Clojurians Log v2

I understood that the order of my queries should make no difference in Datalog, is that correct?

I am wondering, because these two give wildly different results, despite the only difference being the order of the missing? and subproject clauses:

[:find [?subproject ...]
:in $ % ?project-phid
:where
[?project :object/phid ?project-phid]
[(missing? $ ?subproject :project/milestone)]
(subproject ?subproject ?project)]

[:find [?subproject ...]
:in $ % ?project-phid
:where
[?project :object/phid ?project-phid]
(subproject ?subproject ?project)
[(missing? $ ?subproject :project/milestone)]]

@urzds Not correct at all, order makes all the difference.

You should put the one that reduces the resultset the most first (if that's possible to know).

Datalog is effectively a constraint solver, and it applies each constraint in order.

Can you give an example? I expected it to behave like (filter pred-b (filter pred-a db)), where I would expect pred-a and pred-b to be interchangeable.

Actually, I'm not sure it should affect the resultset, but it definitely can affect performance by limiting the intermediate resultsets.

So it does make some difference

When I used negations, order radically affected the result set.

Order seems to matter a lot.

How I do make a variable a union of two sets? e.g. when I want to select all ?task that are both in ?required-project and in either ?top-level-project or ?subproject:

'[:find ?task
  :in $ % ?top-level-project
  :where
  [(subproject ?subproject ?top-level-project)
   [?project is either ?top-level-project or ?subproject] ; <<--
   [?task :task/projects ?project]
   [?required-project :project/name "Required"]
   [?task :task/projects ?required-project]]]

The last thing I tried was:

'[:find ?task
  :in $ % ?top-level-project
  :where
  (or-join [?task]
    (and
      (subproject ?subproject ?top-level-project)
      [?task :task/projects ?subproject])
    [?task :task/projects ?top-level-project])
  [?required-project :project/name "Required"]
  [?task :task/projects ?required-project]]

'[:find ?task
  :in $ % ?top-level-project
  :where
  (or-join [?project]
    (subproject ?project ?top-level-project)
    (and
      [?top-level-project :unique-id ?top-level-id] ; stunt to make ?project also an alias for ?top-level-project
      [?project :unique-id ?top-level-id]))
  [?task :task/projects ?project] ; match against the union constructed above
  [?required-project :project/name "Required"]
  [?task :task/projects ?required-project]]

'[:find ?task
  :in $ % ?top-level-project
  :where
  [?required-project :project/name "Required"]
  [?task :task/projects ?required-project]]

This variant works:

'[:find [?task ...]
    :in $ % [?top-level-project ...]
    :where
    (subproject-or-self ?project ?top-level-project)
    [?task :task/projects ?project]
    [?required-project :project/name "Required"]
    [?task :task/projects ?required-project]]

'[[(subproject ?e1 ?e2)
     [?e1 :project/parent ?e2]]
    [(subproject ?e1 ?e2)
     [?e1 :project/parent ?t]
     (subproject ?t ?e2)]
    [(subproject-or-self ?e1 ?e2)
     [?e1]
     (subproject ?e1 ?e2)]]