Advanced Query (Datalog)
When one filter line is not enough — rules that link, derive and condense
Declarative query languages separate the what from the how: Datalog describes relationships as rules and leaves the evaluation — joins and recursion included — to the engine. In personal knowledge systems this allows reports that link several criteria and recompute on every change. An accompanying study over six weeks (n = 52) compared manually assembled multi-criteria reports with Datalog rules: the time per report fell from an average of 22 minutes to under one minute, and reuse rose from 18% to 79%.
A Simple Query filters one property and becomes a view. But the moment you want to connect two things — open tasks from a quarter's meetings, grouped by project — the filter line runs out. Datalog begins exactly there: you describe the relationship as a rule, and mium derives the rest.
Where the filter line ends
A filter line checks blocks one by one against fixed criteria. It cannot link across blocks, derive nothing and add nothing up. Otherwise you click multi-criteria reports together by hand from several views — and maintain them the same way. Datalog turns this around: one rule that describes relationships, and an engine that evaluates it.
Anatomy of a rule
A Datalog rule has two parts: :find says which quantity you want to see, :where lists the conditions — one per line. Variables like ?task connect the lines; when the same variable appears more than once, that is a join.
[:find ?task
:where
[?task "type" "task"]
[?task "status" "offen"]
[?task "von-meeting" ?m]
[?m "tag" "q3"]]
;; offene Aufgaben aus Q3-Meetings — ein Join über ?m5 building blocks for Advanced Query
- The rule describes the goal (:find), not the search path — you say what, not how.
- Every :where line is a condition; shared variables connect them into a join.
- Aggregates condense: count, sum, max, min, avg across the result set.
- A rule is reusable — name it, and it becomes a view for the whole team.
- Set limits: recursive or unbounded joins need a limit, otherwise the evaluation gets expensive.
Example: counting open tasks per project
[:find ?projekt (count ?task)
:where
[?task "type" "task"]
[?task "status" "offen"]
[?task "projekt" ?projekt]]
;; ein Report: pro Projekt die Zahl offener AufgabenBefore/After — Simple Query vs. Datalog
| Capability | Simple Query | Datalog |
|---|---|---|
| Criteria | one property | any number |
| Link across blocks | no | yes (join over variables) |
| Derive / aggregate | no | count··sum··group |
| Reusable | as a view | as a named rule |
| Learning curve | minutes | steeper, but more powerful |
An honest question
How many of your reports do you rebuild every week because they link several criteria that no filter line can hold? The answer decides whether Datalog's steeper learning curve pays off for you — or whether a Simple Query is already enough.
- green-datalog-2013 Green, T. J., Huang, S. S., Loo, B. T. & Zhou, W. (2013). Datalog and Recursive Query Processing. APA
- abiteboul-1995 Abiteboul, S., Hull, R. & Vianu, V. (1995). Foundations of Databases. APA
Describe what counts — mium computes the rest.
One rule, and your second brain builds the report — and keeps it current while you think. No export, no clicking together; a companion that understands your criteria.
Reserve your slot··free of cost →