feat: add benchmarks, justfile, optimize throughput, harden quality gates

Benchmarking & comparison:
- Add criterion benchmarks (ucobs vs cobs 0.3 vs corncobs 0.1)
- Add Python cobs benchmark script
- Add performance comparison section to README with full results

Performance:
- Rewrite encode to scan-and-bulk-copy via copy_from_slice
- Rewrite decode to bulk-copy via copy_from_slice
- ~2-4x throughput improvement; now matches corncobs on most workloads

Quality hardening:
- Add #![forbid(unsafe_code)], #![warn(missing_docs)], #[must_use]
- Add [lints.rust] unsafe_code = "forbid" in Cargo.toml
- Add rust-version (MSRV) = "1.56"
- Add docs.rs metadata
- 106 tests: 80 unit/corpus + 25 property tests (25k cases each = 625k total)
- External test vectors from: Wikipedia, Cheshire & Baker paper, nanocobs,
  cobs-c, cobs2-rs, Jacques Fortier, Python cobs package
- Add fuzz_small_dest fuzz target (encode/decode with undersized buffers)
- Improve doc examples to show sentinel append/strip convention

Infrastructure:
- Add justfile with test (nextest), check, bench, fuzz, miri, fmt targets
- cargo-nextest for parallel test execution
- Update .gitignore for Python cache files

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

.gitignore

@@ -3,3 +3,5 @@
 /fuzz/corpus
 /fuzz/artifacts
 Cargo.lock
+__pycache__/
+*.pyc

CLAUDE.md

@@ -12,7 +12,7 @@
 2. **Interoperable** — proven byte-for-byte compatibility with `corncobs`, `cobs` crates and Python `cobs` packages.
 3. **Insanely tested** — canonical vectors, property-based tests (proptest), fuzz targets (cargo-fuzz), cross-crate interop, randomized payloads. If you think testing is done, you're 10% done.
 4. **Fast** — benchmark against every competing implementation. Must be at least as fast, ideally fastest. Benchmarks are CI-enforced.
-5. **Tiny** — `no_std`, zero-alloc, zero runtime dependencies. ~140 lines of implementation. Runs on 8-bit MCUs and servers alike.
+5. **Tiny** — `no_std`, zero-alloc, zero runtime dependencies. ~120 lines of implementation. Runs on 8-bit MCUs and servers alike.
 6. **Trustworthy** — small enough to audit by hand. No unsafe. No panics in release. Fuzz-hardened.
 
 ## Architecture

Cargo.toml

@@ -2,6 +2,7 @@
 name = "ucobs"
 version = "0.1.0"
 edition = "2021"
+rust-version = "1.56"
 license = "MIT"
 description = "μCOBS — Consistent Overhead Byte Stuffing. no_std, zero-alloc, fastest, most-tested COBS implementation."
 authors = ["Stephen Waits <steve@waits.net>"]
@@ -15,7 +16,19 @@ categories = ["encoding", "no-std::no-alloc", "embedded"]
 [features]
 default = []
 
+[lints.rust]
+unsafe_code = "forbid"
+
+[package.metadata.docs.rs]
+all-features = true
+
 [dev-dependencies]
+cobs = "0.3"
 corncobs = "0.1"
+criterion = { version = "0.5", features = ["html_reports"] }
 proptest = "1"
 rand = "0.8"
+
+[[bench]]
+name = "bench"
+harness = false

README.md

@@ -20,6 +20,70 @@ allowing `0x00` to serve as an unambiguous frame delimiter. Overhead is exactly
 - **Insanely tested** — canonical vectors, property-based tests, fuzz targets,
   cross-crate interop, randomized payloads, and more coming
 
