Dwayne McDaniel did some thorough reporting of deprecated code use in all Drupal 8 contributed modules in March. Ultimately this kind of reporting would be best to have on drupal.org but while that is figured out, Dwayne’s data set provides a very nice way to mine data about Drupal 9 readiness of contributed modules and to inform our tooling to improve the process. His original numbers showed that almost 44% of contributed modules had no deprecated code use at the time. What I was interested in was how to help the rest of the 56%.
Dwayne created an updated process and a new repository this week with fresh data. I was still curious so I delved right into the data and started mining it. A key question I was interested in is how much of the most widespread deprecations are actionable right now. An actionable deprecation is something core deprecated in a previous version that is not supported anymore, so you can update your code to remove the use of that API. Currently Drupal 8.6 and 8.7 are supported, so deprecations there should only be acted on for your custom code. However deprecations in and before 8.5 are entirely fine to act on.
First I counted the top list of deprecated APIs used from Dwayne’s data across all of contributed projects. Then I wrote a script to collate api.drupal.org documentation to the deprecation notices. Ideally phpstan itself would report these messages directly and Matt Glaman is working on that. However since that is still blocked on the phpstan side, one needs a different data source to find the deprecation documentation for each occurrence, so I took to api.drupal.org to get that for now. Once that is found, we can categorize the deprecations into actionable, not actionable and actionable for custom code only. For the later case you know which core version you are using, and that should be an up to date minor version. So you don’t need to deal with what core branches the community supports otherwise.
The results look really promising so far in terms of how much contributed modules can make progress on even today. If all already actionable deprecations get resolved, there will be very little left at least of the deprecations we already know.
A month ago Dezső Biczó created a set of proof of concept Rector fixes to automate some of these deprecation fixes, so I opened an issue with this new data set to try and cover the top ones that are not just actionable but approachable to automate.
As with all interesting data sets, this summary is just the tip of the iceberg. There is huge potential to mine this data set for other uses, such as finding modules that potential contributors at an event could contribute fixes to. I don’t know yet how much I can continue to work with this data myself, and of course others doing analysis of their own would be more than welcome.
Are you a drupal.org project maintainer? Now would be a good time to fill in your Drupal 9 porting information in your project, so you can let contributors know how to best engage with your in the process towards Drupal 9.
https://t.co/hf2ENvlZSo projects can now specify Drupal 9 porting information, so *you* can direct *your* contributors to provide the most valuable help on the way to Drupal 9, fund the process or just step back (for now). Edit your project to help your contributors help you! pic.twitter.com/l1OWwOllBK
— The Drop is Always Moving (@DropIsMoving) May 21, 2019
Disclaimer: The data is based on the state of contributed projects on May 20, 2019 based on Drupal core’s 8.8.x development branch on May 20, 2019. The lookup on api.drupal.org was not perfect and I found some bugs there that are being resolved as well so the data gets more accurate. Also, as contributed modules will get updated, there will be less uses of deprecated APIs. As core will introduce more deprecations, the data could get worse. There may also be phpstan Drupal integration bugs or missing features, such as not finding uses of deprecated global constants yet. This is a snapshot as the tools and state of code on all sides evolve, the resulting data will be different.