fix(oracle): probe validators with their real lengths#107
Merged
Conversation
The oracle's `inferArb` fell back to `digs(10)` when a validator
omitted `lengths`, so the gate fed 10-digit strings to 8-digit (and
other) formats. 89/170 validators are affected; about ten of them
sit behind gate-tier oracle pairings and silently reported "0
disagreements / 0 valid" because every random sample failed the
length check before reaching the checksum.
Two fixes:
* `inferArb` now derives lengths from `examples` (in compact form)
when `lengths` is undeclared, so every validator with seed
examples gets coverage at its actual length.
* The gate-tier validators that were broken get explicit
`lengths`: dk.vat, fi.vat, hu.vat, lt.vat, lu.vat, mt.vat,
pt.vat, ro.cnp, se.vat, si.vat.
Proper coverage exposes real cross-implementation differences that
the broken arbs were hiding. python-stdnum (the canonical reference)
agrees with us on every newly-disagreeing sample, so the affected
jsvat/stdnum-js pairings move to survey tier with explanatory notes:
* jsvat:{dk,fi,gr,mt,pt}.vat and jsvat:it.iva — jsvat skips
no-leading-zero, checksum, or province-code rules required by
the official country sources.
* stdnum-js:nl.bsn — stdnum-js's `validatePerson("NL", ...)` also
matches Onderwijsnummer, returning valid for non-BSNs.
* stdnum-js:{lt.asmens,ro.cnp} — stdnum-js does not model the
21st/22nd-century gender digits the same way the official specs do.
Also fixes a real bug in ro.cnp uncovered along the way:
`VALID_COUNTIES` listed 49 and 50, which are not allocated county
codes per cnp-spec (40-48 cover Bucuresti sectors, 51-52 are
Calarasi/Giurgiu; 49 and 50 are gaps).
Gate result: 250,000 samples across 25 oracle pairings, 0
disagreements.
|
Warning You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again! |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 0ff81eb56d
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
Addressing Codex review on PR #107: the example-based length fallback under-probes validators whose `candidatePattern` spans a range wider than the example set. ro.vat is the canonical case (accepts 2-10 digits; only example is 8 digits), so the gate was only sampling `digs(8)` (plus 7/9 edge cases) and never exercising the 2-7 or 9-10 length paths. Declares explicit `lengths` on the variable-length validators that previously relied on the example fallback: * ro.vat — [2..10] * bg.vat — [9, 10] * cz.dic — [8, 9, 10] (cz.dic also has a CUSTOM_ARB; the lengths field is informational for non-oracle consumers) * gb.vat — [9, 12] (digit-only standard format; GD/HA letter prefixes are handled directly by `validate`) Also updates the `lengthsFromExamples` comment to flag that the fallback is only reliable for fixed-length validators. Gate result unchanged: 0 disagreements; ro.vat now probes the full digit-length range (895/10000 valid).
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 8305888af3
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
Addressing Codex follow-up on PR #107: gb.vat's `validate` also accepts letter-prefixed GD/HA government and health-authority forms (5 chars). Putting `lengths: [9, 12]` on the validator made the public metadata incomplete and changed `toRegex(gb.vat)` from the grouped form (`GB[\s\-./]?\d{3}[\s\-./]?\d{4}[\s\-./]?\d{2}`, which matches both "GB980780684" and "GB 980 7806 84") to a flat `GB[\s\-./]?\d{9,12}` that no longer matches the spaced form. Reverts `gb.vat.lengths` and adds a CUSTOM_ARB entry in oracle.ts so the oracle still probes the mod-97 digit path at lengths 9 and 12 without leaking the constraint into consumer-facing metadata.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
inferArbin the oracle script fell back todigs(10)whenever a validator omittedlengths, so ~10 gate-tier oracle pairings (dk.vat, fi.vat, hu.vat, lt.vat, lu.vat, mt.vat, pt.vat, ro.cnp, se.vat, si.vat, plus iban) silently reported0 disagreements / 0 valid— every random sample failed the length check before any checksum logic ran.inferArbnow derives lengths fromexamples(in compact form) whenlengthsis undeclared. Covers 89/170 validators automatically.lengthsto the gate-affected validators above so the contract is self-documenting.ro.cnpaccepted county codes 49 and 50 which aren't allocated per cnp-spec (40–48 cover Bucuresti sectors; 51–52 are Calarasi/Giurgiu).Test plan
bun test(4239 pass)bun run oracle(gate clean)