[NspxMiguel]

Closes #35

Adds strictPrefix?: boolean to createId and createSortableId. When true, the prefix must match /^[A-Za-z0-9]{1,63}$/, otherwise a TypeError names the first offending character, e.g. prefid: prefix "us er" contains an invalid character " " at index 2.

• Check lives in assertValidPrefix (src/internal/prefix.ts)
• Tests: space, punctuation, non-ASCII, 63 chars allowed, 64 rejected, default mode unchanged; both generators
• README section and CHANGELOG entry under Unreleased
• npm run verify passes; default behavior is unchanged

[suhailopensource]

Thanks @NspxMiguel, this is a really clean PR. It's small, focused, and follows the issue closely.

What I checked:

• npm run verify passes (128 tests, coverage gate green), and the branch is up to date with main.
• The error text matches the issue exactly: prefid: prefix "us er" contains an invalid character " " at index 2.
• Emoji are handled well. "ab😀" reports "😀" at index 2, not half a character. Nice use of codePointAt.
• Default mode really is unchanged. I generated the same seeded IDs on main and on this branch and got byte-identical output.
• The option is on both IdOptions and SortableIdOptions in the built types. It also works on Node 14 to 24, Bun and Deno.
• Your tests caught the mutations I tried: 63 changed to 64, the strict check removed, and the default flipped to strict.

A few small changes before I merge:

1. Make strict a required parameter in src/internal/prefix.ts. Use

   strict: boolean,

   instead of strict = false,. Both callers already pass it, so nothing else changes. The reason: the two open batch PRs (feat: add batch ID generation #44, feat: add many() for batch ID generation #45) add a .many() method that calls assertValidPrefix(prefix, separator) with only two arguments. With the default value that still compiles, and createId({ strictPrefix: true }).many("us er", 3) would quietly return IDs instead of throwing. With a required parameter, TypeScript flags those calls straight away.

2. Add a test for an emoji prefix. The code already handles this correctly, but no test covers it. If I swap the String.fromCodePoint(prefix.codePointAt(...)) line for prefix[bad.index], every test still passes, and the message would print half an emoji. This test in test/generate.test.ts, right after the non-ASCII test, locks it in:

   it("names a whole emoji, not half of it", () => {
     expect(() => strict("ab😀")).toThrow(/invalid character "😀" at index 2/);
   });

3. README: show the real error text. The example comment leaves out the prefid: part, so people searching for the message won't find it. Suggested line:

   id("us er"); // throws TypeError: prefid: prefix "us er" contains an invalid character " " at index 2.

Optional, happy either way: the 64-character error says prefid: prefix must be at most 63 characters in strict mode. It would help a little more if it said how long the prefix was, for example prefid: prefix is 64 characters long; strict mode allows at most 63. The 64-character test could then check that exact message instead of only TypeError.

Thanks again, this is very close!

[NspxMiguel]

Thanks for the careful review. Pushed: strict is now a required parameter of assertValidPrefix; the emoji test (ab😀); the README line shows the full TypeError: prefid: … text; and the 64-character error now says prefid: prefix is 64 characters long; strict mode allows at most 63., with the test checking that exact message. Rebased on current main. npm run verify passes (129 tests).

For #45 I'll pass strictPrefix into the many() prefix checks and add a test once this one lands.