The json entity parser. This is a cython library that can be used from python or cython to do fast parsing of json entities.
Docs? Nada. But you can get started by looking at how the benchmarks are written.
Installation instructions: follow the script in .gitlab-ci.yml.
- every item you parse has the same fields or fields from a small-ish, known set
- object structures aren't extremely deep
- for a given field (array position or dictionary key) the type is usually the same
- not to be confused with SGML (XML, HTML) entities like
&
- loading lots of json items into cython objects (cython objects for memory efficiency and initialization speed)
- when json items have known schema
- search of a single field in a large json file may be significantly faster than other options
- this doesn't validate and there are known bad strings that it will probably parse
- not tested on comprehensive inputs
- error handling isn't perfect; wouldn't be impossible for this to have an infinite loop or to ignore exceptions.
It's marginally faster than ujson in a contrived benchmark but may be significantly faster if your goal is to load data from JSON to cython structs.
Benchmark:
| library | condition | seconds |
|---|---|---|
| json | from cpython | 0.933 |
| ujson | from cpython | 0.165 |
| ujson | from cython | not tested |
| entparse | from cpython | 0.147 |
| entparse | from cython | 0.125 |
Note: this isn't a fair benchmark. entparse isn't doing anything but bounds-checking at the outer level.