fix: normalize extreme exponents in non-scientific decimal formatter

[JuaniRios]

Problem

The non-scientific path in LibFormatDecimalFloat.toDecimalString reverts with UnformatableExponent when the exponent is outside [-76, 76], because 10 ** uint256(abs(exponent)) overflows uint256 at that range.

However, the library's own arithmetic operations (add, sub) can produce Floats with exponents outside this range through normal operations. Specifically, catastrophic cancellation when subtracting nearly-equal maximized values:

1. add/sub call maximizeFull on both inputs, pushing coefficients to ~76 digits with very negative exponents (e.g., -76)
2. Subtraction of nearly-equal maximized values produces a small coefficient at the same large negative exponent
3. packLossy stores this as-is (small coefficient fits in int224)
4. The resulting Float is valid for all operations except non-scientific formatting

Discovery

This was found in production: a market-making bot (st0x.liquidity) accumulated 605 position events via Float add/sub. The net position ended up with exponent -77, which crashed the application during serialization with UnformatableExponent.

Fix

Instead of reverting when the exponent is outside [-76, 76], truncate trailing coefficient digits by dividing by 10^(abs(exponent) - 76) and clamping the exponent to the boundary. This loses insignificant trailing precision (at the 10^-77+ scale) but produces a valid decimal string for any Float the arithmetic can create.

Testing

• Updated existing tests that expected UnformatableExponent reverts — they now verify successful formatting with truncation
• Added test for the exact case from production: coefficient=9999999910959448, exponent=-77
• Added test for exponent=-78 (two digits of truncation)
• All 395 non-fork tests pass (8 pre-existing fork test failures from missing CI env vars)
• Fuzz tests (testFormatDecimalRoundTripNonNegative, testFormatDecimalRoundTripNegative) pass with 5096 runs each

[coderabbitai]

Important

Review skipped

Draft detected.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Pro

Run ID: b10468c9-65a4-4366-ab81-65e75de17bc4

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/normalize-extreme-exponents-in-non-scientific-formatter

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[JuaniRios]

• fix: normalize extreme exponents in non-scientific decimal formatter #180 👈 (View in Graphite)
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.