Sequential display in an agent map


(Leif Brown) #1

What I have:
• A collection of notes with the prototype “Sequential” with key attributes “IsSequential”(boolean), “Prerequisites”(list), and “Xmin”(number).
• Within their container, the Sequential notes are all linked with links of type “requires”.
• An agent guaranteeing that every Sequential note has a unique name.
• An agent guaranteeing that (un)linking two Sequential notes is mirrored in the Prerequisites list.
• An agent guaranteeing that subsequent notes are always to the right of their prerequisites.
• An agent “Sequence” that finds all notes with prototype “Sequential”

The result is easy to read, easy to use, self-maintaining, and resembles a timeline.

But, now I have to specify other relationships between notes, and the beautiful sequential flow is an obstacle.

I could just turn of the positioning agents, and only activate them when needed, and that might be easiest.

But I think I want that sequential, positioned view in an agent, viewed as a map. Ideally, I’d even like to be able to add new links in the agent map view, and have those update the list property of the original note.

This seems overly complicated, so I thought I should ask for alternatives, or approach nuggets.


(Paul Walters) #2

It would be interesting to see a sample document.

“Other relationships”, I assume, means links other than “Requires”?

What is the obstacle? Something visual to do with the link lines?


(Leif Brown) #3

Paul,
the obstacle is that the sequential map view spreads notes over a large horizontal distance, obscuring other relationships.

Rearranging maps seems to be one of my best prompts for creative insight on a project, but some arrangements are worth preserving as views, but not worth treating as the only correct arrangement, nor worth recreating from scratch (as in, duplicating aliases in a separate note and linking & arranging the aliases manually).

I’ll see about whipping up a sample document. Is there any particular point you’re trying to clarify?

Thanks,
Leif


(Mark Anderson) #4

One of the first things I recall unlearning on the way to more effective use of Tinderbox was that a tidier map isn’t necessarily a better map. We end up wasting a lot of time making maps - that only we will likely view - tidier because it feels right to do so. Also some maps will never sequence (untangle) unless they truly are describing sequential things. To me, the main joy of maps is spatial relationships - grouping things I’m not ready to formally link.

That said after a while, you find that you can capture that nuance in user attributes leaving you free on the constraint of a single map. I use large datasets so maps aren’t really practical (even with a bit screen). However, even if working in other views, I’m thinking in map terms of associations and linkage.

If doing something like your sequential agents is complex with a smallish data-set, I’d wager it will get more complex/fragile as the dataset (number of notes) grows. I think Tinderbox’s greatest strength is in finding and recording association and links regardless of the way that’s visualised.

I’d caution in investing too much in using a particular view, lest the tail wag the dog. Rather than do what (you believe) the view allows, think of the associations you’re trying to investigate and where/how they may be brought to the fore.

Also, if you need to output your findings, then creating the structures necessary to feed that process with give you some useful general constraints.

Above all, experiment and use small experiment TBXs to investigate a technique you can then apply to your actual work. Tinderbox is atypically forgiving at adding structure incrementally as needed. however, the less rubbish you have from failed half-understood earlier efforts, the easier it is to add new structure.


(Pat Maddox) #5

I’m always loathe to hear “that’s not how you should do it” so I hope I don’t come across that way here :slight_smile:

It seems to me like this would be a good use for Outline View, or Timeline view or maybe Chart view? (there are a lot of views you that you might try out). Outline view is straightforward – your sequential notes go top to bottom (assuming you have the agent set to sort that way).

Same. I think it would help us help you if we can see & play with what you’re trying to do.

Something just clicked for me in reading this. I have a strong aversion to tagging (and stuff like it, like user attributes) – it seems like I have to process every note to get any meaning (and in fact, my DEVONthink databases are completely unstructured and untagged, I pretty much rely on search to find stuff).

But… moving notes close together in a map is loosely, unofficially “tagging” them. And moving them into a container is “tagging” them with a single tag.

So – especially with large data sets – I think it might be useful to get more serious with user attributes after all.


(Paul Walters) #6

Actually, I am intrigued by your set of agents and how they work together. I always like to see how clever things like that work.

I wonder if the newer “Dance” feature in 7.5.x might help here. It will take apart your carefully constructed visual, but it might also enable you to see the other links/relationships more clearly.

I’ve sometimes attempted to do maps similar to the horizontally-arranged relationship map you described. Doing it automatically with an agent adds a new twist. In any case, personally I usually end up letting that kind of strict formalism fall apart because links and visual structure don’t always interact well.


(Mark Anderson) #7

Yes. For me, that move was a gateway to using bigger sets of notes efficiently but I’d re-iterate I did so only once I understood the spatial hypertext paradigm of Tinderbox. I work mostly in Outline view (simply due to scale) but I think in map view terms - albeit after some 14 years of heavy Tinderbox use.

The ‘tyranny of search’! If the magic box can’t find it, we’re sunk. In the 1990s the hypertext community were talking about the ‘tyranny of the link’ as typed links were the vogue and so linking wasn’t a trivial task. Fast forward to the Web and its early technical limitations. Now single box Search is ever-present. Which is fine until it yields nothing about something you know is there as Search can’t tell you what it doesn’t know.

Tinderbox attribute values can make elegant proxy links - or rather singular enough data from which to posit or actually make links - and to mark groupings.

All that said, I still find tinkering around in maps the most fun - even if I too can’t resist the feeling things ought to be tidier.


(Leif Brown) #8

In this particular case, I’m planning a game. I’ve dumbed the actual data way down, for clearer illustration.

This arrangement of notes in “/data/” emphasizes the “user’s experience”. Grouping all the red elements together and all the blue elements together would emphasize the underlying “model of the world”. Both are necessary and valid, but they don’t mix well.

I could use two parallel sets of notes, but that creates its own problems.

Possibly, a dancing map might work, if I could mirror links…

simulation.tbx (119.6 KB)