Ximin Luo

> An important point, I guess, is if Vector and MVector has both mapM then the two structures should be interchangeable Yes my suggestion breaks this, in favour of emphasising...

> Unfortunately no, since an applied vector might be a slice. You can't change the constituent type of the first half of a vector while the second half staying the...

IMO this should not be the intended behaviour and the search box should just look in all accounts, and not change its behaviour based on the currently-selected folder. It should...

> honest I don't specially need it either. > > However I can imagine how someone with 2000+ or more bookmarks could benefit from this, don't you think ? UIs...

Because of the lack of browser extensions on mobile, the floccus mobile UI should be trying to replace a browser address bar, and this is already a compromise. I can't...

> I know it's in that folder so I could navigate up to root and then down into folder "work" and open the bookmark directly. But that takes time. +1,...

That sounds good to me yeah. Ideally the separator between the sections wouldn't take up too much screen space.

See #117. I've tested it locally and it works well. Note that latexmath doesn't support inline output at the moment, but it will do once they merge my PR plurimath/latexmath#44....

Apparently plurimath is the successor of latexmath. However, it has [some regressions](https://github.com/plurimath/plurimath/issues/84) from latexmath, so that is not reassuring either. I added a commit to #117 to instead use python-docutil's...

I forgot to mention, some parts could be tidied up, for example I exposed some internal functions. I could move these to an `Internal` module to keep the current API...