Performance¶
python-rapidjson tries to be as performant as possible while staying compatible with
the json module.
Tables¶
The following tables show a comparison between this module and other libraries with different data sets. Last row (“overall”) is the total time taken by all the benchmarks.
Each number shows the factor between the time taken by each contender and
python-rapidjson (in other words, they are normalized against a value of 1.0 for
python-rapidjson): the lower the number, the speedier the contender.
In bold the winner.
Serialization¶
serialize |
|
|
|
|
simdjson5 |
orjson6 |
ujson7 |
simplejson8 |
stdlib9 |
|---|---|---|---|---|---|---|---|---|---|
100 arrays dict |
1.00 |
0.99 |
0.77 |
0.76 |
1.97 |
0.26 |
1.09 |
4.00 |
1.95 |
100 dicts array |
1.00 |
0.99 |
0.81 |
0.76 |
2.06 |
0.34 |
1.06 |
4.96 |
2.04 |
256 Trues array |
1.00 |
1.00 |
1.17 |
1.04 |
2.38 |
0.38 |
1.52 |
2.70 |
2.35 |
256 ascii array |
1.00 |
1.00 |
1.02 |
1.00 |
0.90 |
0.26 |
0.50 |
1.06 |
0.89 |
256 doubles array |
1.00 |
0.99 |
0.99 |
0.99 |
0.84 |
0.06 |
0.21 |
0.88 |
0.83 |
256 unicode array |
1.00 |
0.81 |
0.81 |
0.82 |
0.70 |
0.09 |
0.51 |
0.65 |
0.69 |
apache.json |
1.00 |
1.00 |
1.01 |
1.01 |
1.66 |
0.33 |
1.18 |
2.51 |
1.65 |
canada.json |
1.00 |
1.00 |
0.99 |
1.00 |
1.06 |
0.09 |
0.34 |
1.72 |
1.06 |
complex object |
1.00 |
1.00 |
0.93 |
0.92 |
1.58 |
0.26 |
0.83 |
2.41 |
1.57 |
composite object |
1.00 |
0.99 |
0.75 |
0.72 |
1.91 |
0.36 |
1.12 |
2.44 |
1.87 |
ctm.json |
1.00 |
1.00 |
0.76 |
0.76 |
2.11 |
0.33 |
1.43 |
4.81 |
2.13 |
github.json |
1.00 |
1.00 |
0.98 |
0.97 |
1.34 |
0.30 |
0.98 |
1.72 |
1.32 |
instruments.json |
1.00 |
1.01 |
0.80 |
0.80 |
1.75 |
0.34 |
1.15 |
2.28 |
1.75 |
mesh.json |
1.00 |
0.99 |
0.90 |
0.90 |
0.94 |
0.14 |
0.35 |
1.09 |
0.93 |
truenull.json |
1.00 |
1.02 |
1.03 |
1.01 |
1.87 |
0.41 |
1.69 |
1.89 |
1.86 |
tweet.json |
1.00 |
0.99 |
0.99 |
0.94 |
1.66 |
0.33 |
1.02 |
2.33 |
1.65 |
twitter.json |
1.00 |
1.00 |
0.96 |
0.95 |
1.39 |
0.32 |
1.05 |
1.65 |
1.36 |
overall |
1.00 |
0.99 |
0.83 |
0.82 |
1.71 |
0.22 |
0.90 |
3.33 |
1.70 |
Deserialization¶
deserialize |
|
|
|
|
simdjson |
orjson |
ujson |
simplejson |
stdlib |
|---|---|---|---|---|---|---|---|---|---|
100 arrays dict |
1.00 |
1.05 |
0.90 |
0.95 |
0.86 |
0.56 |
0.86 |
1.07 |
0.97 |
100 dicts array |
1.00 |
0.99 |
0.87 |
0.85 |
0.63 |
0.40 |
0.68 |
1.27 |
1.02 |
256 Trues array |
1.00 |
1.02 |
1.11 |
0.99 |
0.92 |
0.48 |
0.83 |
1.51 |
1.36 |
256 ascii array |
1.00 |
0.99 |
1.02 |
1.01 |
0.46 |
0.45 |
0.82 |
0.84 |
0.81 |
256 doubles array |
1.00 |
1.00 |
0.27 |
0.26 |
0.21 |
0.16 |
0.54 |
