Kuro5hin.org: technology and culture, from the trenches
create account | help/FAQ | contact | links | search | IRC | site news
[ Everything | Diaries | Technology | Science | Culture | Politics | Media | News | Internet | Op-Ed | Fiction | Meta | MLP ]
We need your support: buy an ad | premium membership

[P]
Cushion Treemaps: The Future of Data Navigation?

By Eloquence in Technology
Sun Mar 11, 2001 at 03:28:36 AM EST
Tags: Software (all tags)
Software

Yesterday I first heard about SequoiaView. It is a Win32 program developed at the Eindhoven University of Technology (Netherlands) that uses "cushion treemaps" to visualize the structure of your partitions. What are cushion treemaps, and how could they be relevant to the future of data navigation and visualization?


When you first start SequoiaView, it scans the harddrive it is installed on. You will then see the directory you have installed the program in as a "cushion treemap". Each file is displayed as a single block. The size of the block depends on the file size. The blocks on the root directory are arranged vertically, if they are in a subdirectory, they are arranged horizontally within a rectangle that represents their directory (depending, again, on the file size), if they are in another subdirectory they are arranged horizontally again and so on.

To make it easier to separate the blocks from each other, they are rendered with a light source on the top so they look like little mountains. (You can also use a non-cushion treemap in SequoiaView, but it is much less useful.)

In the program, you can easily navigate within your directory structure (using the right-click-menu or hotkeys) and open files. When you navigate to the root directory, you immediately get an overview of all files on your harddisk. As you move your mouse over the map, the filenames appear under the mouse pointer. You can immediately see which files take up how much space, and where they are located. The blocks are arranged alphabetically so you can easily find individual files. You can filter by name or date and even colorize certain filetypes on the map. Makes finding those lost porn images much easier ;-).

SequoiaView is a very powerful implementation of the treemap concept (its rendering and harddisk scanning speeds are very impressive, but unfortunately, the source code is closed). I have made some suggestions in their forum on further improving file navigation and selection.

I believe that a program like SequoiaView could make finding data and navigating directories easier than ever before -- and I am sure the concept can be improved a lot, and extended to a variety of other areas. Linux might gain a real competitive advantage by building something like this into its file managers. While the concept is unusual at first, it may make organizing and finding files easier than ever. What do you think?

Sponsors

Voxel dot net
o Managed Hosting
o VoxCAST Content Delivery
o Raw Infrastructure

Login

Poll
Cushion treemaps ..
o .. are definitely the future of data navigation 6%
o .. sound like a neat idea 27%
o .. may work, but still have problems to overcome 14%
o .. are too cumbersombe to be useful 9%
o I can't say, I haven't understood the program or your explanation 8%
o I can't say, I only use [OS here] and your explanation is not sufficient 8%
o Who needs them when there's [favorite shell here]? 16%
o Yay! Cushion fight! 8%

Votes: 61
Results | Other Polls

Related Links
o SequoiaVie w
o suggestion s
o Also by Eloquence


Display: Sort:
Cushion Treemaps: The Future of Data Navigation? | 31 comments (31 topical, editorial, 0 hidden)
This might be useful (2.00 / 2) (#1)
by Zeram on Sat Mar 10, 2001 at 12:44:02 PM EST

But somehow just doesn't seem all that interesting. Maybe I'd have to see it in action.
<----^---->
Like Anime? In the Philly metro area? Welcome to the machine...
A useful gadget, but not a make-or-break feature (2.33 / 3) (#2)
by Some call me Tim on Sat Mar 10, 2001 at 12:59:51 PM EST

While it looks like a useful tool to have around, I don't agree that it would add anything like a competitive advantage to Linux.

If it were one of a suite of features (~50-100) that made maintaining a system easier, then maybe. This is an old topic discussed here and elsewhere, and I won't bore you with a rehash here, but suffice it to say Linux needs a profound ease-of-use overhaul before it becomes a serious desktop contender.

Tim

I want something like this... (none / 0) (#3)
by daystar on Sat Mar 10, 2001 at 01:48:24 PM EST

but not for files, which I don't have a problem organizing, for web navigation. I haven't figured out the right way to do it though. hmm.

I guess if I owned a windows box I'd check ou tthe product...

--
There is no God, and I am his prophet.

