What causes "expected float, but got '%s'" in nats-io/nats-server?

Correct the value to valid Go float syntax: digits with optional '.', sign, exponent (e.g. 3.14, 1e5).

expected float, but got '%s'

Error message

expected float, but got '%s'

What it means

An item classified as itemFloat could not be parsed by strconv.ParseFloat at all — the error was not ErrRange, so the literal is not a valid float syntax. The parser reports the offending text verbatim.

Source

Thrown at conf/parse.go:356

		case "tb", "ti", "tib":
			setValue(it, num*1024*1024*1024*1024)
		case "p":
			setValue(it, num*1000*1000*1000*1000*1000)
		case "pb", "pi", "pib":
			setValue(it, num*1024*1024*1024*1024*1024)
		case "e":
			setValue(it, num*1000*1000*1000*1000*1000*1000)
		case "eb", "ei", "eib":
			setValue(it, num*1024*1024*1024*1024*1024*1024)
		}
	case itemFloat:
		num, err := strconv.ParseFloat(it.val, 64)
		if err != nil {
			if e, ok := err.(*strconv.NumError); ok &&
				e.Err == strconv.ErrRange {
				return fmt.Errorf("float '%s' is out of the range", it.val)
			}
			return fmt.Errorf("expected float, but got '%s'", it.val)
		}
		setValue(it, num)
	case itemBool:
		switch strings.ToLower(it.val) {
		case "true", "yes", "on":
			setValue(it, true)
		case "false", "no", "off":
			setValue(it, false)
		default:
			return fmt.Errorf("expected boolean value, but got '%s'", it.val)
		}

	case itemDatetime:
		dt, err := time.Parse("2006-01-02T15:04:05Z", it.val)
		if err != nil {
			return fmt.Errorf(
				"expected Zulu formatted DateTime, but got '%s'", it.val)
		}

View on GitHub (pinned to 3a66a489d2)

Solutions

Correct the value to valid Go float syntax: digits with optional '.', sign, exponent (e.g. 3.14, 1e5).
Replace decimal commas with dots.
Remove units/symbols from the value; express units via the key name or a separate field.
Quote the value as a string and parse it in application code if non-standard syntax is required.

Example fix

// before
threshold = 3,14%
// after
threshold = 3.14

Defensive patterns

Strategy: validation

Validate before calling

if _, err := strconv.ParseFloat(raw, 64); err != nil {
	return fmt.Errorf("%q is not a valid float: %w", raw, err)
}

Prevention

Use dots, not commas, for decimal separators.
Keep units and symbols out of numeric config values.
Validate config values with strconv.ParseFloat in CI before deployment.

When it happens

Trigger: processItem (from parse) reaches case itemFloat with it.val like '1,5', '1_000', '0x1f' in an unquoted context, '3.14.15', or empty text lexed as a float.

Common situations: Locale-formatted decimal commas, thousand separators, trailing units ('100ms', '5%'), typo'd strings where a number was expected, or a lexer misclassification.

Related errors

AI-assisted analysis of nats-io/nats-server@3a66a489d2 (2026-09-02). Data as JSON: /api/errors/6e004beaba9732ce. Report an issue: GitHub.