Add normalized hash-based test IDs to draft2020-12/enum.json (POC for #698)

[AnirudhJindal]

Context

This is a small proof-of-concept for #698, implementing option 3 from our discussion:

an MD5 hash of the normalized schema + test data + valid flag.

The goal is:

• Same test across dialects → same ID
• IDs change automatically when the semantics of a test change
• Fully checkable in CI by recomputing the ID and comparing

normalize.js

• Imported and adapted the AST-builder gist from @jdesrosiers.
• Added a handler for the dialect-specific keyword draft-2020-12/dynamicRef.
• Stripped top-level file: $id during normalization so registerSchema() doesn’t reject test schemas.
• Exposed normalize() to return a stable, dialect-aware, $ref-resolved schema representation.

load-remotes.js

• Recursively loaded all remotes/ JSON files and registered them under http://localhost:1234/....
• Added deduplication so the same remote URL is never registered twice.
• Ensured remotes only load for matching dialects.

add-test-ids.js

• For each test missing an id, normalized its schema.
• Generated an ID using md5(normalizedSchema + data + valid).
• Wrote the updated tests back to disk (in this PR: only enum.json).

check-test-ids.js

• Re-normalized schema and recomputed IDs for each test.
• Reported missing, mismatched, or duplicate IDs.
• Exited with error when inconsistencies were found (CI-ready behavior).

enum.json

• Added id fields to each test using the MD5-based formula.
• No other changes to schema structure or test contents.

How to run

# Add IDs to enum.json
node scripts/add-test-ids.js tests/draft2020-12/enum.json

# Verify IDs within draft2020-12
node scripts/check-test-ids.js tests/draft2020-12

For now, only enum.json has IDs. The checker will still report "Missing ID" for other files in draft2020-12

Questions for reviewers

If this direction looks good, I can:

Extend the tooling as needed based on feedback

Add a small GitHub Action that runs check-test-ids.js on the touched version

Then start updating more files, one small PR at a time.

[jdesrosiers]

This should have been opened as draft PR. I've converted it this time. Next time you open a PR that's a work in progress for the purpose of getting feedback, a draft PR is the way to go.

[AnirudhJindal]

Hey @jdesrosiers, Really sorry for the radio silence! I'm currently drowning in university exams and intern interview prep, so it's been a bit chaotic over here. I'll need a couple days to review and fix up the mistakes. I skimmed through and saw your notes about defensive programming I'll polish it up and circle back soon. Thanks!

[AnirudhJindal]

Hey @jdesrosiers, I’ve updated add-test-ids.js to incorporate the defensive programming points you mentioned. Some of the earlier checks came from initially making it work for a single file, so I’ve now refined and cleaned those parts up.

I’d really appreciate any further feedback when you get a chance.