Fuzz#

The fuzz command evaluates an expression thousands of times and reports output statistics: type breakdown, null rate, min/max/mean/median/stddev, a distribution histogram, and unique value count. Use it to catch distribution surprises before seeding millions of rows.

edg fuzz "norm(50, 10, 0, 100)"

Expression: norm(50, 10, 0, 100)
Iterations: 10000

Type          Count     Pct
float64       10000  100.0%

Nulls:  0 (0.0%)
Panics: 0
Errors: 0
Unique: 71

Min:    14
Max:    86
Mean:   50.01
Median: 50
Stddev: 10.02
p50:    50
p90:    63
p95:    67
p99:    73

Distribution:
        14 -     21.2   25 (0.2%)
      21.2 -     28.4  █ 134 (1.3%)
      28.4 -     35.6  ███████ 586 (5.9%)
      35.6 -     42.8  ████████████████████ 1539 (15.4%)
      42.8 -       50  ██████████████████████████████████ 2510 (25.1%)
        50 -     57.2  ████████████████████████████████████████ 2937 (29.4%)
      57.2 -     64.4  ████████████████████ 1530 (15.3%)
      64.4 -     71.6  ███████ 584 (5.8%)
      71.6 -     78.8  █ 131 (1.3%)
      78.8 -       86   24 (0.2%)

Flags#

Loading Config#

When you need access to globals, references, or user-defined expressions, pass a config file:

edg fuzz "norm(warehouses * 5, warehouses, 0, warehouses * 10)" --config workload.edg -n 5000

Null Rate Checking#

Use fuzz to verify that nullable() produces the expected null rate:

edg fuzz "nullable(0.1, gen('name'))" -n 10000

The output will show Nulls: ~1000 (10.0%) confirming the 10% null rate.

Distribution Comparison#

Compare distribution shapes by fuzzing different expressions:

# Uniform - flat histogram
edg fuzz "uniform(1, 1000)"

# Normal - bell curve centered at 50
edg fuzz "norm(50, 10, 0, 100)"

# Normal float with 2 decimal places
edg fuzz "norm_f(50, 10, 0, 100, 2)"

# N unique normal values
edg fuzz "norm_n(50, 10, 0, 100, 3, 5)"

# Zipfian - heavy left skew (hot keys)
edg fuzz "zipf(2.0, 1.0, 1000)"

# Exponential - recent bias
edg fuzz "exp(0.5, 0, 100)"

# Exponential float with 2 decimal places
edg fuzz "exp_f(0.5, 0, 100, 2)"

# Log-normal
edg fuzz "lognorm(1, 0.5, 0, 100)"

# Log-normal float with 2 decimal places
edg fuzz "lognorm_f(1, 0.5, 0, 100, 2)"

# Pareto - power-law, values near 0
edg fuzz "pareto(2.0, 1000)"