Avoid using Array.prototype.find

[tinydylan]

💡 Idea

constantFrom uses Array.prototype.find, which is not supported on IE11.

Motivation

TinyMCE is a popular open source product and we have switched from jsc to fast-check in the last year. We still support IE11.

[tinydylan]

We hit this issue when using the webUrl arbitrary.

[dubzzz]

Good point, I think we can switch to Array.prototype.indexOf while keeping the same behaviour.

Marking the issue as 'good first issue'

[tinydylan]

Array.prototype.find takes a predicate, whereas Array.prototype.indexOf takes a normal value.

Would you be ok with me adding something like this:

export function findOrNull<T> (ts: ArrayLike<T>, f: (t: T) => boolean): T | null {
  for (let i = 0; i < ts.length; i++) {
    const t = ts[i];
    if (f(t)) return t;
  }
  return null;
} 

in a file like src/check/arbitrary/helpers/ArrayHelper.ts ?

[tinydylan]

PR is up: #583

Happy to adjust the code whatever the project expects.

[dubzzz]

Thanks a lot for the contribution. I started a release of 1.24.2 that will contain your fix. I may backport the fix on older versions if needed (but as I follow semver, bumping to 1.24.2 should not be problematic)

[dubzzz]

TinyMCE is a popular open source product and we have switched from jsc to fast-check in the last year. We still support IE11.

@tinydylan Whaouu great news :) Please let me know if you miss some features that were provided by jsc. Or what you liked more or less in fast-check in general

[tinydylan]

:)

At risk of sending this issue off-topic:

At Tiny, our engineering team uses functional programming heavily. We write in Scala, Haskell, TypeScript and ReasonML, and use FP stuff all through it. We're no strangers to property testing - we use QuickCheck and ScalaCheck, and in our TypeScript code we were using Jsc.

TinyMCE was ported from JS to TS over the last few years and we've been gradually adding types all through it. Unfortunately, Jsc's TS bindings were a bit lacking, so we switched to fast-check. We're starting to move more of TinyMCE to ReasonML and one of our engineers has made ReasonML bindings for fast-check.

I'm a huge fan of fast-check, to be honest. fast-check does pretty much everything I expect, the way I expect it. You've really done a great job.

The IE support has been the biggest issue so far - hence the PR. We also hit some issues because fast-check uses Symbol - again, not supported by IE. We've polyfilled that in our testing tooling for now, but will probably log a ticket for that soon.

I logged a ticket about non-empty lists and non-empty strings a while ago - and I was happy with the answer. I find with a lot of these property test tools that it helps to have quite a wide range of generators built-in. Non empty strings, ascii strings, uppercase ascii strings, alphanumeric strings, hex strings, non-empty hex strings, non-empty uppercase hex strings, number-except-for-NaN, positive ints, negative ints, natural numbers, rationals, domain names, tcp ports, html tags. I like there to be lots of generators.

A liftA2/liftA3 etc could be handy - maybe something like fp-ts's pipe or fp-ts-contrib's attempt at do notation. I really wish we could use proper do notation in TS - that would make composing generators much easier.

Anyway - I'll log separate tickets for these if they become important. For now - I think we can close this ticket, and on behalf of Tiny, we thank you and your contributors for helping us crush edge cases in TinyMCE.

[tinydylan]

BTW - I don't think I've seen anything in jsc that is missing from fast-check. We still have some code to port - mostly tests that generate complex HTML.

[dubzzz]

Is it working fine now? If so can we close the issue?

[tinydylan]

Testing now. Will get back to you soon.

[tinydylan]

Yep. Looks good.