英文:

Running a subroutine provided by a docker image within the main script

问题

The best practice for running a library provided by a Docker image in a Python program is to use Docker containers to isolate the library's environment. To use the library from https://github.com/oseledets/ttpy in your Python script:

1. Pull the Docker image:

   docker pull oseledets/ttpy
   

2. Create a Docker container from the image and run it:

   docker run -it oseledets/ttpy
   

3. Inside the Docker container, you can use the ttpy library as needed.

To pass a tensor from your main Python script to the Dockerized ttpy library and perform operations on it, you would typically use some form of inter-process communication, such as sockets or RESTful APIs, to send data between your main script and the Docker container.

You don't necessarily need "Python on Whales" specifically for this task; regular Python scripts and Docker commands should suffice.

英文:

What is the best practice when wanting to run a library that is given by a docker image in a python program?

I'm trying to use the library: https://github.com/oseledets/ttpy , in one of my python scripts but it seems like installing it from pip is not working due to some fortran issue, one of the suggested fixes is using a docker image, the docker image works but how would I for example take a tensor in my main python script and perform a rounding operation on it using the docker version of the ttpy library?

Basically, do I use something like python on whales to call it from my main script and pass it a tensor?

Thanks.

答案1

得分: 1

A Docker container has an isolated filesystem. If some library is only installed in a container, you can't import it from a non-container process. Typically a Docker image will contain a complete application and not just a library.

在Docker容器中，有一个隔离的文件系统。如果某个库只安装在容器中，您无法从非容器进程中导入它。通常，Docker镜像将包含一个完整的应用程序，而不仅仅是一个库。

In the particular GitHub repository you link to, the Docker image appears to principally provide a Python REPL for experimentation. The Dockerfile is in the repository, and you can see what it does: it starts FROM python:3.10, installs some dependencies, and installs the ttpy library from the local sources, but it doesn't itself contain a CMD.

在您提供的特定GitHub存储库中，Docker镜像似乎主要提供了一个用于实验的Python REPL。Dockerfile在存储库中，您可以看到它的内容：它以FROM python:3.10为基础，安装了一些依赖项，并从本地源安装了ttpy库，但它本身不包含CMD指令。

If this is a reasonable base for your application and you can use exactly the Python version this image packages then you could build an image FROM daskol/ttpy, but it might be more maintainable to repeat the installation steps from that Dockerfile in your own image.

如果这对于您的应用程序是一个合理的基础，并且您可以使用此镜像打包的确切Python版本，那么您可以构建一个镜像FROM daskol/ttpy，但在您自己的镜像中重复Dockerfile中的安装步骤可能更易于维护。

英文:

A Docker container has an isolated filesystem. If some library is only installed in a container, you can't import it from a non-container process. Typically a Docker image will contain a complete application and not just a library.

In the particular GitHub repository you link to, the Docker image appears to principally provide a Python REPL for experimentation. The Dockerfile is in the repository, and you can see what it does: it starts FROM python:3.10, installs some dependencies, and installs the ttpy library from the local sources, but it doesn't itself contain a CMD.

If this is a reasonable base for your application and you can use exactly the Python version this image packages then you could build an image FROM daskol/ttpy, but it might be more maintainable to repeat the installation steps from that Dockerfile in your own image.