ssrjson-benchmark 0.0.2__tar.gz → 0.0.4__tar.gz

{ssrjson_benchmark-0.0.2 → ssrjson_benchmark-0.0.4}/CMakeLists.txt

@@ -59,7 +59,7 @@ message("CMAKE_SYSTEM_PROCESSOR = ${CMAKE_SYSTEM_PROCESSOR}")
 
 # set ssrjson_benchmark src files
 set(SRC_FILES
-    src/benchmark.c
+    src/_ssrjson_benchmark.c
 )
 
 add_library(ssrjson_benchmark SHARED ${SRC_FILES})

{ssrjson_benchmark-0.0.2/src/ssrjson_benchmark.egg-info → ssrjson_benchmark-0.0.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: ssrjson-benchmark
-Version: 0.0.2
+Version: 0.0.4
 Summary: benchmark of ssrJSON
 Author-email: Eritque Arcus <eritque-arcus@ikuyo.dev>, Antares <antares0982@gmail.com>
 License: MIT License
@@ -40,6 +40,7 @@ Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: ssrjson
 Requires-Dist: orjson
+Requires-Dist: ujson
 Requires-Dist: matplotlib
 Provides-Extra: pdf
 Requires-Dist: svglib; extra == "pdf"
@@ -84,7 +85,7 @@ python -m ssrjson_benchmark
 
 ## Notes
 
-* This repository conducts benchmarking using json, orjson, and ssrJSON. The `dumps` benchmark produces str objects, comparing three operations: `json.dumps`, `orjson.dumps` followed by decode, and `ssrjson.dumps`. The `dumps_to_bytes` benchmark produces bytes objects, comparing three functions: `json.dumps` followed by encode, `orjson.dumps`, and `ssrjson.dumps_to_bytes`.
-* When orjson handles non-ASCII strings, if the cache of the `PyUnicodeObject`’s UTF-8 representation does not exist, it invokes the `PyUnicode_AsUTF8AndSize` function to obtain the UTF-8 encoding. This function then caches the UTF-8 representation within the `PyUnicodeObject`. If the same `PyUnicodeObject` undergoes repeated encode-decode operations, subsequent calls after the initial one will execute more quickly due to this caching. However, in real-world production scenarios, it is uncommon to perform JSON encode-decode repeatedly on the exact same string object; even identical strings are unlikely to be the same object instance. To achieve benchmark results that better reflect practical use cases, we employ `ssrjson.run_unicode_accumulate_benchmark` and `benchmark_invalidate_dump_cache` functions, which ensure that new `PyUnicodeObject`s are different for each input every time. (ref: [orjson#586](https://github.com/ijl/orjson/issues/586))
+* This repository conducts benchmarking using json, [orjson](https://github.com/ijl/orjson), [ujson](https://github.com/ultrajson/ultrajson), and [ssrJSON](https://github.com/Antares0982/ssrjson). The `dumps` benchmark produces str objects, comparing three operations: `json.dumps`, `orjson.dumps` followed by decode, and `ssrjson.dumps`. The `dumps_to_bytes` benchmark produces bytes objects, comparing three functions: `json.dumps` followed by encode, `orjson.dumps`, and `ssrjson.dumps_to_bytes`.
+* When orjson handles non-ASCII strings, if the cache of the `PyUnicodeObject`’s UTF-8 representation does not exist, it invokes the `PyUnicode_AsUTF8AndSize` function to obtain the UTF-8 encoding. This function then caches the UTF-8 representation within the `PyUnicodeObject`. If the same `PyUnicodeObject` undergoes repeated encode-decode operations, subsequent calls after the initial one will execute more quickly due to this caching. However, in real-world production scenarios, it is uncommon to perform JSON encode-decode repeatedly on the exact same string object; even identical strings are unlikely to be the same object instance. To achieve benchmark results that better reflect practical use cases, we employ `ssrjson.run_unicode_accumulate_benchmark` and `_benchmark_invalidate_dump_cache` functions, which ensure that new `PyUnicodeObject`s are different for each input every time. (ref: [orjson#586](https://github.com/ijl/orjson/issues/586))
 * The performance of JSON encoding is primarily constrained by the speed of writing to the buffer, whereas decoding performance is mainly limited by the frequent invocation of CPython interfaces for object creation. During decoding, both ssrJSON and orjson employ short key caching to reduce the number of object creations, and this caching mechanism is global in both cases. As a result, decoding benchmark tests may not accurately reflect the conditions encountered in real-world production environments.
 * The files simple_object.json and simple_object_zh.json do not represent real-world data; they are solely used to compare the performance of the fast path. Therefore, the benchmark results should not be interpreted as indicative of actual performance.

