Query-Limited Facets

Earlier we mentioned a fifth argument to cts:element-attribute-values(). The fifth argument is a cts:query object that limits the returned results to those taken from documents matching the given cts:query. The following finds the top senders of mails matching a specific query.

let $lists := ("httpd", "firefox") let $types := ("announcements", "general", "development") let $word := "release" let $query := cts:and-query(( cts:element-attribute-word-query( xs:QName("message"), xs:QName("list"), $lists ), cts:element-attribute-value-query( xs:QName("message"), xs:QName("type"), $types ), cts:or-query(( cts:element-attribute-word-query( xs:QName("message"), xs:QName("list"), $word, "unstemmed", 3 ), cts:element-word-query( xs:QName("subject"), $word, "stemmed", 2 ), cts:element-word-query( xs:QName("para"), $word, "stemmed", 1 ), cts:element-word-query( xs:QName("quotepara"), $word, "stemmed", 0.5 ) )) )) return for $value in cts:element-attribute-values( xs:QName("from"), xs:QName("personal"), "", "frequency-order", $query )[1 to 100] return concat($value, ": ", cts:frequency($value))

We're extracting values from the from/@personal attributes, which holds the human-readable names of senders. We start at "" (the beginning), ask for results in frequency order (so the most common occurrances are first), and limit results to the $query. We then print the value and count for the top 100. These are the folks from Apache and Firefox who talk a lot about releases.

We can also extract message/@year-month and sort by value (value is the default) and what we'll see is a traffic-by-month report, the basis for the MarkMail histogram chart:

let $query := "satellite" for $value in cts:element-attribute-values( xs:QName("message"), xs:QName("year-month"), "", (), $query )[1 to 100] return concat($value, ": ", cts:frequency($value))

We've now seen almost all the building blocks of MarkMail.org. But how do we take the user's typed string in the search box and convert it to cts:query objects? For that we can use the Search API.