Getting the nearest instances from the past and the future - postgresql

Question

I've got a table that contains appointment records, and the user selects a date range (begin_date, end_date). I would like to get the appointment that falls in this date range, as well as the closest instances in the past and future that fall outside of this date range (aka, the previous occurrence and the next occurrence).
I think the best way to approach this is using CTEs and self joins, but I'm open to another strategy. This is my current query:
WITH present AS
(SELECT appt.ewssubject, appt.ewsstart::DATE, appt.ewsend::DATE
FROM appointment appt
WHERE (appt.ewsstart, appt.ewsend) OVERLAPS (begin_date::DATE, end_date::DATE))
SELECT
present.ewssubject, present.ewsstart, present.ewsend,
past.ewssubject AS pastsubject, past.ewsstart::DATE AS paststart,past.ewsend::DATE AS pastend,
future.ewssubject AS futuresubject, future.ewsstart::date AS futurestart, future.ewsend::date AS futureend
FROM present
LEFT JOIN appointment AS past USING (ewssubject)
LEFT JOIN appointment AS future USING (ewssubject)
WHERE
present.ewssubject = past.ewssubject AND
present.ewssubject = future.ewssubject AND
past.ewsend < present.ewsstart AND
future.ewsstart > present.ewsend
ORDER BY present.ewsstart ASC
I'm getting a huge list of appointments, and there are a lot of repeats -- like so:
subject start end last_start last_end next_start next_end
DINNER 2015-09-18 2015-09 18 2015-09-17 2015-09-17 2015-09-19 2015-09-19
DINNER 2015-09-18 2015-09 18 2015-09-17 2015-09-17 2015-09-19 2015-09-19
... // more repeats! :(
All I want to do is reduce the number of duplicates, like this format:
subject start end last_start last_end next_start next_end
DINNER 2015-09-18 2015-09-18 2015-09-17 2015-09-17 2015-09-19 2015-09-19
DINNER 2015-09-21 2015-09-21 2015-09-18 2015-09-18 2015-10-02 2015-10-02
... // and so on
n.b. An appointment can span multiple days.
How can I fix my query? Or is there another one I can write?

Answer 1

You did not enter details about your data so I'm not sure whether this is a good idea, but you can probably use window functions:
select
ewssubject, ewsstart, ewsend,
lag(ewsstart) over (partition by ewssubject order by ewstart) prior_start,
lag(ewsend) over (partition by ewssubject order by ewstart) prior_end,
lead(ewsstart) over (partition by ewssubject order by ewstart) next_start,
lead(ewsend) over (partition by ewssubject order by ewstart) next_end
from appointment
order by ewstart;