{ssrjson_benchmark-0.0.2 → ssrjson_benchmark-0.0.4}/README.md

@@ -34,7 +34,7 @@ python -m ssrjson_benchmark
 
 ## Notes
 
-* This repository conducts benchmarking using json, orjson, and ssrJSON. The `dumps` benchmark produces str objects, comparing three operations: `json.dumps`, `orjson.dumps` followed by decode, and `ssrjson.dumps`. The `dumps_to_bytes` benchmark produces bytes objects, comparing three functions: `json.dumps` followed by encode, `orjson.dumps`, and `ssrjson.dumps_to_bytes`.
-* When orjson handles non-ASCII strings, if the cache of the `PyUnicodeObject`’s UTF-8 representation does not exist, it invokes the `PyUnicode_AsUTF8AndSize` function to obtain the UTF-8 encoding. This function then caches the UTF-8 representation within the `PyUnicodeObject`. If the same `PyUnicodeObject` undergoes repeated encode-decode operations, subsequent calls after the initial one will execute more quickly due to this caching. However, in real-world production scenarios, it is uncommon to perform JSON encode-decode repeatedly on the exact same string object; even identical strings are unlikely to be the same object instance. To achieve benchmark results that better reflect practical use cases, we employ `ssrjson.run_unicode_accumulate_benchmark` and `benchmark_invalidate_dump_cache` functions, which ensure that new `PyUnicodeObject`s are different for each input every time. (ref: [orjson#586](https://github.com/ijl/orjson/issues/586))
+* This repository conducts benchmarking using json, [orjson](https://github.com/ijl/orjson), [ujson](https://github.com/ultrajson/ultrajson), and [ssrJSON](https://github.com/Antares0982/ssrjson). The `dumps` benchmark produces str objects, comparing three operations: `json.dumps`, `orjson.dumps` followed by decode, and `ssrjson.dumps`. The `dumps_to_bytes` benchmark produces bytes objects, comparing three functions: `json.dumps` followed by encode, `orjson.dumps`, and `ssrjson.dumps_to_bytes`.
+* When orjson handles non-ASCII strings, if the cache of the `PyUnicodeObject`’s UTF-8 representation does not exist, it invokes the `PyUnicode_AsUTF8AndSize` function to obtain the UTF-8 encoding. This function then caches the UTF-8 representation within the `PyUnicodeObject`. If the same `PyUnicodeObject` undergoes repeated encode-decode operations, subsequent calls after the initial one will execute more quickly due to this caching. However, in real-world production scenarios, it is uncommon to perform JSON encode-decode repeatedly on the exact same string object; even identical strings are unlikely to be the same object instance. To achieve benchmark results that better reflect practical use cases, we employ `ssrjson.run_unicode_accumulate_benchmark` and `_benchmark_invalidate_dump_cache` functions, which ensure that new `PyUnicodeObject`s are different for each input every time. (ref: [orjson#586](https://github.com/ijl/orjson/issues/586))
 * The performance of JSON encoding is primarily constrained by the speed of writing to the buffer, whereas decoding performance is mainly limited by the frequent invocation of CPython interfaces for object creation. During decoding, both ssrJSON and orjson employ short key caching to reduce the number of object creations, and this caching mechanism is global in both cases. As a result, decoding benchmark tests may not accurately reflect the conditions encountered in real-world production environments.
 * The files simple_object.json and simple_object_zh.json do not represent real-world data; they are solely used to compare the performance of the fast path. Therefore, the benchmark results should not be interpreted as indicative of actual performance.

