A Hyperbolic Mess

beck · February 3, 2021, 6:37pm

Hi folks, I have a document that’s organized like this (arrows indicate linkages, childcount is the number in parens):

So for example, note “10” in “Participants” links to five notes in “Data Items” (01-10, 02-10, 03-10, 04-10, 05-10). Those five notes link to 12 Data Extracts, for example, here are nine of them from Data Item 02-10.

And those data extracts are coded in all sorts of ways, for example, here is one data extract from 02-10 that links to three codes.

I wish to view these codes in an exploratory manner that I think hyperbolic view is designed for. For example, I’d love to see my codes (red in the photo below) and their relationship to the extracts (blue) like this:

But hyperbolic view slows my machine to a non-responsive crawl and looks like this:

Is there any hope for using this view with so many items, or is this just a limitation I’ll have to accept and move on from?

Thanks,
Beck

beck · February 3, 2021, 7:04pm

For example, would it render more elegantly if the data were organized in a different way? Instead of:

Maybe this?

mwra · February 3, 2021, 10:23pm

I thikn that’s where the view is right now, but innovations are planned. The view was just getting going when we all got diverted onto ziplinks and the like.

As you show, the problem is the view plots all connected notes leading to visual overkill and (cpu overkill!). One of the suggested ideas is being able to filter to only show some link type(s), whereas at present you can only emphasise the lines for one type. Not so helpful when you’ve a complex doc—it won’t surprise you’ve I’ve docs that look rather like your grab.

Another suggestion is a limited scope view, e.g. everything/filter things within N links of the focus note.

Anyway if you see things of interest in this view, I’m sure @eastgate would be interested to hear your ideas so as to feed into current work.

Until then, I’m afraid it’s “No jam today; jam tomorrow.”

beck · February 3, 2021, 10:17pm

Is this something I can do now?

mwra · February 3, 2021, 10:23pm

Not at present. I was just noting some of the suggestions that have been made of late. I think there are two separate design challenges here:

what is plotted, i.e. filtering in various forms
visualising what is plotted. As the number of items grows the load created by amount of data drawn to screen needs consideration.

satikusala · February 5, 2021, 1:07pm

Beck, Yes, there is, but it takes a little thinking about what you want to do. You can re-engineer the links you want to focus on by create notes, agents, and linkTo actions. I could try to experiment with this now, but it might be easier to go over this together on our call Sunday. We can then back that into training for the community. NOTE: I’ve had success doing this with other files.

mwra · February 5, 2021, 1:21pm

If you’re thinking of the mechanism you’ve used in your thesis doc, I wonder if that’s flexible enough for ad hoc use. If nothing else, the discussion here makes me look forward to hyperbolic view improvements; I think the view has yet to really show it’s potential (due to display complexity/filter issues—which I’m not assuming are trivial from a design perspective).

rkaplan · February 5, 2021, 1:34pm

What a great picture to illustrate the issue

This question of the practical utility of a graph of links has been discussed quite a bit in the Devonthink and Obsidian forums lately. So far they all come down to the same thing - it seems like an amazing idea at first, but in practice it is hard to get the filtering of what you want to see to a level that still keeps the big picture usable.

I agree with the goal, I agree with the concerns raised - not sure yet what the answer is but I agree it is likely in the area of how to effectively filter what you want to see.

satikusala · February 5, 2021, 2:26pm

Sort of, in, I think we can use the “node” linking method that I used in my “analyzing massive spreadsheet” work. I, quickly, ran into the same problem that @beck faced, and with a little noodling was able to come up with a centralized node, linkTo strategy. I. don’t want to get ahead of myself, though, and presume this will work for Beck. We’ll discuss it Sunday.

mwra · February 5, 2021, 2:37pm

Sure, and I think it’s a good thing for the Saturday meet-up. Many eyes, etc.

beck · February 25, 2021, 4:19pm

An update on my progress here. I ended exporting data from Tinderbox and opening it up in Obsidian and while there are definitely shortcomings, it’s a serviceable option for exploration until Hyperbolic shapes up. I need both apps up at once to annotate, verify, etc.

eastgate · February 25, 2021, 4:35pm

There’s a fairly extensive research literature on the subject. I think the map view, originating in Storyspace and improved over the years, remains the most broadly useful exemplar. Hyperbolic view will benefit from filters.

And there will always be messes, because some topics are messy. I remember a summer job when I was in school, working on computational models for walking. It seems simple enough. Start with the left hip at {0,0}, extend a femur of length L and angle {theta, phi), figure out where the knee is. Then extend the fiber of length M… The equations were godawful. I couldn’t simplify them. I was certain I’d made a ghastly mistake. But no: nature had no particular reason to make this simple, especially not in Cartesian space, and it simply isn’t.

satikusala · February 25, 2021, 8:15pm

Great quote…but Tinderbox can help you get the mess under control.