Maybe I'm missing something (5.00 / 1) (#4)
by ism on Sat Mar 10, 2001 at 01:55:32 PM EST

but it seems like this is just another view type. It seems similar to a pie chart in that it subdivides area based on data. I'm not sure what the "cushion" is, either. Is it simply adding a proportional gradient to an area to more easily delineate areas? That's what it seems to say in the paper.

It looks like the main goals of treemaps are to maximize the use of rectangular space and to present a view of hierarchal information, which a pie chart wouldn't be too good at. Certainly an advantage, but I don't think it's revolutionary. I took a look at some of the links and came across the stock portfolio site, and frankly, a pie chart would be better suited for that kind of information. This seems constrained to visualizing hierachal file systems.

Aside from the fact that your article suddenly veered off to a Linux tangent, I don't see how TreeMaps in the file manager would give it an advantage. I tried SequoiaView out, and it's aesthetically pleasing, and maybe I need to get used to it, but if I wanted to find the largest files, it's easier for me to do a du to find how much is in each directory and sort directories by filesize.

Interesting, but "the future?" I doubt it.

I downloaded it and it's NEAT! (4.66 / 3) (#5)
by R343L on Sat Mar 10, 2001 at 04:45:31 PM EST

I downloaded it (for silly reasons I'm stuck on my Windows partition for internet access). Anyway, I installed it and opened it up. It took about a minute or two for it to read my disk in its entirety. I have about 7 Gigs of space in my windows partition (about 3 used). I have plenty of personal documents on it and random things that I download -- I'm a packrat when it comes to documents I find that are interesting that I don't want to have go download again (government papers, etc.) The reason it's cool (but not necessarily useful) is that it shows all of my hard disk at once, with relative sizes of files. When you put your mouse over a rectangle it tells you what file it is and the status bar at the bottom tells you releveant info about it (size, percentage of disk, percentage of map, how much space the current directory takes). The cushioning seems to help because it makes it easier on the eyes to look at. Also the non-cushion view makes it hard to see the little tiny files way down in the tree.

It has an export as bitmap function so I put up a webpage with some png's and jpg's of my harddrive (one annotated). My Harddrive with Sequoia.

I'm not sure this would actually help a person (there's really too much info) but its neat for seeing where space is being taken up. For example, I never realized that \Windows was so big! Or that most of the Redhat RPMS are about the same size. Or the varying sizes of the mp3s I've ripped off my CDs (in the My Documents area-- all those big rectangles are mp3s). Sequoia View also gave me an idea how deep the file structure is in some places.

As for this doing anything for linux popularity, I don't think it would. But it would be neat just to have it so I could look at my Linux side like this!

Rachael
"Like cheese spread over too much cantelope, the people I spoke with liked their shoes." Ctrl-Alt-Del

Try coloring and filtering (none / 0) (#7)
by Eloquence on Sat Mar 10, 2001 at 05:51:19 PM EST

The cool thing is, you can assign colors to different file types. Because all colors are "cushioned", this looks really neat. You have a green landscape of MP3s, some TXT hills, as rosebed of images.. And you can filter so that you only see certain extensions, or only files added after/before a certain date.

As for the usefulness, if you try out the navigation functions and practice a bit, you can navigate quite quickly to individual files. I'm seriously considering replacing Explorer with this; it just needs a few more shortcuts. You have problems with folders where you have lots of large files and some small ones; in these cases, the small ones may not actually be accessible through Sequoia. Different view modes would be useful here, perhaps an extension of the current filename view under the mousepointer using a dynamically expending panel with multiple filenames..
--
Copyright law is bad: infoAnarchy Pleasure is good: Origins of Violence
spread the word!
[ Parent ]

Re: Try coloring and filtering (none / 0) (#15)
by R343L on Sun Mar 11, 2001 at 03:48:51 AM EST

The cool thing is, you can assign colors to different file types. Because all colors are "cushioned", this looks really neat. You have a green landscape of MP3s, some TXT hills, as rosebed of images.. And you can filter so that you only see certain extensions, or only files added after/before a certain date.

I saw it...I was just trying to get something up for other people to look at as I didn't see anything on the webpage or in the story--and it is pretty cool.

I was actually thinking it might be fun to write this for linux (since they aren't distributing a linux version or distributing source). I wonder if its patented...which would be annoying, but possible.

