agora inbox for pgsql-sql@postgresql.org  

help / color / mirror / Atom feed

From: Adam Jensen <hanzer@riseup.net>
To: pgsql-sql@lists.postgresql.org
Subject: Re: interval origami
Date: Fri, 30 Nov 2018 19:04:10 -0500
Message-ID: <80ea3708-11be-e7e3-9f49-011a682e3bf4@riseup.net> (raw)
In-Reply-To: <6d32b24f-2d3e-df79-81d2-3b1abb8cc4b4@riseup.net>
References: <20181130175719.l7izv65wjwgkls35@alvherre.pgsql>
	<d05565aa-551d-fac3-ba88-8215c8cbe0d4@riseup.net>
	<6d32b24f-2d3e-df79-81d2-3b1abb8cc4b4@riseup.net>

On 11/30/18 4:02 PM, Adam Jensen wrote:
> On 11/30/18 3:19 PM, Adam Jensen wrote:
>> The 'numrange' type with the 'overlaps' and 'intersection' operators
>> seem to cover the fundamental computations in a very natural way.
> 
> Actually, those operators might not be entirely sufficient. Given two
> ranges like this:
> 
> 10.0|39.0|interesting
> 15.0|21.0|fail
> 
> Something like the negative or inverse of the intersection is needed:
> 
> 10.0|15.0|interesting
> 21.0|39.0|interesting

I've mapped out nine time segment overlap scenarios:

1. good(10, 40) | bad(05, 15) -> good(15, 40)
2. good(10, 40) | bad(10, 15) -> good(15, 40)
3. good(10, 40) | bad(20, 30) -> good(10, 20), good(30, 40)
4. good(10, 40) | bad(20, 40) -> good(10, 20)
5. good(10, 40) | bad(20, 45) -> good(10, 20)
6. good(10, 40) | bad(05, 40) -> good()
7. good(10, 40) | bad(05, 45) -> good()
8. good(10, 40) | bad(10, 40) -> good()
9. good(10, 40) | bad(10, 45) -> good()

Letting gs/gf and bs/bf represent "good start-time"/"good finish-time"
and so on, pseudo-code to remove the bad segments looks like this:

find overlap: good(gs, gf) | bad(bs, bf)

CASE
	WHEN ((bs <= gs) AND (bf <  gf)) THEN  # 1 & 2
		-> (bf, gf)
	WHEN ((bs >  gs) AND (bf <  gf)) THEN  # 3
		-> (gs, bs), (bf, gf)
	WHEN ((bs >  gs) AND (bf >= gf)) THEN  # 4 &
		-> (gs, bs)
	WHEN ((bs <= gs) AND (bf >= gf)) THEN  # 6 & 7 & 8 & 9
		-> ()
END CASE;

And my first attempt at writing a PostgreSQL function looks like this:

CREATE FUNCTION find_overlap(gs REAL, gf REAL, bs REAL, bf REAL)
	RETURNS TABLE (start REAL, stop REAL) AS $$
	BEGIN
		CASE
			WHEN ((bs <= gs) AND (bf <  gf)) THEN
				RETURN NEXT (bf, gf);
				RETURN;
			WHEN ((bs >  gs) AND (bf <  gf)) THEN
				RETURN NEXT (gs, bs);
				RETURN NEXT (bf, gf);
				RETURN;
			WHEN ((bs >  gs) AND (bf >= gf)) THEN
				RETURN NEXT (gs, bs);
				RETURN;
			WHEN ((bs <= gs) AND (bf >= gf)) THEN
				RETURN;
		END CASE;
	END; $$
LANGUAGE plpgsql;

It results in:

ERROR:  RETURN NEXT cannot have a parameter in function with OUT parameters
LINE 6: RETURN NEXT (bf, gf);

Page 83 of the book "PostgreSQL Server Programming" mentions this
situation but doesn't actually describe or explain anything; nor does it
present a working example...

Any ideas?

view thread (10+ messages)  latest in thread

Message-ID: <80ea3708-11be-e7e3-9f49-011a682e3bf4@riseup.net>
Permalink:  ../80ea3708-11be-e7e3-9f49-011a682e3bf4@riseup.net/
Also on:    postgresql.org/message-id/80ea3708-11be-e7e3-9f49-011a682e3bf4@riseup.net

reply

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Reply to all the recipients using the --to and --cc options:
  reply via email

  To: pgsql-sql@postgresql.org
  Cc: hanzer@riseup.net, pgsql-sql@lists.postgresql.org
  Subject: Re: interval origami
  In-Reply-To: <80ea3708-11be-e7e3-9f49-011a682e3bf4@riseup.net>

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

This inbox is served by agora; see mirroring instructions
for how to clone and mirror all data and code used for this inbox