[hugomallet]

Observable defects

1. formatErrors: the bracket-to-dot .replace(/\[([0-9]+)\]/, '.$1') lacked the g flag, so only the first array index was converted — an AJV error on items[0].tags[1] produced field = "items.0.tags[1]", and <FieldError name="items.0.tags.1"> never matched.
2. filterByFieldNameWithWildcard: the wildcard prefix went raw into new RegExp — the dot in user.* matched any character, so userX was wrongly matched. A prefix with an unclosed bracket (e.g. items[0*) even threw SyntaxError: Unterminated character class.

Fix

• Add the g flag to the index-conversion replace.
• Escape regex metacharacters in the wildcard prefix before building the regex (also fixes the SyntaxError crash).

Tests

• New test: items[0].tags[1] formats to items.0.tags.1 (fails on master with items.0.tags[1]).
• New test: user.* matches user.email but not userX (fails on master with userX included).
• New test: items[0* filters without crashing (fails on master with SyntaxError: Unterminated character class).
• CI=true yarn test:runtime: 6 suites, 66 tests (63 before + 3 new), 5 snapshots — all green.

Heads-up for consumers

If you worked around the old bug by writing bracket-notation paths in <FieldError name> (e.g. name="items.0.tags[1]"), those names will no longer match: error paths are now fully dot-normalized (items.0.tags.1).

[hugomallet]

Tooled self-review (review53 skill) — PR authored by this account, so GitHub refuses an APPROVE; posting as COMMENT. Verdict: mergeable.

Verified

• formatErrors g flag: correct. Multi-digit indices (a[12].b[345] → a.12.b.345) and mixed paths convert properly; the [0-9]+ class leaves AJV's quoted-key notation (['weird-key']) untouched.
• Wildcard prefix escaping: the escape set is the standard MDN one and handles [, (, +, \ correctly (checked by direct evaluation). Bonus not stated in the PR description: the old code threw SyntaxError: Unterminated character class for a prefix containing an unbalanced [ (e.g. items[0.*) — the escape fixes that latent crash too.
• Both call sites in src/lib/Form/Form.js (errors + touchedFields filtering) are consistent with the new behavior; scope is strictly helpers.js + helpers.test.js as announced.
• New tests bite (fail on master), full suite green: 6 suites, 65 tests. Lint: the 6 jest/no-test-callback errors in helpers.test.js are pre-existing (untouched getFieldValue blocks), not introduced here.

Minor (grouped, non-blocking)

1. The new wildcard test only exercises .; one extra case with a [ prefix (e.g. items[0].*) would lock in the SyntaxError fix above, which is currently untested.
2. Consumers who worked around bug 1 by writing bracket paths in <FieldError name> (e.g. items.0.tags[1]) will see those stop matching — intentional, but worth a line in the changelog; same for the SyntaxError fix.