nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

Search Performance in Documentum 6.5 SP2.. how to solve this?

parimalraj

Hi All,

We use Documentum 6.5 sp2 + RM + RPS.

We have customized the product as per our requirements. In this process, we have created two object types A, B as sub-types of dm_document.

B is linked to A. Even though A is not made a virtual doc by flagging r_is_virtual_doc to true, its r_link_cnt is >=1 making it a virtual doc. And therefore we

have many B objects linked to 1 A object.

The requirement is to execute a full text search in both A and its linked B objects and their content files but to display only A.

That is, even if the search term is present only in A or present only in B and not present in its linked A, A has to be displayed on results page.

Please suggest me if there is any way to do that.

We tried to cusomize the search code by adding separate DQL queries like:

Query 1: Select r_object_id from B search document contains '****'

and then:

Query 2: Select r_object_id_of_A from dm_dbo.mapping_table where r_object_id_of_B = 'r_object_id of B from above query'

and then:

Query 3: Select r_object_id from A search document contains '****'

Filter the r_object_ids and collect them in IDfList.

The above is our custom code to retrieve all the r_object_id s of all As having the search term in them or in their linked Bs.

This is only half way as we need IDfResultsSet to display results in Search results page. Therefore, we allowed % symbol as Search term to Search API so that the idfQueryProcessor.getResults() method gives us all the As in the system (say 100,000).

We now compare the r_object_id s of As in IDfResultsSet to those in IDfList and if they are matching, then we will create a new IDfModifiableResultsSet to be displayed on to the screen.

In this entire process, the execution time to retrieve all the As from the system (100,000 in our case) is taking long time (of course it is obvious).

And other issue is, the Query 2 in our custom code is breaking when the count of Bs in Query 1 is high (>10,000 approx).This makes our search cumbersome.

Please suggest any tips/ideas to improve this.

Thanks and regards,

Parimal

Find more posts tagged with

Comments

bacham2

This is quite challenging, especially with virtual documents because there is no direct relationship between the root document and descendent beyond the first level. Furthermore, a given document of type B may theoretically be included in multiple virtual documents (which is one of the strengths of VD): so which parent do you want to display in this case?

If your virtual documents have only one level, you could perform an outer query on dmr_containment:

select distinct parent_id from dmr_containment where component_id in (<your fulltext search here>)

The documents of type A require no special handling because there is also a dmr_containment object representing the root VD.

If your VDs all have assemblies, you could do something similar using dm_assembly instead of dmr_containment and it would work even if your VDs have multiple levels (because this time there is a direct relationship between the parent and children of any depth).

Otherwise, you may have to create a database view (and register it) which performs a recursive query against dmr_containment_s in order to retrieve the parent regardless of the depth.

parimalraj

Hi Bacham2,

Thanks to you for your reply.

Furthermore, a given document of type B may theoretically be included in multiple virtual documents (which is one of the strengths of VD): so which parent do you want to display in this case?

In our case we have only one to one relationship i.e one child belongs to one parent only.

The problem im facing even using this query is: it is taking lot of time to execute if the search term is present in large number of objects/documents. And moreover when my search query is on dm_document (this is because A and B are different and if i need to get them both in single query i need to use its parent which is dm_document) the query is also retrieving other child objects of dm_document which are not needed. If i've to eliminate unwanted child types by using where clause in fulltext query, then it is further increasing time.

I also have a question. can we use inner queries in Search API?

Thanks and regards,

Parimal

bacham2

Are you saying that adding one the following where clauses is slowing down the search?

select * from dm_document search document contains ... where r_object_type in ('A','B')

or equivalent:

select * from dm_document search document contains ... where type(A) or type(B)

parimalraj

Yes Bacham2!

I've not tried the second query you mentioned above but the first one.

My query is as follows:

select distinct parent_id from dmr_containment where component_id in (select r_object_id from dm_document search document contains '****' where r_object_type in ('A1', 'A2','A3', 'A4', 'B'))

- A1,A2,A3,A4 are sub-types of A.

And also would you help me to know if we can call any DFC method that creates IDfResultEntry object so that I'll search for dm_document through Search API and once i get the results through idfQueryProcessor.getResults(),

i would identify the type of the object in the ResultEntry and if it is of type B then query out its corresponding A and make a ResultEntry object of A and add it to IDfModifiableResultsSet.

This is because by using the custom code (mentioned in original post) i'm not able to utilize the operator parsing capability of Search API rather I've to write a custom code to parse the user input string and then give it as an input in the fulltext query.

Thanks and regards,

Parimal