Replace URL.canParse() with better supported code

[aloisklink]

URL.canParse was only pretty recently created, and is not available in many environments.

We can instead wrap a URL creation in a try..catch, as listed in MDN. Unfortunately, this is a bit slower, but at least it works almost everywhere:

This is a fast and easy alternative to constructing a URL within a try...catch block.

Taken from https://github.com/mdn/content/blob/392ce991114e55e2187510b640ab545d09258a16/files/en-us/web/api/url/canparse_static/index.md?plain=1#L13

Fixes #78

---

I've also updated the CI to run on Node.JS v18.0, instead of Node.JS v18.x. This is because #78 would have caught this issue, as URL.canParse() was only added in Node.JS v18.17.

[aloisklink]

If it helps, @ibooker said in #78 (comment) that this bug has an internal tracking ID of BTWEB-171.

[dcousens]

return URL.canParse?.(url) ?? (new URL(url) ? true : false)

[aloisklink]

This won't work, since new URL(url) throws an exception if it fails, so it needs to be wrapped in a try..catch.

But, we could do something like:

  if (URL.canParse) {
    return URL.canParse(url);
  }
  try {
    // We can replace this with URL.canParse() once it's more widely available
    new URL(url); // eslint-disable-line no-new -- We're only checking for exceptions
    return true;
  } catch (error) {
    return false;
  }

However, I'm not a fan of this. Sure, it's slightly faster in environments that do support URL.canParse, but the code is more complicated, and it's much harder to test, as we'd need environments that both support URL.canParse and don't support it.

[dcousens]

to be clear, I didn't suggest to remove the try catch

[dcousens]

I guess it's a useless suggestion though, when the documentation says:

It returns true for the same values for which the URL() constructor would succeed, and false for the values that would cause the constructor to throw.

If that holds true in perpetuity, then don't worry about my suggestion

[Slokilla]

Why do you pin it ? I have read your explanation, but I don't understand more.

[aloisklink]

It's because URL.canParse doesn't exist in Node.JS v18.0, since it was only added in Node.JS v18.17.

So this test will make sure that we don't accidentally add anything that breaks with old versions of Node.JS v18. (if we do, we'd ideally want to make a new MAJOR change and state that the minimum supported of Node.JS has changed in the release notes).

For example, Amazon Linux 2023 only comes with Node.JS v18.12.

[Slokilla]

Oh, that's a clever move. Thanks for explanation.

[adrmachado-public]

closing this as this overlaps with #93 which is more recently updated and has tests