Rachael
"Like cheese spread over too much cantelope, the people I spoke with liked their shoes." Ctrl-Alt-Del
[ Parent ]

Probably not (1.00 / 1) (#16)
by Eloquence on Sun Mar 11, 2001 at 09:34:23 AM EST

I wonder if its patented...which would be annoying, but possible.

There are no software patents in Europe. Perhaps they have patented part of the idea, but I don't think so.
--
Copyright law is bad: infoAnarchy Pleasure is good: Origins of Violence
spread the word!
[ Parent ]

Re: Probably not (none / 0) (#19)
by R343L on Sun Mar 11, 2001 at 03:38:59 PM EST

But wasn't MP3 patented in the US by a German company or something (I don't remember what country--only the name of the institute sounded to me German, which is not any second language I know). In other words, it could be patented in the US. I suppose its actually a US company created by a foreign one that has to be the patent owners. But you get the idea.

Rachael
"Like cheese spread over too much cantelope, the people I spoke with liked their shoes." Ctrl-Alt-Del
[ Parent ]

Algorithms vs. Code (none / 0) (#20)
by Eloquence on Sun Mar 11, 2001 at 07:15:22 PM EST

Actually, some of the mathematical algorithms behind MP3 compression are patented. I don't believe that you could mathematical patents for treemaps, though. In the US, OTOH, you could probably get patents for everything (need I say "one click"?), but that has probably not been done in this case.
--
Copyright law is bad: infoAnarchy Pleasure is good: Origins of Violence
spread the word!
[ Parent ]
Thanks (none / 0) (#9)
by davidduncanscott on Sat Mar 10, 2001 at 07:04:48 PM EST

As I was reading the article I was thinking that it would be good to have a link directly to some screen shots (presumably there are some in the Sequoia site, but one typically has to wade through some stuff first) for such a visual product, and lo! you had put some up.

[ Parent ]
Re: Thanks (none / 0) (#14)
by R343L on Sun Mar 11, 2001 at 03:43:52 AM EST

Thank you!
"Like cheese spread over too much cantelope, the people I spoke with liked their shoes." Ctrl-Alt-Del
[ Parent ]

Hrm sweet (1.66 / 6) (#11)
by delmoi on Sat Mar 10, 2001 at 08:40:53 PM EST

But I see from your website that you like Ayn Rand. I hope some day you grow up.
--
"'argumentation' is not a word, idiot." -- thelizman
[ Parent ]
Re: Hrm sweet (3.00 / 1) (#13)
by R343L on Sun Mar 11, 2001 at 03:43:06 AM EST

But I see from your website that you like Ayn Rand. I hope some day you grow up.

Hmm...and I hope you grow up someday so that someone's choice in novelist doesn't automatically make you think they are childish. It's not like I had "And I really think those Goosebump stories are just the coolest!" on my webpage.

Rachael
"Like cheese spread over too much cantelope, the people I spoke with liked their shoes." Ctrl-Alt-Del
[ Parent ]

Too bad (3.00 / 1) (#6)
by Demona on Sat Mar 10, 2001 at 04:57:22 PM EST

Natrificial decided to piss in the patent pool with their rather nifty product, The Brain.

cushions and trees (none / 0) (#8)
by danny on Sat Mar 10, 2001 at 06:39:16 PM EST

I'd like to be able to access data by lying on a pile of cushions and having them massage me appropriately. And I'd like to visualise data at the same time through watching the branches of a tree overhead waving in the wind.

Danny.
[900 book reviews and other stuff]

Nice, but... (4.00 / 1) (#10)
by scriptkiddie on Sat Mar 10, 2001 at 07:43:12 PM EST

Nice, but it has some problems. The biggest is that files are graphed both horizontally and vertically (in fact the direction alternates by depth). It's hard to compare two areas if they extend in different directions.

Second, I don't see any reason for the "cushion" part of the treemap. All it does is make it look nice!

This reminds me a lot of Apple's HotSauce project. It was a second way to look at data - for example, you could go to Apple's Web site and see pages clustered together and connected by lines. I thought it was cool, but current "visualization" techniques with icons and such seem to have a lot of momentum.

