Say No To Poetry

Did you grow up using just pip or conda? Are you being forced to use poetry everywhere at work? They try to convince you poetry just works? You spiraled down into the conclusion that poetry is better than pip?

Its dependency resolution is shit slow. If one of the package fails to install (yes looking at you uwsgi), poetry starts screaming like kid (printing out errors, which requires additional sheets), further won't install any of the other packages.

poetry is good only on good days, it is horrible on bad days. days where I need to conda install -c conda-forge uwsgi just for uwsgi. and since poetry gave up at uwsgi, I need to manually install other dependencies.

So?

Use this binary instead,

say_no_to_poetry pyproject.toml

It converts your pyproject.toml into a requirements.txt.

Example usage:

~/demo-sntp$ ls
hypothetical.pyproject.toml  say_no_to_poetry

~/demo-sntp$ cat hypothetical.pyproject.toml 
[tool.poetry]
name = "hypothetical"
version = "0.5.2"
description = "hypothetical utilities"
authors = []

[tool.poetry.dependencies]
python = "^3.8"
docopt = "^0.6.2"
scikit-learn = "^0.24.2"
stanza = { version = "^1.2", optional = true }
levenshtein = "^0.12.0"
tqdm = "^4.61.0"
pandas = "^1.3.1"
pydash = "^5.0.2"
furo = "^2021.4.11-beta.34"
pytz = "^2020.4"

[tool.poetry.extras]
asr = ["stanza"]

[tool.poetry.dev-dependencies]
pytest = "^6.2.4"
mypy = "^0.812"
jupyter = "^1.0.0"
black = "^20.8b1"


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

~/demo-sntp$ ./say_no_to_poetry hypothetical.pyproject.toml 

~/demo-sntp$ ls
hypothetical.pyproject.toml  requirements.txt  say_no_to_poetry

~/demo-sntp$ cat requirements.txt 
black>=20.8b1
scikit-learn>=0.24.2
furo>=2021.4.11-beta.34
pytz>=2020.4
mypy>=0.812
pandas>=1.3.1
stanza>=1.2
tqdm>=4.61.0
pytest>=6.2.4
jupyter>=1.0.0
levenshtein>=0.12.0
docopt>=0.6.2
pydash>=5.0.2

Caveats

Only after finishing this code, I realized there is an export support in poetry itself.

So if you already have poetry globally/any-venv you can directly use it.

Also regarding this binary itself, complete syntax of how version can be defined is here for poetry. the binary handles only the ^ and exact ones. not the ~, * ones. Also doesn't set the upper bound.

So is this useless?

Yeah whatever. It works, but pointless at this point of time. If you don't like poetry, don't have it in your system. use this binary, if you frequently need requirements.txt for setting up stuff.

It also helps in a situation where you want to use somebody's repo in your particular project/random environment. And if they happen to have pyproject.toml, convert it into requirements.txt, you can install dependencies without going through poetry, poetry install and poetry managing its own virtual environment for that repo.

tldr: you don't want to use poetry, happy with pip. Use this.

Learnings

• my very first rust project.
• I should google more before starting a project. re-search.
• It could have been easier if I had taken time, to write down the steps and functions that'll make the project complete. And think if that's the most efficient way to do it. I ended up writing regex based file extraction (with look-aheads), to extract version for all possible version definitions. Without realzing it is just an ordinary .toml file. Later used the toml parser crate to do most of the hardwork.
• lazy_static is interesting, to avoid compiling regex patter several times, it is just done once during later stages while running. Therefore if there is a need to compile same pattern repeatedly, one can use this to save up runtime speed.
• I tried to write code such that each function is being told what to do directly, and given only the relevant information. no additional useless stuff.