fix(utils): allow integer prices written with trailing decimal zeros in formatPrice

[Nexory]

Bug

formatPrice documents that "integer prices are always allowed regardless of significant figures" and bypasses the 5-significant-figure truncation when the input matches /^-?\d+$/. That check only runs on the raw input, so a price that is integer-valued but written with a trailing decimal (e.g. "100001.0") fails the regex, falls through to toFixedTruncate (which strips the .0, yielding "100001"), and is then truncated by the sig-fig step to "100000".

The same integer price produces different output depending on how it was formatted:

formatPrice("100001",    0) // "100001"   ok
formatPrice("100001.0",  0) // "100000"   wrong, off by 1
formatPrice("123456.0",  0) // "123450"   wrong
formatPrice("-123456.0", 0) // "-123450"  wrong
formatPrice("100001.00", 0) // "100000"   wrong

A float-derived price submitted as a string (e.g. (100001).toFixed(1) -> "100001.0", or any value that carries a trailing zero) therefore places an order at a different price than intended - a direct mispricing on high-value assets.

Fix

Re-check the integer-bypass condition after toFixedTruncate, so a value that has become an integer is treated the same as a raw integer input. A value that truncated all the way to zero is excluded, so the existing "price too small" RangeError still fires.

price = StringMath.toFixedTruncate(price, maxDecimals);

// Integer prices are always allowed, including values that became integers
// after decimal truncation (e.g. "100001.0" -> "100001").
if (/^-?\d+$/.test(price) && !/^-?0+$/.test(price)) {
  return formatDecimalString(price);
}

price = StringMath.toPrecisionTruncate(price, 5);

Tests

Added regression cases to tests/utils/format.test.ts:

assertEquals(formatPrice("100001.0", 0), "100001");
assertEquals(formatPrice("123456.0", 0), "123456");
assertEquals(formatPrice("100001.00", 0), "100001");
assertEquals(formatPrice("-123456.0", 0), "-123456");
assertEquals(formatPrice("100001.0", 0, "spot"), "100001");

Verification (local)

• The new cases fail on main (return "100000" / "123450") and pass with this change.
• The full formatPrice + formatSize suite stays green (46 steps, including the reference-asset validation against PURR/DYDX/SOL/BNB/ETH/BTC and the spot pairs).
• The "price too small -> RangeError" edge case still throws (formatPrice("0.0000001", 0)), and non-integer truncation is unchanged (formatPrice("100000.5", 0) -> "100000", formatPrice("12345.6", 0) -> "12345").
• deno fmt --check and deno lint are clean.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.