A high performance JSON library written in ANSI C.
- Fast: can read or write gigabytes per second JSON data on modern CPU.
- Portable: compliant with ANSI C (C89).
- Standard: compliant with RFC 8259 and ECMA-404 standard.
- Safe: complete JSON form, number format and UTF-8 validation.
- Accuracy: can process
int64,uint64anddoublenumber accurately. - No Limit: support large data size, unlimited JSON level,
\u0000string. - Extendable: support comments, trailing commas, nan/inf, custom memory allocator.
- Developer Friendly: only one
hand onecfile, easy to use API.
Benchmark project and dataset: yyjson_benchmark
| twitter.json | parse (GB/s) | stringify (GB/s) |
|---|---|---|
| yyjson(insitu) | 1.82 | 1.51 |
| yyjson | 1.74 | 1.46 |
| simdjson | 1.52 | 0.60 |
| sajson | 1.17 | |
| rapidjson(insitu) | 0.78 | |
| rapidjson(utf8) | 0.26 | 0.39 |
| cjson | 0.31 | 0.16 |
| jansson | 0.04 | 0.09 |
| twitter.json | parse (GB/s) | stringify (GB/s) |
|---|---|---|
| yyjson(insitu) | 3.50 | 2.43 |
| yyjson | 2.41 | 1.97 |
| simdjson | 2.23 | 0.81 |
| sajson | 1.76 | |
| rapidjson(insitu) | 0.76 | |
| rapidjson(utf8) | 0.30 | 0.58 |
| cjson | 0.48 | 0.33 |
| jansson | 0.09 | 0.23 |
More benchmark reports with interactive charts (update 2020-11-15)
| Platform | CPU | Compiler | OS | Report |
|---|---|---|---|---|
| Intel NUC 8i5 | Core i5-8259U | msvc 2019 | Windows 10 2004 | Charts |
| Intel NUC 8i5 | Core i5-8259U | clang 10 | Ubuntu 20.04 | Charts |
| Intel NUC 8i5 | Core i5-8259U | gcc 10.2 | Ubuntu 20.04 | Charts |
| Intel NUC 8i5 | Core i5-8259U | gcc 9.3 | Ubuntu 20.04 | Charts |
| AWS EC2 c5a.large | AMD EPYC 7002 | gcc 9.3 | Ubuntu 20.04 | Charts |
| AWS EC2 t4g.medium | Graviton2 (ARM64) | gcc 9.3 | Ubuntu 20.04 | Charts |
| Apple iPhone 12 Pro | A14 (ARM64) | clang 12 | iOS 14 | Charts |
- A modern processor with:
- high instruction level parallelism
- excellent branch predictor
- low penalty for misaligned memory access
- A modern compiler with good optimization.
Just copy yyjson.h and yyjson.c to your project and start using it.
Since yyjson is ANSI C compatible, no other configuration is needed typically.
We have tested yyjson with the following compilers: gcc, clang, msvc, icc, tcc. If you find a compile error, please report a bug.
Clone repository and create build directory:
git clone https://github.com/ibireme/yyjson.git
mkdir build
cd buildBuild static library:
cmake ..
cmake --build .Build static library and run tests:
cmake .. -DYYJSON_BUILD_TESTS=ON
cmake --build .
ctestBuild shared library:
cmake .. -DBUILD_SHARED_LIBS=ON
cmake --build .Supported CMake options:
-DYYJSON_BUILD_TESTS=ONBuild all tests.-DYYJSON_ENABLE_COVERAGE=ONEnable code coverage for tests.-DYYJSON_ENABLE_VALGRIND=ONEnable valgrind memory checker for tests.-DYYJSON_ENABLE_SANITIZE=ONEnable sanitizer for tests.-DYYJSON_DISABLE_READER=ONDisable JSON reader if you don't need it.-DYYJSON_DISABLE_WRITER=ONDisable JSON writer if you don't need it.-DYYJSON_DISABLE_FP_READER=ONDisable custom double number reader to reduce binary size.-DYYJSON_DISABLE_FP_WRITER=ONDisable custom double number writer to reduce binary size.-DYYJSON_DISABLE_COMMENT_READER=ONDisable non-standard comment support at compile time.-DYYJSON_DISABLE_INF_AND_NAN_READER=ONDisable non-standard nan/inf support at compile time.
