I'm trying to get accurate unfiltered results using the following 2 records. I
have the default fragmentation and really do not want to further fragment the
document (as I realize that could help fix my problem but at the expense of
many, many fragments). From way back when, I seem to recall a trick where you
could 'nest' element-query elements to leverage a document structure and still
get accurate unfiltered results. But, the tricks (as I remember them) do not
appear to be working. My guess is that the tricks don't work when a range
index is involved because the concept of 'word position' (for the element/value
associated with the range index) within the document has been lost. Of course,
I could also be completely wrong.
Here is my simple XML for the 2 records.
<metadata>
<doc>1</doc>
<manifestation>
<fmt>xml</fmt>
<ts>2012-01-13T13:56:12.484283-05:00</ts>
</manifestation>
<manifestation>
<fmt>html</fmt>
<ts>2012-01-11T13:56:12.484283-05:00</ts>
</manifestation>
</metadata>
<metadata>
<doc>2</doc>
<manifestation>
<fmt>xml</fmt>
<ts>2012-01-10T13:56:12.484283-05:00</ts>
</manifestation>
<manifestation>
<fmt>html</fmt>
<ts>2012-01-12T13:56:12.484283-05:00</ts>
</manifestation>
</metadata>
I created an element range index of type xs:dateTime on the ts element. Here
is my query.
cts:search(collection(),
cts:element-query(QName("","manifestation"),
cts:and-query(
(cts:element-value-query((QName("","fmt")),"xml"),
cts:element-range-query((QName("","ts")), ">",
xs:dateTime("2012-01-11T13:56:12.484283-05:00")))
)
),
("unfiltered")
)
I get 2 results for this query (the correct result is 1). I then tried to
modify the document structure as follows:
<metadata>
<doc>1</doc>
<manifestation>
<fmt ts="2012-01-13T13:56:12.484283-05:00">xml</fmt>
</manifestation>
<manifestation>
<fmt ts="2012-01-11T13:56:12.484283-05:00">html</fmt>
</manifestation>
</metadata>
<metadata>
<doc>2</doc>
<manifestation>
<fmt ts="2012-01-10T13:56:12.484283-05:00">xml</fmt>
</manifestation>
<manifestation>
<fmt ts="2012-01-12T13:56:12.484283-05:00">html</fmt>
</manifestation>
</metadata>
I created an element attribute range index of type xs:dateTime on the fmt
element and ts attribute. Here is my query.
cts:search(collection(),
cts:element-query(QName("","manifestation"),
cts:and-query(
(cts:element-value-query((QName("","fmt")),"xml"),
cts:element-attribute-range-query((QName("","fmt")),
(QName("","ts")), ">", xs:dateTime("2012-01-11T13:56:12.484283-05:00")))
)
),
("unfiltered")
)
Once again, I get 2 results for this query. I then tried to modify the
document structure as follows:
<metadata>
<doc>1</doc>
<xml>
<ts>2012-01-13T13:56:12.484283-05:00</ts>
</xml>
<html>
<ts>2012-01-11T13:56:12.484283-05:00</ts>
</html>
</metadata>
<metadata>
<doc>2</doc>
<xml>
<ts>2012-01-10T13:56:12.484283-05:00</ts>
</xml>
<html>
<ts>2012-01-12T13:56:12.484283-05:00</ts>
</html>
</metadata>
I created an element range index of type xs:dateTime on the ts element. Here
is my query.
cts:search(collection(),
cts:element-query(QName("","xml"),
cts:element-range-query((QName("","ts")), ">",
xs:dateTime("2012-01-11T13:56:12.484283-05:00"))
),
("unfiltered")
)
Once again, 2 results.
I realize that I should be able to fix this if I modified my XML as follows:
<metadata>
<doc>2</doc>
<xml ts="2012-01-10T13:56:12.484283-05:00"/>
<html ts="2012-01-12T13:56:12.484283-05:00"/>
</metadata>
But, that is really not preferable as I may not necessarily know all of the
possible values for 'fmt' (xml, html, etc.) in advance. So, I'm curious if
anyone else has any other ideas that I might try (I'm using 4.2-8) so if
perhaps something like this works in 5.0 please let me know.
As an aside, a nice feature request might be to allow the specification of a
range index if the local-name matches a 'pattern' instead of an exact match.
Something like SOLR allows with the specification of 'dynamic fields'.
Thanks.
Darin.
_______________________________________________
General mailing list
General@developer.marklogic.com
http://developer.marklogic.com/mailman/listinfo/general
