Simple instalation without cache

• I have searched the issues of this repo and believe that this is not a duplicate.
• I have searched the documentation and believe that my question is not covered.

Feature Request

TL;DR

I’d like to use pyproject.toml for instaling dependencies for project with minimal impact on size where the docker container provides sufficient isolation. E.G.:

poetry install --no-root --no-cache --no-interaction

Details, blockers and other similar issues

I’d like to use poetry in docker as simple as pip install --no-cache-dir -r requirements.txt (with cleared cache), but have some trouble. I have a local dependency which is wanted to be editable in the same local folder.

...
[tool.poetry.dependencies]
python = "^3.8"
numpy = "^1.19.4"
pandas = "^1.1.4"
jupyterlab = "^2.2.9"

[tool.poetry.dev-dependencies]
# is installed as "python setup.py develop" and can be edited in mypackage folder
mypackage = {path = "./mypackage", develop = true}

Installation in docker:

FROM python:3.8-slim-buster

WORKDIR /lab

RUN pip install poetry --no-cache-dir && \
    poetry config virtualenvs.create false
COPY poetry.lock pyproject.toml .

# my local dependency
COPY ./mypackage ./mypackage

# too big
RUN poetry install --no-root --no-interaction

EXPOSE 9180
CMD ["jupyter", "lab", "--allow-root", "--ip=0.0.0.0", "--port=9180", "."]

I’ve tried also RUN poetry install --no-root --no-interaction && poetry cache clear --no-interaction --all pypi, but size is the same.

I found https://github.com/python-poetry/poetry/issues/1879#issuecomment-592133519, but there is too many rows and dockerfile is being complex.

I’ve also tried https://github.com/python-poetry/poetry/issues/1879#issuecomment-650384931:

RUN poetry export -f requirements.txt --dev --without-hashes | pip install --no-cache-dir -r /dev/stdin

This is fine workaround but its not suitable for mypackage (local package) dependency. Export row of package is mypackage @ /home/vh/test-docker/mypackage; python_version >= "3.8" and python_version < "4.0" and pip searching this package on pypi and fail.

When I remove the local dependency and run previous command without --dev param:

RUN poetry export -f requirements.txt -v --without-hashes | pip install --no-cache-dir -r /dev/stdin && \
         cd mypackage && \
         poetry install --develop mypackage && \
         cd ..

Also crashes with error that --develop param was removed with release poetry v1.0.0. But I found opened issue: https://github.com/python-poetry/poetry/issues/2572.

[cbrochtrup]

Another option to keep the cache out of the final image is to use docker BuildKit’s cache feature. I’ve used it with success.

Example

# syntax=docker/dockerfile:1.2

COPY poetry.lock /
RUN --mount=type=cache,target=/home/.cache/pypoetry/cache \
    --mount=type=cache,target=/home/.cache/pypoetry/artifacts \
    poetry install

This way, docker keeps your poetry cache around for quick builds, layer sharing, etc. but not in the final image.

[sinoroc]

Personal opinion:

I see how it might be a helpful feature. On the other hand I can’t help but think that it should be quite easy to just delete the cache and other build artefacts and what not right after the installation:

RUN poetry install ... && rm -rf ~/.cache/pypoetry/{cache,artifacts}

We might spare us adding complexity into poetry’s code base. But maybe also it is not as simple as that and a --no-cache feature makes sense.