Add Containerfile

[xxyzz]

Run code in a container could solve all the environment errors. We could publish the container image to GitHub Packages so the users don't need to build the image. This could also relief the burden of compatible with older Python verisons.

[xxyzz]

The command of installing package from pypi.org is deleted because that package is so outdated.

[kristian-clausal]

This seems good for those users who wish to use containers, I don't see why we couldn't have this. Is the image generation possible to automate so that upkeep of a GitHub package is hassle-free?

[xxyzz]

Maybe just need to run two podman commands to build and upload the container image. I'll try the commands in my fork repo.

[kristian-clausal]

Thanks, this seems great.

[xxyzz]

The published container image page in my fork repo: https://github.com/xxyzz/wiktextract/pkgs/container/wiktextract

[empiriker]

I always appreciate containerized applications when I just want to run an application as is. So this is a good addition. 👍

However, I don't think it solves all issues with versioning, in general. In the past, I have used wiktextract as a dependency and then overwritten or customized parts of the application and I always struggled with using a specific version of wiktextact.

In particular, I used to fix the version of wiktextract by specifying a certain commit, e.g.: wiktextract @ git+https://github.com/tatuylonen/wiktextract.git@a3665b8779ad78b045406cc7b1e9ce31876bc6b6.

The problem is that the install script of wiktextract always installs the latest version of wikitextprocessor from the repository. After a while a fresh install will result in incompatibile versions of wikitextprocessor and the commit of wiktextract.

As I understand, the containerized application won't have that problem. But in the described use case, where the wiktextract isn't run as is but used as a dependency, the container won't help.

IMO, the ideal solution would be to introduce proper versioning for wiktextract's dependecies.

[xxyzz]

You could use pip freeze and install the packages with the --no-deps option.

I don't think we could make releases for both packages, because Wiktionary changes every day and our parser is unstable.