how to avoid mixture of \ and / in file paths when joining paths in Docker containerized Python code

Question

As far as I'm aware I'm using best practices to define paths (using raw strings) and how I go about joining them (using os.path.join()), e.g.

import os

fdir = r'C:\Code\...\samples'

fpath = os.path.join(fdir, 'fname.ext')

and doing so has not caused me any problems when running my code within a Python or command shell. If I print fpath to the console I get consistent use of \s in the path:

C:\Code...\samples\fname.ext

But when I run a Docker containerized version of the code and run the image I get the error:

FileNotFoundError: [Errno 2] No such file or directory: 'C:\Code\...\samples/fname.ext'

I don't understand why os.path.join() has used a / to join fdir and fname.ext when the rest of the path included \\. It doesn't do this when I run the code outside of the container.

I have tried using os.path.normpath():

fpath = os.path.join(fdir, 'fname.ext')
fpath = os.path.normpath(fpath)

as discussed here, and os.sep.join():

fpath = os.sep.join([fdir, 'fname.ext'])

as covered here, and Path().joinpath():

from pathlib import Path

fpath = Path(fdir).joinpath('fname.ext')

as well as Path() / 'path_to_add':

fpath = Path(fdir) / 'fname.ext'

as discussed here, but in every case I end up with the same result using os.path.join().

Can someone please help me to understand what is going on and how to create consistent paths that will work whether I run the code in Python in a Windows environment, or in a Docker container?

Update Nov. 16:

In trying to keep my question brief I think I've left out details that are crucial. Apologies to those who have kindly taken the time to offer suggestions based on my incomplete description of the problem.

My code needs to import/export files from/to directories that are defined within a user-specified configuration file.

So the configuration file has a section of code where the user defines variables and paths, e.g.

samplesDir = r"path-to-samples-directory"

The variables are stored in a dictionary of dictionaris and stored as a .json.

At the start of the code the user defines the key that selects the dictionary of interest so that at various parts in my code when a file needs to be imported/exported, the paths are at hand.

So back to my example, samplesDir is stored in the configuration dictionary, cfgDict, so all I need to do is append the file name:

sampleFpath = os.path.join(sampleDir, sampleFname)

and sampleFname is determined based on other variables.

Because of the dynamic nature of the variables (including directory paths and file paths), I think it rules out the use of static path defined in a .yml with Docker Compose.

Update Nov. 18:

It may help to include a few more details and some screenshots.

The above screenshot shows the file and folder structure of the src directory containing the source code, the main app.py script for command-line use, the Dockerfile, etc.

The configs folder contains JSON files that includes variables, paths to directories and files. The user can create configuration files either by copying an existing one and modifying the entries, or configuration files can be generated by calling config.py.

Within config.py I have pre-set variables and paths, so that the directory path to the configuration files (configs), sample files (sample_DROs) and others (e.g. fiducials) are all within src.

I don't anticipate any reason why the user would want to store the config files anywhere else, nor do I expect them to want to use different sample files (or move them elsewhere). However, they will undoubtedly create their own fiducials and may decide not to store them in the fiducials directory (i.e. somewhere not within the src directory).

Likewise I have pre-set the download directory (based on the parameters stored within the configuration files, files are fetched from a server and downloaded) to be the default Downloads directory:

rootDownloadDir = os.path.join(Path.home(), "Downloads", "xnat_downloads")

Those files are later imported, processed, and the outputs are (by default) exported into sub-directories within rootDownloadDir.

Within Dockerfile I set the working directory of the container to be that of the source code and copy all of the contents of src (with the exception of some directories defined in .dockerignore):

WORKDIR C:/Code/WP1.3_multiple_modalities/src
...
COPY . .

so that the structure of the container mimics that of WORKDIR:

Hence I have allowed for flexibility in import/export directories, and they are by default a combination of paths within and outside of the src directory. And so, the code executed within the container will need to access files both within and outside of src.

That said, I don't know what rootDownloadDir will look like when os.path.join(Path.home(), "Downloads", "xnat_downloads") is run within the container.

This has got me thinking - Is it bad practice to set the download directory outside of src?

Returning to the original error:

the sample file is in the container:

Answer 1

From the actual behavior I can suppose that the container is based on Unix-like image. Path separator is / in such systems. To build an environment-independent path which works inside and outside of the container you need the following steps:

1. Mounting of host folder to container directory.
2. Environment variable inside and outside the container.

I can show an example of how this is achievable via docker-compose tool and its configuration file docker-compose.yml:

# docker-compose.yml file
version: '3'

services:
  <service_name>: # your service name here
    image: <image_name> # name of image your container is built on
    environment:
      - SAMPLES_PATH=/samples
    volumes:
      - C:\Code\somepath\samples:/samples

In your python code you can use the following structure:

import os

fdir = os.getenv('SAMPLES_PATH', r'C:\Code\...\samples')

fpath = os.path.join(fdir, 'fname.ext')