Two changes that together cut allocs/call from 15 to 13 (client-internal
bench) and per-call CPU from 600ns to 455ns (-24%) on the no-HTTP path:
1. Codec gets an optional BodyEncoder extension (MarshalTo io.Writer).
When present, encodeJSONBody stream-encodes the request directly into
a pooled *bytes.Buffer instead of allocating a [2-step] Marshal+Reader
pair. DefaultCodec implements it via goccy/go-json.NewEncoder.
2. *Bot caches the parsed base URL on construction. buildRequest skips
net/http.NewRequestWithContext for the common case and constructs
*http.Request manually — clones the URL by value, sets the method
path, and populates ContentLength + GetBody from the body's concrete
type so RetryDoer's body-replay across attempts still works.
Cross-library bench (sendMessage round-trip vs httptest.Server): -2
allocs/call (104 -> 102), bytes -1.2%, time within noise (real HTTP
plumbing dominates). The CPU win is visible on synthetic stub-doer
benches and translates to lower GC pressure on sustained-throughput
workloads.
Slow-path fallback preserved for codecs that don't implement BodyEncoder
and for *Bot instances where url.Parse on the configured base failed —
they take the original NewRequestWithContext path.
Adds test/benchmarks/ as a separate Go module so competitor deps
(go-telegram-bot-api/v5, telebot.v3, go-telegram/bot, telego,
echotron/v3) stay out of the root go.mod.
Hot paths covered:
- Webhook decode (small Update -> typed Update struct)
- Large unmarshal (Update with entities + reply markup + photo array)
- API round-trip (sendMessage against httptest.Server)
- Dispatch route (20 handlers, last-registered matches)
Results on Apple M4 Max / go1.26.2: ours wins 3 of 4 paths and is
2nd of 5 in the round-trip path. Full report at
docs/benchmarks/2026-05-10-comparison.md, raw output committed under
test/benchmarks/results/.
Caveats called out in the report:
- codec asymmetry (we ship goccy/go-json; competitors mostly stdlib)
- echotron call bench skipped — built-in rate limiter not externally
configurable; would measure throttling, not the library
- dispatch bench limited to libs with a public sync entry point
(ours, telebot, gobot); gotba has no dispatcher, telego/echotron
use channel/per-chat paradigms not directly comparable
Also gitignores docs/superpowers/ (local brainstorm/spec scratch)
and regenerates docs/reference/dispatch.md after the new
Router.Process method.
lukaszraczylo