A useful (maybe) tool for admins (none / 0) (#12)
by briandunbar on Sat Mar 10, 2001 at 11:17:19 PM EST

More than once I've had arrays of data on a server, with no clear idea exactly how it's arranged - DOS tree views being unsat and ditto for Explorer when WinX came around.

ANY tool that makes a sys admin job easier is welcome, and this will be one such. It's only limitation being that it's Win32 only.

I'll bet I can examine my Samba partition and use it there, but it'd be nice to have something like this for unix.


Feed the poor, eat the rich!

Maybe there is a similiar tool for X11. (4.00 / 1) (#18)
by DuckSauce on Sun Mar 11, 2001 at 11:06:44 AM EST

I'm not sure if this is exactly similiar (I don't have a Win9x box), but you can check out fsv.

[ Parent ]
Author of the X-Windows version (5.00 / 2) (#17)
by lastwolf on Sun Mar 11, 2001 at 10:16:47 AM EST

Well, I've read this paper on "visualizing hierachal file systems" which was suggested by ism in this comment. I quote: "John Stasko at Georgia Tech produced a nice X-Windows version"

His homepage is located here. He's got some nice visualization projects going on. I couldn't find his treemap program though. It exist, I know, because a few screenshots can be seen here, and in this detailed PDF explaining the workings of the treemap technology.

If you're really interested in a X version of it, you could send an email to John Stasko.

LastWOLF
"Take your wings, go out and fly.
Learn, read and soar the sky."


Java (X, MS, Mac) version exists on SourceForge (none / 0) (#31)
by ulysees2001 on Mon Apr 02, 2001 at 07:13:32 PM EST

http://sourceforge.net/projects/treemap Implements treemaps only, but you get the source, so someone might create cushion treemaps from that.

[ Parent ]
nothing new (3.50 / 2) (#21)
by eMBee on Mon Mar 12, 2001 at 04:25:55 AM EST

this doesn't look any different from xdu to me, ok, it may look nicer, but it's nothing revolutionary.

greetings, eMBee.
--
Gnu is Not Unix / Linux Is Not UniX

hmm. filesystem visualization (1.00 / 1) (#22)
by exa on Mon Mar 12, 2001 at 05:10:06 AM EST

this didn't look like some state-of-the-art research to me. people have been doing visualization research for decades and you're saying some lame tree visualization of your file system changes the way we navigate data? sad.

yep. data visualization will someday be the norm, and will change the way we compute but it looks like there is some great knowledge barrier between the researchers and users. at comp. sci. departments people work on very advanced things, and when there is some lame implementation out there people think it's magic. really very boring. kind of reminds me terms like 'peer-to-peer', you know as if distributed people hadn't existed at all. and about the 'intellectual average' of kuro5hin, i don't think it rates that high on the cerebral scale.. this place is supposed to be about technology, and technology demands some wit.

thanks,

__
exa a.k.a Eray Ozkural
There is no perfect circle.

hmm. filesystem visualization (3.00 / 1) (#23)
by jackvw on Mon Mar 12, 2001 at 03:58:42 PM EST

I want to react to this. I must warn that I'm biased, since I am invented the cushion treemap idea and headed the sequoiaview development team. Some comments:
> this didn't look like some state-of-the-art research to me.
It was published in '99 in IEEE Information Visualization, the leading conference in this area. Ben Shneiderman (leading in this field) is very fond of the idea.
> people have been doing visualization research for decades
The startpoint of SciVis is '87, of InfoVis (in the modern interpretation) is about '90. So decades...?
> some lame tree visualization
What do you mean? Lame in what sense? Did you download and use it yourself? Please, give us feedback how we can improve it. We get tens of very positive user-reactions each day. So far 20,000 downloads of this lame visualization.

> at comp.sci departments ... when there is some lame implementation out there people think it's magic. really very boring.
What do you mean again? Don't you think we read the literature? I am reviewer for IEEE TVCG, IEEE Vis, and IEEE InfoVis, for instance. What do you know about visualization research?

> this place is supposed to be about technology, and technology demands some wit.

I thought this place was for people who know what they're talking about.

