Cross-library benchmarks

Apples-to-apples micro-benchmarks comparing lukaszraczylo/go-telegram against the five most-starred Go Telegram libraries:

github.com/go-telegram-bot-api/telegram-bot-api/v5
gopkg.in/telebot.v3 (tucnak)
github.com/go-telegram/bot
github.com/mymmrac/telego
github.com/NicoNex/echotron/v3

Lives in its own Go module so competitor dependencies don't leak into the main repo's go.mod.

Run

go test -count=10 -bench=. -benchmem | tee results/raw.txt
go install golang.org/x/perf/cmd/benchstat@latest    # one-time
benchstat results/raw.txt > results/benchstat.txt

Hot paths covered

File	Path
`webhook_bench_test.go`	Decode small text-message Update from JSON
`unmarshal_bench_test.go`	Decode large Update (entities + reply markup + photo array)
`call_bench_test.go`	`sendMessage` round-trip against `httptest.Server`
`dispatch_bench_test.go`	Route an Update through 20 registered handlers (worst-case match)

Fixtures

shared/fixtures.go defines the JSON payloads and the mock HTTP server. Every library decodes the same bytes; every round-trip hits the same canned response.

Latest results

See ../../docs/benchmarks/2026-05-10-comparison.md for the rendered comparison.