Time series tweaks
We’ve built on the time series work from last sprint, adding support for the new structure to the time series explorer tool, and improving the metadata returned in our search results. We’ve also completed the testing and migration of time series content to the new structure in our test and production environments.
We spent two days researching some changes to our front-end web service. While the existing codebase was successful in delivering our goals during the beta, we took the opportunity to look at what we’d built to ensure its sustainable and extensible going forward – particularly as we’ll soon be building on our platform to support small area and geospatial data.
We built a number of alternative implementations of our front-end service, focusing on different parts of the codebase (for example templating, design patterns and web frameworks). This helped uncover some complexity in our existing design, and the lessons we’ve learned from the spike will be carried forward into future development work.
Our next steps will be to review the output of the spike, identify specific areas which need improvement, and integrate those changes into our technical roadmap for the months ahead.
Git flow, pull requests and code reviews
Throughout the alpha and beta we’ve been a bit too relaxed about our development practices, and we’ve been caught out a couple of times with bigger than intended releases. As our permanent team is now in place we felt it was a good time to start making some improvements.
From the start of our next sprint we’ll be using the Gitflow branching model, and as part of that we’ll be using GitHub pull requests to manage our workflow, with all pull requests tested and code reviewed by another member of the team.
Bug fixes and iterations
- We’ve made some additional improvements to our PDF downloads, adding support for data tables to the PDFs of some additional page formats, and pre-generating PDFs for all the formats.
- Our Excel data downloads now correctly format number values as numbers, instead of returning them as a text value. This will apply to all Excel time series data we publish going forward.