Search code examples
marklogicmarklogic-8marklogic-7

How to define relationship between xmls in MarkLogic and use it while search API?

These are the documents I have inserted in the database.

<biblio>
    <biblioid>123</biblioid>
    <mediaid>456</mediaid>
    <name>dixit</name>
    <title>title</title>
</biblio>

<biblio>
    <biblioid>456</biblioid>
    <mediaid>789</mediaid>
    <name>singla</name>
    <title>title1</title>
</biblio>

<media>
    <mediaid>456</mediaid>
    <mediaName>media1</mediaName>
    <title>hello</title>
</media>

<media>
    <mediaid>789</mediaid>
    <mediaName>media1</mediaName>
    <title>hello</title>
</media>

<media>
    <mediaid>384</mediaid>
    <mediaName>media2</mediaName>
    <title>hello</title>
</media>

I want to search those documents having <name> in <biblio> as "Dixit" and <medianame> in <media> as "media1".

But it should check of only those documents having <mediaid> is same as in biblio/mediaid

As in our case, the result will be 

<biblio>
    <biblioid>123</biblioid>
    <mediaid>456</mediaid>
    <name>dixit</name> <!-- name is matching ("dixit")-->
    <title>title</title>
</biblio>

<media>
    <mediaid>456</mediaid> <!-- mediaid is same as in biblio/mediaid  -->
    <mediaName>media1</mediaName> <!-- medianame is matching ("media1") -->
    <title>hello</title>
</media>

I can achieve this, first by getting all the biblio docs matching the name ("dixit") then from the results I will extract the mediaid's(456) and then will query the media documents with fetched mediaid's(456) and medianame(media1).

But I want to achieve this by search:search API.

Is there any way to make this happen and Is there any place where I can define the relationship between elements of xmls.

Solution

  • If you put a range index on the media:id, then you will be able to do some super-fast joins. That is 1/2 the battle.

    The other half is: If I wanted to do this problematically, I would do it via cts and xQuery.

    BUT: If I did want to do this as part of search:search, I would do it with custom constraints - then naturally call them via the search:search library.

    A possible custom constraint: biblio-by-media-name

    • search for media with the name in question
    • resolve this to the biblio references
    • return a ct:document-query* of the biblio docs

    Then if I also had a constraint on the biblio-name called "biblio-name" then I could have a search something like the below example - all delivered via the search api in the end:

    biblio-name:Dixit AND biblio-by-media-name:media1

    * Note that the document-query is one approach. However, I would possibly build this not on the biblio URI, but perhaps on a range-query on the mediaid element in the biblio files. But that is all about tuning. The results will be the same in the more simplified example above.

    https://docs.marklogic.com/guide/search-dev/search-api#id_49750