Fix to avoid minification problem with esbuild

[diegovdev]

esbuild will minify this to 1.toString (removing the trailing zero) which will result in a syntax error in the browser. A trailing zero will be removed during minification but a trailing 1 wont.

See image below:

[diegovdev]

Proof that it's the same resulting toString method, in the browser console:

[0f-0b]

esbuild actually minifies 1.0.toString to 1 .toString (note the space), so it must be some other tool that misprinted the code. The bug is in that tool, not core-js.

[zloirock]

This is a bug on your minificator side. I already added some workarounds for such kind of parser / minificator bugs to this line (for example, 1..toString -> 1.0.toString), but I don't think we should encourage these bugs any further. I can accept this PR for compatibility with old versions of this tool (because it costs just 1 byte of compressed code), but only in case if this bug will be fixed in this tool itself.

[diego-carvallo]

@zloirock we use Vite (https://vite.dev/) which under the hood uses esbuild. We tried enforcing esbuid configs and nothing worked, we tried to use terser as minifyer but we got the same error as with esbuild. It makes sense that some tooling on the road will remove unecesary spaces in 1 .toString, you shouldn't be relying on a space to be a solution.

The fix I proposed here is just preventing using a zero 0 so that you don't run any more into minifyers trying to modify the expression. For a minifyer 1.0 is safely reduced to just 1, they have no configs to prevent this behavior.

It's funny how such a small and efficient fix is just rejected, this for sure would have prevented other people to run into this annoyance. Our only solution now is to have this ugly plugin for Vite that just pollutes every project we have that depends on core-js.

function corejsMinificationFix() {
  return {
    name: 'corejsMinificationFix',
    enforce: 'post',
    apply: 'build',
    transform(code, id) {
      if (id.includes('xxxx/node_modules/core-js/internals/uid.js')) {
        return {
          code: code.replace('1.0.toString', '1.1.toString'),
          map: null,
        }
      }
    },
  }
}

[zloirock]

@diego-carvallo I'm curious: Did you miss my comment above? It's not a core-js bug, but I'm ready to accept this PR as a workaround in case you at least report it to the source of this bug. Accepting this PR without that is irresponsible since it will cause the same problems in case of such a valid JS expression in other code — and there were already examples of this mentioned above. Is reporting this to the source of the problem so hard to do and easier to throw a tantrum?

[diego-carvallo]

Here is the issue reported to ESBuild evanw/esbuild#3975