New command twine preview?

[tillahoffmann]

Twine already renders the long description as part of the twine check command. Is there interest to add a twine preview command that saves the output to disc for inspection prior to pushing to pypi? Happy to submit a PR if yes.

Possible interface options:

• new command twine preview or twine render.
• optional flag --output [path] for twine check specifying where to save the rendered output.

[bhrutledge]

Thanks for sharing the idea. Can you give more details about the motivation and value of this feature? For example, what problem are you trying to solve? What output are you expecting? HTML?

Two quick notes:

• Twine only calls render for RST files
• Twine already supports custom commands via entry_points; i.e. you could implement the command in your own project

[sigmavirus24]

--output for twine check would indicate to me that the printed results of check should be written to that file, not what we attempted to render (if, in fact, we attempted to render the long description).

twine preview doesn't seem to have many benefits over uploading to test.pypi.org because - if I remember correctly - the check that we can render something could still fail on PyPI/TestPyPI proper.

Still, I find myself asking the same question as @bhrutledge - What's the motivation?

Also worth noting that this is an example of https://xyproblem.info/

[tillahoffmann]

Good point re-motivation: the output on PyPI is sometimes not quite as expected (e.g. because some directives are not supported). twine check can of course pick up those directives. However, developers are rewriting the long description to strip out such directives (see here for further details), and it's sometimes difficult to know what the rendered output will look like even when it technically passes the check. It would be useful to view the rendered output to check it before pushing to PyPI, leading to additional commits and versions pushed (see pypi/warehouse#2170 for additional discussion). Being able to render the output to HTML and inspect it before pushing to GitHub/PyPI could shorten the feedback loop.

[bhrutledge]

the output on PyPI is sometimes not quite as expected
It would be useful to view the rendered output to check it before pushing to PyPI

As @sigmavirus24 suggested, I would solve this problem by publishing to TestPyPI. That doesn't address the feedback loop and additional commits, but it is the source of truth for rendering. I think anything else would be an approximation, and therefore not a complete solution to the problem.

Also, it looks like Readme Renderer has a command line interface that does what you're suggesting.

[bhrutledge]

Also, if you're concerned about version numbers, I would take advantage of pre- and post-releases.

[sigmavirus24]

Closing as the consensus seems to be that we won't accept this