1.09 |
1.07 |
256 unicode array |
1.00 |
1.00 |
1.00 |
1.00 |
1.10 |
0.39 |
0.80 |
4.49 |
1.29 |
apache.json |
1.00 |
1.00 |
1.01 |
1.01 |
0.65 |
0.48 |
0.85 |
0.81 |
0.83 |
canada.json |
1.00 |
1.00 |
0.34 |
0.34 |
0.29 |
0.25 |
0.49 |
1.04 |
0.96 |
complex object |
1.00 |
1.00 |
0.84 |
0.84 |
0.65 |
0.37 |
0.80 |
1.15 |
1.02 |
composite object |
1.00 |
0.99 |
0.83 |
0.81 |
0.61 |
0.45 |
0.59 |
1.37 |
1.07 |
ctm.json |
1.00 |
1.01 |
0.92 |
0.92 |
0.74 |
0.55 |
1.08 |
1.20 |
1.14 |
github.json |
1.00 |
1.00 |
0.98 |
0.98 |
0.62 |
0.41 |
0.76 |
0.86 |
0.79 |
instruments.json |
1.00 |
1.00 |
0.91 |
0.90 |
0.63 |
0.37 |
0.72 |
1.12 |
0.92 |
mesh.json |
1.00 |
1.01 |
0.51 |
0.51 |
0.44 |
0.36 |
0.63 |
1.44 |
1.03 |
truenull.json |
1.00 |
0.99 |
1.00 |
1.00 |
0.58 |
0.39 |
0.84 |
0.98 |
0.87 |
tweet.json |
1.00 |
1.02 |
0.99 |
0.98 |
0.66 |
0.44 |
0.84 |
1.14 |
1.03 |
twitter.json |
1.00 |
1.00 |
0.99 |
0.99 |
0.65 |
0.42 |
0.86 |
1.01 |
1.00 |
overall |
1.00 |
1.05 |
0.85 |
0.90 |
0.80 |
0.53 |
0.83 |
1.08 |
0.97 |
ASCII vs UTF-8 Serialization¶
serialize |
|
|
|
|
|
|
|
|
|---|---|---|---|---|---|---|---|---|
Long ASCII string |
1.00 |
0.46 |
0.25 |
0.41 |
0.63 |
0.93 |
0.57 |
1.49 |
Long Unicode string |
1.00 |
0.62 |
0.63 |
0.58 |
0.71 |
0.70 |
0.80 |
0.66 |
overall |
1.00 |
0.57 |
0.51 |
0.53 |
0.69 |
0.78 |
0.73 |
0.92 |
- 1
rapidjson.dumps()- 2
rapidjson.Encoder()- 3
rapidjson.dumps(number_mode=NM_NATIVE)- 4
rapidjson.Encoder(number_mode=NM_NATIVE)- 5
- 6
- 7
- 8
- 9
Python 3.10.5 standard library
json- 10
rapidjson.loads()- 11
rapidjson.Decoder()- 12
rapidjson.loads(number_mode=NM_NATIVE)- 13
rapidjson.Decoder(number_mode=NM_NATIVE)- 15
rapidjson.dumps(ensure_ascii=True)- 16
rapidjson.dumps(ensure_ascii=False)- 17
ujson.dumps(ensure_ascii=True)- 18
ujson.dumps(ensure_ascii=False)- 19
simplejson.dumps(ensure_ascii=True)- 20
simplejson.dumps(ensure_ascii=False)- 21
stdlib json.dumps(ensure_ascii=True)- 22
stdlib json.dumps(ensure_ascii=False)
DIY¶
To run these tests yourself, clone the repo and run:
$ make benchmarks
to focus only on RapidJSON or:
$ make benchmarks-others
to get full comparison against other engines.
To reproduce the tables above, run:
$ make benchmarks-tables
Compare different versions¶
pytest-benchmark implements an handy feature that allows you to weight the impact of a particular change. For example, you may start from a released version and execute:
$ make benchmarks PYTEST_OPTIONS=--benchmark-autosave
After applying whatever change to the code base, you can get a differential view by executing:
$ make benchmarks PYTEST_OPTIONS=--benchmark-compare=0001