Message-ID: <4FE28AD5.5090702@ringerc.id.au>
Date: Thu, 21 Jun 2012 10:45:41 +0800
From: Craig Ringer <ringerc@ringerc.id.au>
User-Agent: Mozilla/5.0 (X11; Linux x86_64;
	rv:13.0) Gecko/20120605 Thunderbird/13.0
MIME-Version: 1.0
To: sthomas@optionshouse.com
CC: Eyal Wilde <eyal@impactsoft.co.il>, pgsql-performance@postgresql.org
Subject: Re: index-only scan is missing the INCLUDE feature
References: 
 <CAMiEbcjxN7cRYGtRMZuxEoztEOMB4us+pSdanwSkRY=ocJiQ8Q@mail.gmail.com>
	<4FE1D9F9.1020805@ringerc.id.au>
	<4FE1ED03.3090709@optionshouse.com>
In-Reply-To: <4FE1ED03.3090709@optionshouse.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit

On 06/20/2012 11:32 PM, Shaun Thomas wrote:
> On 06/20/2012 09:11 AM, Craig Ringer wrote:
>
>> For those of us who don't know MS-SQL, can you give a quick
>> explanation of what the INCLUDE keyword in an index definition is
>> expected to do, or some documentation references?
>
> He's talking about what MS SQL Server commonly calls a "covering 
> index." In these cases, you can specify columns to be included in the 
> index, but not actually part of the calculated hash. This prevents a 
> trip to the table data, so selects can be serviced entirely by an 
> index scan.

Oh, OK, so it's a covering index with added fields that don't form part 
of the searchable index structure to make the index a little less 
expensive than a fully covering index on all the columns of interest. 
Fair enough. Thanks for the explanation.

Eyal, you'll get a better response to questions about other DBMSs if you 
explain what you need/want to do with the desired feature and what that 
feature does in the other DBMS.
>
> PostgreSQL is about half way there by allowing index-only scans, 
> though I've no idea if they intend on adding further functionality 
> like this.

There's certainly lots of interest in adding more, but not that many 
people with the expertise to be able to do it - and fewer still who're 
paid to work on Pg so they have time to focus on it. Covering indexes 
with Pg's MVCC model seem to be particularly challenging, too.

--
Craig Ringer