Weekly update – 24 September 2021

[ryan]

This week I've been focusing on the dev branch and some low level code optimization, especially as it relates to the PW boot process, translating URLs to pages and identifying the page, language, URL segments, page numbers, etc. from the request URL. I'm breaking down much of the logic (currently in the ProcessPageView module) into more focused parts that can become part of PW's $pages API, increasing reusability of related code. (Don't worry, this won't affect any existing hooks). At the same time, I'm looking for opportunities for optimization and performance improvement, and have already found a few. For instance, the LanguageSupportPageNames module introduces overhead into some page finding operations and there's good improvement potential there, among others. I'm pretty much still in the middle of it all though, so am going to hold off on committing any of this to the dev branch until it's further along. But since we've got a quiet commit log this week, I just wanted to keep you up-to-date. Hopefully some of these fairly significant updates will be stable enough to commit to the dev branch next week. By the way, these updates will also enable the SessionAllow module (from last week) to be configurable by page, as PW will now identify the requested page before starting the session. More soon. Have a great weekend! 

[Ivan Gretsky]

That's intriguing! Waiting to see how it will turn out. Are we going to have some sort of middleware ??

[ryan]

There's a couple things that made me think this part of the core could be better (and nothing related to middleware). First is that PW didn't originally support multi-language URLs, and so the LanguageSupportPageNames module was built to hook in and add support for that. There are benefits (performance, simplicity, etc.) to having that logic lower-level in the core and part of the same logic that identifies non-multi-language URLs. So that's something I've been interested in for a long time, but it was always more than I wanted to get into.

The other thing is that I was working on that SessionAllow module and found it was a real drawback not to be able to know what Page was requested before deciding on things like whether to allow a session or not. Granted, we can still know the request URL before we know the Page, but it just seemed a lot more flexible for the core to know the requested Page much earlier in the boot process. That way, we can do things like configure whether sessions are allowed on a per-template basis, where such an option seems to make the most sense. Another example is 404 pages—maybe we don't need a session on a 404 page. But without knowing whether the URL will resolve to a Page or not, we didn't have that choice. Following these updates, we will.

Lastly, I just like to revisit the core logic for things and improve it when/where possible, which helps to keep it fresh on the mind. But I don't like to get into replacing this kind of major core code when it'll soon be merged to the master, as this is the kind of stuff that needs more testing over a longer period of time. So with a comfortably stable master version available right now, it seemed like a good time to work on this.