Developing on Bedrock

Managing Dependencies

For Python we use pip-compile from pip-tools to manage dependencies expressed in our requirements files. pip-compile is wrapped up in Makefile commands, to ensure we use it consistently.

If you add a new Python dependency (eg to requirements/prod.in or requirements/dev.in) you can generate a pinned and hash-marked addition to our requirements files just by running:

make compile-requirements

and committing any changes that are made. Please re-build your docker image and test it with make build test to be sure the dependency does not cause a regression.

Similarly, if you upgrade a pinned dependency in an *.in file, run make compile-requirements then rebuild, test and commit the results

To check for stale Python dependencies (basically pip list -o but in the Docker container):

make check-requirements

For Node packages we use NPM, which should already be installed alongside Node.js.

Front-end Dependencies

Our team maintains a few dependencies that we serve on Bedrock’s front-end.

Because they are all published on NPM, install the packages and keep up-to-date with the latest version of each dependency by running an npm install. For further documentation on installing NPM packages, check out the official documentation.

Asset Management and Bundling

Bedrock uses Webpack to manage front-end asset processing and bundling. This includes processing and minifying JavaScript and SCSS/CSS bundles, as well as managing static assets such as images, fonts, and other common file types.

When developing on bedrock you can start Webpack by running make run when using Docker, or npm start when running bedrock locally.

Once Webpack has finished compiling, a local development server will be available at localhost:8000. When Webpack detects changes to a JS/SCSS file, it will automatically recompile the bundle and then refresh any page running locally in the browser.

Webpack Configuration

We have two main Webpack config files in the root directory:

The webpack.static.config.js file is responsible for copying static assets, such as images and fonts, from the /media/ directory over to the /assets/ directory. This is required so Django can serve them correctly.

The webpack.config.js file is responsible for processing JS and SCSS files in the /media/ directory and compiling them into the /assets/ directory. This config file also starts a local development server and watches for file changes.

We use two separate config files to keep responsibilities clearly defined, and to make the configs both shorter and easier to follow.

Note

Because of the large number of files used in bedrock, only JS and SCSS files managed by webpack.config.js are watched for changes when in development mode. This helps save on memory consumption. The implication of this is that files handled by webpack.static.config.js are only copied over when Webpack first runs. If you update an image for example, then you will need to stop and restart Webpack to pick up the change. This is not true for JS and SCSS files, which will be watched for change automatically.

Asset Bundling

Asset bundles for both JS and SCSS are defined in ./media/static-bundles.json. This is the file where you can define the bundle names that will get used in page templates. For example, a CSS bundle can be defined as:

"css": [
    {
        "files": [
            "css/firefox/new/basic/download.scss"
        ],
        "name": "firefox_new_download"
    }
]

Which can then be referenced in a page template using:

{{ css_bundle('firefox_new_download') }}

A JS bundle can be defied as:

"js": [
    {
        "files": [
            "protocol/js/protocol-modal.js",
            "js/firefox/new/basic/download.js"
        ],
        "name": "firefox_new_download"
    }
]

Which can then be referenced in a page template using:

{{ js_bundle('firefox_new_download') }}

Once you define a bundle in static-bundles.json, the webpack.config.js file will use these as entrypoints for compiling JS and CSS and watching for changes.

Writing JavaScript

Bedrock’s Webpack configuration supports some different options for writing JavaScript:

Default Configuration

Write example-script.js using ES5 syntax and features. Webpack will bundle the JS as-is, without any additional pre-processing.

Babel Configuration

Write example-script.es6.js using ES2015+ syntax. Webpack will transpile the code to ES5 using Babel. This is useful when you want to write modern syntax but still support older browsers.

Important

Whilst Babel will transpile most modern JS syntax to ES5 when suitable fallbacks exist, it won’t automatically include custom polyfills for everything since these can start to greatly increase bundle size. If you want to use Promise or async/await functions for example, then you will need to load polyfills for those. This can be done either at the page level, or globally in lib.js if it’s something that multiple pages would benefit from. But please pick and choose wisely, and be concious of performance.

For pages that are served to Firefox browsers only, such as /whatsnew, it is also possible to write native ES2015+ syntax and serve that directly in production. Here there is no need to include the .es6.js file extension. Instead, you can simply use .js instead. The rules that which files you can do this in are defined in our ESLint config.

