@[email protected] looks good! I am fluent in regex and SQL and I know some Clojure, but these datalog queries are still a bit of mystery to me… that’s the thing I need to visualise!
@Deebster I came a different route from data analysis (C, R, Python, Rust) but I’m less familiar with databases. I’m still trying to remind myself that Clojure, datalog, and Datomics are different … And some syntax is not available in Logseq so there is that … 😂
I did a lot of tinkering around recently to get an advanced query working for me which ended up being quite tricky to work through. I have Project pages (eg [[12335]] ) and on journal pages I have job note blocks for specific jobs ie #12335 Notes with a :job property so the block title can change if needed. There are multiple levels of notes / subnotes / tasks here and I was attempting to do the below query before I learned or-join, but the query was fragile & failing if tasks weren’t at a specific indent level. I ended up spending a Sunday afternoon deep diving into this stuff to figure this out.
As I understand it, the datomic data model is just a HUUGE list of ‘datoms’ which are super basic [element-id|attribute|value] rows for everything.
There is some concept of ‘unifying’ which is a variable that appears twice in a :where represents the same value across all clauses.
Something like (or-join) allows you to control this unification to selected sub items.
My visualization on the query is a graph of conditions
The :find (?task) element is absolutely required
There are ‘facts’ you want to satisfy [(get ?prop :job) ?job][(contains? #{"TODO""WAITING""DOING"} ?marker)].
?task → ?prop (through or-join) → ?prop must contain :job with value :current-page
. ↳ ?marker -> must be one of TODO / WAITING / DOING
#+BEGIN_QUERY
{
:title [:h3 "📅 Outstanding Tasks"]
:inputs [:current-page]
:query [
:find (pull ?task [*])
:in $ ?job
:where
(or-join [?task ?prop] ; only care that ?task and ?prop are 'unified' with rest of clauses
(and
[?task :block/page ?page]
[?page :block/properties-text-values ?prop] ; does page have :job property?
)
(and
[?task :block/parent ?tp]
[?tp :block/properties-text-values ?prop] ; does task parent have :job property?
)
(and
[?task :block/parent ?tp]
[?tp :block/parent ?tpp]
[?tpp :block/properties-text-values ?prop] ; does task grand-parent contain :job prop?
)
(and
[?task :block/parent ?tp]
[?tp :block/parent ?tpp]
[?tpp :block/parent ?tppp]
[?tppp :block/properties-text-values ?prop] ; does task great-grand-parent contain :job prop?
)
)
[(get ?prop :job) ?job] ; does one-of ?props from above contain :job <%current page%>?
[?task :block/marker ?marker]
[(contains? #{"TODO" "WAITING" "DOING"} ?marker)] ; ?task:block/marker must match one of these
]
:table-view? false
:result-transform (fn [result]
(sort-by (fn [m]
(get m :block/marker)) > result
)
)
:breadcrumb-show? false
:collapsed? false
}
#+END_QUERY
@[email protected] looks good! I am fluent in regex and SQL and I know some Clojure, but these datalog queries are still a bit of mystery to me… that’s the thing I need to visualise!
@Deebster I came a different route from data analysis (C, R, Python, Rust) but I’m less familiar with databases. I’m still trying to remind myself that Clojure, datalog, and Datomics are different … And some syntax is not available in Logseq so there is that … 😂
I did a lot of tinkering around recently to get an advanced query working for me which ended up being quite tricky to work through. I have Project pages (eg
[[12335]]
) and on journal pages I have job note blocks for specific jobs ie#12335 Notes
with a:job
property so the block title can change if needed. There are multiple levels of notes / subnotes / tasks here and I was attempting to do the below query before I learned or-join, but the query was fragile & failing if tasks weren’t at a specific indent level. I ended up spending a Sunday afternoon deep diving into this stuff to figure this out.:where
represents the same value across all clauses.[(get ?prop :job) ?job]
[(contains? #{"TODO" "WAITING" "DOING"} ?marker)]
.