Skip to content

Latest commit

 

History

History
36 lines (33 loc) · 4.07 KB

FULCRO_MIGRATION_NOTES.md

File metadata and controls

36 lines (33 loc) · 4.07 KB
Raw

Migrating to Fulcro

I initially wrote the ClojureScript web interface in 2015/2016, using Om Next. It was fun, but I didn't particularly enjoy dealing with Om. Looking at this again in 2021, it's clear that Om is a moribund framework, with newer takes on the same ideas having come to the forefront, namely Fulcro and re-frame. Because I'm now working in an environment where Fulcro is widely used, I'm going to port Snergly's web interface to Fulcro as a way of exploring and learning it. (I may later port it to re-frame as well, because it is also used in my workplace, although not as much as Fulcro.)

Steps:

  • Make sure the tests still run. (As I recall, I was in the process of converting them to use clojure.spec and I'm not sure I left them in a working state.)
  • Finish converting property-based tests to use spec rather than test.check, to the degree that this makes sense.
  • Add notes to the README about commands for various things (running tests, building, etc.)
  • Create a deps.edn file that points to up-to-date dependencies, including Fulcro. (I'll probably have to also include Om libraries, so that core.cljs will compile.)
  • Create a shadow-cljs.edn file and configure it to build with Fulcro. (It would be nice if I could also get it to work with the existing Om build, but that's not required.)
  • Get the CLJC tests running in ClojureScript as well.
  • Build a Fulcro front-end in fulcro.cljs (with a fulcro.html file to keep things completely separated).
  • Break the implementation up into separate namespaces to clean the design up a bit. The single-file implementation is difficult to navigate.
  • Update the push-pages script to deploy the Fulcro version as index.html (but also leaving it as fulcro.html).
  • Build a re-frame front-end in re-frame.cljs (with a re-frame.html file to keep things completely separated).
  • Update the push-pages script to deploy the re-frame version alongside the Fulcro version.
  • Do something with the old Om version. Either:
    • Rename it to om-next.cljs/om-next.html and get it to work with shadow-cljs and the new push-pages script; or
    • Just delete it since Om isn't being supported anymore.
  • Delete project.clj, .lein*, figwheel-server.log, script/figwheel.clj, and any other remnants of Leiningen, Figwheel, and (maybe) Om.
  • Convert animation.cljs to a CLJC file. (I can use reader conditionals to choose the correct version to require if that distinction is still required.)
    • Why do this? Because I don't think it will be hard, and there's nothing in the file that really requires ClojureScript, so it shouldn't be a cljs file. This would allow writing tests against this code that run in the JVM alongside the algorithm tests and the like. Additionally, while I'm not aware of a JVM-based GUI framework I would want to use to write a non-web version of the animated interface, that doesn't mean there isn't one; I've just been out of that world for so long. So why leave this as an apparent barrier to that if I can easily move it over?
  • Revisit my use of core.async based on Tim Baldridge's Core.Async In Use talk. (In particular, his "just don't use async" section at the 13:00 mark that uses a transducer instead might allow me to get away from using core.async at all.)
  • Better dynamic status display (not just a single message).
    • Total cells
    • Cells carved
    • Cells flooded
    • Length of longest path
    • Algorithm-specific stats (number of backtracks, total length of backtracking, number of deadends for hunt/kill, number of successful/failed start attempts for wilsons)
  • Add pause/resume buttons for the animation.
  • Keep stats over several runs (mainly length of longest path)
  • Allow backing up through history?
  • Keep on going through other algorithms and tricks in Jamis' book?
  • Add descriptions of the algorithms
  • Other analyses/stats on the mazes? Are there ways I can determine and or quantify bias or other attributes? Average number of connections per cell, broken down by NSEW, perhaps?