{ssrjson_benchmark-0.0.2 → ssrjson_benchmark-0.0.4}/pyproject.toml

@@ -5,7 +5,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "ssrjson-benchmark"
-version = "0.0.2"
+version = "0.0.4"
 authors = [
     { name = "Eritque Arcus", email = "eritque-arcus@ikuyo.dev" },
     { name = "Antares", email = "antares0982@gmail.com" },
@@ -27,7 +27,7 @@ classifiers = [
 # license-files = ["LICENSE"]
 # setuptools < 77 but deprecated
 license = { file = "LICENSE" }
-dependencies = ["ssrjson", "orjson", "matplotlib"]
+dependencies = ["ssrjson", "orjson", "ujson", "matplotlib"]
 
 [project.optional-dependencies]
 pdf = ["svglib", "reportlab"]

{ssrjson_benchmark-0.0.2 → ssrjson_benchmark-0.0.4}/setup.py

@@ -51,7 +51,7 @@ setup(
     ext_modules=[
         Extension(
             "_ssrjson_benchmark",
-            sources=["src/benchmark.c"],
+            sources=["src/_ssrjson_benchmark.c"],
             language="c",
         )
     ],
@@ -62,18 +62,6 @@ setup(
         "ssrjson_benchmark._files": ["*.json"],
     },
     include_package_data=True,
-    install_requires=[
-        "ssrjson",
-        "orjson",
-        "matplotlib",
-    ],
-    extras_require={
-        "all": [
-            "svglib",
-            "reportlab",
-            "py-cpuinfo",
-        ],
-    },
     cmdclass={
         "build_ext": CMakeBuild,
     },

ssrjson_benchmark-0.0.2/src/benchmark.c → ssrjson_benchmark-0.0.4/src/_ssrjson_benchmark.c

@@ -76,15 +76,15 @@ usize perf_counter(void) {
 
 #endif
 
-typedef struct PyUnicodeNewCallArg {
+typedef struct PyUnicodeCopyInfo {
     Py_ssize_t size;
     int kind;
     Py_UCS4 max_char;
     bool valid;
-} PyUnicodeNewCallArg;
+} PyUnicodeCopyInfo;
 
-PyObject *_copy_unicode(PyObject *unicode, PyUnicodeNewCallArg *call_arg) {
-    if (!call_arg->valid) {
+PyObject *_copy_unicode(PyObject *unicode, PyUnicodeCopyInfo *unicode_copy_info) {
+    if (!unicode_copy_info->valid) {
         // create copy of unicode object.
         int kind = PyUnicode_KIND(unicode);
         Py_UCS4 max_char;
@@ -98,15 +98,16 @@ PyObject *_copy_unicode(PyObject *unicode, PyUnicodeNewCallArg *call_arg) {
             max_char = 0xff;
         }
         //
-        call_arg->size = PyUnicode_GET_LENGTH(unicode);
-        call_arg->kind = kind;
-        call_arg->max_char = max_char;
-        call_arg->valid = true;
+        unicode_copy_info->size = PyUnicode_GET_LENGTH(unicode);
+        unicode_copy_info->kind = kind;
+        unicode_copy_info->max_char = max_char;
+        unicode_copy_info->valid = true;
     }
 
-    PyObject *unicode_copy = PyUnicode_New(call_arg->size, call_arg->max_char);
+    PyObject *unicode_copy = PyUnicode_New(unicode_copy_info->size, unicode_copy_info->max_char);
+    if (!unicode_copy) return NULL;
     memcpy(PyUnicode_DATA(unicode_copy), PyUnicode_DATA(unicode),
-           call_arg->size * call_arg->kind);
+           unicode_copy_info->size * unicode_copy_info->kind);
     return unicode_copy;
 }
 
@@ -129,61 +130,33 @@ PyObject *_parse_additional_args(PyObject *additional_args) {
     return new_args;
 }
 
