F-06 (Medium): Guard the bigint↔number bridge at the storage boundary#6
Merged
Conversation
SqlStorage cannot bind bigint, so every amount crosses an IEEE 754 number boundary on write, read, and SUM aggregation. Number() and BigInt() do not error on precision loss — above 2^53 minor units they silently round, which for a ledger means corrupted money with no signal anywhere. The ceiling (~$90T at scale 2) is documented but was unenforced, and the README's advice to raise SCALE for high-precision currencies shrinks it to ~90M major units at scale 8 — well inside plausible balances. - toStorageInt(minor): throws RepositoryError beyond 2^53 - 1 instead of silently rounding; applied to the write path, precomputed before the insert transaction opens - fromStorageInt(value, context): rejects non-safe integers (silent SUM overflow) and non-integers (float contamination via non-library SQL) on every read path — row hydration and both SUM aggregates Audit finding F-06 (Medium). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
![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: 28860c7fe7
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".
The PR states toStorageInt/fromStorageInt are exported for third-party Repository ports, but only the internal module exported them — the package exports map exposes only src/index.ts, so consumers could not import them. Addresses PR #6 review (P2). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
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 join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Finding
F-06 (Medium) from the ledger audit: domain math is exact
bigint, but storage is not — the write path boundNumber(amount.minor), and every read (includingSUM()aggregates) returned a JSnumberre-wrapped viaBigInt(...). Neither conversion errors on precision loss:Number(9007199254740993n)yields9007199254740992(verified in the audit, probe P6c). Above 2^53 minor units, amounts and aggregates silently corrupt.The
2^53 ceiling ($90T at scale 2) was documented in a comment but enforced nowhere — a policy, not a control. It matters because the README explicitly recommends raisingSCALEfor higher-precision currencies (8 for crypto), which collapses the ceiling to ~90 million major units — well inside plausible balances. At that pointbalanceByTypeandtrialBalancewould drift by ±1 minor unit with no error anywhere.Fix
Exactness is now enforced, not assumed, at the one seam where it can be lost:
toStorageInt(minor)(write path): throwsRepositoryErrorbeyondNumber.MAX_SAFE_INTEGERinstead of silently rounding. Conversions are precomputed before the insert transaction opens, so an out-of-range amount fails cleanly with nothing written.fromStorageInt(value, context)(read paths — row hydration, per-account and per-typeSUMs): rejects unsafe integers (silent aggregate overflow) and non-integer values (float contamination reaching the column via non-library SQL — complements the DB-leveltypeof(amount) = 'integer'trigger in F-04 (High) + F-14: Append-only triggers and row-validity checks #4).Both helpers are exported for third-party
Repositoryports, which face the identical driver problem.Testing
New
storage-int.spec.ts(written first, watched fail): exact pass-through up toMAX_SAFE_INTEGER, loud failure at2^53and beyond in both directions, and float/NaN rejection on the read path.Full suite: 83 passed.
tsc --noEmitandbiome checkclean.🤖 Generated with Claude Code