selfrefactor/rambda

R.uniq is slower than Ramda for big arrays

bovine3dom opened this issue · 5 comments

bovine3dom commented

Minimal example:

// Make an array of 10,000 strings with some repetition
let strings = R.map(() => Math.floor(Math.random() * 1000) + "", R.range(0,10000))

// Time R.uniq
const t0 = new Date().getTime()
R.uniq(strings)
const t1 = new Date().getTime()

// Store the result
const result = t1-t0
console.log(t1 - t0)
  • in Ramda: ~1ms
  • in Rambda: ~600ms

Related downstream issue: tridactyl/tridactyl#3772

: )

selfrefactor commented

I will run the code through Rambda benchmark and I'll get back to you.

selfrefactor commented

Thank you @bovine3dom as this is a serious issue, which I can confirm. I will report again when there is a progress on this problem.

  • ❤️2
selfrefactor commented

I have pinned the reason for this issue. Expect a fix with version 6.9.0.

  • ❤️1
selfrefactor commented

6.9.0 was released with improved performance for R.uniq and several other related methods. I am closing the issue, but please feel free to comment further.

selfrefactor commented

As a side note I am adding a bit more information about the result.

During developing the fix, I realized that underlying issue was lack of correct benchmarking process. As a result, there is now a new format for Rambda benchmark(https://github.com/selfrefactor/rambda/blob/master/source/benchmarks/uniq.js) which can produce dev-only benchmark result: https://github.com/selfrefactor/rambda-scripts/blob/master/scripts/run-benchmarks/benchmark-results/uniq.json

UNIQ-ALL key refers to the result used in README, while UNIQ-0, UNIQ-1, etc. refers to the index of modes list used to generate the benchmark.

  • 🎉2