Add string benchmark results

README.md

@@ -199,19 +199,27 @@ Lower is better.
 
 ![Integer result, chart](benchmark/runtime/results/int.png)
 
-| Test   |   `scn::scan` | `scn::scan_value` | `std::stringstream` | `sscanf` | `strtol` | `std::from_chars` |
-|:-------|--------------:| ----------------: | ------------------: |---------:|----------|------------------:|
-| Test 1 | 55.9          | 50.9              | 142                 |     81.6 |  21.2    |              13.3 |
-| Test 2 | 56.3          | 51.8              | 59.1                |      512 |  49.7    |              13.9 |
+| Test   |   `scn::scan` | `scn::scan_value` | `stringstream` | `sscanf` | `strtol` | `from_chars` |
+|:-------|--------------:| ----------------: | -------------: |---------:|----------|-------------:|
+| Test 1 | 55.9          | 50.9              | 142            |     81.6 |  21.2    |         13.3 |
+| Test 2 | 56.3          | 51.8              | 59.1           |      512 |  49.7    |         13.9 |
 
 #### Floating-point number parsing (`double`)
 
 ![Float result, chart](benchmark/runtime/results/float.png)
 
-| Test   | `scn::scan` | `scn::scan_value` | `std::stringstream` | `sscanf` | `strtod`  | `std::from_chars` |
-|:-------| ----------: | ----------------: | ------------------: |---------:|-----------|------------------:|
-| Test 1 | 83.6        | 76.6              | 403                 |      206 |     50.1  |              28.7 |
-| Test 2 | 79.8        | 73.9              | 268                 |      716 |     87.2  |              30.4 |
+| Test   | `scn::scan` | `scn::scan_value` | `stringstream` | `sscanf` | `strtod`  | `from_chars` |
+|:-------| ----------: | ----------------: | -------------: |---------:|-----------|-------------:|
+| Test 1 | 83.6        | 76.6              | 403            |      206 |     50.1  |         28.7 |
+| Test 2 | 79.8        | 73.9              | 268            |      716 |     87.2  |         30.4 |
+
+#### String "word" (whitespace-separated character sequence) parsing (`string` and `string_view`)
+
+![String result, chart](benchmark/runtime/results/string.png)
+
+| `scn::scan<string>` | `scn::scan<string_view>` | `scn::scan_value<string>` | `scn::scan_value<string_view>` | `stringstream` | `sscanf` |
+| ------------------: | -----------------------: | ------------------------: | -----------------------------: | -------------: | -------: |
+| 35.6                | 35.4                     | 27.8                      | 26.4                           | 173            | 69.5     |
 
 #### Conclusions
 
@@ -231,6 +239,7 @@ Above,
    which contains multiple values in their text representations, separated by spaces.
    The time used for creating any state needed for the scanner is not included.
    This test is called `"repeated"` in the benchmark sources.
+ * The string test is an exception: strings are read one after another from a sample of Lorem Ipsum.
 
 The difference between "Test 1" and "Test 2" is most pronounced when using a `stringstream`,
 which is relatively expensive to construct, and seems to be adding around ~100ns of runtime.
@@ -242,7 +251,8 @@ These benchmarks were run on a Fedora 37 machine, running Linux kernel version 6
 with an AMD Ryzen 7 5700X processor, and compiled with gcc version 12.3.1,
 with `-O3 -DNDEBUG -march=haswell` and LTO enabled.
 C++20 was used, with the standard library (libstdc++) `<ranges>` implementation.
-These benchmarks were run on 2023-08-24 (commit fdefc09).
+These benchmarks were run on 2023-08-24 (commit fdefc09), except for the string-benchmark,
+which was run on 2023-10-28 (commit 586e093).
 