Jack van Wijk


[ Parent ]
tree vis (none / 0) (#24)
by exa on Tue Mar 13, 2001 at 03:02:33 AM EST

I have yet to see how it works, but after all it is just 'tree visualization'. It's not even graph visualization, or visualization of hierarchical clustering. Therefore, I was unable to regard it as state-of-the-art. The fact that it has been published and it has become somewhat popular doesn't mean that it will 'change the future of data navigation'. Not all data are realized as files in a filesystem, and not all navigation is filesystem navigation.

So, if you are one of the researchers, why do you let someone else announce it under a misleading title? You know, as if cushion treemaps are the only method in visualization.

I think it's been more than a decade since the first vis paper got published, and the idea has been around for more than 2 decades. I'm sure we could find some ancient papers on visualization if we searched.

What do I know about visualization? I've done graph partitioning and graph clustering, so I guess I could write a graph visualizer in practically no time. The more relevant is a DVR which I'm working on. It will hopefully be a non polygon thingie for the masses. It's got shear-warp for regular grids now, but I want to extend it with my own ideas (Sorry, just can't mention that yet ;)

You don't have to take this as a personal attack. I was mainly criticizing the article, not poor gradutate students [like me] who did it, or their hard working advisors. I do know that the work tends to feel a bit special for the researcher.

Thanks,

__
exa a.k.a Eray Ozkural
There is no perfect circle.

[ Parent ]
just tree vis? (5.00 / 1) (#25)
by jackvw on Tue Mar 13, 2001 at 06:43:22 AM EST

I agree, it is just 'tree visualization'. What you shouldn't underestimate is scalability: Large trees (>10,000 nodes) are still a hard problem, and we are convinced we made an interesting advance. So, in this sense it certainly is state-of-the-art. I do agree that the idea is unlikely to change the future of filesystem navigation. It is no more and also no less than about getting insight in large tree structures - which is a relevant and ubiquitous problem. That also explains why people take the initiative themselves to post enthusiastic stories about it: They found out that it is really useful and really works - which they want to share with others.

I look forward to your graph visualizer! Partitioning and clustering are important steps, but only part of the puzzle. Graphical representation and interaction are equally important, and it is hard to predict what will work and what not.

I didn't see your first message as a personal attack. I just had problems with unfair, unjustified, wrong qualifications.


[ Parent ]
tree vis (none / 0) (#26)
by exa on Tue Mar 13, 2001 at 02:31:49 PM EST

I'm sure the implementation is neat. That's why it must have been impressive. Scalability is a big problem in any kind of visualization problem. It's very significant that you've solved it for trees.

Graph layout in general is a very difficult problem so I suppose it would mean that complex trees are hard to visualize, with varying bf and imbalance. Probably some of your techniques could be extended to graphs if they work for trees. Though, I guess you'd appreciate that you can't view a graph as it is but you can view a tree more like it is. That is, a tree is a hierarchical structure in the first place and you don't have to invent a hierarchy for it. In graphs, you would have to display a large graph in a multi-level way. It's the more general case.

In addition, I would rather think that a filesystem tree is a quite regular tree. (people tend to create more regular trees, one of the principles many software also follows). If I created large random trees, would your algorithm handle them just as well? (say, |V| > 1.000.000)

Then, as you say there is the problem of interaction. How do you minimize cognitive load for arbitrary query/update operations to a tree by a human? When human gets in an equation, things get 1000 times harder.

The static layout problem is similar to graph partitioning/clustering (as in VLSI layout...), but making it dynamic would cost a lot.

That's why I think I need to take a look at the program! Too bad I don't run windows that frequently. ;) Last time I tried your page, it didn't load, but I'm sure the server will take a break anytime soon.

I have a multilevel bisection routine I'd written for the senior project, and right now I'm doing hypergraph partitioning, so I should be able to try my hand at graph vis whenever this code finishes. graphvis just doesn't cut it any more :)

Good luck with your work,

Regards,
__
exa a.k.a Eray Ozkural
There is no perfect circle.

[ Parent ]
Trees and graphs (none / 0) (#27)
by jackvw on Tue Mar 13, 2001 at 03:05:56 PM EST

Right, trees are much easier than graphs. One solution to handle graphs is therefore to reduce them to trees (f.i. spanning trees), and to add the other edges later on. But this is of course not always a neat solution. In general, I think that to visualize large graphs effectively you must exploit special properties of the graph in some way. For instance, a student of mine has succesfully attacked graphs describing Finite State Machines. Such graphs, resulting from Process Algebra f.i., typically exhibit some form of symmetry and regularity. He did a great job in developing a 3D visualization method that shows this symmetry.

