directory-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Emmanuel Lecharny <>
Subject Re: Question regarding FilterVisitor implementation
Date Sat, 09 Sep 2006 18:08:52 GMT
Stefan Zoerner a écrit :

> Hi all!
> I use ApacheDS 1.0 RC4 and have a question regarding FilterVisitor 
> from, and related stuff.
> I have programmed a simple interceptor which determines attribute 
> types most frequently used in search filters, and periodically writes 
> a report about them to disk. I plan to use it for index optimization 
> in my partition definition. Everything works well, but one question 
> remained unsolved for me.

Very interesting piece of work ! I think this will be very usefull in 
the soon future.

> I have overwritten the search method of BaseInterceptor and created an 
> inner class which implements FilterVisitor and looks like this:
>    class AttributeDetector implements FilterVisitor {
>        public boolean canVisit(ExprNode node) {
>            return true;
>        }
>        public ArrayList getOrder(BranchNode branchNode, ArrayList 
> nodes) {
>            return nodes;
>        }
>        public boolean isPrefix() {
>            return true;
>        }
>        public void visit(ExprNode node) {                           if 
> (node instanceof LeafNode) {
>                LeafNode leaf = (LeafNode) node;
>                addAttributeOccurence(leaf.getAttribute());
>            }                  }
>    }
> I use it within search to analyze the filter.
> My question is about the result of leaf.getAttribute() (which is a 
> String). In most cases it contains the OID of the attribute type, but 
> sometimes, it contains the name ("cn" instead of ""). It seems 
> that the name is returned in case of substring expressions.

Strange. From what I understand of the code - and sure Alex can give you 
a better explanation -, there is a 'visit' method in the 
NormalizingVisitor which is applied to all the attributes found in a 
filter into the It seems that for some 
attributes, like ObjectClass and cn, there is no transformation to an 
OID form. I don't know if it's a bug, or if the search fails because of 
this behavior, but definitively this need some further investigation.

> For instance in "(&(objectClass=*)(sn=Amos)(cn=Tor*))" for the first 
> two operands within the AND expression, the OID is returned,for the 
> third one "cn". I have created a work around with the help of 
> AttributeTypeRegistry. Lucky me -- it is available in interceptors. 
> But I am curios: is it a bug, that sometime not the OID but the (a) 
> attribute name is returned?

I really don't know, but I can reasonably say that it should not work 
the way it does :) Let's open a JIRA !

> Thanks in advance,
>    Stefan


View raw message