[PYTHON] Recommendation of Poetry

Python packaging mechanism

--wheel: Python package format --Files containing Python code and shared libraries (* .whl) --If it depends on other packages, it is written in the metadata

• PyPI: The Python Package Index --Anyone can upload by logging in --You can also set up your own Index for internal use, etc. --pip: Package installer --The package is searched from PyPI and installed. --The packages that the packages depend on will be installed together. --Basically install on the system (/usr/lib/python3.x/site-package)

pip install numpy

Semantic Versioning

https://semver.org/lang/ja/

Uses X.Y.Z format versioning

1. Major version (X) if API changes are incompatible,
2. If it is backward compatible and functionality is added, the minor version (Y),
3. If you fix a bug with backward compatibility, increase the patch version (Z).

pip install "numpy==1.15"

--List the packages that depend on the file requirements.txt and install them all together.

pip install -r requirements.txt

Develop multiple projects

• Project A
  • tensorflow == 1.15 --It does not work on 2.0 series
• Project B
  • tensorflow == 2.0 --It does not work on 1.0 series

You cannot have multiple versions of a package installed on your system at the same time!

venv (virtualenv)

--I want to install a package for each project --Entered the standard from Python 3.3 --There was a package called virtualenv for Python 2.x

python3 -m venv path/to/new/venv  #Create venv
source path/to/new/venv/bin/activate  #activation

--Venv itself will be managed manually --Typically, create a venv directory directly under the project

pyproject.toml

--Official file format for Python project management - PEP(Python Enhancement Proposals) 518 --Integrate packaging settings setup.py, setup.cfg and list of dependent packages requirements.txt --Supported from pip 19.2 --You can pip install from pypoetry.toml without setup.py --You can create a namespace tool. $ {Toolname} for each tool and write your own settings. --You no longer have to create a separate configuration file for each tool

poetry

--A tool that manages projects based on pyproject.toml --poetry init will generate pyproject.toml ――Creates and manages venv without permission --poetry run [shell script] runs the script under venv --There is also an upload function to PyPI

[tool.poetry]
name = "hello"
version = "0.1.0"
description = "Hello Poetry!"

[tool.poetry.dependencies]
python = "^3.7"   #You can also write a version of Python itself (just check if it meets)
numpy = "1.14"    #I will write packages that depend on here

[tool.poetry.dev-dependencies]

[build-system]
requires = ["poetry>=0.12"]
build-backend = "poetry.masonry.api"