Concerning very large random trees... Well, there you have problems. First, the number of leaves exceeds the number of pixels... Second, if the patterns are purely random, you can't expect much to see. A nice aspect of Treemaps (and SequoiaView) is that regularities pop up quite nice. Each type of directory (be it a game, a bunch of DLL's, a development environment) has its own structural characteristics, and you develop an eye in catching them.

Dynamic trees and especially graphs are indeed much harder to attack. Optimizing algorithms for visualization may lead to wild jumps from one local optimum to another, destroying visual continuity.

Anyhow, success in your work on graphs, and concerning their visualization, there is plenty of room for good, innovative ideas!

Regards,

Jack van Wijk

[ Parent ]
Some feedback on sequoiaview (none / 0) (#28)
by exa on Sun Mar 18, 2001 at 10:50:12 AM EST

First off: I love the name. I've always been fond of those large trees with all sorts of repercussions about indians. :)

I've checked the program and it's rather cool. I may offer a few suggestions.

I tried it on a 24GB partition with lots of files. program installations, devel directories, mp3s, docs ans such. The first thing I noticed was that the file size variance is not handled good enough. Some of the files occupy a large percentage of the screen while others are hard to click on (i use 1600x1200 resolution). You might want to employ a logarithmic scale to compress those sizes.

Second, I think using a lightsource is good, but in some way a bit confusing since it gives each file area the same feel. Using different shading/texture methods for different types of files/directories would be more informative. Even color coding would be an instant improvement.

Third, I think the program simply puts too many subdivisions on a single screen when there are many files it's supposed to handle. Some kind of "coarse" view would be most useful. Coarsening can be made for instance by defining a simplex of the tree (say 3 dirs / 30 files makes a simplex and viewed as a single node). More advanced coarsening schemes are at your disposal ;)

Fourth, I would like to recommend a programmatic improvement. Would it be possible in any way to extend this program to a fully functional windows shell? Like litestep for instance? (a port of afterstep to win32? Making it some kind of shell extension might be a good idea, too!

Congratulations on your work! I hope that this feedback proves useful.

Regards,
__
exa a.k.a Eray Ozkural
There is no perfect circle.

[ Parent ]
Re: Some feedback on SequoiaView (none / 0) (#29)
by jackvw on Wed Mar 21, 2001 at 03:10:05 AM EST

Thanks for the comments! Some feedback:
> Some of the files occupy a a large percentage of the screen.
This is not a bug, this is a feature. We just try to show which files are big and which are not. Nevertheless, log-scaling could be interesting.

> Using different shading/texture

We are already thinking about that, hold on.

> Even color coding would be an instant improvement.

When you click on the second button on the left, you'll get your colors.

> Too many subdivisions

Select 'Options', tabblad 'View', and enable 'Maximum Depth', then you'll get what you want.

> fully functional windows shell?

The most frequently asked question is: "Where is the delete button?" We consciously did not add this, the risk of errors seemed too big for us. Nevertheless, we offer an escape, via the right-button menu Explorer can be started. And furthermore, we are interested in visualization, not in making new shells.

[ Parent ]
Subdivisions (none / 0) (#30)
by exa on Wed Mar 21, 2001 at 06:49:28 AM EST

Cutting the tree at a pre-defined depth would work, but a tree contraction scheme might be useful, too.

Thanks,

__
exa a.k.a Eray Ozkural
There is no perfect circle.

[ Parent ]
Cushion Treemaps: The Future of Data Navigation? | 31 comments (31 topical, 0 editorial, 0 hidden)
Display: Sort:

kuro5hin.org

[XML]
All trademarks and copyrights on this page are owned by their respective companies. The Rest 2000 - Present Kuro5hin.org Inc.
See our legalese page for copyright policies. Please also read our Privacy Policy.
Kuro5hin.org is powered by Free Software, including Apache, Perl, and Linux, The Scoop Engine that runs this site is freely available, under the terms of the GPL.
Need some help? Email help@kuro5hin.org.
My heart's the long stairs.

Powered by Scoop create account | help/FAQ | mission | links | search | IRC | YOU choose the stories!