-PyObject *run_unicode_accumulate_benchmark(PyObject *self, PyObject *args,
-                                           PyObject *kwargs) {
-    PyObject *callable;
-    usize repeat;
-    PyObject *unicode;
-    PyObject *additional_args = NULL;
-    static const char *kwlist[] = {"func", "repeat", "unicode", "args", NULL};
-    if (!PyArg_ParseTupleAndKeywords(args, kwargs, "OKO|O", (char **)kwlist,
-                                     &callable, &repeat, &unicode,
-                                     &additional_args)) {
+PyObject *copy_unicode_list_invalidate_cache(PyObject *self, PyObject *args, PyObject *kwargs) {
+    static const char *kwlist[] = {"s", "size", NULL};
+    PyObject *s;
+    usize size;
+    if (!PyArg_ParseTupleAndKeywords(args, kwargs, "OK", (char **)kwlist, &s, &size)) {
         PyErr_SetString(PyExc_TypeError, "Invalid argument");
-        goto fail;
+        return NULL;
     }
-    if (!PyCallable_Check(callable)) {
-        PyErr_SetString(PyExc_TypeError, "First argument must be callable");
-        goto fail;
+    if (!PyUnicode_CheckExact(s)) {
+        PyErr_SetString(PyExc_TypeError, "First argument must be str, not other types or subclass of str");
+        return NULL;
     }
-    if (!PyUnicode_Check(unicode)) {
-        PyErr_SetString(PyExc_TypeError, "Third argument must be unicode");
-        goto fail;
+    PyObject *ret = PyList_New(size);
+    if (!ret) {
+        return NULL;
     }
-    PyUnicodeNewCallArg call_arg;
-    call_arg.valid = false;
-    usize total = 0;
-    for (usize i = 0; i < repeat; i++) {
-        // create copy of unicode object.
-        PyObject *new_args = _parse_additional_args(additional_args);
-        if (!new_args)
-            goto fail;
-        PyObject *unicode_copy = _copy_unicode(unicode, &call_arg);
-        if (!unicode_copy) {
-            Py_DECREF(new_args);
-            goto fail;
-        }
-        PyTuple_SET_ITEM(new_args, 0, unicode_copy);
-        usize start = perf_counter();
-        PyObject *result = PyObject_Call(callable, new_args, NULL);
-        usize end = perf_counter();
-        assert(unicode_copy->ob_refcnt == 1);
-        Py_DECREF(new_args);
-        unicode_copy = NULL;
-        new_args = NULL;
-        if (unlikely(!result)) {
-            if (!PyErr_Occurred()) {
-                PyErr_SetString(PyExc_RuntimeError, "Failed to call callable");
-            }
-            goto fail;
-        } else {
-            Py_DECREF(result);
+    PyUnicodeCopyInfo unicode_copy_info;
+    unicode_copy_info.valid = false;
+    for (usize i = 0; i < size; i++) {
+        PyObject *s_copy = _copy_unicode(s, &unicode_copy_info);
+        if (!s_copy) {
+            Py_DECREF(ret);
+            return NULL;
         }
-        total += end - start;
+        PyList_SET_ITEM(ret, i, s_copy);
     }
-    return PyLong_FromUnsignedLongLong(total);
-fail:;
-    return NULL;
+    return ret;
 }
 
 PyObject *run_object_accumulate_benchmark(PyObject *self, PyObject *args,
@@ -197,6 +170,7 @@ PyObject *run_object_accumulate_benchmark(PyObject *self, PyObject *args,
         PyErr_SetString(PyExc_TypeError, "Invalid argument");
         goto fail;
     }
+    //
     if (!PyCallable_Check(callable)) {
         PyErr_SetString(PyExc_TypeError, "First argument must be callable");
         goto fail;
@@ -205,6 +179,7 @@ PyObject *run_object_accumulate_benchmark(PyObject *self, PyObject *args,
         PyErr_SetString(PyExc_TypeError, "Third argument must be tuple");
         goto fail;
     }
+    //
     usize total = 0;
     for (usize i = 0; i < repeat; i++) {
         usize start = perf_counter();
@@ -308,7 +283,7 @@ fail:;
 }
 
