public inbox for [email protected]  

help / color / mirror / Atom feed

From: Heikki Linnakangas <[email protected]>
To: Tom Lane <[email protected]>
To: Greg Stark <[email protected]>
Cc: Bruce Momjian <[email protected]>
Cc: Gavin Flower <[email protected]>
Cc: David E. Wheeler <[email protected]>
Cc: Robert Haas <[email protected]>
Cc: Andrew Dunstan <[email protected]>
Cc: Peter Geoghegan <[email protected]>
Cc: [email protected] <[email protected]>
Subject: Re: default opclass for jsonb (was Re: Call for GIST/GIN/SP-GIST opclass documentation)
Date: Sat, 10 May 2014 23:42:34 +0300
Message-ID: <[email protected]> (raw)
In-Reply-To: <[email protected]>
References: <[email protected]>
	<[email protected]>
	<[email protected]>
	<[email protected]>
	<[email protected]>
	<[email protected]>
	<[email protected]>
	<[email protected]>
	<[email protected]>
	<[email protected]>
	<[email protected]>
	<CAM-w4HOQ+ynm70Cgr4BwMqMc0jfL_8EZAwG2UD6r=EehjH3OOw@mail.gmail.com>
	<[email protected]>
List-Unsubscribe: <mailto:[email protected]?body=unsub%20pgsql-hackers>

On 05/09/2014 11:44 PM, Tom Lane wrote:
> Greg Stark <[email protected]> writes:
>> Well the question seems to me to be that if we're always doing recheck
>> then what advantage is there to not hashing everything?
>
> Right now, there's not much.  But it seems likely to me that there will be
> more JSON operators in future, and some of them might be able to make use
> of the additional specificity of unhashed entries.  For example, it's only
> a very arbitrary definitional choice for the exists operator (ie, not
> looking into sub-objects) that makes jsonb_ops lossy for it.  We might
> eventually build a recursive-exists-check operator for which the index
> could be lossless, at least up to the string length where we start to
> hash.

Back to the naming:

The main difference between the two opclasses from a user's standpoint 
is not whether they hash or not. The big difference is that one indexes 
complete paths from the root, and the other indexes just the "leaf" 
level. For example, if you have an object like '{"foo": {"bar": 123 } 
}', one will index "foo", "foo->bar", and "foo->bar->123" while the 
other will index "foo", "bar" and "123".

Whether the opclasses use hashing to shorten the key is an orthogonal 
property, and IMHO not as important. To reflect that, I suggest that we 
name the opclasses:

json_path_ops
json_value_ops

or something along those lines.

- Heikki


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers

view thread (65+ messages)  latest in thread

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: [email protected]
  Cc: [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected]
  Subject: Re: default opclass for jsonb (was Re: Call for GIST/GIN/SP-GIST opclass documentation)
  In-Reply-To: <[email protected]>

* 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