Kiwi TCMS roadmap for 2018

cleaner code, improved UI/UX and more tests

Posted by Kiwi TCMS Team on Mon 22 January 2018 under community , roadmap

Hello everyone. As you know Kiwi TCMS has been around for a while and it has gone through some big changes in the last year! It started as an abandoned Django 1.6 project with broken test suite and is now currently running on the latest Django 2.0.1 with Python 3.5! It is clearly a legacy code base!

We, the Kiwi TCMS team, have learned a lot more about the project and this blog post describes our roadmap for 2018 in terms of technical work and community engagement. The general technical direction is cleaner/simpler code, improved UI/UX and more tests!

Make code easier to maintain

The current code is a bunch of very large modules and functions and classes bundled together. It is also old and sometimes looks like spaghetti code. CodeClimate gives us a "D" rating for maintainability with a 1 year estimated effort to fix that. There are 600+ code smells and 600+ duplications detected by CodeClimate. Our goal is to get this number down to zero!

Use pylint and pylint-django

pylint is the standard static analyzer for Python and currently it reports over 4000 errors and warnings when executed against Kiwi TCMS. This is a huge number and it needs to become zero! We've also identified interesting patterns that will make it into pylint and pylint-django plugins!

Render HTML, return JSON

The current state of affairs is a mix and match of everything. There are views that render HTML, others which return pure JSON, other which return HTML encoded as JSON string and probably everything in between. Views that render pages need to do just that and views that are used with AJAX calls from the UI need to return pure JSON! A lot of these are hiding in plain sight and will come to light when we start overhauling the user interface.

Submit forms, post JSON, GET clean URLs

There are lots of forms in Kiwi TCMS. Sometimes they are submitted by the user and other times they are serialized and POSTed by some JavaScript code. This isn't very easy to understand plus this entire home-grown utility code doesn't bring anything useful to the project. All of these need to be identified and cleaned up. JavaScript code needs to send or consume JSON, nothing else!

There are also lots of places where Kiwi issues GET requests with a number of query parameters. Our goal is to minimize these and where possible have the parameters as part of the Django urls scheme, not as query strings.

API layer

The current API module is a bit disorganized. API namespaces don't match the names of the underlying DB models and the API client classes don't match any of the other two! The way we pass parameters and what these parameters are named should match between the client, the RPC method and the underlying model!

In earlier releases we've removed duplicate or similar RPC methods and we think there are more of these that need our love.

Documentation

We've started producing documentation from doc-strings and most of the RPC methods have such. However it is unformatted and barely readable, sometimes even inaccurate. Previous releases made progress on this but there's a lot more to cover.

All RPC methods should be documented first and then the rest of Kiwi's internals to make development easier!

No vendored JavaScript libraries

There are 11 vendored-in JavaScript files that we carry around in Kiwi TCMS. Most notable are jQuery plus a few addons and Handlebars.js. To the extent possible our goal is to use jQuery provided by Django or installed via NPM dependencies!

Less HTML templates with better organization

There are over 100 HTML templates in Kiwi TCMS. There are also email and even JavaScript templates. For example there are get_cases.html and get_review_cases.html which are essentially the same. This is kind of also true for templates used in new and edit views! Such templates should be unified!

Those JavaScript templates need to be totally gone!

All templates are stuffed together in a single directory and their names are not very predictable. They need to be split per application and follow some kind of naming convention.

Modern interface with Patternfly

The UI already uses the Patternfly library for some of its widgets. Most notably the main navigation header. Our goal is for the entire UI to be ported to Patternfly by updating widgets HTML and CSS where needed. This will also help clean things up a lot. At the same time we'll be rethinking how information is organized on screen and how the person interacts with it! Usability suggestions are very welcome!

This move will also help us get rid of the Handlebars dependency which is now used for pop-up dialogs.

JavaScript updates and front-end testing

There's lots of JavaScript code on the front-end and honestly we don't quite know what it does or how it is organized. There are also no tests on the front-end whatsoever!

It is our goal for this to change with the intention to not overdo the JavaScript part and keep things very minimal. At the moment we're thinking about vanilla jQuery that will act as a proxy between the browser and the back-end!

Community efforts

A year ago hardly anybody knew what Kiwi TCMS was, the project didn't even have this name, there was 1 active contributor and hardly any users! Now the community has started to slowly revitalize, we're seeing some contributions from our junior team members (more on this in another blog post) and also seeing folks installing and using Kiwi TCMS and reporting interesting bugs and feature requests around it!

For the upcoming year our goal is to seek active engagement with other open source projects that could make use of Kiwi TCMS and work with them. Kudos to openSUSE for being the first to propose such integration! We're also planning couple of appearances at a few conferences but there's nothing confirmed yet.

Every other contribution in terms of bug reports, new users and feature requests is of course welcome but we're very conscious of the fact that there's tons of work to be done and the team is still very small!

Team wise we'd like to get the existing team members up to speed and tackle the above tasks with priority. This will also help us introduce bug fixes and new features more quickly!

Happy testing!