Add best-effort support for RegExp flags in `z.string().regex()` validations

[Spappz]

Summary

Reworked the string parser to attempt to mutate provided regexes into a flag-independent form. When the applyRegexFlags option is falsy, nothing actually changes; the regex.source that was previously fed into addPattern's arguments now just happens within addPattern itself.

The following flags are supported:

• i case insensitive
• m multiline (^ and $ anchors adjacent to newlines)
• s dot-all (. matches newlines)

I introduced a new option applyRegexFlags because, on inspection, it's not mutually exclusive to patternStrategy: the former applies for z.string().regex(), but latter applies only for .includes(), .startsWith(), and .endsWith(). I've reworded the description for patternStrategy to highlight this a little more clearly—I honestly didn't properly understand it until I took a look at the code!

To make the code cleaner, I also changed zodPatterns to be the full regexes from Zod (made case-insensitive). I also updated a couple that were out-of-date relative to Zod itself 😅

Example

import { z } from "zod";
import { zodToJsonSchema } from "zod-to-json-schema";
const zodSchema = z.string().regex(/foo .\* (ba[r-z_])+$/ims).startsWith("@foo").min(5);
const jsonSchema = zodToJsonSchema(zodSchema, { applyRegexFlags: true });
console.log(JSON.stringify(jsonSchema, null, "   "));

{
   "type": "string",
   "allOf": [
      {
         "pattern": "[fF][oO][oO] [.\r\n]\\* ([bB][aA][r-zR-Z_])+($|(?=[\r\n]))"
      },
      {
         "pattern": "^\\@foo"
      }
   ],
   "minLength": 5,
   "$schema": "http://json-schema.org/draft-07/schema#"
}

Caveats

• Case-insensitive transformations work perfectly for basic Latin as far as I can tell, except in the atypical case where a character range crosses 'character classes'. For example, /[Z-a]/ is a valid regex which would match "_" but not "A". However, "A" would be matched by /[Z-a]/i; with my additions, this package would unfortunately transform it to /[z-aZ-A]/, which would cause an error (range out of order) on parsing. If an invalid RegExp would be created, the package warns the user and falls back to the old regex.source.
• Extended Latin doesn't work: /é/i would not be transformed to /[éÉ]/, it would be left unchanged. I don't know how to do this without going mad, so I haven't.

---

Sorry for the delay; life suddenly got a little busy!

Resolves #116

[StefanTerdell]

Don't have time for a thorough review right now, but this PR just makes me all warm and fuzzy. Issue->Discussion->PR with a great description, tests included, and readme updated. Bravo!

[StefanTerdell]

Sorry the wait! Great work.

[Spappz]

Thank you! Glad to have helped.

Though, as a nit, I think you mistook me for another guy with an open PR in 23195b3 and 9275d61 😅

[StefanTerdell]

@Spappz I'm an idiot. Fixed though :) https://github.com/StefanTerdell/zod-to-json-schema/commit/3e7c648e626a73517bccb846869d15ef66531a00