Writing URL Patterns

URL patterns should be the entire URL you desire, minus any prefixes from URLs files importing this one, and including a trailing slash. You should also give the URL a name so that other pages can reference it instead of hardcoding the URL. Example:

path("channel/", channel, name="mozorg.channel")

If you only want to render a template and don’t need to do anything else in a custom view, Bedrock comes with a handy shortcut to automate all of this:

from bedrock.mozorg.util import page
page("channel/", "mozorg/channel.html")

You don’t need to create a view. It will serve up the specified template at the given URL (the first parameter. see the Django docs for details). You can also pass template data as keyword arguments:

page("channel/", "mozorg/channel.html",
     latest_version=product_details.firefox_versions["LATEST_FIREFOX_VERSION"])

The variable latest_version will be available in the template.

Finding Templates by URL

General Structure

Bedrock follows the Django app structure and most templates are easy to find by matching URL path segments to folders and files within the correct app.

URL: https://www.mozilla.org/en-US/firefox/features/private-browsing/
Template path: bedrock/bedrock/firefox/templates/firefox/features/private-browsing.html

To get from URL to template path:

  • Ignore https://www.mozilla.org and the locale path segment /en-US. The next path segment is the app name /firefox.

  • From the root folder of bedrock, find the app’s template folder at bedrock/{app}/templates/{app}

  • Match remaining URL path segments (/features/private-browsing) to the template folder’s structure (/features/private-browsing.html)

Note

mozorg is the app name for the home page and child pages related to Mozilla Corporation (i.e. About, Contact, Diversity).

Whatsnew and Firstrun

These pages are specific to Firefox browsers, and only appear when a user updates or installs and runs a Firefox browser for the first time. The URL and template depend on what Firefox browser and version are in use.

Note

There may be extra logic in the app’s views.py file to change the template based on locale or geographic location as well.

Firefox release

Version number is digits only.

Firefox Nightly

Version number is digits and a1.

Firefox Developer

Version number is digits and a2.

Release Notes

Release note templates live here: https://github.com/mozilla/bedrock/tree/main/bedrock/firefox/templates/firefox/releases

Note

Release note content is pulled in from an external data source.

Optimizing Images

Images can take a long time to load and eat up a lot of bandwidth. Always take care to optimize images before uploading them to the site.

The script img.sh can be used to optimize images locally on the command line:

  1. Before you run it for the first time you will need to run npm install to install dependencies

  2. Add the image files to git’s staging area git add *

  3. Run the script ./bin/img.sh

  4. The optimized files will not automatically be staged, so be sure to add them before commiting

The script will:

  • optimize JPG and PNG files using tinypng (
    • this step is optional since running compression on the same images over and over degrades them)

    • you will be prompted to add a TinyPNG API key

  • optimize SVG images locally with svgo

  • check that SVGs have a viewbox (needed for IE support)

  • check that images that end in -high-res have low res versions as well

Embedding Images

Images should be included on pages using one of the following helper functions.

Primary image helpers

The following image helpers support the most common features and use cases you may encounter when coding pages:

static()

For a simple image, the static() function is used to generate the image URL. For example:

<img src="{{ static('img/firefox/new/firefox-wordmark-logo.svg') }}" alt="Firefox">

will output an image:

<img src="/media/img/firefox/new/firefox-wordmark-logo.svg" alt="Firefox">

resp_img()

For responsive images, where we want to specify multiple different image sizes and let the browser select which is best to use.

The example below shows how to serve an appropriately sized, responsive red panda image:

resp_img(
    url="img/panda-500.png",
    srcset={
        "img/panda-500.png": "500w",
        "img/panda-750.png": "750w",
        "img/panda-1000.png": "1000w"
    },
    sizes={
        "(min-width: 1000px)": "calc(50vw - 200px)",
        "default": "calc(100vw - 50px)"
    }
)

This would output:

<img src="/media/img/panda-500.png"
     srcset="/media/img/panda-500.png 500w,/media/img/panda-750.png 750w,/media/img/panda-1000.png 1000w"
     sizes="(min-width: 1000px) calc(50vw - 200px),calc(100vw - 50px)" alt="">'

In the above example we specified the available image sources using the srcset parameter. We then used sizes to say:

  • When the viewport is greater than 1000px wide, the panda image will take up roughly half of the page width.

  • When the viewport is less than 1000px wide, the panda image will take up roughly full page width.

The default image src is what we specified using the url param. This is also what older browsers will fall back to using. Modern browsers will instead pick the best source option from srcset (based on both the estimated image size and screen resolution) to satisfy the condition met in sizes.

Note

The value default in the second sizes entry above should be used when you want to omit a media query. This makes it possible to provide a fallback size when no other media queries match.

Another example might be to serve a high resolution alternative for a fixed size image:

resp_img(
    url="img/panda.png",
    srcset={
        "img/panda-high-res.png": "2x"
    }
)

This would output:

<img src="/media/img/panda.png" srcset="/media/img/panda-high-res.png 2x" alt="">

Here we don’t need a sizes attribute, since the panda image is fixed in size and small enough that it won’t need to resize along with the browser window. Instead the srcset image includes an alternate high resolution source URL, along with a pixel density descriptor. This can then be used to say:

  • When a browser specifies a device pixel ratio of 2x or greater, use panda-high-res.png.

  • When a browser specifies a device pixel ration of less than 2x, use panda.png.

The resp_img() helper also supports localized images by setting the 'l10n' parameter to True`:

resp_img(
    url="img/panda-500.png",
    srcset={
        "img/panda-500.png": "500w",
        "img/panda-750.png": "750w",
        "img/panda-1000.png": "1000w"
    },
    sizes={
        "(min-width: 1000px)": "calc(50vw - 200px)",
        "default": "calc(100vw - 50px)"
    },
    optional_attributes={
        "l10n": True
    }
)

This would output (assuming de was your locale):

<img src="/media/img/l10n/de/panda-500.png"
     srcset="/media/img/l10n/de/panda-500.png 500w,/media/img/l10n/de/panda-750.png 750w,/media/img/l10n/de/panda-1000.png 1000w"
     sizes="(min-width: 1000px) calc(50vw - 200px),calc(100vw - 50px)" alt="">'

Finally, you can also specify any other additional attributes you might need using optional_attributes:

resp_img(
    url="img/panda-500.png",
    srcset={
        "img/panda-500.png": "500w",
        "img/panda-750.png": "750w",
        "img/panda-1000.png": "1000w"
    },
    sizes={
        "(min-width: 1000px)": "calc(50vw - 200px)",
        "default": "calc(100vw - 50px)"
    },
    optional_attributes={
        "alt": "Red Panda",
        "class": "panda-hero",
        "height": "500",
        "l10n": True,
        "loading": "lazy",
        "width": "500"
    }
)

picture()

For responsive images, where we want to serve different images, or image types, to suit different display sizes.

The example below shows how to serve a different image for desktop and mobile sizes screens:

picture(
    url="img/panda-mobile.png",
    sources=[
        {
            "media": "(max-width: 799px)",
            "srcset": {
                "img/panda-mobile.png": "default"
            }
        },
        {
            "media": "(min-width: 800px)",
            "srcset": {
                "img/panda-desktop.png": "default"
            }
        }
    ]
)

This would output:

<picture>
    <source media="(max-width: 799px)" srcset="/media/img/panda-mobile.png">
    <source media="(min-width: 800px)" srcset="/media/img/panda-desktop.png">
    <img src="/media/img/panda-mobile.png" alt="">
</picture>

In the above example, the default image src is what we specifed using the url param. This is also what older browsers will fall back to using. We then used the sources parameter to specify one or more alternate image <source> elements, which modern browsers can take advantage of. For each <source>, media lets us specify a media query as a condition for when to load an image, and srcset lets us specify one or more sizes for each image.

Note

The value default in the srcset entry above should be used when you want to omit a descriptor. In this example we only have one entry in srcset (meaning it will be chosen immediately should the media query be satisfied), hence we omit a descriptor value.

A more complex example might be when we want to load responsively sized, animated gifs, but also offer still images for users who set (prefers-reduced-motion: reduce):

picture(
    url="img/dancing-panda-500.gif",
    sources=[
        {
            "media": "(prefers-reduced-motion: reduce)",
            "srcset": {
                "img/sleeping-panda-500.png": "500w",
                "img/sleepinng-panda-750.png": "750w",
                "img/sleeping-panda-1000.png": "1000w"
            },
            "sizes": {
                "(min-width: 1000px)": "calc(50vw - 200px)",
                "default": "calc(100vw - 50px)"
            }
        },
        {
            "media": "(prefers-reduced-motion: no-preference)",
            "srcset": {
                "img/dancing-panda-500.gif": "500w",
                "img/dancing-panda-750.gif": "750w",
                "img/dancing-panda-1000.gif": "1000w"
            },
            "sizes": {
                "(min-width: 1000px)": "calc(50vw - 200px)",
                "default": "calc(100vw - 50px)"
            }
        }
    ]
)

This would output:

<picture>
    <source media="(prefers-reduced-motion: reduce)"
            srcset="/media/img/sleeping-panda-500.png 500w,/media/img/sleeping-panda-750.png 750w,/media/img/sleeping-panda-1000.png 1000w"
            sizes="(min-width: 1000px) calc(50vw - 200px),calc(100vw - 50px)">
    <source media="(prefers-reduced-motion: no-preference)"
            srcset="/media/img/dancing-panda-500.gif 500w,/media/img/dancing-panda-750.gif 750w,/media/img/dancing-panda-1000.gif 1000w"
            sizes="(min-width: 1000px) calc(50vw - 200px),calc(100vw - 50px)">
    <img src="/media/img/dancing-panda-500.gif" alt="">
</picture>

In the above example we would default to loading animated gifs, but if a user agent specified (prefers-reduced-motion: reduce) then the browser would load static png files instead. Multiple image sizes are also supported for each <source> using srcset and sizes.

Another type of use case might be to serve different image formats, so capable browsers can take advantage of more efficient encoding:

picture(
    url="img/red-panda.png",
    sources=[
        {
            "type": "image/webp",
            "srcset": {
                "img/red-panda.webp": "default"
            }
        }
    ]
)

This would output:

<picture>
    <source type="image/webp" srcset="/media/img/red-panda.webp">
    <img src="/media/img/red-panda.png" alt="">
</picture>

In the above example we use sources to specify an alternate image with a type attribute of image/webp. This lets browsers that support WebP to download red-panda.webp, whilst older browsers would download red-panda.png.

Like resp_img(), the picture() helper also supports L10n images and other useful attributes via the optional_attributes parameter:

picture(
    url="img/panda-mobile.png",
    sources=[
        {
            "media": "(max-width: 799px)",
            "srcset": {
                "img/panda-mobile.png": "default"
            }
        },
        {
            "media": "(min-width: 800px)",
            "srcset": {
                "img/panda-desktop.png": "default"
            }
        }
    ],
    optional_attributes={
        "alt": "Red Panda",
        "class": "panda-hero",
        "l10n": True,
        "loading": "lazy",
    }
)

Which image helper should you use?

This is a good question. The answer depends entirely on the image in question. A good rule of thumb is as follows:

  • Is the image a vector format (e.g. .svg)?
    • If yes, then for most cases you can simply use static().

  • Is the image a raster format (e.g. .png or .jpg)?
    • Is the same image displayed on both large and small viewports? Does the image need to scale as the browser resizes? If yes to both, then use resp_img() with both srcset and sizes.

    • Is the image fixed in size (non-responsive)? Do you need to serve a high resolution version? If yes to both, then use resp_img() with just srcset.

  • Does the source image need to change depending on a media query (e.g serve a different image on both desktop and mobile)? If yes, then use picture() with media and srcset.

  • Is the image format only supported in certain browsers? Do you need to provide a fallback? If yes to both, then use picture() with type and srcset.

Secondary image helpers

The following image helpers are less commonly used, but exist to support more specific use cases. Some are also encapsulated as features inside inside of primary helpers, such as l1n_img().

l10n_img()

Images that have translatable text can be handled with l10n_img():

<img src="{{ l10n_img('firefox/os/have-it-all/messages.jpg') }}">

The images referenced by l10n_img() must exist in media/img/l10n/, so for above example, the images could include media/img/l10n/en-US/firefox/os/have-it-all/messages.jpg and media/img/l10n/es-ES/firefox/os/have-it-all/messages.jpg.

qrcode()

This is a helper function that will output SVG data for a QR Code at the spot in the template where it is called. It caches the results to the data/qrcode_cache directory, so it only generates the SVG data one time per data and box_size combination.

qrcode("https://accounts.firefox.com", 30)

The first argument is the data you’d like to encode in the QR Code (usually a URL), and the second is the “box size”. It’s a parameter that tells the generator how large to set the height and width parameters on the XML SVG tag, the units of which are “mm”. This can be overriden with CSS so you may not need to use it at all. The box_size parameter is optional.

Using Large Assets

We don’t want to (and if large enough GitHub won’t let us) commit large files to the bedrock repo. Files such as large PDFs or very-high-res JPG files (e.g. leadership team photos), or videos are not well-tracked in git and will make every checkout after they’re added slower and this diffs less useful. So we have another domain at which we upload these files: assets.mozilla.net

This domain is simply an AWS S3 bucket with a CloudFront CDN in front of it. It is highly available and fast. We’ve made adding files to this domain very simple using git-lfs. You simply install git-lfs, clone our assets.mozilla.net repo, and then add and commit files under the assets directory there as usual. Open a pull request, and once it’s merged it will be automatically uploaded to the S3 bucket and be available on the domain.

For example, if you add a file to the repo under assets/pdf/the-dude-abides.pdf, it will be available as https://assets.mozilla.net/pdf/the-dude-abides.pdf. Once that is done you can link to that URL from bedrock as you would any other URL.

Writing Migrations

Bedrock uses Django’s built-in Migrations framework for its database migrations, and has no custom database routing, etc. So, no big surprises here – write things as you regularly would.

However, as with any complex system, care needs to be taken with schema changes that drop or rename database columns. Due to the way the rollout process works (ask for details directly from the team), an absent column can cause some of the rollout to enter a crashloop.

To avoid this, split your changes across releases, such as below.

For column renames:

  • Release 1: Add your new column

  • Release 2: Amend the codebase to use it instead of the old column

  • Release 3: Clean up - drop the old, deprecated column, which should not be referenced in code at this point.

For column drops:

  • Release 1: Update all code that uses the relevant column, so that nothing interacts with it any more.

  • Release 2: Clean up - drop the old, deprecated column.

With both paths, check for any custom schema or data migrations that might reference the deprecated column.

Writing Views

You should rarely need to write a view for mozilla.org. Most pages are static and you should use the page function documented above.

If you need to write a view and the page is translated or translatable then it should use the l10n_utils.render() function to render the template.

from lib import l10n_utils

from django.views.decorators.http import require_safe


@require_safe
def my_view(request):
    # do your fancy things
    ctx = {"template_variable": "awesome data"}
    return l10n_utils.render(request, "app/template.html", ctx)

Make sure to namespace your templates by putting them in a directory named after your app, so instead of templates/template.html they would be in templates/blog/template.html if blog was the name of your app.

The require_safe ensures that only GET or HEAD requests will make it through to your view.

If you prefer to use Django’s Generic View classes we have a convenient helper for that. You can use it either to create a custom view class of your own, or use it directly in a urls.py file.

# app/views.py
from lib.l10n_utils import L10nTemplateView

class FirefoxRoxView(L10nTemplateView):
    template_name = "app/firefox-rox.html"

# app/urls.py
urlpatterns = [
    # from views.py
    path("firefox/rox/", FirefoxRoxView.as_view()),
    # directly
    path("firefox/sox/", L10nTemplateView.as_view(template_name="app/firefox-sox.html")),
]

The L10nTemplateView functionality is mostly in a template mixin called LangFilesMixin which you can use with other generic Django view classes if you need one other than TemplateView. The L10nTemplateView already ensures that only GET or HEAD requests will be served.

Variation Views

We have a generic view that allows you to easily create and use a/b testing templates. If you’d like to have either separate templates or just a template context variable for switching, this will help you out. For example.

# urls.py

from django.urls import path

from bedrock.utils.views import VariationTemplateView

urlpatterns = [
    path("testing/",
         VariationTemplateView.as_view(template_name="testing.html",
                                       template_context_variations=["a", "b"]),
         name="testing"),
]

This will give you a context variable called variation that will either be an empty string if no param is set, or a if ?v=a is in the URL, or b if ?v=b is in the URL. No other options will be valid for the v query parameter and variation will be empty if any other value is passed in for v via the URL. So in your template code you’d simply do the following:

{% if variation == 'b' %}<p>This is the B variation of our test. Enjoy!</p>{% endif %}

If you’d rather have a fully separate template for your test, you can use the template_name_variations argument to the view instead of template_context_variations.

# urls.py

from django.urls import path

from bedrock.utils.views import VariationTemplateView

urlpatterns = [
    path("testing/",
         VariationTemplateView.as_view(template_name="testing.html",
                                       template_name_variations=["1", "2"]),
         name="testing"),
]

This will not provide any extra template context variables, but will instead look for alternate template names. If the URL is testing/?v=1, it will use a template named testing-1.html, if v=2 it will use testing-2.html, and for everything else it will use the default. It simply puts a dash and the variation value between the template file name and file extension.

It is theoretically possible to use the template name and template context versions of this view together, but that would be an odd situation and potentially inappropriate for this utility.

You can also limit your variations to certain locales. By default the variations will work for any localization of the page, but if you supply a list of locales to the variation_locales argument to the view then it will only set the variation context variable or alter the template name (depending on the options explained above) when requested at one of said locales. For example, the template name example above could be modified to only work for English or German like so

# urls.py

from django.urls import path

from bedrock.utils.views import VariationTemplateView

urlpatterns = [
    path("testing/",
         VariationTemplateView.as_view(template_name="testing.html",
                                       template_name_variations=["1", "2"],
                                       variation_locales=["en-US", "de"]),
         name="testing"),
]

Any request to the page in for example French would not use the alternate template even if a valid variation were given in the URL.

Note

If you’d like to add this functionality to an existing Class-Based View, there is a mixin that implements this pattern that should work with most views: bedrock.utils.views.VariationMixin.

Geo Template View

Now that we have our CDN configured properly, we can also just swap out templates per request country. This is very similar to the above, but it will simply use the proper template for the country from which the request originated.

from bedrock.base.views import GeoTemplateView

class CanadaIsSpecialView(GeoTemplateView):
    geo_template_names = {
        "CA": "mozorg/canada-is-special.html",
    }
    template_name = "mozorg/everywhere-else-is-also-good.html"

For testing purposes while you’re developing or on any deployment that is not accessed via the production domain (www.mozilla.org) you can append your URL with a geo query param (e.g. /firefox/?geo=DE) and that will take precedence over the country from the request header.

Other Geo Stuff

There are a couple of other tools at your disposal if you need to change things depending on the location of the user. You can use the bedrock.base.geo.get_country_from_request function in a view and it will return the country code for the request (either from the CDN or the query param, just like above).

from bedrock.base.geo import get_country_from_request

def dude_view(request):
    country = get_country_from_request(request)
    if country == "US":
        # do a thing for the US
    else:
        # do the default thing

The other convenience available is that the country code, either from the CDN or the query param, is avilable in any template in the country_code variable. This allows you to change anything about how the template renders based on the location of the user.

{% if country_code == "US" %}
    <h1>GO MURICA!</h1>
{% else %}
    <h1>Yay World!</h1>
{% endif %}

Reference:

Metrics Collection with Markus

Markus is a metrics library that we use in our project for collecting and reporting statistics about our code’s operation. It provides a simple and consistent way to record custom metrics from your application, which can be crucial for monitoring and performance analysis.

Markus supports a variety of backends, including Datadog, Statsd, and Logging. This means you can choose the backend that best fits your monitoring infrastructure and requirements. Each backend has its own set of features and capabilities, but Markus provides a unified interface to all of them.

Once the metrics are collected by Markus they are then forwarded to Telegraf. Telegraf is an agent for collecting and reporting metrics, which we use to process and format the data before it’s sent to Grafana.

Grafana is a popular open-source platform for visualizing metrics. It allows us to create dashboards with panels representing the metrics we’re interested in, making it easy to understand the data at a glance.

Here’s an example of how to use Markus to record a metric:

from bedrock.base import metrics

# Counting events
metrics.incr("event_name")

# Timing events
metrics.timing("event_name", 123)

# Or timing events with context manager
with metrics.timer("event_name"):
    # code to time goes here

In addition to recording the metric values, Markus also allows you to add tags to your metrics. Tags are key-value pairs that provide additional context about the metric, making it easier to filter and aggregate the data in Grafana. For example, you might tag a metric with the version of your application, the user’s country, or the result of an operation. To add tags to a metric in Markus, you can pass them as a dictionary to the metric recording method. Here’s an example:

# Counting events with tags
metrics.incr("event_name", tags=[f"version:{version}", f"country:{country}"])

For more information, refer to the Markus documentation.

Coding Style

Bedrock uses the following open source tools to follow coding styles and conventions, as well as applying automatic code formatting:

  • ruff for Python style, code quality rules, and import ordering.

  • black for Python code formatting.

  • Prettier for JavaScript code formatting.

  • ESLint for JavaScript code quality rules.

  • Stylelint for Sass/CSS style and code quality rules.