+## Performance
+
+Benchmarked against the two other `no_std` COBS crates using
+[Criterion](https://github.com/bheisler/criterion.rs) across three payload
+patterns: **all zeros** (worst case — every byte triggers a code emit),
+**no zeros**, and **mixed** (0x00–0xFF cycling). Throughput in MB/s, higher
+is better.
+
+### Encode
+
+| Payload | Pattern | μCOBS | `cobs` 0.3 | `corncobs` 0.1 |
+|---------|---------|------:|-----------:|---------------:|
+| 64 B    | zeros   |   595 |        748 |            180 |
+| 64 B    | mixed   | 1,853 |      1,011 |          1,734 |
+| 64 B    | nonzero | 1,884 |      1,012 |          2,110 |
+| 256 B   | zeros   |   654 |        817 |            184 |
+| 256 B   | mixed   | 1,888 |      1,002 |          1,887 |
+| 256 B   | nonzero | 1,964 |      1,004 |          1,968 |
+| 4096 B  | zeros   |   675 |        845 |            188 |
+| 4096 B  | mixed   | 1,734 |      1,016 |          1,746 |
+| 4096 B  | nonzero | 1,775 |      1,020 |          2,245 |
+
+### Decode
+
+| Payload | Pattern | μCOBS | `cobs` 0.3 | `corncobs` 0.1 |
+|---------|---------|------:|-----------:|---------------:|
+| 64 B    | zeros   |   595 |        440 |            732 |
+| 64 B    | mixed   | 7,041 |        504 |          7,991 |
+| 64 B    | nonzero | 9,900 |        502 |          9,807 |
+| 256 B   | zeros   |   656 |        472 |            742 |
+| 256 B   | mixed   |17,024 |        645 |         16,207 |
+| 256 B   | nonzero |18,456 |        647 |         19,549 |
+| 4096 B  | zeros   |   675 |        483 |            768 |
+| 4096 B  | mixed   |17,522 |        656 |         22,959 |
+| 4096 B  | nonzero |37,987 |        623 |         40,005 |
+
+### Code size
+
+Measured from `.text` section of release-optimized symbols (`encode` + `decode`
+combined):
+
+| Crate          | Code size |
+|----------------|----------:|
+| μCOBS          |   541 B   |
+| `cobs` 0.3     |   537 B   |
+| `corncobs` 0.1 |   637 B   |
+
+All three crates are tiny. The difference is negligible for any target.
+
+### Crate properties
+
+| Property       | μCOBS  | `cobs` 0.3       | `corncobs` 0.1 |
+|----------------|:------:|:-----------------:|:--------------:|
+| `no_std`       | always | opt-in[^1]        | always         |
+| Zero-alloc     | yes    | opt-in[^2]        | yes            |
+| `unsafe`-free  | yes    | yes               | yes            |
+| Implementation | ~120 LOC | ~790 LOC        | ~645 LOC       |
+
+[^1]: Requires disabling the default `std` feature.
+[^2]: `alloc` is enabled by default via the `std` feature.
+
+> Measured on Intel N100, Rust 1.93.1, Linux 6.12. Run `cargo bench` to
+> reproduce. Full Criterion reports are generated in `target/criterion/report/`.
+
 ## Usage
 
 ```rust

bench_python.py

@@ -0,0 +1,67 @@
+#!/usr/bin/env python3
+"""Benchmark Python cobs encode/decode vs ucobs (reference timing only).
+
+Requires: pip install cobs
+"""
+
+import time
+
+
+def payload_zeros(n: int) -> bytes:
+    return b"\x00" * n
+
+
+def payload_nonzero(n: int) -> bytes:
+    return bytes((i % 255 + 1) for i in range(n))
+
+
+def payload_mixed(n: int) -> bytes:
+    return bytes(i % 256 for i in range(n))
+
+
+def bench(func, data, iterations=10_000):
+    start = time.perf_counter()
+    for _ in range(iterations):
+        func(data)
+    elapsed = time.perf_counter() - start
+    return elapsed
+
+
+def main():
+    try:
+        from cobs import cobs as cobs_mod
+    except ImportError:
+        print("ERROR: python cobs package not installed. Run: pip install cobs")
+        return
+
+    sizes = [0, 1, 10, 64, 254, 255, 256, 1024, 4096]
+    payloads = [
+        ("zeros", payload_zeros),
+        ("nonzero", payload_nonzero),
+        ("mixed", payload_mixed),
+    ]
+    iterations = 10_000
+
+    print(f"{'operation':<30} {'size':>6} {'iters':>7} {'total_ms':>10} {'per_op_us':>10} {'MB/s':>10}")
+    print("-" * 80)
+
+    for name, make in payloads:
+        for size in sizes:
+            data = make(size)
+
+            # Encode
+            elapsed = bench(cobs_mod.encode, data, iterations)
+            per_op_us = (elapsed / iterations) * 1_000_000
+            mbps = (size * iterations / elapsed / 1_000_000) if elapsed > 0 and size > 0 else 0
+            print(f"{'python-cobs encode/' + name:<30} {size:>6} {iterations:>7} {elapsed*1000:>10.2f} {per_op_us:>10.3f} {mbps:>10.1f}")
+
+            # Decode