 The source code for these benchmarks can be found in the `benchmark` directory.
 You can run these benchmarks yourself by enabling the CMake variable `SCN_BENCHMARKS`.

benchmark/runtime/string/string_bench.cpp

@@ -65,6 +65,39 @@ BENCHMARK(bench_string_scn<wchar_t, std::wstring, unicode_tag>);
 BENCHMARK(bench_string_scn<wchar_t, std::wstring_view, lipsum_tag>);
 BENCHMARK(bench_string_scn<wchar_t, std::wstring_view, unicode_tag>);
 
+template <typename SourceCharT, typename DestStringT, typename Tag>
+static void bench_string_scn_value(benchmark::State& state)
+{
+    auto input = get_benchmark_input<SourceCharT, Tag>();
+    auto subr = scn::ranges::subrange{input};
+    for (auto _ : state) {
+        if (auto result = scn::scan_value<DestStringT>(subr)) {
+            benchmark::DoNotOptimize(result->value());
+            subr = result->range();
+        }
+        else if (result.error() == scn::scan_error::end_of_range) {
+            subr = scn::ranges::subrange{input};
+        }
+        else {
+            state.SkipWithError("Failed scan");
+            break;
+        }
+    }
+}
+
+BENCHMARK(bench_string_scn_value<char, std::string_view, lipsum_tag>);
+BENCHMARK(bench_string_scn_value<char, std::string_view, unicode_tag>);
+BENCHMARK(bench_string_scn_value<char, std::string, lipsum_tag>);
+BENCHMARK(bench_string_scn_value<char, std::string, unicode_tag>);
+BENCHMARK(bench_string_scn_value<char, std::wstring, lipsum_tag>);
+BENCHMARK(bench_string_scn_value<char, std::wstring, unicode_tag>);
+BENCHMARK(bench_string_scn_value<wchar_t, std::string, lipsum_tag>);
+BENCHMARK(bench_string_scn_value<wchar_t, std::string, unicode_tag>);
+BENCHMARK(bench_string_scn_value<wchar_t, std::wstring, lipsum_tag>);
+BENCHMARK(bench_string_scn_value<wchar_t, std::wstring, unicode_tag>);
+BENCHMARK(bench_string_scn_value<wchar_t, std::wstring_view, lipsum_tag>);
+BENCHMARK(bench_string_scn_value<wchar_t, std::wstring_view, unicode_tag>);
+
 template <typename CharT, typename Tag>
 static void bench_string_sstream(benchmark::State& state)
 {
@@ -89,4 +122,48 @@ BENCHMARK(bench_string_sstream<char, unicode_tag>);
 BENCHMARK(bench_string_sstream<wchar_t, lipsum_tag>);
 BENCHMARK(bench_string_sstream<wchar_t, unicode_tag>);
 
-BENCHMARK_MAIN();
+namespace {
+    const char* sscanf_impl(const char* input, std::string& value)
+    {
+        auto r = std::sscanf(input, " %255s", value.data());
+        if (r != 1) {
+            return nullptr;
+        }
+        auto len = std::strlen(value.c_str());
+        value.resize(len);
+        return input + len;
+    }
+    const wchar_t* sscanf_impl(const wchar_t* input, std::wstring& value)
+    {
+        auto r = std::swscanf(input, L" %255s", value.data());
+        if (r != 1) {
+            return nullptr;
+        }
+        auto len = std::wcslen(value.c_str());
+        value.resize(len);
+        return input + len;
+    }
+}  // namespace
+
+template <typename CharT, typename Tag>
+static void bench_string_scanf(benchmark::State& state)
+{
+    auto input = get_benchmark_input<CharT, Tag>();
+    const CharT* begin = input.data();
+    for (auto _ : state) {
+        std::basic_string<CharT> val{};
+        val.resize(256);
+        auto p = sscanf_impl(begin, val);
+        if (!p) {
+            begin = input.data();
+            continue;
+        }
+        begin = p;
+        benchmark::DoNotOptimize(val);
+    }
+}
+
+BENCHMARK(bench_string_scanf<char, lipsum_tag>);
+BENCHMARK(bench_string_scanf<char, unicode_tag>);
+BENCHMARK(bench_string_scanf<wchar_t, lipsum_tag>);
+BENCHMARK(bench_string_scanf<wchar_t, unicode_tag>);