splitWordsRegex fails on iOS 16

[acorn1010]

Safari iOS below 16.3 doesn't support certain group specifiers in the below RegEx, probably lookbehind: https://caniuse.com/?search=lookbehind.

const splitWordsRegex = new RegExp(
    '[^\\dA-Za-z]' +  // match any character that is not a letter or a digit
    '|' +                    // or
    '(?<=[a-z])' +           // lookbehind for a lowercase letter
    '(?=[A-Z])' +            // lookahead for an uppercase letter
    '|' +                    // or
    '(?<=[A-Z])' +           // lookbehind for an uppercase letter
    '(?=[A-Z][a-z])'         // lookahead for an uppercase letter followed by a lowercase letter
);

I haven't fully tested it, but what about replacing this splitWords function with something like:

const splitWordsRegex = /([\dA-Z][\da-z]+)|([\dA-Z]+)|([\da-z]+)/g;

export function splitWords(str: string): string[] {
  return [...str.matchAll(splitWordsRegex)].map(match => match[0]);
}

Other issues I've encountered with switching from Lodash to moderndash on a medium-size project:

• range(0, 3) is inclusive instead of exclusive (e.g. returning [0, 1, 2, 3] instead of [0, 1, 2]). This is different from Lodash, and resulted in some bugs.
• merge({}, b) doesn't modify the first input. While it's typically best practice to not modify inputs, I believe it makes sense here because otherwise a shallow copy (e.g. {...{}, ...b}) could be used in most cases. I've personally used Lodash's merge in areas of my code specifically because it modifies the input, cutting down on garbage collection.

[Maggi64]

Hi @acorn1010, thanks for the suggestion. I will answer this step by step. Your regex doesn't account for this case:

AssertionError: expected 'some_cruelw_orld' to be 'some_cruel_world' // Object.is equality
 ❯ test/string/snakeCase.test.ts:10:45
     10|         expect(snakeCase('someCRUELWorld')).toBe('some_cruel_world');
       |                                             ^
     11|     });
     12|

  - Expected   "some_cruel_world"
  + Received   "some_cruelw_orld"

But I agree that we should at least make it compatible with Safari 15. I will see what good replacements are out there when I have the time.

[Maggi64]

range(0, 3) is inclusive instead of exclusive (e.g. returns [0, 1, 2, 3] instead of [0, 1, 2]). This is different from Lodash and caused some bugs.

I decided to do this to have a more readable syntax. All ranges in moderndash include the last parameter. In randomInt and randomFloat the last parameter is also included in the range.
ModernDash is not intended to be a drop-in replacement. Instead, I tried to make the syntax as readable as possible. Therefore I would like to keep the range function as it is.

[Maggi64]

@acorn1010

The issue got autoclosed with merging the PR.
I added a fallback for browsers that don't support positive lookbehind in regex. Let me know if it works! :)

We can still discuss if it makes sense to modify inputs in merge.

[acorn1010]

Sorry about the original RegEx—just noticed there were tests! This one should cover everything in the tests, plus some other edge cases like "enable6HFormat" -> ["enable", "6H", "Format"] while still maintaining a small file size:

const splitWordsRegex = /(\d*[a-z]+)|([A-Z][a-z]+)|(\d*[A-Z]+(?=[^a-z]|$))|(\d+)/g;

(I got the enable6HFormat from here: https://github.com/lodash/lodash/blob/4.17.11/test/test.js#L2351)

P.S. The fallback works!

[Maggi64]

Thanks, this is a much simpler solution. I deployed it in 3.3.1. It is 10% slower than my lookbehind solution, but the
"enable6HFormat" -> ["enable", "6H", "format"] case was handled differently. It makes sense to match lodash's behaviour and its still 2x faster. 🥳

[Maggi64]

@acorn1010 opened a new issue for the array mutations, where we can discuss this further. ⬆️