+            encoded = cobs_mod.encode(data)
+            elapsed = bench(cobs_mod.decode, encoded, iterations)
+            per_op_us = (elapsed / iterations) * 1_000_000
+            mbps = (size * iterations / elapsed / 1_000_000) if elapsed > 0 and size > 0 else 0
+            print(f"{'python-cobs decode/' + name:<30} {size:>6} {iterations:>7} {elapsed*1000:>10.2f} {per_op_us:>10.3f} {mbps:>10.1f}")
+
+
+if __name__ == "__main__":
+    main()

benches/bench.rs

@@ -0,0 +1,149 @@
+use criterion::{black_box, criterion_group, criterion_main, BenchmarkId, Criterion, Throughput};
+
+// ── Test payloads ───────────────────────────────────────────────────
+
+fn payload_zeros(len: usize) -> Vec<u8> {
+    vec![0u8; len]
+}
+
+fn payload_nonzero(len: usize) -> Vec<u8> {
+    (0..len).map(|i| (i % 255 + 1) as u8).collect()
+}
+
+fn payload_mixed(len: usize) -> Vec<u8> {
+    (0..len).map(|i| (i % 256) as u8).collect()
+}
+
+// ── ucobs ───────────────────────────────────────────────────────────
+
+fn ucobs_encode(data: &[u8]) -> Vec<u8> {
+    let mut buf = vec![0u8; ucobs::max_encoded_len(data.len()) + 1];
+    let n = ucobs::encode(data, &mut buf).unwrap();
+    buf.truncate(n);
+    buf
+}
+
+fn ucobs_decode(encoded: &[u8], out: &mut [u8]) -> usize {
+    ucobs::decode(encoded, out).unwrap()
+}
+
+// ── corncobs ────────────────────────────────────────────────────────
+
+fn corncobs_encode(data: &[u8]) -> Vec<u8> {
+    let mut buf = vec![0u8; corncobs::max_encoded_len(data.len())];
+    let n = corncobs::encode_buf(data, &mut buf);
+    buf.truncate(n);
+    buf
+}
+
+fn corncobs_decode(encoded: &[u8], out: &mut [u8]) -> usize {
+    // corncobs expects trailing sentinel
+    corncobs::decode_buf(encoded, out).unwrap()
+}
+
+// ── cobs crate ──────────────────────────────────────────────────────
+
+fn cobs_encode(data: &[u8]) -> Vec<u8> {
+    let mut buf = vec![0u8; cobs::max_encoding_length(data.len())];
+    let n = cobs::encode(data, &mut buf);
+    buf.truncate(n);
+    buf
+}
+
+fn cobs_decode(encoded: &[u8], out: &mut [u8]) -> usize {
+    // unwrap_or(0): the cobs crate returns Err(EmptyFrame) for empty input,
+    // which is valid COBS. We treat it as 0 decoded bytes for benchmarking.
+    cobs::decode(encoded, out).unwrap_or(0)
+}
+
+// ── Benchmarks ──────────────────────────────────────────────────────
+
+const SIZES: &[usize] = &[0, 1, 10, 64, 254, 255, 256, 1024, 4096];
+
+fn bench_encode(c: &mut Criterion) {
+    let payloads: Vec<(&str, Box<dyn Fn(usize) -> Vec<u8>>)> = vec![
+        ("zeros", Box::new(payload_zeros)),
+        ("nonzero", Box::new(payload_nonzero)),
+        ("mixed", Box::new(payload_mixed)),
+    ];
+
+    for (name, make) in &payloads {
+        let mut group = c.benchmark_group(format!("encode/{name}"));
+        for &size in SIZES {
+            let data = make(size);
+            group.throughput(Throughput::Bytes(size as u64));
+
+            group.bench_with_input(BenchmarkId::new("ucobs", size), &data, |b, d| {
+                let mut buf = vec![0u8; ucobs::max_encoded_len(d.len()) + 1];
+                b.iter(|| {
+                    black_box(ucobs::encode(black_box(d), &mut buf).unwrap());
+                });
+            });
+
+            group.bench_with_input(BenchmarkId::new("corncobs", size), &data, |b, d| {
+                let mut buf = vec![0u8; corncobs::max_encoded_len(d.len())];
+                b.iter(|| {
+                    black_box(corncobs::encode_buf(black_box(d), &mut buf));
+                });
+            });
+
+            group.bench_with_input(BenchmarkId::new("cobs", size), &data, |b, d| {
+                let mut buf = vec![0u8; cobs::max_encoding_length(d.len())];
+                b.iter(|| {
+                    black_box(cobs::encode(black_box(d), &mut buf));
+                });
+            });
+        }
+        group.finish();
+    }
+}
+
+fn bench_decode(c: &mut Criterion) {
+    let payloads: Vec<(&str, Box<dyn Fn(usize) -> Vec<u8>>)> = vec![
+        ("zeros", Box::new(payload_zeros)),
+        ("nonzero", Box::new(payload_nonzero)),
+        ("mixed", Box::new(payload_mixed)),
+    ];
+
+    for (name, make) in &payloads {
+        let mut group = c.benchmark_group(format!("decode/{name}"));
+        for &size in SIZES {
+            let data = make(size);
+            let ucobs_enc = ucobs_encode(&data);
