Statistical shape

Real data isn't uniform. Synth draws from distributions so your test data stresses the same paths production does.

Via tags:

type Txn struct {
    Amount   float64 `synth:"amount,dist=lognormal,mu=10,sigma=1"` // long tail
    Status   string  `synth:"enum,choices=settled|pending|failed,weights=0.94|0.05|0.01"`
    Category string  `synth:"enum,choices=a|b|c|d,dist=zipf,s=1.2"` // hot keys
}

Or in code:

synth.Make[Txn](1_000_000, synth.Weighted("Status", map[string]float64{
    "settled": 0.94, "pending": 0.05, "failed": 0.01,
}))

Distributions: normal, lognormal, exp (numeric fields) and zipf / explicit weights (enums).

Long tails, hot keys, and skew are what break partitioning and query planners — uniform fakers never surface those bugs.

Correlated fields

Numeric columns don't have to be independent. derive makes one a linear function of another in the same row, so a scatter plot has a shape instead of a cloud:

age:    { kind: int, min: 25, max: 65 }
income: { kind: float, derive: age, slope: 1200, intercept: 20000, noise: 0.1 }

Time series

kind: timeseries makes a column follow a curve over time — trend plus a seasonal cycle plus noise — for metrics and IoT data:

ts:  { kind: time, min: 2026-01-01T00:00:00Z, max: 2026-02-01T00:00:00Z }
cpu: { kind: timeseries, axis: ts, base: 40, trend: 0.5, amplitude: 20, period: 24h, noise: 3, min: 0, max: 100 }

Both are pure functions of the same row, so they generate in the same streaming, deterministic pass as everything else.

Learning the shape instead of declaring it

Profiling reads the distribution off a real export, and constraint mining learns the cross-column invariants a per-column distribution cannot see.