Measure per byte in benchmarks and remove some of them

benchmarks/bench_commit_request.cpp

@@ -13,9 +13,10 @@ int main() {
   // One-shot parsing benchmarks
   auto bench = ankerl::nanobench::Bench()
                    .title("CommitRequest One-Shot Parsing")
-                   .unit("parse")
+                   .unit("byte")
                    .warmup(100);
 
+  bench.batch(SIMPLE_JSON.size());
   // Simple JSON parsing
   bench.run("Simple JSON (3 fields)", [&] {
     CommitRequest request;
@@ -27,6 +28,7 @@ int main() {
     ankerl::nanobench::doNotOptimizeAway(request.leader_id());
   });
 
+  bench.batch(MEDIUM_JSON.size());
   // Medium complexity JSON parsing
   bench.run("Medium JSON (2 preconditions, 2 operations)", [&] {
     CommitRequest request;
@@ -38,6 +40,7 @@ int main() {
     ankerl::nanobench::doNotOptimizeAway(request.leader_id());
   });
 
+  bench.batch(COMPLEX_JSON.size());
   // Complex JSON parsing
   bench.run("Complex JSON (3 preconditions, 5 operations)", [&] {
     CommitRequest request;
@@ -52,6 +55,7 @@ int main() {
   // Large batch operations
   for (int num_ops : {10, 50, 100, 500}) {
     std::string large_json = generate_large_json(num_ops);
+    bench.batch(large_json.size());
     bench.run("Large JSON (" + std::to_string(num_ops) + " operations)", [&] {
       CommitRequest request;
       JsonCommitRequestParser parser;
@@ -66,7 +70,8 @@ int main() {
   // Streaming parsing benchmarks
   auto streaming_bench = ankerl::nanobench::Bench()
                              .title("CommitRequest Streaming Parsing")
-                             .unit("parse")
+                             .unit("byte")
+                             .batch(MEDIUM_JSON.size())
                              .warmup(50);
 
   // Streaming with different chunk sizes
@@ -102,67 +107,5 @@ int main() {
         });
   }
 
-  // Reset and reuse benchmarks
-  auto reuse_bench = ankerl::nanobench::Bench()
-                         .title("CommitRequest Reset and Reuse")
-                         .unit("operation")
-                         .warmup(50);
-
-  reuse_bench.run("Parse -> Reset -> Parse cycle", [&] {
-    static CommitRequest request; // Static to persist across invocations
-    static JsonCommitRequestParser parser;
-
-    std::string mutable_json1 = SIMPLE_JSON;
-    bool result1 =
-        parser.parse(request, mutable_json1.data(), mutable_json1.size());
-
-    request.reset();
-
-    std::string mutable_json2 = MEDIUM_JSON;
-    bool result2 =
-        parser.parse(request, mutable_json2.data(), mutable_json2.size());
-
-    ankerl::nanobench::doNotOptimizeAway(result1);
-    ankerl::nanobench::doNotOptimizeAway(result2);
-    ankerl::nanobench::doNotOptimizeAway(request.leader_id());
-  });
-
-  // Base64 decoding performance
-  auto base64_bench = ankerl::nanobench::Bench()
-                          .title("Base64 Decoding Performance")
-                          .unit("decode")
-                          .warmup(50);
-
-  // JSON with lots of base64 encoded data
-  std::string base64_heavy_json = R"({
-        "leader_id": "base64-test-leader",
-        "read_version": 12345,
-        "operations": [)";
-
-  for (int i = 0; i < 20; ++i) {
-    if (i > 0)
-      base64_heavy_json += ",";
-    base64_heavy_json += R"(
-            {
-                "type": "write",
-                "key": "VGhpc0lzQUxvbmdCYXNlNjRFbmNvZGVkS2V5V2l0aExvdHNPZkRhdGFGb3JUZXN0aW5nUHVycG9zZXM=",
-                "value": "VGhpc0lzQW5FdmVuTG9uZ2VyQmFzZTY0RW5jb2RlZFZhbHVlV2l0aEV2ZW5Nb3JlRGF0YUZvclRlc3RpbmdUaGVCYXNlNjREZWNvZGluZ1BlcmZvcm1hbmNlT2ZUaGVQYXJzZXI="
-            })";
-  }
-  base64_heavy_json += R"(
-        ]
-    })";
-
-  base64_bench.run(
-      "Heavy Base64 JSON (20 operations with long encoded data)", [&] {
-        CommitRequest request;
-        JsonCommitRequestParser parser;
-        std::string mutable_json = base64_heavy_json;
-        bool result =
-            parser.parse(request, mutable_json.data(), mutable_json.size());
-        ankerl::nanobench::doNotOptimizeAway(result);
-        ankerl::nanobench::doNotOptimizeAway(request.leader_id());
-      });
-
   return 0;
 }

