pulsar
Pulsar
Pulsar is a simple Crystal library for publishing and subscribing to events. It also has timing information for metrics. So what does that mean in practice?
You can define an event and any number of subscribers can subscribe to the event and do whatever they need with it.
For example, in Lucky, we use Pulsar to create events for things like requests being processed, queries being made, before and after pipes running. Then we subscribe to these events to write to the logs. We also use this internally to log debugging information in an upcoming UI called Breeze that let's users debug development information.
Installation
-
Add the dependency to your
shard.yml
:dependencies: pulsar: github: luckyframework/pulsar
-
Run
shards install
How to use Pulsar
Let's say we're writing a library to charge a credit card and we may want to let people run code whenever a charge is made. Here's how you can do that with Pulsar.
Create and publish an event
class PaymentProcessor::ChargeCardEvent < Pulsar::Event
def initialize(@amount : Int32)
end
end
class PaymentProcessor
def charge_card(amount : Int32)
# Run code to charge the card...
# Then fire an event
PaymentProcessor::ChargeCardEvent.publish(amount)
end
end
Subscribe to it and do whatever you want with it
Now you can subscribe to the event and do whatever you want with it. For example, you might log that a charge was made, or you might send an email to the sales team.
PaymentProcessor::ChargeCardEvent.subscribe do |event|
puts "Charged: #{event.amount} at #{event.started_at}"
end
Recording timing information
You can also time how long it takes to run an event by inheriting from
Pulsar::TimedEvent
. You define them in the same way, but when you subscribe
you must also accept a second argument:
class Database::QueryEvent < Pulsar::TimedEvent
end
Database::QueryEvent.subscribe do |event, duration|
# Do something with the event and duration
end
Database::QueryEvent.publish do
# Run a query, run some other code, etc.
end
Add more information to the event
To add more information to the event you can use initialize
like you would
with any other Crystal class.
For example, we can record the database query in the event from above
class Database::QueryEvent < Pulsar::TimedEvent
getter :query
def initialize(@query : String)
end
end
Database::QueryEvent.subscribe do |event, duration|
puts event.query
end
Database::QueryEvent.publish(query: "SELECT * FROM users") do
# Run a query, run some other code, etc.
end
Testing Pulsar events
If you want to test that events are published you can use Pulsar's built-in test mode.
# Typically in spec/spec_helper.cr
# Must come *after* `require "spec"`
Pulsar.enable_test_mode!
This will enable an in-memory log for published events and will set up a hook to clear the events before each spec runs.
You can access events using {MyEventClass}.logged_events
.
# Create an event
class QueryEvent < Pulsar::TimedEvent
def initialize(@query : String)
end
end
def run_my_query(query)
# Publish the event
QueryEvent.publish(query: query) do
# Run the query somehow
end
end
it "publishes an event when a SQL query is executed" do
run_my_query "SELECT * FROM users
the_published_event = QueryEvent.logged_events.first
the_published_event.query.should eq("SELECT * FROM users")
end
Pulsar.elapsed_text
Pulsar.elapsed_text
will return the time taken (Time::Span
) as a human readable String.
Database::QueryEvent.subscribe do |event, duration|
puts Pulsar.elaspted_text(duration) # "2.3ms"
end
This method can be used with any Time::Span
.
Performance gotchas
Subscribers are notified synchronously in the same Fiber as the publisher. This means that if you have a subscriber that takes a long time to run, it will block anything else from running.
If you are doing some logging it is probably fine, but if you are doing something more time-intensive or failure prone like making an HTTP request or saving to the database you should pay special attention.
Example of a problematic subscriber
MyEvent.subscribe do |event|
sleep(5)
end
MyEvent.publish
puts "I just took 5 seconds to print!"
Oops. To get around this you can spawn a new fiber:
MyEvent.subscribe do |event|
# Now the `sleep` will run in a new Fiber and will not block this one
spawn do
sleep(5)
end
end
MyEvent.publish
puts "This will print right away!"
Potential solutions
As described above you could run long running code in a new Fiber with spawn
.
You could also use a background job library like https://github.com/robacarp/mosquito.
Be aware that running things in a Fiber will lose the current Fiber's context. This is
important for logging since Log.context
only works for the current Fiber.
So if you plan to log using the built-in Logger, you likely do not want to
spawn a new fiber. It is fast enough to just log like normal.
Contributing
- Fork it (https://github.com/luckyframework/pulsar/fork)
- Create your feature branch (
git checkout -b my-new-feature
) - Commit your changes (
git commit -am 'Add some feature'
) - Push to the branch (
git push origin my-new-feature
) - Create a new Pull Request
Contributors
- paulcsmith - creator and maintainer