aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorOnur Gündüz <ogunduz@gmail.com>2026-09-29 14:15:13 +0200
committerOnur Gündüz <ogunduz@gmail.com>2026-09-29 14:15:13 +0200
commita8a89b3ed3101981d4d87ce50f924ea499c369fd (patch)
treef3a3b51fa2eb6612652dfbc26bd9ee8527be2822
parented979a49940b3054468b48ad1a9f97fab5a2fd81 (diff)
downloadele-a8a89b3ed3101981d4d87ce50f924ea499c369fd.tar.gz
ele-a8a89b3ed3101981d4d87ce50f924ea499c369fd.zip
Optimize release throughput and add reproducible CLI benchmarks
-rw-r--r--Cargo.toml2
-rw-r--r--README.md19
-rw-r--r--bench/benchmark.py77
3 files changed, 97 insertions, 1 deletions
diff --git a/Cargo.toml b/Cargo.toml
index 6e2d5e4..ed2ec34 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -8,7 +8,7 @@ regex = "1"
serde_json = "1"
[profile.release]
-opt-level = "z"
+opt-level = 3
lto = true
codegen-units = 1
panic = "abort"
diff --git a/README.md b/README.md
index 056f76c..f0d487d 100644
--- a/README.md
+++ b/README.md
@@ -52,6 +52,25 @@ echo '{"user": {"settings": {"theme": "dark"}}}' | \
aip-filter "user.settings.theme = dark"
```
+## Performance
+
+Release builds optimize for speed rather than minimum executable size. The evaluator reuses compiled regexes and literals, borrows JSON collections, and searches overlapping phrases in linear time. JSON parsing still allocates memory for each record.
+
+To compare two release binaries, use Python 3:
+
+```bash
+python3 bench/benchmark.py ./target/baseline ./target/release/aip-filter \
+ --records 100000 --runs 5
+```
+
+The benchmark generates JSON lines under `target/benchmarks`, checks that both binaries produce identical output, and reports median elapsed times for seven workloads. It alternates execution order and discards output during timing. Results depend on the machine and data.
+
+For a smaller binary, Cargo also supports a size-oriented build without changing the configuration:
+
+```bash
+CARGO_PROFILE_RELEASE_OPT_LEVEL=z cargo build --release --locked
+```
+
## API
Implement `Filterable` for your type, compile the expression once with `Expr::compile`, and reuse `CompiledFilter::evaluate`. Compilation returns errors for unsupported expressions before any records are read. `serde_json::Value` implements `Filterable` directly and borrows arrays and objects without copying them.
diff --git a/bench/benchmark.py b/bench/benchmark.py
new file mode 100644
index 0000000..4fe208d
--- /dev/null
+++ b/bench/benchmark.py
@@ -0,0 +1,77 @@
+#!/usr/bin/env python3
+"""Compare CLI throughput and verify identical output on generated JSON lines."""
+
+import argparse
+import hashlib
+import json
+from pathlib import Path
+import statistics
+import subprocess
+import time
+
+ROOT = Path(__file__).resolve().parent.parent
+CASES = {
+ "bare": "needle",
+ "equality": "state = open",
+ "numeric": "priority > 50 AND score < 3.5",
+ "nested": "user.settings.theme = dark",
+ "list": "tags:beta",
+ "regex": 'msg =~ "[a-z]+ needle[0-9]+$"',
+ "regex_group": 'msg =~ ("[a-z]+ needle[0-9]+$" OR "^skip")',
+}
+
+
+def digest(binary, expression, data):
+ result = subprocess.run(
+ [str(binary), expression, str(data)], capture_output=True, check=True
+ )
+ return hashlib.sha256(result.stdout).hexdigest()
+
+
+def main():
+ parser = argparse.ArgumentParser(description=__doc__)
+ parser.add_argument("baseline", type=Path)
+ parser.add_argument("candidate", type=Path)
+ parser.add_argument("--records", type=int, default=100_000)
+ parser.add_argument("--runs", type=int, default=5)
+ args = parser.parse_args()
+ if args.records <= 0 or args.runs <= 0:
+ parser.error("records and runs must be positive")
+ binaries = [args.baseline.resolve(), args.candidate.resolve()]
+ data = ROOT / "target" / "benchmarks" / "records.jsonl"
+ data.parent.mkdir(parents=True, exist_ok=True)
+ with data.open("w") as output:
+ for i in range(args.records):
+ record = {
+ "id": i,
+ "msg": f"hello needle{i}" if i % 5 == 0 else "skip this record",
+ "state": "open" if i % 2 else "closed",
+ "priority": i % 100,
+ "score": (i % 10) / 2,
+ "user": {"settings": {"theme": "dark" if i % 3 else "light"}},
+ "tags": ["stable", "beta" if i % 4 else "alpha"],
+ }
+ output.write(json.dumps(record, separators=(",", ":")) + "\n")
+ print(f"{args.records:,} records; {data.stat().st_size:,} bytes; median of {args.runs} runs")
+ print("case baseline(s) candidate(s) speedup")
+ for label, expression in CASES.items():
+ assert digest(binaries[0], expression, data) == digest(binaries[1], expression, data), label
+ samples = [[], []]
+ for run in range(args.runs):
+ for index in ([0, 1] if run % 2 == 0 else [1, 0]):
+ start = time.perf_counter()
+ subprocess.run(
+ [str(binaries[index]), expression, str(data)],
+ stdout=subprocess.DEVNULL,
+ stderr=subprocess.PIPE,
+ check=True,
+ )
+ samples[index].append(time.perf_counter() - start)
+ before, after = (statistics.median(sample) for sample in samples)
+ print(f"{label:13} {before:11.4f} {after:12.4f} {before / after:6.2f}x")
+ for binary in binaries:
+ print(f"{binary.name}: {binary.stat().st_size:,} bytes")
+
+
+if __name__ == "__main__":
+ main()