borkdude / advent-of-cljc Goto Github PK

I'm the creator of Advent of Code. A system similar to this one was recently proposed on Reddit; it makes me nervous for a few reasons which I listed in a reply:

https://www.reddit.com/r/adventofcode/comments/9yylms/advent_of_code_where_submitted_code_is_tested_on/ea5ck28/?context=3

Please don't create a database of inputs. Doing this would make it easy for someone to rip and re-host my work.

Doing a beauty contest, size comparison, etc don't require a common runtime; you can already compare the code of many solutions on the megathreads or even GitHub.

Some puzzles (intentionally) use a lot of CPU or memory, which could make hosting this service expensive.

Without having a human check the submitted source code, there's not a good way to verify that the program doesn't simply print the answer and immediately exit.

Do not duplicate the puzzle descriptions on such a site. Doing so could remove adventofcode.com entirely from users' workflows and effectively steal my work and use it to also steal my traffic. (In fact, I'm already nervous about this even without duplicating puzzle descriptions.)

Basically: I spend hundreds and hundreds of hours building this stuff. Please be careful not to ruin it.

@mfikes said:

An aside on the script topic: Previously coal-mine used bash scripts. At some point I converted it to simply use Clojure that is in the source tree (the coal-mine.script namespace). For me this had two benefits:

1. I actually found it easier to deal with logic in the "script" when written instead in Clojure
2. This made it so that you can run coal-mine as a GitDep. In other words, there is no need to even manually clone it. (This capability is, for example, used in speculative.)

This, and this commit to Tubular by David mfikes/tubular@e8b5691
really solidified my thinking that having a -main for logic in a codebase, driven by deps aliases, is really a great way to arrange things.

I think that's a good idea!

Currently only different namespaces compared to master are tested on CI.
The time comparison compared to others for the same day is now missing.

TODO:

• discover which days have been committed.
• run tests for all of these days regardless of user

Most namespaces use lazy seq and delays or memoized functions at the top level. This accumulates used memory and eventually causes problems for the later tests.

One approach to solve this could be: delete the namespace after the last test using a global fixture.

If you run

script/test-one 2017 1 borkdude

everything will be fine. But then add the following

(defprotocol IFoo)

as the first form in the aoc.y2017.d01.borkdude namespace (after the ns form).

This will result in

WARNING - incomplete alias created for namespace aoc.y2017.d01.borkdude

where otherwise everything else works.

In the end, this is likely going to be an issue in the ClojureScript compiler itself (presuming we can come up with a minimal repro), but logging it here as it may be occuring due to some arrangement in the Advent of CLJC project.