benchmarks/bench_parser_comparison.cpp

@@ -463,7 +463,8 @@ int main() {
   // Simple JSON comparison
   auto simple_bench = ankerl::nanobench::Bench()
                           .title("Simple JSON Parsing Comparison")
-                          .unit("parse")
+                          .unit("byte")
+                          .batch(SIMPLE_JSON.size())
                           .warmup(100)
                           .minEpochIterations(1000);
 
@@ -542,7 +543,8 @@ int main() {
   // Medium complexity JSON comparison
   auto medium_bench = ankerl::nanobench::Bench()
                           .title("Medium JSON Parsing Comparison")
-                          .unit("parse")
+                          .unit("byte")
+                          .batch(MEDIUM_JSON.size())
                           .warmup(100)
                           .minEpochIterations(500);
 
@@ -621,7 +623,8 @@ int main() {
   // Complex JSON comparison
   auto complex_bench = ankerl::nanobench::Bench()
                            .title("Complex JSON Parsing Comparison")
-                           .unit("parse")
+                           .unit("byte")
+                           .batch(COMPLEX_JSON.size())
                            .warmup(100)
                            .minEpochIterations(200);
 
@@ -697,16 +700,16 @@ int main() {
     ankerl::nanobench::doNotOptimizeAway(handler.operations.size());
   });
 
-  // Large batch operations comparison
-  auto large_bench = ankerl::nanobench::Bench()
-                         .title("Large JSON Parsing Comparison")
-                         .unit("parse")
-                         .warmup(50)
-                         .minEpochIterations(100);
-
   for (int num_ops : {50, 100, 500}) {
     std::string large_json = generate_large_json(num_ops);
     std::string bench_name = std::to_string(num_ops) + " operations";
+    // Large batch operations comparison
+    auto large_bench = ankerl::nanobench::Bench()
+                           .title("Large JSON Parsing Comparison")
+                           .unit("byte")
+                           .batch(large_json.size())
+                           .warmup(50)
+                           .minEpochIterations(100);
 
     large_bench.run("WeaselDB Parser (" + bench_name + ")", [&] {
       CommitRequest request;
@@ -783,133 +786,6 @@ int main() {
         });
   }
 
