Pip install certain version6/20/2023 ![]() A package without classifiers should be installed with a warningįinally I think that pip (and particularly the latest work on it) improved the packaging situation enormously.A flag should be provided for the old behavior."compatible with a previous Python version" refers to within the same series.If none of these is possible, returns an error. The most recent version of the package which is advertised as compatible with the used Python version.The latest version of the package if it is advertised as compatible with a Python version older to the one that is being used.The latest version of the package if it is advertised as compatible with the Python version being used.If no specific version is provided, pip should install one of the following: PyPI does not support querying for the latest package compatible with certain Python version (not sure if this is the case): If this is the case, a very inefficient way will be to do this client side fetching all metadata. I think a reasonable heuristic can be devised (see below) But I think most errors occur when compatibility with older versions is dropped. Packages do not update their classifiers when a new Python version comes out: True. I think that being a little more stringent with experienced users (those uploading packages) will provide a better user experience for the new users (those installing packages without the knowledge of what changed in each Python version). Moreover, I think PyPI should reject packages without certain classifiers. ![]() Many packages do not have classifiers: True, but is becoming less common. I understand that there might be a few problems to achieve this: But when dealing with users (particularly new users) that should install packages using the command line it is convenient and less error prone to have pip resolve things for them (or at least provide useful feedback before their program/tutorial ends with a weird error message). I understand that the version of the dependencies can be specified. travis.yaml to deal with this but it made me wonder why is pip not checking the advertised compatibility of the package. Just to make it clear: the PRs were ok but pip was installing for Python 3.2 the latest coverage (4.0) and not the latest compatible one (3.7.1). ![]() Recently coverage dropped Python 3.2 support and several PRs of many different projects started failing on travis. (sorry if this issue has been raised before)
0 Comments
Leave a Reply. |