gerby-project / gerby-website Goto Github PK

When checking things in Chrome's "different devices" option I get behaviour similar to [https://physics.meta.stackexchange.com/questions/9843/problem-in-viewing-mathjax-formatted-writing-in-smartphone](this MathJax problem). As I cannot check with an actual smartphone, I cannot check whether this problem is there on an actual device. I'm putting it here as an issue to remind myself to check this.

This happens even if you change from "search all tags" to "search only statements".

Additionally, if you just searched using "search only statements" and then you click the 50 box, then it jumps back to "search all tags".

This is with commit 51b3562

We should do benchmarks, and see whether there are certain bottlenecks.

It's not as bad as in #41, but there is a huge right margin.

I tried to implement this in a clean way, but failed.

Also, give some information on the result on the page itself, because the breadcrumb isn't the most obvious place to look for information.

It kind of would be fun if we could ask contributors for latex expressions for their names, e.g., for chinese names using characters 李时璋 or 李時璋 for my student Shizhang Li. Maybe this should be an issue for stacks/stacks-project instead?

Tag 03B6 should have all subsections displayed. It doesn't. Should be easy to fix.

On the table of contents page I would (personally) like to have it collapsed. Is there a way to remember that I collapsed it?

Or edit all comments using macros.

Using my iPad, on tag 0BEL (for which I have a special affinity given my last name), MathJax is reflowing the page, causing the layout to become narrower. Hence the top bar isn't 100% anymore.

Testing such a thing is extremely annoying, and we need more test cases (and more devices).

The current setup using python-bibtexparser apparently adds the eprinttype to the eprint field as a prefix.

Yesterday I was showing Jacob what JavaScript insertion means, and that it shouldn't be possible. Turns out it was possible. Somewhere I'm messing up, and I need to fix this.

Consider tag 00P7 of Higher Topos Theory. The MathJax error (which of course we shouldn't have) gives a reflow of the layout, forcing the sidebar to the bottom.

We should try to implement a treeview sidebar as in a good pdf viewer. Actually, a good source for inspiration might be the Dash documentation viewer.

For some reason the side-by-side editing feature isn't working properly when creating your own toolbar.

It would be nice to have these. The icon for inline could be fa-usd by the way, for display mode I haven't found a good symbol yet.

The problem I have is that I cannot define a method like drawTable, as some of the necessary methods are hidden.

Also, the filename is wrong, always.

In Chrome it seems to be working fine. In Firefox (where I preview things by using "Open PDF in Preview" on my Mac) the following seem to be wrong:

• the breadcrumb is looking very sad because it's alone on the first page
• the second page is fine, but everything that should go on the next pages is cut off, with an empty thir page to make up for it

We would like to have a line count for the Stacks project, but given that Gerby never sees the LaTeX code we don't have that straight away now. We could sum up the number of HTML lines, but that feels like a crude approximation.

We also need a page count. Again, not easily accessible from Gerby right now.

I'm just looking for an excuse to use Bootstrap's badges probably.

There is an issue with tag 01LR (also on the current website actually):

• it isn't numbered (also not on the section page 01LQ)
• it is described as eqnarray on tag 01LE, which is not really pretty

This is probably plasTeX-related too.

I know that my queries aren't perfect, but this is certainly surprising :).

I think the issue is the line that creates the variable results: the << operator probably can't handle several thousand tags at this point.

I made the mistake of allowing URL's of the form /chapter/55. These should redirect (or maybe display a warning?) when accessed in the new website.

Check Nakayama!

Right now there is a __gt__ method for the Tag model in Peewee. But it fails for appendices, as it cannot convert strings to integers.

It should be possible to run the entire setup on (paid) cloud hosting. We should try this.

So don't show them.

I just tested all the existing comments in the new comments system. Scrolling through, I noticed the following issues (which are not supposed to be fixed by our code):

• turn all \ref{label} into \ref{tag} on 112 (00DC), 394 (00DE, 0038), 459 (00DV), 672 (0052), 1163 (02OK), 1912 (0A30), 3002 (0101)
• comment 176 has an escaped underscore: just make it mathematics
• comment 2177 has escaping gone wrong, remove it
• comments 240, 657 and 1977 have <i> and <b>, turn this into Markdown markup
• typos and other things
  • \mathrak on 686
  • \codts on 3012
  • comment 1504 is copy-pasted TeX code

See e.g. tag 02M4, or 0BEL, and probably many many others. Should be easy to fix, it's pure CSS.

Right now the preview isn't MathJax aware

We could also use a dedicated statistics page, so that the Statistics pane on the front page would actually link to something. Besides the obvious statistics, we could have things like

• most complex tag
• most used tag
• most commented tag
• ...

There's some broken HTML. Easy peasy.

This is due to the table of contents at the end of each chapter. We just need a separate TeX creation script for the website.

People keep on referring to numbers in comments, which makes perfect sense, as the only way to read off the tag of a lemma inside a section is by hovering over the link. When writing a comment (and possibly somewhere else, maybe even keep track of their preferences using a cookie) we can let them toggle between these.

Solution: just enclose the number in a <span> which contains the tag as a data-tag attribute. Then in JavaScript we can toggle between these two pieces of information in a completely uniform way.

For instance searching for -and.

With the new website setup, it will be hard to have the sloganerator integrated in the same way we do now. Given that all the logic and layout has been done already, we could easily port it to Python.

This has very low priority though, we can do without it as no-one really uses it.

Higher Topos Theory has a theorem 7.1.0.1. So make sure that the breadcrumb can deal with missing subsections.

Once someone leaves a comment, we should save their information and use it on all pages.

Because jquery-bonsai changes the way all list items look, the preview of a search result is borked. I played around for a few minutes, but I couldn't figure out a good solution straightaway. Maybe we should decorate all list items on which Bonsai is supposed to act or something?

This can be done by encapsulating the default renderer, catching \ref's where they appear.

For some reason it's not triggering. I don't feel like spending any time on it today though.

Currently the sidebar gives just the commits, but with the information in the database it is possible to say which tags were actually altered (provided an update has happened). In the new website, we could try to say which tags were updated inside a particular commit displayed in the sidebar.

Eventually we will run Flask inside Apache (at least for the Stacks project). We should run it that way on the test server, to get familiar with the setup.

As per title.

As the Stacks project maintainer, I think it makes sense to have one database for the mathematical content of the site and a separate one for the comments. Namely, everything in the database except for the comments is currently generated by scripts acting on the Stacks project. Splitting the database into two databases would mean that if we update the Stacks project we can run all the update scripts locally and then simply copy over the database to the live server without touching the comments database.

I've done something stupid there, the fix should be easy.

It should be easy to run a local version of the website, by just running a download script for all the external code.

1. move comments to top
2. online changes in index