Default search sorting

DEVONthink
1

Couple of short questions I can’t find the answer to. Is it possible to always show search results by Date Modified? Whenever I search it always gives me the results sorted by whatever that heatbar represents. I’m not sure what goes into that, but it doesn’t weight recency as much as I’d like, so I have to manually sort by date modified each time.

Second question, when I do sort by Date Modified why does it put all tags on top? It means my non-tag results get pushed down the page, even if the tag hasn’t been modified in several years.

Couldn’t see a reference to this in the manual, or anything in preferences. The chat window (via ChatGPT) suggested using the View→Sort menu, but that didn’t work.

DT 4.1.1

2

The date doesn’t matter. However, there’s a hidden preference PersistentSortingOfSearchResults to remember the last used sorting, see hidden preferences in help.

Is the option to Keep groups on top when sorting enabled, see Settings > General > Interface?

3

Great, that kept the Search sort setting.

On the other point, I didn’t realise tags were groups, but that makes sense. Turning it off worked. However I prefer to have that setting on in general use, while I’d rather it were off in Search Results. If it can’t be turned off just for Search, perhaps that could be added to the (no doubt very long) list of feature requests.

many thanks

4

Longer than could be accomplished in our combined lifetimes :wink:

1 Like
5

Can I ask for some more details on what goes into the heat map?

I was looking for a file I’d been reading less than a week ago called “Head teacher report December 2025 draft.docx” but when I search for “head teacher report december 2025”, it’s only 22nd in the list. It has 17 hits. If I sort by hits, its first in the list.

4th in the list when sorted by heatmap is a report I added 8 months ago, called “[My company name] Annual Report 2005-06”. It has 4 hits, and no references to Head, Teacher, December or 2025.

I can’t see to see the rationale in how in how the latter is so far above the former. Any insights you can share?

thanks

6

I can’t offer any insight into how results are scored, but you’ll probably get more out of the search if you improve your queries.

Your example translates to:

head AND teacher AND report AND december AND 2025

across all possible attributes – name, text content, comments, tags, labels, custom metadata fields etc. What search options are enabled also plays a role – fuzzy word comparison, related words, partial matches while typing?

I suggest reading up on the various search operators and prefixes in the manual; DEVONthink’s search is really powerful.

If you’re searching for a file by name, you would likely see huge improvements just from using the name prefix. And if you remember a part of the name exactly, put that in quotes to make it even better. Try these variations and see:

name:report 2025

name:"head teacher report" 2025

name:"head teacher" rep* 2025

name:"head teacher report" dec* 2025

If you want to restrict your search to recently added files, try:

report added:#30

name:report added:#30

Since you were reading this file less than a week ago:

name:"head teacher report" opened:#7

Or, say you have a global smart group for recent files. Open it and type in the search bar – you can use it as the search scope to pre-filter results.

2 Likes
7

@troejgaard beat me to it, but I agree your search is not optimal.

8

I know I can manually add criteria. I was hoping that Devonthink’s smarts would offer me some assistance. Hence I’m trying to understand how the heatmap works, if it is at all for that purpose.

9

My understanding is that the heat-map in the results list is just a visual representation of the search score – i.e. how well an item matches the query. (I might be wrong.) I don’t know how that score is calculated exactly, and I would think the algorithm is a company secret. But the score would naturally depend on the query…

The heat-map/score in the See Also inspector is different, since there is no user query. I think it’s based on similarity.

10

What you say with your search is basically “I don’t care where these terms occur, just give me anything containing at least one of them”. Which DT does, and the order will be whatever DT finds appropriate.
It’s alo about specificity. So, if you search for a particular name, use the name operator and quotes.

1 Like
11

I assume that’s true. And this “score” may be useful for some people. It’s of no use to me. I wish there was a way to hide this column. For me, it just takes up unnecessary space.

12

I’m all for customization, but I truly don’t understand you here – and I actually think it’s sensible that it can’t be hidden.

First of all, it clearly telegraphs that you are viewing search results contra any other item list.

But more importantly: How is it not useful to see how well a search result matches the query?? I really can’t wrap my head around that.

  • For example, a search might return only low-scored results. That’s very useful to see at a glance. Either you need to carefully comb through the results – or maybe you need to write a better query.
  • It lets you sort results by any other column while still seeing the score against the query – giving you different angles of approach.
3 Likes
13

If you don’t know why a search query is resulting in a low heatmap score, then its not giving you any information you can improve your search query with.

14

That should be a very compelling reason. And even if I understood what DT does, it wouldn’t be what I want.

I always sort the results by date (hidden preference PersistentSortingOfSearchResults).

In addition, I search so precisely that I don’t get an infinite number of results.

All I get telegraphed every time is that I don’t need this column :slightly_smiling_face: :man_shrugging:

But as I said, I understand that others find it useful. I just would like this column to be optional so it can be hidden. That’s not going to happen, and I can live with that :slightly_smiling_face:

1 Like