This new release secures the class mechanism and improves project tools.
New version
Here comes the new version:
- Stories and bugs implemented
- Sources
- Documentation
Improved $super keyword
When writing the article My own super implementation I found several issues that were immediately fixed.
Better tooling
The following tools were modified:
- It is now possible to remove unused files from the sources page
- Each file modification triggers one pass of tests (it takes less than 1 second to execute). This way, I know as soon as possible if something goes wrong
- fs middleware is secured to limit access to the project files only. It now supports the DELETE verb
- The watch and serve tasks are monitoring the sources.json file modifications to update the list of linted files. This way, it is no more required to restart the grunt task.
More flavors for browser testing
Selenium was upgraded to version 3 and the detection has been fixed to make it more reliable.
On top of Selenium, a command line execution wrapper combined with a caching service (to store & grab the results) allows testing of non-automated browsers (such as Safari on Windows). Once the tests are done, the command line is killed.
Automated release
Releasing a version has never been so easy, a script using github-api module to call the GitHub API implements the following steps:
- Check version number
- Update package.json (if needed)
- Check GitHub milestones to identify the milestone details and check that all issues are closed
- Update README.md
- Grunt make
- Copy tmp/plato/report.history. to build/ (grunt copy:releasePlatoHistory)
- commit & push
- Create a new release on GitHub
- Copy build/tests.js into test/host/legacy/{version}.js
- commit & push
However, once last step was forgotten: closing the milestone. An incident is created.
Lessons learned
Documenting features makes them better
I will task the risk of repeating myself here but the article about super made me realize several mistakes in the implementation of the $super equivalent. Furthermore, taking the time to explore the ECMAScript super keyword gave me a better understanding of the feature.
In general, it is valuable to step back from the code and document the intent behind a feature.
Better Selenium detection
One of the reasons why I wanted to remove Selenium was the buggy detection. Indeed, there are some failures which are not encapsulated properly in a Promise. As a result, the whole process fails when it happens.
After digging on the web, I found this excellent thread on NodeJS Exception handling. It allowed me to handle those unmanaged exceptions the proper way and it secured the detection.
Next release
The next release will introduce the interface concept.
No comments:
Post a Comment