The reviewer can test the change locally to do the review. once the reviewer is happy with it:
- Make sure your branches are up to date
- Check out the development branch and do a
git pull origin develop - Check out the master branch and do a
git pull origin master
- Check out the development branch and do a
- Finish the hotfix with
git flow hotfix finishThis command will make sure that the fix is merged into the master and develop branches. Making sure the change is in both places, means you can test it on dev and your change won't get clobbered later. - Push to the develop branch, that now has the fix to trigger a deploy to the dev enviornment
- Check that the update deployed correctly.
- Once sastisfied, go to GitHub to approve and merge the PR
- That will trigger a new build with the fix. Check that the fix worked
The git-flow cheatsheet is a great explainer of the git-flow tool.
For more technical documentation see the docs
- A11y testing plan
- Branching strategy
- Maintenance or infrequent tasks
- Pull request instructions
- Monitoring
To prevent form autocomplete on an application instance, add FORM_AUTOCOMPLETE_OFF=True as an environment variable. We are using that in staging to help mask personal data during usability testing.
To restore default behavior of allowing form autocomplete, remove the FORM_AUTOCOMPLETE_OFF flag.
These are some technologies we are using in the build, here are some links for background.
Pipenv This is what we use to manage python packages
Postgres Here are the database docs
This is a tool for for interfacing with postgres pgcli
Docker We are using containers for local development.
USWDS We are using 2.0 as our base
Django This is the web framework
Cloud.gov This is the PaaS the app is on
Thanks for reading all the way!
![dependabot[bot] avatar](https://avatars.githubusercontent.com/in/29110?v=4 "dependabot[bot]")
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent