fedbadges 2.2.0

Triggers

Every BadgeRule must carry the following minimum set of metadata:

# This is some metadata about the badge
name:           Like a Rock
description:    You have pushed 500 or more bodhi updates to stable status.
creator:        ralph

# This is a link to the discussion about adopting this as a for-real badge.
discussion: http://github.com/fedora-infra/badges/pull/SOME_NUMBER

# A link to the image for the badge
image_url: http://somelink.org/to-an-image.png

Here’s a simple example of a trigger:

trigger:
  category: bodhi

The above will match any bodhi message on any of the topics that come from the bodhi update system.

Triggers may employ a little bit of logic to make more complex filters. The following trigger will match any message that comes from either the bodhi update system or the fedora git package repos:

trigger:
  category:
    any:
      - bodhi
      - git

At present triggers may directly compare themselves against only the category or the topic of a message. In the future we’d like to add more comparisons.. in the meantime, here’s an example of comparing against the fully qualified message topic. This will match any message that is specifically for editing a wiki page:

trigger:
  topic: org.fedoraproject.prod.wiki.article.edit

There is one additional way you can specify a trigger. If you need more flexibility than topic and category allow, you may specify a custom filter expression with a lambda filter. For example:

trigger:
  lambda: "a string of interest" in json.dumps(message.body)

The above trigger will match if the string "a string of interest" appears anywhere in the incoming message. fedbadges takes the expression you provide it and compiles it into a python callable on initialization. Our callable here serializes the message to a JSON string before doing its comparison. Powerful!

Previous

As mentioned above in the architecture section, we currently only support datanommer as a queryable backend for previous queries. We hope to expand that in the future.

Datanommer queries are composed of two things:

• A filter limits the scope of the query to datanommer.

• An operation defines what we want to do with the filtered query. Currently, we can count the results or run them through a lambda function that will return an integer (the number of matched messages).

Here’s an example of a simple previous definition:

previous:
  filter:
    topics:
    - message.topic
  operation: count

The above previous query will return the number of messages in datanommer with the same topic as the incoming message being handled. Here, message.topic is a lambda function that has the incoming message in scope.

The above example doesn’t make much sense – we’d never use it for a real badge. The previous query would be true if there were two of any message kicked off by any user at any time in the past. Pretty generic. Here’s a more interesting previous query:

previous:
  filter:
    topics:
    - org.fedoraproject.prod.git.receive
    users:
    - message.body["commit"]["username"]
  operation: count

This previous query would return the number of messages of the topic "org.fedoraproject.prod.git.receive" that were also kicked off by whatever user is listed in the message.body['commit']['username'] field of the message being currently processed. In other words, this query would return the number of pushes to the fedora git repos by the user.

Condition

You can do some fancy things with the condition field. Here’s a list of the possible comparisons you can make:

• "is greater than or equal to" or alternatively "greater than or equal to"

• "greater than"

• "is less than or equal to" or alternatively "less than or equal to"

• "less than"

• "equal to" or alternatively "is equal to"

• "is not" or alternatively "is not equal to"

As you can see, some of them are synonyms for each other.

If any of those don’t meet your needs, you can specify a custom expression by using the lambda condition whereby fedbadges will compile whatever statement you provide into a callable and use that at runtime. For example:

condition:
  lambda: value != 0 and ((value & (value - 1)) == 0)

Who knows why you would want to do this, but the above condition check will succeed if the number of messages that matched in the past is exactly a power of 2.

Specifying Recipients

By default, if the trigger and condition match, fedbadges will award badges to the user returned by the message’s agent_name property. This usually corresponds with “which user is responsible” for this message. That is usually what we want to award badges for.

There are some instances for which that is not what we want.

Take the org.fedoraproject.prod.bodhi.update.comment message for example. When user A comments on user B’s update, user A is returned by the message’s agent_name property.

Imagine we have a “Received Comments” badge that’s awarded to packagers that received comments on their updates. We don’t want to inadvertently award that badge to the person who commented, only to the one who created the update.

To allow for this scenario, badges may optionally define a recipient in dotted notation that tells fedbadges where to find the username of the recipient in the originating message. For instance, the following would handle the fas case we described above:

trigger:
  topic: org.fedoraproject.prod.bodhi.update.comment
condition:
  greater than or equal to: 1
previous:
  filter:
    topics:
    - message.topic
    users:
    - recipient
  operation: count
recipient: message.body["update"]["user"]["name"]