const char *json = "{\"name\":\"Mash\",\"star\":4,\"hits\":[2,2,1,3]}";
yyjson_doc *doc = yyjson_read(json, strlen(json), 0);
yyjson_val *root = yyjson_doc_get_root(doc);
yyjson_val *name = yyjson_obj_get(root, "name");
printf("name: %s\n", yyjson_get_str(name));
yyjson_val *star = yyjson_obj_get(root, "star");
printf("star: %d\n", (int)yyjson_get_int(star));
yyjson_val *hits = yyjson_obj_get(root, "hits");
size_t idx, max;
yyjson_val *hit;
yyjson_arr_foreach(hits, idx, max, hit) {
printf("hit%d: %d\n", (int)idx, (int)yyjson_get_int(hit));
}
yyjson_doc_free(doc);yyjson_mut_doc *doc = yyjson_mut_doc_new(NULL);
yyjson_mut_val *root = yyjson_mut_obj(doc);
yyjson_mut_doc_set_root(doc, root);
yyjson_mut_obj_add_str(doc, root, "name", "Mash");
yyjson_mut_obj_add_int(doc, root, "star", 4);
int hits_arr[] = {2, 2, 1, 3};
yyjson_mut_val *hits = yyjson_mut_arr_with_sint32(doc, hits_arr, 4);
yyjson_mut_obj_add_val(doc, root, "hits", hits);
const char *json = yyjson_mut_write(doc, 0, NULL);
if (json) {
printf("json: %s\n", json); // {"name":"Mash","star":4,"hits":[2,2,1,3]}
free((void *)json);
}
yyjson_mut_doc_free(doc);yyjson_read_err err;
yyjson_read_flag flg = YYJSON_READ_ALLOW_COMMENTS | YYJSON_READ_ALLOW_TRAILING_COMMAS;
yyjson_doc *doc = yyjson_read_file("/tmp/test.json", flg, NULL, &err);
if (doc) {
yyjson_val *obj = yyjson_doc_get_root(doc);
yyjson_obj_iter iter;
yyjson_obj_iter_init(obj, &iter);
yyjson_val *key, *val;
while ((key = yyjson_obj_iter_next(&iter))) {
val = yyjson_obj_iter_get_val(key);
printf("%s: %s\n", yyjson_get_str(key), yyjson_get_type_desc(val));
}
yyjson_doc_free(doc);
} else {
printf("read error: %s at position: %ld\n", err.msg, err.pos);
}yyjson_mut_doc *doc = yyjson_mut_read_file("/tmp/test.json", 0, NULL, NULL);
yyjson_mut_val *obj = yyjson_mut_doc_get_root(doc);
yyjson_mut_obj_iter iter;
yyjson_mut_obj_iter_init(obj, &iter);
yyjson_mut_val *key, *val;
while ((key = yyjson_mut_obj_iter_next(&iter))) {
val = yyjson_mut_obj_iter_get_val(key);
if (yyjson_mut_is_null(val)) {
yyjson_mut_obj_iter_remove(&iter);
}
}
yyjson_write_err err;
yyjson_write_flag flg = YYJSON_WRITE_PRETTY | YYJSON_WRITE_ESCAPE_UNICODE;
yyjson_mut_write_file("/tmp/test.json", doc, flg, NULL, &err);
if (err.code) {
printf("write error: %d, message: %s\n", err.code, err.msg);
}
yyjson_mut_doc_free(doc);- Add documentation page.
- Add GitHub workflow for CI and codecov.
- Add more tests: valgrind, sanitizer.
- Add fuzzer.
- Support JSON Pointer to query value from document.
- Add documentation for performance.
- Optimize performance for 32-bit processor.
- Optimize performance of double number reader/writer with other algorithms [1] [2].
This project is released under the MIT license.