Fix benchmark script to be generic and not specific to batch process

BENCHMARKING.md

@@ -1,7 +1,7 @@
 # Benchmarking phpcs-changed
 
 `benchmark.sh` measures the wall-clock time of running `phpcs-changed --git-staged`
-against a set of changed PHP files, and compares the current branch to `trunk`.
+against a set of changed PHP files, comparing the **current branch** to `trunk`.
 
 It uses [hyperfine](https://github.com/sharkdp/hyperfine) for statistical
 measurement and a self-contained throwaway git repo for reproducible test inputs.
@@ -66,37 +66,25 @@ The worktree is created on the first run and reused on subsequent runs.
 
 ### What is being compared
 
-| Branch | phpcs invocations per run |
-|--------|--------------------------|
-| trunk | up to `2 × N_FILES` — one per file per version (modified + unmodified) |
-| feature branch | 1 — all file versions batched into a single phpcs call |
-
-The dominant cost in each phpcs invocation is **process startup** (~250–400 ms
-on typical hardware). The batch approach eliminates `2 × N_FILES − 1` of those
-startups.
+The script runs the same `phpcs-changed --git-staged` command against the same
+set of test files using two builds: the current branch and `trunk`. Hyperfine
+runs both commands back-to-back the same number of times, so any difference in
+wall-clock time reflects a real performance difference between the two builds.
 
 ## Interpreting results
 
 Hyperfine reports mean time, standard deviation, and a relative speedup ratio.
 Example output for `N_FILES=10`:
 
 ```
-Benchmark 1: batch (a75e2c2): 1 phpcs call
+Benchmark 1: my-feature-branch (a75e2c2)
   Time (mean ± σ):      2.44 s ±  0.05 s
-Benchmark 2: trunk (5c9f6b2): 20 phpcs calls
+Benchmark 2: trunk (5c9f6b2)
   Time (mean ± σ):      5.74 s ±  0.11 s
 
-Summary: batch ran 2.36 ± 0.06 times faster than trunk
+Summary: my-feature-branch ran 2.36 ± 0.06 times faster than trunk
 ```
 
-The speedup scales roughly linearly with `N_FILES`. A project with 20 changed
-files should see approximately twice the speedup of a 10-file project.
-
-The batch variant carries its own overhead (temp directory creation, one file
-write per file version, one phpcs invocation on all files). This overhead
-appears as a baseline cost that does not scale with `N_FILES`, so the
-crossover point where batching wins is low — roughly 2 or more files.
-
 ## Artifacts
 
 The following files are created outside the git index and are gitignored:

benchmark.sh

@@ -1,5 +1,5 @@
 #!/usr/bin/env bash
-# benchmark.sh — compare batch-phpcs-invocations branch vs trunk
+# benchmark.sh — compare the current branch vs trunk
 #
 # Usage:
 #   ./benchmark.sh
@@ -17,7 +17,7 @@ set -euo pipefail
 
 # ── Paths ──────────────────────────────────────────────────────────────────
 SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
-BATCH_BIN="$SCRIPT_DIR/bin/phpcs-changed"
+CURRENT_BIN="$SCRIPT_DIR/bin/phpcs-changed"
 TRUNK_WORKTREE="$SCRIPT_DIR/.bench-trunk"
 TRUNK_BIN="$TRUNK_WORKTREE/bin/phpcs-changed"
 PHPCS="$SCRIPT_DIR/vendor/bin/phpcs"
@@ -51,8 +51,8 @@ git -C "$BENCH_REPO" config user.email "bench@example.com"
 git -C "$BENCH_REPO" config user.name "Benchmark"
 
 # Create initial PHP files with PSR2 violations (uppercase TRUE / FALSE).
-# Both the committed and staged versions have violations so trunk will always
-# scan both file versions — giving it 2×N_FILES phpcs invocations.
+# Both the committed and staged versions have violations so both builds must
+# scan both file versions — this is the worst-case and ensures a fair comparison.
 echo "→ Preparing $N_FILES PHP test files …"
 FILES=()
 for i in $(seq 1 "$N_FILES"); do
@@ -98,26 +98,27 @@ done
 
 # ── 3. Run hyperfine ──────────────────────────────────────────────────────
 FILES_STR="${FILES[*]}"
-BATCH_SHA="$(git -C "$SCRIPT_DIR" rev-parse --short HEAD)"
+CURRENT_BRANCH="$(git -C "$SCRIPT_DIR" branch --show-current)"
+CURRENT_SHA="$(git -C "$SCRIPT_DIR" rev-parse --short HEAD)"
 TRUNK_SHA="$(git -C "$TRUNK_WORKTREE" rev-parse --short HEAD)"
 
-BATCH_CMD="php '$BATCH_BIN' --git-staged --phpcs-path='$PHPCS' --standard=PSR2 --always-exit-zero $FILES_STR"
+CURRENT_CMD="php '$CURRENT_BIN' --git-staged --phpcs-path='$PHPCS' --standard=PSR2 --always-exit-zero $FILES_STR"
 TRUNK_CMD="php '$TRUNK_BIN' --git-staged --phpcs-path='$PHPCS' --standard=PSR2 --always-exit-zero $FILES_STR"
 
 echo ""
 printf "Benchmark: %d staged PHP files, --standard=PSR2\n" "$N_FILES"
-printf "  batch  %s  — 1 phpcs invocation (all files in one shot)\n" "$BATCH_SHA"
-printf "  trunk  %s  — up to %d phpcs invocations (2 per file)\n" "$TRUNK_SHA" "$((N_FILES * 2))"
+printf "  %-30s  %s\n" "$CURRENT_BRANCH" "$CURRENT_SHA"
+printf "  %-30s  %s\n" "trunk" "$TRUNK_SHA"
 echo ""
 
 cd "$BENCH_REPO"
 # shellcheck disable=SC2086
 hyperfine \
   --warmup "$WARMUP" \
   --runs   "$RUNS" \
-  -n "batch ($BATCH_SHA): 1 phpcs call" \
-  -n "trunk ($TRUNK_SHA): $((N_FILES * 2)) phpcs calls" \
-  "$BATCH_CMD" \
+  -n "$CURRENT_BRANCH ($CURRENT_SHA)" \
+  -n "trunk ($TRUNK_SHA)" \
+  "$CURRENT_CMD" \
   "$TRUNK_CMD" \
   --export-markdown "$RESULTS_FILE"