Tags misbehaving

Ok. Very concerned. There are files missing from my database after I tried to move the tags in one of them into the tags list of my global inbox (as I’d like to have the same capacity to tag in both, it seems silly to have this be somehow limited). Now I have files in tags (which suddenly act like groups) that exist, but the same files in the home directory (House Icon, I don’t want to use groups just yet) cannot be found. What is going on here? Why would this work this way at all? The entire tags vs. groups distinction is very odd; It adds a complexity whose purpose I don’t understand. One part of DEVONthink I’d like to embrace is its flexibility to be used how I want to use it; I essentially want one database that works more like Yojimbo. Same interaction paradigm, all that. Then I want one database that’s highly organized and structured that contains my giant academic papers database and annotations and commentary on them, a situation where I can leverage the similarity functions of DT. I’d like one simple, the other complex. One anything bucket, one very purpose-built. I don’t want to buy Yojimbo, but it’s starting to seem like a better option (especially if someone develops a script to export to DT while including tags; I’d buy Yojimbo in a heartbeat).

I abandoned DT a few years ago because of some kind of data corruption back what the database used to be unified; I lost entire words from a huge library of quotations. This confusion with tags is sparking a similar feeling.

So, in closing, question: Why, conceptually, would it even be possible to have an item exist in a tag and not elsewhere in the library, and why would there still be an entry in my database of where the file once existed, while still giving me a File missing error?

I fixed this by selecting all my tags, moving them to the root of my database, and deleting all the duplicates. This seems kind of silly.

Two comments to make:

  1. I had been using DTPO, briefly went to EagleFiler, but came back when DTPO 2 came out - it has too many features that EF does not. I also used Yojmbo as my snippet library, it’s fast to get stuff into it. BUT I can’t use it as a general purpose information manager because the “folders” can’t be hierarchical. Aside from these one-dimensional folders (collections) you rely only on tags and tag collections. Too limiting for me. Plus, again, DTPO’s features far outnumber Yojumbo’s, and that’s what I need for a full featured information manager.

  2. I’m also struggling to get DT tags to work “my way.” I’ve been reading the new eBook on version 2, and the mantra seems to be that tags are groups and groups are tags. Well, almost. Yes, you can have a file in a tag group and not in any other group, just as you can have a file in a group but not in a tag group. I have set all my libraries so that groups do not add tags to contained flags. I want to organize files by “folder” and then tag them for finer granularity and with multiple tags if needed. You can do that, but you also need to recognize that the tag groups “look” very much like groups.

I pretty much decided to give up on DT as an anything bucket. I bought Yojimbo. DT will be used folders-only, for topics, tags will be subjects, and groups will be either how I encountered them or what they’re very generally about.

I’ve always been amused by the term ‘anything bucket’, and have never used DEVONthink that way, even in the early days when DT Personal with its single database was the only option.

I’ve never put all the files on my computer into DEVONthink. I use topical databases that reflect special interests and needs. Some of them are quite large (big enough to choke Yojimbo or EagleFiler), but still have a lot of headroom for continued expansion.

I’ve been working with computerized information resources since the 1960’s. In the old days, one was forced to file or tag every item up front as it was added to a database, else it was lost and couldn’t be found. People had to be hired to add keyword tags. If a tag was wrong, the file would be incorrectly pulled. If an important keyword wasn’t added, the file wouldn’t be found by a search for that keyword tag. It was at that point that I lost all respect for the enterprise of accurately and reliably tagging documents a priori (up front), as tagging in practice is subjective, inconsistent and usually incomplete unless an inordinate amount of time is spent on each document. And of course there are similar problems with attempting to file items into a detailed organizational structure.

The ability to do full-text searches of documents was revolutionary, although it is common now. That doesn’t mean, however, that the ability to search for text necessarily results in fast search results. For example, putting large text fields into a relational database such as FileMaker can be a very frustrating experience; FileMaker is a pretty lousy document manager. Even a big and powerful database such as Oracle can be agonizingly slow if it is told to search for text that wasn’t specifically set up for fast retrieval. I’ve seen an Oracle database running on a big mainframe take hours to sift out text that it hadn’t been programmed to index and find.

Nowadays indexing of text for fast searching and retrieval has become common on personal computers. Spotlight does that, Yojimbo and EagleFiler do that, and there are many other examples.

DEVONthink doesn’t index text in the same way. There are AI routines at the very core of the database that not only ‘know’ where every occurrence of a word is, for fast retrieval, but also ‘analyze’ contextual relationships of the words in the documents. That’s the basis of the ‘Classify’, ‘See Also’, ‘See Related Text’ and Search ranking features of DEVONthink, and they set DEVONthink apart from other document managers. As a database continues to grow (especially if it’s topically designed, i.e., with a degree of coherence in content) these special features of DEVONthink become more and more useful.

What this means is that in the databases in which I spend most of my time for research and writing, I don’t spend much time in group organization of the content and almost never bother to tag items as they are added. I use those special features of DEVONthink to explore the content of the database for useful information when I embark on a new project, and they help me look at that content in new ways that wouldn’t have happened if I had depended on finding information just by an organization or tagging scheme. As I’m working on that project, I’ll probably create a new group for it and will probably tag some items - usually removing project-specific tags when the project is finished, so that they won’t get in the way of the next project.

On the other hand, there are some databases in which I do quite detailed up front organization or tagging of items as they are added, such as my database holding financial information — banking and investments reports and forms, invoices and receipts that are important at tax filing time, etc. As I will almost always use that content in the same way, it makes sense to ‘catalog’ it up front. I won’t spend much time exploring and thinking about a receipt; it’s what it is, and so can easily be grouped or tagged.