If the DN substr index is not available, does this mean that
searches will find everything that matches the search attributes,
rather than in the "subtree" of the base DN to search for ?
Also, is enabling the DN subindex achieved by simply adding an
appropriate entry in the slapd.conf file ? And is it also subject
to the allIds limit (i.e. if there are more than 4K entries in the
database, it becomes an ALLID block and is useless in searches as
far as performance goes ?)
Thanks, Vinod.
-----Original Message-----
From: Kurt D. Zeilenga [mailto:Kurt@OpenLDAP.org]
Sent: Tuesday, May 16, 2000 11:59 AM
To: Vinod Nair
Cc: 'openldap-software@openldap.org'; Vinod Nair
Subject: Re: tuning searches
At 10:16 AM 5/15/00 -0500, Vinod Nair wrote:
>In a compound search, such as
> (&(attr1=value1)(attr2=value2)(attr3=value3))
>
>is the order of evaluation left-to-right ?
In OpenLDAP, yes.
>I'm assuming the first _expression_ prunes the search
>space for the 2nd, and so forth. In that case, it
>would make sense to place the "most restrictive"
>_expression_ up front (and have it indexed).
Basically, yes.
>Also, how is the "search base" used to prune the search ?
Depends upon the version of OpenLDAP and configuration.
In some versions, DN sub index was enabled by default. However,
I believe it is off in current versions.
>Is this used to get an appropriate "range" from the underlying btree ?
No. back-ldbm treates the underlying datastore as a simple
key/value system. It may or may not be a btree.
>Or is the base DN being used in some other way to prune the search space ?
If a substr DN index is available, it is used.