Skip to content
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

✨ Using sphinx toc functions #219

Merged
merged 13 commits into from
Jan 23, 2021

Conversation

choldgraf
Copy link
Collaborator

@choldgraf choldgraf commented Jun 27, 2020

This PR refactors our navigation bar, sidebar, and table of contents logic to use Sphinx toc functions instead of our own custom parsers. However, the output shouldn't change because it uses beautifulsoup to parse the HTML that Sphinx gives us in order to add the right classes and structure.

In implementing this, I also noticed some other bugs that our old parsing code had, that we didn't notice. (e.g., in our demo page, clicking the api documentation sidebar link doesn't expand that page's children). This should fix those bugs too since we're just using the basic Sphinx sidebar functionality, nothing custom.

Pros

  • This code is (IMO) much, much simpler and more maintainable than using a ton of custom jinja templates + python code to parse docutils
  • We will be more resistant to bugs over time, because we depend on Sphinx for handling the logic of toc / sidebar / etc
  • We'll get more predictable behavior for "toctree" features (such as :numbered:)
  • This theme should now behave more consistently with other sphinx themes
  • image

Cons

  • We add a dependency on beautifulsoup (which I hope isn't too bad, since that is such a standard in the Python world)

@jorisvandenbossche would love to know what you think here

resolves #220 #110
also can resolve #218 #135 (but needs tests)

@choldgraf choldgraf force-pushed the missing_sidebar branch 3 times, most recently from 647eff3 to b381823 Compare June 29, 2020 15:22
@choldgraf choldgraf changed the title fixing single-page header bug Using sphinx toc functions Jun 29, 2020
@choldgraf choldgraf force-pushed the missing_sidebar branch 8 times, most recently from 6a10768 to 7027c90 Compare June 29, 2020 15:37
@@ -283,35 +288,26 @@ table.field-list {
}
}

// For page headers with toctrees under them
li.page-header {
Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This just simplifies the CSS a little bit - we were using a bunch of level-specific selectors for our rules, but I think we can get the same effect with the generic rules here

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Do you remember where/when this "page-header" class is used? I don't directly find it in the output HTML somewhere

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@choldgraf I don't see this "page-header" class use anywhere, so for now going to remove it again. If it actually had a purpose, please explain and I will add it back :)

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

go for it - I have no idea what I was thinking here because this PR is from 7 months ago :-)

</ul>

</div>
Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I guess we weren't closing out this div before 🤷

@choldgraf
Copy link
Collaborator Author

choldgraf commented Jun 30, 2020

Friendly ping on this - this PR is also fixing a pretty big bug that we're experiencing in the Jupyter docs (pages that are > 2 levels deep don't show up in the sidebar).

I could try to implement the fix with our current docutils approach but it seemed like this PR would actually be a simpler way to solve the bug and solve ourselves some headache in the future as well...

@jorisvandenbossche
Copy link
Member

(I am only just back, will try to take a look at this tomorrow, but from the description I am not fully fond of the idea ... but maybe we should have a call about it one of the coming days to easier discuss it?)

@choldgraf
Copy link
Collaborator Author

choldgraf commented Jul 2, 2020

I'm happy to take a call to chat it through. This PR was mostly borne out of me being frustrated with debugging issues over and over again in our custom docutils parsing code :-)

My personal opinion is that this approach is much, much simpler than what we're currently doing. Rather than replicating 80% of Sphinx's toctree functionality with our own custom docutils parsing, we instead use the Sphinx toctree function and then just add a few classes here and there.

Basically, here's what we do now:

- Use sphinx TocTree class that outputs docutils objects
- Manually parse docutils to pull out the things we want
  - Replicate edge-cases like "toctrees that are underneath headers", and "multiple toctrees per page"
  - make lots of assumptions about the docutils tree that we have discovered are often not true
- Manually write HTML converters for the resulting docutils objects
- Write a bunch of custom jinja templates to build the HTML we want

and this PR makes it:

- Use sphinx's toctree function that outputs HTML
- Manually parse the HTML to add classes and remove the pieces we don't want

(FWIW, I'm also happy to chat on a quick call now if you wanna chat it through)

@choldgraf
Copy link
Collaborator Author

choldgraf commented Jul 8, 2020

@jorisvandenbossche I just pushed an update to this PR that adds the ability for the get_nav_object function to return a raw BeautifulSoup object with minor modifications, in case theme editors would like to use it.

I also tested out how this could be used by a theme by trying it out in the sphinx-book-theme. You can see that in action in this PR: executablebooks/sphinx-book-theme#130

In particular, see how I'm calling get_nav_object here:

https://github.com/executablebooks/sphinx-book-theme/pull/130/files#diff-95c3c14f442154a7a77462f1e36e81e7R56

I'm then just doing some minor editing to that BeautifulSoup object and then returning my own HTML. What do you think? I can develop in this direction a bit further if you think it's a nice compromise.

@choldgraf
Copy link
Collaborator Author

Just a note that I'm just gonna implement this for the SBT theme. I'd really prefer to upstream this but I can't wait super long to get it in. Here's the PR that we can use for reference in the future:

executablebooks/sphinx-book-theme#130

Note that as I was implementing that PR, I also realized that sphinx-style page / section numbering also doesn't work in this theme (unless we manually add it in), so we should document that (or get this PR merged)

@jorisvandenbossche
Copy link
Member

Just a note that I'm just gonna implement this for the SBT theme. I'd really prefer to upstream this but I can't wait super long to get it in.

