Input: Thank You project Phase 1: Part 1

Summary

Beginning

When users click on "Submit feedback..." in Firefox, they end up on our Input site where they can let us know whether they're happy or sad about Firefox and why. This data gets collected and we analyze it in aggregate looking for trends and correlations between sentiment and subject matter, releases, events, etc. It's one of the many ways that users directly affect the development of Firefox.

One of the things that's always bugged me about this process is that some number of users are leaving feedback about issues they have with Firefox that aren't problems with the product, but rather something else or are known issues with the product that have workarounds. It bugs me because users go out of their way to leave us this kind of feedback and then they get a Thank You page that isn't remotely helpful for them.

I've been thinking about this problem since the beginning of 2014, but hadn't had a chance to really get into it until the end of 2014 when I wrote up a project plan and some bugs.

In the first quarter of 2015, Adam worked on this project with me as part of the Outreachy program. I took the work that he did and finished it up in the second quarter of 2015.

Surprise ending!

The code has been out there for a little under a month now and early analysis suggests SUCCESS!

But keep reading for the riveting middle!

This blog post is a write-up for the Thank You project phase 1. It's long because I wanted to go through the entire project beginning to end as a case study.

Read more…

Ditching ElasticUtils on Input for elasticsearch-dsl-py

What was it?

ElasticUtils was a Python library for building and executing Elasticsearch searches. I picked up maintenance after the original authors moved on with their lives, did a lot of work, released many versions and eventually ended the project in January 2015.

Why end it? A bunch of reasons.

It started at PyCon 2014, when I had a long talk with Rob, Jannis, and Erik about ElasticUtils the new library Honza was working on which became elasticsearch-dsl-py.

At the time, I knew that ElasticUtils had a bunch of architectural decisions that turned out to be make life really hard. Doing some things was hard. It was built for a pre-1.0 Elasticsearch and it would have been a monumental effort to upgrade it past the Elasticsearch 1.0 hump. The code wasn't structured particularly well. I was tired of working on it.

Honza's library had a lot of promise and did the sorts of things that ElasticUtils should have done and did them better--even at that ridiculously early stage of development.

By the end of PyCon 2014, I was pretty sure elasticsearch-dsl-py was the future. The only question was whether to gut ElasticUtils and make it a small shim on top of elasticsearch-dsl-py or end it.

In January 2015, I decided to just end it because I didn't see a compelling reason to keep it around or rewrite it into something on top of elasticsearch-dsl-py. Thus I ended it.

Now to migrate to something different.

Read more…