Feature Request: Documents organized in hierarchical folders

I started this thread because when I started storing large files in DEVONthink which have to be used by applications outside of DEVONthink because of DEVONthink’s limited ability to modify these files, I started seeing cracks in DEVONthink’s ability to efficiently and quickly manage these files. In other words, DEVONthink - itself - was slowing me down.

As I pointed out, DEVONthink freezes for a while when apparently reindexing large PDFs which I just minimally highlighted in Adobe. This make is very difficult to spontaneously switch back and forth between numerous large PDFs in DEVONthink.

This is a situation where EagleFiler is actually faster to use than DEVONthink - even for a few large PDFs - say less than 100.

No, those databases do not change the organization of files for performance optimization.

I agree that DEVONthink’s AI as represented by the See Also assistant doesn’t understand concepts or ideas. It has no training in intellectual disciplines. Nevertheless, I find it a great tool for exploring concepts and ideas.

My work also involves a number of scientific disciplines, as well as associated policy issues.

An example I often cite was the result of invoking See Also when I was reading a paper about the impacts of invasive species on the population dynamics of native species in an ecosystem. One of See Also’s suggestions was a paper on factors that affect chemical reaction equilibria such as catalysts, temperature, pressure, etc. Although the terminology used in those two papers was quite different, in analyzing the contextual relationships of terms in documents within that database, which contained some 30,000 documents, See Also “saw” similarity in the dynamic principles discussed in the papers. It was a useful suggestion, and illustrates why I do find See Also a useful tool for exploring ideas.

Because See Also is examining contextual relationships of words throughout the database, it can “bridge” relationships among terms that might not have been detected by an examination of just those two papers.

When I examine a list of possibly similar documents proposed by See Also, many of them are rejected quickly as not interesting. But often enough, a suggestion is found that is unexpected and interesting, providing me with an insight that I hadn’t thought of. The human mind, although perhaps highly trained in one or more disciplines, isn’t wired to simultaneously grasp the contextual relationships of words among tens of thousands of documents. DEVONthink and a Mac, although blissfully unaware of meaning and lacking any understanding of the documents in a human sense, can do that almost instantly. That’s why I regard DEVONthink as the best research assistant I’ve ever had (as well as the least expensive one).

In his recent book, Where Good Ideas Come From: The Natural History of Ideas, Steven Johnson discusses in some detail how he uses DEVONthink’s See Also to explore ideas. It’s a good read.

I’m using DT in “Indexed Database mode”.

Some toughts about the topic:

  1. I don’t see any slow down due to indexed files (as opposit to imported). Maybe a serch action is 3 times slower (taking 66 ms instead of 11 ms) but I prefer to trade that 44 ms with the advantages of a more clear files hierarchy on disk.

  2. I think that DevonTechnologies had introduced the indexed mode because they realized that a lot of people prefer to keep their files in a standard folder structure. But, once started a way, you have to go to the end. Any changes to DT groups hierarchy should be immediately mirrored on the Finder folder structure (as done by other competitors); that’s the only way to go.
    E.g.: if it so easy to perform the 3 steps moving of indexed files (move into datbase + move to another folder + move to external folder), the user can legitimate ask DT to do it automatically (out of the box; no scipt).

  3. Someone says that this way DT group hierarchy simply becomes a duplication of Finder structure. That’s true.
    In any case you can only have ONE hierarchical structure containing your data; it can be either DT abstraction (groups) or physical folder structure on disk. In the latter case it is clear that groups are reduced to a mirror but you’re loosing nothing.
    You still have your unique hierarchical structure.

  4. Replicants. It is clear that they’re not real files and can therefore be managed only inside DT database (indexes) without a counterpart on disk.