Skip to content

Latest commit

 

History

History
 
 

transform

README.md

Overview

Transform is used to create relevant metrics from the data saved in the SQL database. The SQL database contains the issues and the list of events, but we may want to calculate additional metrics that reflect the health of the project. For example, we may want to understand when certain labels were applied, and what happened to the pull-request or issues through its lifetime.

This logic is written as "Plugins". A quick look at the code (plugins.go) explains the interface of a plugin, and the type of parameters it receives.

Plugins either wait for:

  • Changes to issues (they come by order of modification)
  • Events and comments: they come sorted by creation date.

Note that you will always receives changes to issues before receiving events or comments.

The program periodically fetches from the SQL database to find changes and pushes them to each plugin.

Walk-through: Creating a new plugin

To create a plugin, you need to implement the interface defined in plugin.go, and also register its creation there.

Let's review an example plugin, merged.go:

  • At registration time, NewMergedPlugin tries to get the last measurement from the InfluxDB, so that it doesn't have to store data it has already written. Every time a plugin is run, it receives all historical events from the SQL Database, and it processes all of them, but only stores new ones.

  • We need to implement ReceiveIssue, ReceiveComment, and ReceiveIssueEvent even if we don't use all of them.

  • When we receive an Event, in this situation, we want to count how many "merged" events we received, so we just discard all events other than "merged". Then we discard events that are already in the database because they we have already processed them. And finally, if the event is of the correct type and we have not already processed it, we insert the value in the database through the InfluxDB interface.

  • We make sure the plugin is registered in NewCountPlugin in count.go or it's never going to receive any events.

Testing locally

In order to test this program locally, you will need a database set-up in MySQL, populated with data. Refer to ../fetcher documentation to see how to populate your own database.

Once it is set-up, you will also need the grafana-stack set-up locally. Refer to ../grafana-stack to see how to do that.

You can then run transform to connect to your local instances.