error with problem 100 (CLJS comparison of floats vs CLJ comparison of Ratio)

[bo-tato]

with problem 100 https://4clojure.oxal.org/#/problem/100 when I run my solution locally I get all tests passing but on the site only 2/5 pass
when I submit any of the solutions from the solutions archive they also only get 2/5 passing so I think something's wrong with the site's evalutation

[borkdude]

Can you be more specific, e.g. list your solution here, so we can actually reproduce this?

[bo-tato]

sure, my solution was:

(fn lcm
  [head & tail]
  (first (filter
          #(every? (fn [num]
                     (zero? (mod % num)))
                   tail)
          (range head 999 head))))

when I run locally:

(def lcm (fn lcm
  [head & tail]
  (first (filter
          #(every? (fn [num]
                     (zero? (mod % num)))
                   tail)
          (range head 999 head)))))
(== (lcm 2 3) 6)
(== (lcm 5 3 7) 105)
(== (lcm 1/3 2/5) 2)
(== (lcm 3/4 1/6) 3/2)
(== (lcm 7 5/7 2 3/5) 210)

they all return true.
Also rather than having range with a hardcoded max of 999, when I use an infinite lazy seq (iterate (partial + head) head) it runs fine locally but hangs in the browser
I don't think it's something to do with my solution though as I tried submitting 5-10 random solutions from the solutions page https://4clojure.oxal.org/#/problem/100/solutions
and they all also only pass the first two tests (on the site, but work locally)

[borkdude]

@bo-tato The reason is probably that == in JVM Clojure is type-independent numerical comparison, but the solutions in this website are evaluated using CLJS and SCI, in which == has a different meaning. Also 3/4 are Ratio types in Clojure which preserve more than floating point numbers. CLJS doesn't have that concept.

We'll have to think what a proper solution for this website and problem number 100 is.

[bo-tato]

interesting, and I guess the reason the solutions that use an infinite seq and return the first solution hang is without the ratio types they never find any solution that is exactly equal

[bo-tato]

a similar thing in problem 108, in order to get my solution to pass I had to replace = with:

(fn kinda-equal [a b]
                      (< -0.001 (- a b) 0.001))

it could be good to just add a simple note on those problems, saying that in JVM clojure it uses exact ratio types while CLJS is using floating point, to save beginners some confusion why their code works locally but not when they submit

[borkdude]

Interesting library in the context of this issue: https://github.com/microsoft/same-ish