JuliusBrussee/caveman · error

cacheengine: invalid cache economics

Error message

cacheengine: invalid cache economics

What it means

Thrown by validatePlanRequest when profile.EconomicsKnown is true but WriteMultiplier or ReadMultiplier is not a finite, non-negative float (NaN, +Inf, -Inf, or < 0). These multipliers price cache writes vs reads; non-finite economics would poison the net-gain math in breakpointCandidates.

Source

Thrown at cacheengine/engine.go:298

	}
	profile := normalizedProfile(request.Profile)
	if !validIdentity(profile.ID, 256, false) || !validIdentity(profile.Provider, 64, true) || !validIdentity(profile.OptimizerID, 256, true) {
		return errors.New("cacheengine: invalid profile identity")
	}
	if profile.Mode != ModeUnsupported && profile.Mode != ModeImplicit && profile.Mode != ModeAffinity && profile.Mode != ModeExplicit {
		return fmt.Errorf("cacheengine: unknown mode %q", profile.Mode)
	}
	if profile.Mode != ModeUnsupported {
		if profile.MaxBreakpoints <= 0 || profile.MinPrefixTokens < 0 || profile.MaxRPMPerKey < 0 || profile.TTL < 0 {
			return errors.New("cacheengine: invalid cache thresholds")
		}
		switch profile.Attribution {
		case AttributionNone, AttributionOrganic, AttributionAffinity, AttributionCausal:
		default:
			return fmt.Errorf("cacheengine: unknown attribution %q", profile.Attribution)
		}
		if profile.EconomicsKnown && (!finiteNonNegative(profile.WriteMultiplier) || !finiteNonNegative(profile.ReadMultiplier)) {
			return errors.New("cacheengine: invalid cache economics")
		}
	}
	return nil
}

func normalizedProfile(profile Profile) Profile {
	if profile.Mode == ModeUnsupported {
		if profile.ID == "" {
			profile.ID = "unsupported"
		}
		return profile
	}
	if profile.MaxBreakpoints == 0 {
		profile.MaxBreakpoints = 1
	}
	if profile.MaxRPMPerKey == 0 {
		profile.MaxRPMPerKey = 15
	}

View on GitHub (pinned to 27d5a3981a)

Solutions

1. Sanitize multipliers with math.IsNaN/math.IsInf and a >= 0 check before setting EconomicsKnown
2. Fix the division/parse that produced NaN or Inf (guard zero denominators, reject 'NaN'/'Infinity' in config parsing)
3. If you have no reliable pricing data, leave EconomicsKnown false so the check is skipped

Example fix

// before
profile.WriteMultiplier = writeCost / float64(n) // NaN when n == 0
profile.EconomicsKnown = true

// after
if n > 0 {
    profile.WriteMultiplier = writeCost / float64(n)
    profile.ReadMultiplier = readCost / float64(n)
    profile.EconomicsKnown = !math.IsNaN(profile.WriteMultiplier) && !math.IsInf(profile.WriteMultiplier, 0) && profile.WriteMultiplier >= 0
}

func economicsOK(p cacheengine.Profile) bool {
    return !p.EconomicsKnown || (finiteNonNeg(p.WriteMultiplier) && finiteNonNeg(p.ReadMultiplier))
}
func finiteNonNeg(f float64) bool { return !math.IsNaN(f) && !math.IsInf(f, 0) && f >= 0 }

// n/a

When it happens

Trigger: Calling Plan with EconomicsKnown: true and a multiplier loaded from a computation that produced NaN/Inf (division by zero, parsing 'NaN' from JSON), or a negative multiplier from a sign error.

Common situations: Parsing multipliers from user/env config where 'NaN' or 'Infinity' is accepted by encoding/json into float64; computing price = cost/count with count == 0; upstream price API returning garbage that propagates into the profile.

Related errors

• cacheengine: cache economics overflow
• cacheengine: invalid partition key
• cacheengine: negative traffic expectation
• cacheengine: invalid profile identity
• cacheengine: invalid cache thresholds

AI-assisted analysis of JuliusBrussee/caveman@27d5a3981a (2026-08-15). Data as JSON: /api/errors/2d7798ad5b55a85a. Report an issue: GitHub.