Add benchmark_runner info and query commands

benchmarks/Cargo.toml

@@ -44,7 +44,7 @@ anstream = "1.0"
 arrow = { workspace = true }
 async-trait = "0.1"
 bytes = { workspace = true }
-clap = { version = "4.6.1", features = ["derive", "env", "color"] }
+clap = { version = "4.6.1", features = ["derive", "env", "color", "string"] }
 criterion = { workspace = true, features = ["html_reports"] }
 datafusion = { workspace = true, default-features = true }
 datafusion-common = { workspace = true, default-features = true }

benchmarks/benches/sql.rs

@@ -169,10 +169,10 @@ fn handle_run(
     name: &str,
 ) {
     rt.block_on(async {
-        benchmark
+        let _ = benchmark
             .run(ctx, args.validate)
             .await
-            .unwrap_or_else(|err| panic!("Failed to run benchmark {name}: {err:?}"))
+            .unwrap_or_else(|err| panic!("Failed to run benchmark {name}: {err:?}"));
     });
 }
 
@@ -205,7 +205,7 @@ fn handle_verify(
     info!("Verifying benchmark {name} results ...");
 
     rt.block_on(async {
-        benchmark
+        let _ = benchmark
             .run(ctx, true)
             .await
             .unwrap_or_else(|err| panic!("Failed to run benchmark {name}: {err:?}"));
@@ -278,7 +278,8 @@ async fn load_benchmarks(
     for path in paths {
         debug!("Loading benchmark from {path}");
 
-        let benchmark = SqlBenchmark::new(ctx, &path, &*SQL_BENCHMARK_DIRECTORY).await?;
+        let benchmark =
+            SqlBenchmark::new(ctx, &path, &*SQL_BENCHMARK_DIRECTORY, None).await?;
         let entries = benches
             .entry(benchmark.group().to_string())
             .or_insert(vec![]);

benchmarks/sql_benchmarks/tpch/tpch.suite

@@ -1,16 +1,48 @@
-name = "tpch"
 description = "TPC-H SQL benchmarks"
 
+# Query patterns control how numeric QUERY_ID values map to .benchmark files
+# during discovery and command resolution. Use exactly one query-id token:
+# - {QUERY_ID_PADDED}: two-digit ids, such as q01.benchmark
+# - {QUERY_ID}: unpadded ids, such as query-1.benchmark
+# If omitted, this defaults to q{QUERY_ID_PADDED}.benchmark.
+query_pattern = "q{QUERY_ID_PADDED}.benchmark"
+
+# Path replacements define path-like variables used while parsing benchmark
+# files. Relative paths are resolved from this suite file's directory and then
+# passed to SqlBenchmark's replacement mapping, so benchmark SQL can refer to
+# values such as ${DATA_DIR}. For timed runs, the runner's --path/-p option
+# overrides DATA_DIR.
+[path_replacements]
+DATA_DIR = "../../data"
+
 [[options]]
 name = "format"
 short = "f"
+env = "TPCH_FILE_TYPE"
 default = "parquet"
 values = ["parquet", "csv", "mem"]
 help = "Selects the TPC-H data format."
 
 [[options]]
 name = "scale-factor"
 short = "sf"
+env = "BENCH_SIZE"
 default = "1"
 values = ["1", "10"]
 help = "Selects the TPC-H scale factor."
+
+[[examples]]
+command = "cargo run --release --bin benchmark_runner -- run tpch"
+description = "Run all TPC-H queries with the default parquet SF1 configuration."
+
+[[examples]]
+command = "cargo run --release --bin benchmark_runner -- run tpch 15"
+description = "Run TPC-H query 15 with the default parquet SF1 configuration."
+
+[[examples]]
+command = "cargo run --release --bin benchmark_runner -- run tpch 15 -f csv"
+description = "Run TPC-H query 15 against CSV data."
+
+[[examples]]
+command = "cargo run --release --bin benchmark_runner -- run tpch 15 -sf 10"
+description = "Run TPC-H query 15 at scale factor 10."