+            // corncobs expects sentinel in encoded form
+            let corncobs_enc = corncobs_encode(&data);
+            let cobs_enc = cobs_encode(&data);
+
+            group.throughput(Throughput::Bytes(size as u64));
+
+            group.bench_with_input(BenchmarkId::new("ucobs", size), &ucobs_enc, |b, enc| {
+                let mut buf = vec![0u8; size + 1];
+                b.iter(|| {
+                    black_box(ucobs_decode(black_box(enc), &mut buf));
+                });
+            });
+
+            group.bench_with_input(
+                BenchmarkId::new("corncobs", size),
+                &corncobs_enc,
+                |b, enc| {
+                    let mut buf = vec![0u8; size + 1];
+                    b.iter(|| {
+                        black_box(corncobs_decode(black_box(enc), &mut buf));
+                    });
+                },
+            );
+
+            group.bench_with_input(BenchmarkId::new("cobs", size), &cobs_enc, |b, enc| {
+                let mut buf = vec![0u8; size + 1];
+                b.iter(|| {
+                    black_box(cobs_decode(black_box(enc), &mut buf));
+                });
+            });
+        }
+        group.finish();
+    }
+}
+
+criterion_group!(benches, bench_encode, bench_decode);
+criterion_main!(benches);

fuzz/Cargo.toml

@@ -24,3 +24,9 @@ name = "fuzz_roundtrip"
 path = "fuzz_targets/fuzz_roundtrip.rs"
 test = false
 doc = false
+
+[[bin]]
+name = "fuzz_small_dest"
+path = "fuzz_targets/fuzz_small_dest.rs"
+test = false
+doc = false

fuzz/fuzz_targets/fuzz_small_dest.rs

@@ -0,0 +1,36 @@
+//! Fuzz target: encode and decode with undersized destination buffers.
+//! Must never panic — should return None when dest is too small.
+#![no_main]
+use libfuzzer_sys::fuzz_target;
+
+fuzz_target!(|data: &[u8]| {
+    if data.is_empty() {
+        return;
+    }
+
+    // Split fuzz input: first byte selects dest size, rest is payload.
+    let dest_size = data[0] as usize;
+    let payload = &data[1..];
+
+    // Encode with a potentially too-small buffer.
+    let mut enc_buf = vec![0u8; dest_size];
+    let enc_result = ucobs::encode(payload, &mut enc_buf);
+
+    if let Some(enc_len) = enc_result {
+        let encoded = &enc_buf[..enc_len];
+
+        // Encoded output must never contain 0x00.
+        assert!(!encoded.contains(&0x00), "encoded contains sentinel");
+
+        // Decode with a potentially too-small buffer.
+        let mut dec_buf = vec![0u8; dest_size];
+        let dec_result = ucobs::decode(encoded, &mut dec_buf);
+
+        if let Some(dec_len) = dec_result {
+            assert_eq!(
+                &dec_buf[..dec_len], payload,
+                "round-trip mismatch with small buffers"
+            );
+        }
+    }
+});

justfile

@@ -0,0 +1,64 @@
+# μCOBS development tasks
+
+# default: list all recipes
+default:
+    @just --list
+
+# run ALL quality gates (tests, clippy, fmt check, doc, fuzz, miri)
+test: test-unit test-clippy test-fmt test-doc test-fuzz test-miri
+
+# quick compile check (no tests)
+check:
+    cargo check
+
+# unit + integration + proptest tests (via nextest for parallelism)
+test-unit:
+    cargo nextest run
+    cargo test --doc
+
+# clippy with all warnings as errors
+test-clippy:
+    cargo clippy -- -D warnings
+
+# check formatting
+test-fmt:
+    cargo fmt -- --check
+
+# doc build (catches broken doc links)
+test-doc:
+    cargo doc --no-deps
+
+# fuzz all targets (60s each, requires nightly + cargo-fuzz)
+test-fuzz:
+    cd fuzz && cargo +nightly fuzz run fuzz_decode -- -max_total_time=60
+    cd fuzz && cargo +nightly fuzz run fuzz_roundtrip -- -max_total_time=60
+    cd fuzz && cargo +nightly fuzz run fuzz_small_dest -- -max_total_time=60
+
+# run tests under miri (catches UB, requires nightly + miri component)
+# proptest is excluded — its file-persistence layer calls getcwd which miri does not support
+test-miri:
+    cargo +nightly miri test -- --skip proptests
+
+# run criterion benchmarks (ucobs vs cobs vs corncobs)
+bench:
+    cargo bench
+
+# run python cobs benchmarks (requires: pip install cobs)
+bench-python:
+    python3 bench_python.py
+
+# run all benchmarks (rust + python)
+bench-all: bench bench-python
+
+# format code
+fmt:
+    cargo fmt
+
+# build in release mode
+build:
+    cargo build --release
+
+# clean all build artifacts
+clean:
+    cargo clean
+    rm -rf fuzz/target fuzz/corpus fuzz/artifacts