pbwhy

Testing Google's protobuf solution vs betterproto

protobuf cpp

https://github.com/protocolbuffers/protobuf

cd protobuf_cpp
pipenv install --dev
[ pipenv run codegen ]
pipenv run all

=>

protobuf_cpp serialization [10000]: 145.17 ms
protobuf_cpp deserialization [10000]: 36.95 ms

CPU: Intel64 Family 6 Model 142 Stepping 10, GenuineIntel, 8 cpu, 2.0 GHz
OS: win-amd64
Python: 3.8.10, CPython, 3d8993a

protobuf python

cd protobuf_python
pipenv install --dev
[ pipenv run codegen ]
pipenv run all

=>

Loading .env environment variables...
protobuf_python serialization [10000]: 590.71 ms
protobuf_python deserialization [10000]: 348.48 ms

CPU: Intel64 Family 6 Model 142 Stepping 10, GenuineIntel, 8 cpu, 2.0 GHz
OS: win-amd64
Python: 3.8.10, CPython, 3d8993a

betterproto

https://github.com/danielgtaylor/python-betterproto

cd betterproto
pipenv install --dev
[ pipenv run codegen ]
pipenv run all

=>

betterproto serialization [10000]: 613.47 ms
betterproto deserialization [10000]: 962.49 ms

CPU: Intel64 Family 6 Model 142 Stepping 10, GenuineIntel, 8 cpu, 2.0 GHz
OS: win-amd64
Python: 3.8.10, CPython, 3d8993a

Benchmark execution with profiling

Run both serialization & deserialization benchmark optionally with CPU and memory profiling:

pipenv run all [ -c, --cpu_profile ] [ -m, --memory_profile ]

Run both serialization benchmark optionally with CPU and memory profiling:

pipenv run serialization [ -c, --cpu_profile ] [ -m, --memory_profile ]

Run both deserialization benchmark optionally with CPU and memory profiling:

pipenv run deserialization [ -c, --cpu_profile ] [ -m, --memory_profile ]