morning!

good morning!

morning

@ordnungswidrig just makess it tricky to create a series of intervals where the interval switches at one point on the same date

but yeah, I'm not sure there is a good answer

The problem is that were lacking the notion of "rounding" with timestamps. It's clear to everybody that the integer 2 "corresponds" with the rational interval [1.0...2.0)

We have a date notions of the interval as e.g. "2020-05-03" but all libraries I know internally represent this to the revision of the nanosecond.

It's like using floats to represent integer values. Prepare for some weird confusing stuff to happen :-)

Morning

@ordnungswidrig yeah, I think you are right there. I wonder if there is other prior art like the relational algrebra that they borrow that makes this behaviour a bit more principled. I think my worry is that any work around I do could change in a later version if it is undefined

Something like “floating point dates”. One property of FP is the change of absolute precision.

But dates are weird because they are a weird mix of absolute values, intervals, the use of arbitrary, mixed and changing units, UI features which change over time, non-unique representations (DST change) and bad compatibitlity with the binary and decimal system.

java.time is generally pretty good at this stuff though, since they actually recognize that a date is different from a "year-month", is different from a timestamp, etc. for instance LocalDate is really just a year+month+day fields, no unix timestamp in sight https://github.com/frohoff/jdk8u-jdk/blob/master/src/share/classes/java/time/LocalDate.java#L169-L177

the problem to me in the above @otfrom is that you're asking for an interval between dates, but you're getting an interval between date-times

(import '(java.time LocalDate Period))

(Period/between
 (LocalDate/of 2020 01 01)
 (LocalDate/of 2020 01 02))
;; => #object[java.time.Period 0x35860e12 "P1D"]

@plexus java.time has some good stuff in it

Although now I’m confused that LocalDate doesn’t implement a single Interface which is parametrized over a dozen generics 😛

That’s weird public boolean isSupported(TemporalField field) that sounds where you would use types for.

OMG “* Defend against malicious streams.“. I had enough java for today

@plexus tick wraps java.time and has durations and periods

and interval is what I need as it is anchored in calendar dates

interval is an implementation of this: https://en.wikipedia.org/wiki/Allen%27s_interval_algebra

https://www.juxt.land/tick/docs/index.html#_introduction

looks like my real issue is whether or not this behaviour is correct:

(t/bounds (t/today))
  #:tick{:beginning #time/date-time "2020-07-09T00:00", :end #time/date-time "2020-07-10T00:00"}

or if :end should be 23:59 on the same day

(I'll still have the issue as my end of episode 1 and start of episode 2 are on the same day)

if only there was some way I could write some unit tests to make sure the behaviour stays the same as I change the code... 😉

right... I think this is the core of my issue;

(t/end (t/date "2020-07-09"))
  #time/date-time "2020-07-10T00:00"

rather than being 2020-07-09T23:59

so where does the minute between 2020-07-09T23:59 and 2020-07-10T00:00 go (or second or nanosecond)? seems to me to make sense that (= (end-of today) (start-of tomorrow)), it's the same point on the timeline

yeah, I think it is just a conversion problem with where you draw the bounds when you only have something less precise

or that you're deriving date-level semantics for timestamp-level values

should an interval that is {:start 2020-01-02 :end 2020-01-03} and {:start 2020-01-03 :end 2002-01-05} be considered as overlapping or meeting on 3 January?

yeah, the problem is the library is doing the conversion for me rather than comparing the dates

if you're talking about dates, and your intervals are ending exclusive, then those two would meet but not overlap

(t/relation
   #:tick{:beginning #time/date-time "2020-07-09T00:00", :end #time/date-time "2020-07-10T00:00"}
   #:tick{:beginning #time/date-time "2020-07-10T00:00", :end #time/date-time "2020-07-11T00:00"})
  :meets

so when the date times are equal, then the relation meets, but if only dates are supplied and equal then they overlap

so I have to pre-process my dates into suitable interval friendly date times before supplying them to t/relation

yeah looks like it

which is fine as long as the conversion from date to date time bounded interval is stable behaviour

looks like I can use t/at and t/on rather than doing the arithmetic at least

(t/at (t/date "2018-01-01") (t/time "13:00")) => 2018-01-01T13:00

The weird things is that the answer to do these intervals overlap is quite obvious {:beginning 10 :end 15} {:beginning 15 :end 20}

.. it depends! On wether the ends are open or closed.

I think the behaviour of the date time example shows how it works: https://clojurians.slack.com/archives/CBJ5CGE0G/p1594287276141400

I think I'd down to my issue being how dates are converted to intervals, and I think I have a solution to it now

tho it feels a bit hacky

so, I've gone with this in my code and it gives me the :meets behaviour I want when creating relations

(defn episode-interval [{:keys [report-date ceased] :as rec}]
  (if ceased
    (assoc rec
           :episode-interval
           (t/new-interval (t/at report-date "13:00") (t/at ceased "13:00")))
    rec))

the common code for this is going here: https://github.com/MastodonC/witan.cic.driver

tho the code specific to the clients is in a private repo

morning

This is intereting https://www.w3.org/TR/owl-time/

(teaser: it’s not about owls 😞 )

is the Web Ontology Language widely used? I almost never come across it for my work But that might just be because I'm boxed into a niche

will probably have some RDF impact (as they are often the same people) and thus impact on Rich who is a fan of RDF (see schema, datomic, etc)

and it is often well thought through and built on prior art

so it looks like you can't create a LocalDateTime from a LocalDate w/o supplying a time explicitly https://docs.oracle.com/javase/8/docs/api/java/time/LocalDateTime.html#of-java.time.LocalDate-java.time.LocalTime-

which makes sense

so I'm tripping over the sugar in tick then

does anyone work with hiccup here? or other templating engines?

I’m wondering how to pass global objects to the templates… things that shall be rendered on all sites.

I have a few ideas, like using thread-locals or so… or ^:dynamic (it’s technically also a thread local, right?), or injecting some payload into the request object and pass that to the templates, but somehow it doesn’t feel right…