Benchmarks

Measured on 2026-11-02 at commit 194ec36, with Go 1.27.3 on an Intel Core i5-8257U at 1.40 GHz under macOS. Each figure is the median of three runs, and runs of the same benchmark differed by up to 7%. Treat differences smaller than that as ties.

The small struct has four fields and nine rules. The complex one adds a nested struct and a list of addresses, for 18 rules.

Correctly typed JSON, parsed and validated

	gopantic	encoding/json + validator
Small struct	2,217 ns, 10 allocs	2,501 ns, 10 allocs
Complex struct	6,365 ns, 29 allocs	5,006 ns, 19 allocs

These are close enough that speed should not decide between them. gopantic allocates more memory on both: 453 bytes against 186 on the small struct, and 1,531 against 520 on the complex one.

What the extras cost

These use the small struct, against 2,217 ns for correctly typed JSON.

Input	Time	Cost
Numbers sent as strings	7,099 ns	3.2x
YAML	17,381 ns	7.8x (against 2,215 ns for JSON through the same call)

Coercion is the expensive path because the whole input is decoded a second time by the walker. One string in a number field costs the same as several. YAML always takes that path.

Validation alone

Validate on a struct already in memory, against validator's Struct.

	gopantic	validator
Small struct	1,263 ns	1,575 ns
Complex struct	2,891 ns	2,152 ns

Caching

For the same input, a parse took 2,918 ns and a cache hit 686 ns, about 4x less. A hit on this struct allocates once. A struct holding slices or maps pays to copy them on every hit, so its gain is smaller. See the caching guide.

Hostile input

From tests/adversarial_bench_test.go. Each input needs coercion, so each goes through the walker.

Input	Time
20,000 array items	41 ms
50,000 unknown keys	25 ms
Nesting 30 levels deep	72 µs

Running them

go test -run='^$' -bench=. -benchmem ./tests/
cd benchmarks && go test -run='^$' -bench=. -benchmem ./...

The comparisons against validator live in benchmarks/, a separate module, so that validator stays out of your dependency graph.