-
Notifications
You must be signed in to change notification settings - Fork 403
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Remove bokeh plot and replace with simple matplotlib plot #767
Conversation
Signed-off-by: Roman Lutz <rolutz@microsoft.com>
Signed-off-by: Roman Lutz <rolutz@microsoft.com>
Signed-off-by: Roman Lutz <rolutz@microsoft.com>
Signed-off-by: Roman Lutz <rolutz@microsoft.com>
I'm not sure what you mean here @romanlutz |
Let me explain what I initially thought, and then my correction to what I wrote: What I realized just now is that |
Sadly, I was right with the current version of the PR. I cannot remove In summary, I don't have further changes to this and would like to merge as is (or rather: someone else should merge it) unless there are concerns. Definitely want to wait to see if @adrinjalali or @riedgar-ms have more input since you've both worked with multiversion in some capacity (and certainly more than me). |
we should definitely NOT do that lol Regarding sphinx-multiversion, I think we should replace it with a mechanism that builds the different versions completely independently of one another, and that's what we do in sklearn. Each version is completely sandboxed and the dependencies are not leaked into the other ones. We can build the docs for different versions independently, and then put them under different folders on the website, and have those links linking to different versoins. |
@adrinjalali That's interesting. I think the difference is that scikit-learn doesn't use @riedgar-ms do you think there's a point in filing an issue with Regardless of the answer to either of these questions, shall I open an issue in our repo about this point in particular (getting multi-version builds done without having to rely on old dependencies) and we discuss it further there? I think this PR has reached its intended scope and these are just discussions beyond that. Wdyt? |
The scikit-learn doc build system looking very scary was why I was happy to come across Opening an issue to review the versioned documentation build does seem wise. A related issue would be only building 'new' documentation each time, rather than always building every version (particularly if we're going to build each completely independently). |
Our doc builds are taking a while, so it seemed like a good idea to replace the custom bokeh plot with a simpler matplotlib plot just like in #766 . Sadly, we are locked into all dependencies we ever had unless we're willing to go into older versions of Fairlearn, remove them from there, and then move the tag. Still, this simpler plotting code seemed like a step in the right direction.