Object Oriented Code in the Gnus Newsreader

Eric Abrahamsen

Download video

Download compressed .webm video (43.9M)
Download compressed .webm video (21.3M, highly compressed)

The venerable Gnus newsreader has evolved over the years to interface with many different types of news- or mail-like backend programs, presenting all of them using a unified interface. This sort of software often calls for an object-oriented architecture, at least as regards polymorphism, yet Gnus was written well before Emacs lisp acquired the object-oriented tools and libraries – largely borrowed from Common Lisp – that it boasts today.

Yet Gnus needed something "object-oriented-like", and so nnoo.el was born: a rather amazing (and frankly terrifying) implementation of object-oriented behavior using functional code.

This talk will be a brief introduction to how this existing system works, and to the ongoing, incremental effort to port it over to newer Elisp tools like generic functions, structs, and objects.

Questions

Q3: Have you done any other projects using EIEIO and/or defstruct?

Right, EBDB is super deep into EIEIO, and was kind of written as a project for learning it, and the new gnus-search library is a more restrained usage. The search engines are defclasses, and much of the code is shared, which works quite well.

Q2: Is there may activity on maintenance of Gnus today? (and is Lars involved/aware of this work?)

Yes, there's still development going on. I don't think Lars is very focused on Gnus right now, but I run all changes by him first. He fixes bugs, but as far as I know, I'm the only one adding features right now, which is a terrifying thought.

Q1: How much of this 90's funny code can be replaced and how much will have to stay forever?

Eventually I think we can get most of it out of there. I was happy to be able to replace obarrays-as-hashtables with real hashtables, though that was a very painful process

Notes

Famous last words: "Sometimes the only thing that's worse than not knowing why something doesn't work is not knowing why it does work."

Sunday, Nov 29 2020, ~ 3:17 PM - 3:41 PM EST
Sunday, Nov 29 2020, ~12:17 PM - 12:41 PM PST
Sunday, Nov 29 2020, ~ 8:17 PM - 8:41 PM UTC
Sunday, Nov 29 2020, ~ 9:17 PM - 9:41 PM CET
Monday, Nov 30 2020, ~ 4:17 AM - 4:41 AM +08

Back to the schedule
Previous: Lakota Language and Emacs
Next: Maxima a computer algebra system in Emacs