 static PyMethodDef ssrjson_benchmark_methods[] = {
-        {"run_unicode_accumulate_benchmark", (PyCFunction)run_unicode_accumulate_benchmark, METH_VARARGS | METH_KEYWORDS, "Benchmark."},
+        {"copy_unicode_list_invalidate_cache", (PyCFunction)copy_unicode_list_invalidate_cache, METH_VARARGS | METH_KEYWORDS, "Copy unicode list invalidate cache."},
         {"run_object_accumulate_benchmark", (PyCFunction)run_object_accumulate_benchmark, METH_VARARGS | METH_KEYWORDS, "Benchmark."},
         {"run_object_benchmark", (PyCFunction)run_object_benchmark, METH_VARARGS | METH_KEYWORDS, "Benchmark."},
         {"inspect_pyunicode", (PyCFunction)inspect_pyunicode, METH_VARARGS | METH_KEYWORDS, "Inspect PyUnicode."},
@@ -317,7 +292,7 @@ static PyMethodDef ssrjson_benchmark_methods[] = {
 
 static struct PyModuleDef moduledef = {
         PyModuleDef_HEAD_INIT,
-        "_ssrjson_benchmark",       /* m_name */
+        "_ssrjson_benchmark",      /* m_name */
         0,                         /* m_doc */
         0,                         /* m_size */
         ssrjson_benchmark_methods, /* m_methods */

ssrjson_benchmark-0.0.4/src/ssrjson_benchmark/__init__.py

@@ -0,0 +1,19 @@
+from .benchmark_impl import (
+    generate_report_pdf,
+    generate_report_markdown,
+    run_benchmark,
+)
+
+try:
+    from importlib.metadata import version
+
+    __version__ = version("ssrjson-benchmark")
+except Exception:
+    __version__ = "0.0.0"
+
+__all__ = [
+    "run_benchmark",
+    "generate_report_markdown",
+    "generate_report_pdf",
+    "__version__",
+]

ssrjson_benchmark-0.0.4/src/ssrjson_benchmark/__main__.py

@@ -0,0 +1,83 @@
+def main():
+    import argparse
+    import json
+    import os
+    import pathlib
+    import sys
+
+    from .benchmark_impl import (
+        generate_report_markdown,
+        generate_report_pdf,
+        parse_file_result,
+        run_benchmark,
+    )
+
+    parser = argparse.ArgumentParser()
+
+    parser.add_argument(
+        "-f", "--file", help="record JSON file", required=False, default=None
+    )
+    parser.add_argument(
+        "-d",
+        "--in-dir",
+        help="Benchmark JSON files directory",
+        required=False,
+    )
+    parser.add_argument(
+        "-m",
+        "--markdown",
+        help="Generate Markdown report",
+        required=False,
+        action="store_true",
+    )
+    parser.add_argument(
+        "--no-pdf",
+        help="Don't generate PDF report",
+        required=False,
+        action="store_true",
+    )
+    parser.add_argument(
+        "--process-bytes",
+        help="Total process bytes per test, default 1e8",
+        required=False,
+        default=1e8,
+        type=int,
+    )
+    parser.add_argument(
+        "--out-dir",
+        help="Output directory for reports",
+        required=False,
+        default=os.getcwd(),
+    )
+    args = parser.parse_args()
+    if args.file and args.no_pdf and not args.markdown:
+        print("Nothing to do.")
+        sys.exit(0)
+
+    _benchmark_files_dir = args.in_dir
+    if not _benchmark_files_dir:
+        _benchmark_files_dir = os.path.join(
+            os.path.dirname(os.path.abspath(__file__)), "_files"
+        )
+    benchmark_files_dir = sorted(pathlib.Path(_benchmark_files_dir).glob("*.json"))
+    if not benchmark_files_dir:
+        print(f"No benchmark file found using given path: {_benchmark_files_dir}")
+        sys.exit(0)
+
+    if args.file:
+        with open(args.file, "rb") as f:
+            result_ = json.load(f)
+        result = parse_file_result(result_)
+        file = args.file.split("/")[-1]
+    else:
+        result, file = run_benchmark(benchmark_files_dir, args.process_bytes)
+        file = file.split("/")[-1]
+
+    if args.markdown:
+        generate_report_markdown(result, file, args.out_dir)
+    if not args.no_pdf:
+        generate_report_pdf(result, file, args.out_dir)
+
+
+if __name__ == "__main__":
+    main()