For front-end HTML & CSS conventions, bedrock uses Mozilla’s Protocol design system for building components. You can read the Protocol documentation site for more information.

Mozilla also has some more general coding styleguides available, although some of these are now rather outdated:

Test coverage

When the Python tests are run, a coverage report is generated, showing which lines of the codebase have tests that execute them, and which do not. You can view this report in your browser at file:///path/to/your/checkout/of/bedrock/python_coverage/index.html.

When adding code, please aim to provide solid test coverage, using the coverage report as a guide. This doesn’t necessarily mean every single line needs a test, and 100% coverage doesn’t mean 0% defects.

Configuring your Code Editor

Bedrock includes an .editorconfig file in the root directory that you can use with your code editor to help maintain consistent coding styles. Please see editorconfig.org. for a list of supported editors and available plugins.

Working with Protocol Design System

Bedrock uses the Protocol Design System to quickly produce consistent, stable components. There are different methods – depending on the component – to import a Protocol component into our codebase.

One method involves two steps:

  1. Adding the correct markup or importing the appropriate macro to the page’s HTML file.

  2. Importing the necessary Protocol styles to a page’s SCSS file.

The other method is to import CSS bundles onto the HTML file. However, this only works for certain components, which are listed below in the respective section.

Styles and Components

The base templates in Bedrock have global styles from Protocol that apply to every page. When we need to extend these styles on a page-specific basis, we set up Protocol in a page-specific SCSS file.

For example, on a Firefox product page, we might want to use Firefox logos or wordmarks that do not exist on every page.

To do this, we add Protocol mzp- classes to the HTML:

// bedrock/bedrock/firefox/templates/firefox/{specific-page}.html

<div class="mzp-c-wordmark mzp-t-wordmark-md mzp-t-product-firefox">
    Firefox Browser
</div>

Then we need to include those Protocol styles in the page’s SCSS file:

/* bedrock/media/css/firefox/{specific-page}.scss */

/* if we need to use protocol images, we need to set the $image-path variable */
$image-path: '/media/protocol/img';
/* mozilla is the default theme, so if we want a different one, we need to set the $brand-theme variable */
$brand-theme: 'firefox';

/* the lib import is always essential: it provides access to tokens, functions, mixins, and theming */
@import '~@mozilla-protocol/core/protocol/css/includes/lib';
/* then you add whatever specific protocol styling you need */
@import '~@mozilla-protocol/core/protocol/css/components/logos/wordmark';
@import '~@mozilla-protocol/core/protocol/css/components/logos/wordmark-product-firefox';

Note

If you create a new SCSS file for a page, you will have to include it in that page’s CSS bundle by updating static-bundles.json file.

Macros

The team has created several Jinja macros out of Protocol components to simplify the usage of components housing larger blocks of code (i.e. Billboard). The code housing the custom macros can be found in our protocol macros file. These Jinja macros include parameters that are simple to define and customize based on how the component should look like on a given page.

To use these macros in files, we simply import a macro to the page’s HTML code and call it with the desired arguments, instead of manually adding Protocol markup. We can import multiple macros in a comma-separated fashion, ending the import with with context:

// bedrock/bedrock/firefox/templates/firefox/{specific-page}.html

{% from "macros-protocol.html" import billboard with context %}

{{ billboard(
    title='This is Firefox.',
    ga_title='This is Firefox',
    desc='Firefox is an awesome web browser.',
    link_cta='Click here to install',
    link_url=url('firefox.new')
  )}}

Because not all component styles are global, we still have to import the page-specific Protocol styles in SCSS:

/* bedrock/media/css/firefox/{specific-page}.scss */

$brand-theme: 'firefox';

@import '~@mozilla-protocol/core/protocol/css/includes/lib';
@import '~@mozilla-protocol/core/protocol/css/components/billboard';

Import CSS Bundles

We created pre-built CSS bundles to be used for some components due to their frequency of use. This method only requires an import into the HTML template. Since it’s a separate CSS bundle, we don’t need to import that component in the respective page CSS. The CSS bundle import only works for the following components:

  • Split

  • Card

  • Picto

  • Callout

  • Article

  • Newsletter form

  • Emphasis box

Include a CSS bundle in the template’s page_css block along with any other page-specific bundles, like so:

{% block page_css %}
    {{ css_bundle('protocol-split') }}
    {{ css_bundle('protocol-card') }}
    {{ css_bundle('page-specific-bundle') }}
{% endblock %}