Improve Performance of Parsing

Cargo.toml

@@ -86,4 +86,4 @@ debug = false
 strip = true
 
 [profile.bench]
-inherits = "release-minimal"
+inherits = "release"

nemo/Cargo.toml

@@ -39,8 +39,7 @@ ascii_tree = { workspace = true }
 async-trait = { workspace = true }
 serde_json = { workspace = true }
 serde = { workspace = true, features = ["derive"] }
-nom-greedyerror = "0.5.0"
-nom-supreme = "0.8.0"
+smallvec = "1.15"
 enum-assoc = "1.1.0"
 ariadne = { workspace = true, features = ["auto-color"] }
 strum = { workspace = true }
@@ -63,4 +62,9 @@ env_logger = { workspace = true }
 assert_fs = { workspace = true }
 test-log = { workspace = true }
 tokio = { workspace = true, features = [ "rt", "macros" ] }
+criterion = { version = "0.8.2", default-features = false, features = [ "cargo_bench_support" ] }
+
+[[bench]]
+name = "parser"
+harness = false
 

nemo/benches/parser.rs

@@ -0,0 +1,53 @@
+//! Benchmarks for the parser and the AST-to-rule-model translation.
+
+use criterion::{Criterion, Throughput, criterion_group, criterion_main};
+use nemo::{parser::Parser, rule_file::RuleFile, rule_model::programs::handle::ProgramHandle};
+use std::hint::black_box;
+
+/// Program consisting of many rules, each joining many atoms.
+const RULES: &str = include_str!("../../resources/benchmarks/variable-order-stress.rls");
+/// Program consisting of many ground facts.
+const FACTS: &str = include_str!("../../resources/benchmarks/fact-stress.rls");
+
+/// Parse the given program, discarding the result.
+fn parse(input: &str) {
+    let parser = Parser::initialize(input);
+    let _ = black_box(parser.parse());
+}
+
+/// Parse the given program and translate it into the rule model.
+fn parse_and_translate(input: &str) {
+    let file = RuleFile::new(input.to_owned(), String::from("bench"));
+    let _ = black_box(ProgramHandle::from_file(&file));
+}
+
+/// Number of samples per benchmark.
+const SAMPLE_SIZE: usize = 10;
+
+fn benchmark_parse(criterion: &mut Criterion) {
+    let mut group = criterion.benchmark_group("parse");
+    group.sample_size(SAMPLE_SIZE);
+
+    group.throughput(Throughput::Bytes(RULES.len() as u64));
+    group.bench_function("rules", |bencher| bencher.iter(|| parse(black_box(RULES))));
+
+    group.throughput(Throughput::Bytes(FACTS.len() as u64));
+    group.bench_function("facts", |bencher| bencher.iter(|| parse(black_box(FACTS))));
+
+    group.finish();
+}
+
+fn benchmark_translate(criterion: &mut Criterion) {
+    let mut group = criterion.benchmark_group("parse+translate");
+    group.sample_size(SAMPLE_SIZE);
+
+    group.throughput(Throughput::Bytes(RULES.len() as u64));
+    group.bench_function("rules", |bencher| {
+        bencher.iter(|| parse_and_translate(black_box(RULES)))
+    });
+
+    group.finish();
+}
+
+criterion_group!(benches, benchmark_parse, benchmark_translate);
+criterion_main!(benches);