-  // Memory efficiency comparison
-  auto memory_bench = ankerl::nanobench::Bench()
-                          .title("Memory Efficiency Comparison")
-                          .unit("allocation")
-                          .warmup(50)
-                          .minEpochIterations(200);
-
-  memory_bench.run("WeaselDB Parser (arena allocation)", [&] {
-    CommitRequest request;
-    JsonCommitRequestParser parser;
-    std::string mutable_json = COMPLEX_JSON;
-    bool result =
-        parser.parse(request, mutable_json.data(), mutable_json.size());
-    ankerl::nanobench::doNotOptimizeAway(result);
-    ankerl::nanobench::doNotOptimizeAway(request.total_allocated());
-    ankerl::nanobench::doNotOptimizeAway(request.used_bytes());
-  });
-
-  memory_bench.run("nlohmann/json (standard allocation)", [&] {
-    try {
-      nlohmann::json j = nlohmann::json::parse(COMPLEX_JSON);
-      bool result = validate_nlohmann_commit_request(j);
-      ankerl::nanobench::doNotOptimizeAway(result);
-      ankerl::nanobench::doNotOptimizeAway(j.size());
-    } catch (const std::exception &) {
-      ankerl::nanobench::doNotOptimizeAway(false);
-    }
-  });
-
-  memory_bench.run("RapidJSON SAX (standard allocation)", [&] {
-    CommitRequestSaxHandler handler;
-    rapidjson::Reader reader;
-    rapidjson::StringStream ss(COMPLEX_JSON.c_str());
-    bool result = reader.Parse(ss, handler);
-    result = result && handler.validate();
-    ankerl::nanobench::doNotOptimizeAway(result);
-    ankerl::nanobench::doNotOptimizeAway(handler.preconditions.size());
-    ankerl::nanobench::doNotOptimizeAway(handler.operations.size());
-  });
-
-  memory_bench.run("RapidJSON SAX Arena (arena allocation)", [&] {
-    CommitRequestArenaHandler handler;
-    rapidjson::Reader reader;
-    rapidjson::StringStream ss(COMPLEX_JSON.c_str());
-    bool result = reader.Parse(ss, handler);
-    result = result && handler.validate();
-    ankerl::nanobench::doNotOptimizeAway(result);
-    ankerl::nanobench::doNotOptimizeAway(handler.total_allocated());
-    ankerl::nanobench::doNotOptimizeAway(handler.used_bytes());
-  });
-
-  // Reset and reuse comparison
-  auto reuse_bench = ankerl::nanobench::Bench()
-                         .title("Reset and Reuse Comparison")
-                         .unit("cycle")
-                         .warmup(50)
-                         .minEpochIterations(100);
-
-  reuse_bench.run("WeaselDB Parser (reset)", [&] {
-    static CommitRequest request;
-    static JsonCommitRequestParser parser;
-
-    std::string mutable_json1 = SIMPLE_JSON;
-    bool result1 =
-        parser.parse(request, mutable_json1.data(), mutable_json1.size());
-
-    request.reset();
-
-    std::string mutable_json2 = MEDIUM_JSON;
-    bool result2 =
-        parser.parse(request, mutable_json2.data(), mutable_json2.size());
-
-    ankerl::nanobench::doNotOptimizeAway(result1);
-    ankerl::nanobench::doNotOptimizeAway(result2);
-  });
-
-  reuse_bench.run("nlohmann/json (new instance)", [&] {
-    try {
-      nlohmann::json j1 = nlohmann::json::parse(SIMPLE_JSON);
-      bool result1 = validate_nlohmann_commit_request(j1);
-
-      nlohmann::json j2 = nlohmann::json::parse(MEDIUM_JSON);
-      bool result2 = validate_nlohmann_commit_request(j2);
-
-      ankerl::nanobench::doNotOptimizeAway(result1);
-      ankerl::nanobench::doNotOptimizeAway(result2);
-    } catch (const std::exception &) {
-      ankerl::nanobench::doNotOptimizeAway(false);
-    }
-  });
-
-  reuse_bench.run("RapidJSON SAX (reset)", [&] {
-    static CommitRequestSaxHandler handler;
-
-    rapidjson::Reader reader;
-    rapidjson::StringStream ss1(SIMPLE_JSON.c_str());
-    bool result1 = reader.Parse(ss1, handler);
-    result1 = result1 && handler.validate();
-
-    handler.reset();
-
-    rapidjson::StringStream ss2(MEDIUM_JSON.c_str());
-    bool result2 = reader.Parse(ss2, handler);
-    result2 = result2 && handler.validate();
-
-    ankerl::nanobench::doNotOptimizeAway(result1);
-    ankerl::nanobench::doNotOptimizeAway(result2);
-  });
-
-  reuse_bench.run("RapidJSON SAX Arena (reset)", [&] {
-    static CommitRequestArenaHandler handler;
-
-    rapidjson::Reader reader;
-    rapidjson::StringStream ss1(SIMPLE_JSON.c_str());
-    bool result1 = reader.Parse(ss1, handler);
-    result1 = result1 && handler.validate();
-
-    handler.reset();
-
-    rapidjson::StringStream ss2(MEDIUM_JSON.c_str());
-    bool result2 = reader.Parse(ss2, handler);
-    result2 = result2 && handler.validate();
-
-    ankerl::nanobench::doNotOptimizeAway(result1);
-    ankerl::nanobench::doNotOptimizeAway(result2);
-  });
-
   std::cout << "\nBenchmark completed. The WeaselDB parser is optimized for:\n";
   std::cout << "- Arena-based memory allocation for reduced fragmentation\n";
   std::cout << "- Streaming parsing for network protocols\n";