OK, no problem, that way it can be test drived in jupyter-book-theme before upstreaming it ;-)

Sorry for the delay, and leaving now for holidays. But as I said on our call: I am certainly on board with the idea! (but I haven't yet been able to test it with some other projects)

@choldgraf
Copy link
Collaborator Author

@jorisvandenbossche what are your thoughts on this? FWIW we've had this in jupyter book for a few months now and it has worked fine.

@mattip
Copy link

mattip commented Sep 4, 2020

FWIW, this PR changed numpy document build time (without multiprocessing via -j) from 300 secs to 352 secs. Most of the time is spent in page writing. Is that expected?

@choldgraf
Copy link
Collaborator Author

choldgraf commented Sep 4, 2020

I wouldn't expect much performance change unless beautifulsoup is being extremely inefficient somehow...that's weird

that said, I'm on paternity leave now and don't have the time to keep rebasing this PR as the pydata theme changes, and it doesn't seem like folks are super interested in this functionality, so I am not planning to put a lot more work into it (though will note again that this has simplified things a lot in the sphinx-book-theme).

Comment on lines 92 to 97
for i in range(1, context["theme_show_toc_level"] + 1):
for li in soup.select("li.toc-h{}".format(i + 1)):
ul = li.find_parent("ul")
classes = ul.get("class", [])
if "visible" not in classes:
ul["class"] = classes + ["visible"]
Copy link
Member

@jorisvandenbossche jorisvandenbossche Dec 30, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@choldgraf this is "a" way that I found to get show_toc_level working, but I find it kind of hacky ..

The problem with the basic

        for ul in soup("ul"):
            ul["class"] = ...

loop above this loop is that this is a flat iterator, and so you don't know how "deep" you are for each ul.

This "workaround" uses the toc-hN class names we added above, to find the appropriate levels of ul for which to add the class. I didn't directly find a way with beautifulsoup to only search a certain number of levels down.

Copy link
Collaborator Author

@choldgraf choldgraf Dec 31, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

this seems reasonable to me :-), but maybe we should add a comment to explain your reasoning here?

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hmm, not sure why I missed that earlier. But of course the code that adds those toc-hN classes (the classes that I was then using here to navigate the soup as a workaround), is already recursively iterating through the lists keeping track of the level ... :)
So simplified this

@choldgraf
Copy link
Collaborator Author

thanks for the pushes! can you update your "to do" list to reflect the things that you've tackled here?

@choldgraf
Copy link
Collaborator Author

@jorisvandenbossche do you think that we can leave:

sidebar html without skipping first level

to another PR? It feels to me like a feature request whereas this PR is meant to be a more "under the hood" change that doesn't affect end functionality but makes the theme less buggy. Then we can just merge this one now so it's easier to iterate with smaller updates. What do you think?

@choldgraf
Copy link
Collaborator Author

ping @jorisvandenbossche - I wanna get this PR merged before it goes stale again but I don't have the bandwidth to implement the extra logic for the sidebar right now...

@jorisvandenbossche
Copy link
Member

@choldgraf reviving this. This PR certainly doesn't need to do everything that is possible with the new functions.
Some things I noticed while now trying out further:

@jorisvandenbossche
Copy link
Member

The TOC for a page with multiple top-level headers doesn't work anymore

This is actually something you fixed in executablebooks/sphinx-book-theme#215, so copied that part of the PR here.

@choldgraf
Copy link
Collaborator Author

So it sounds like we are good to merge this, and then we'll need to re-implement #135 using the new setup, right?

@jorisvandenbossche jorisvandenbossche changed the title Using sphinx toc functions ✨ Using sphinx toc functions Jan 23, 2021
@jorisvandenbossche jorisvandenbossche merged commit d2211c3 into pydata:master Jan 23, 2021
@jorisvandenbossche
Copy link
Member

OK, this is merged ;)

I further removed the header-page CSS, renamed the new functions that returns the HTML to make the name more correct (generate_nav_html, like in sphinx book theme), and added a quick version of the old functions based on beautifulsoup for some backwards compatibility (we can see if we then want to deprecate them).

So it sounds like we are good to merge this, and then we'll need to re-implement #135 using the new setup, right?

Indeed, next steps are adding support for captions, and I also want to add support for collapsing subsections in the sidebar.
I have a messy branch experimenting with those on top of this PR, will try to extract some PRs from it.

@choldgraf
Copy link
Collaborator Author

hooray! what a journey :-)

are there issues for collapsible sidebar items etc? We should make sure not to lose those threads (and many may be upstreamable from the book theme)

@ChaiByte
Copy link
Contributor

ChaiByte commented Feb 7, 2021

OK, this is merged ;)

I further removed the header-page CSS, renamed the new functions that returns the HTML to make the name more correct (generate_nav_html, like in sphinx book theme), and added a quick version of the old functions based on beautifulsoup for some backwards compatibility (we can see if we then want to deprecate them).

So it sounds like we are good to merge this, and then we'll need to re-implement #135 using the new setup, right?

Indeed, next steps are adding support for captions, and I also want to add support for collapsing subsections in the sidebar.
I have a messy branch experimenting with those on top of this PR, will try to extract some PRs from it.

Support for captions and collapsing subsections~ Wohh sounds wonderful. 👍

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
4 participants