The XML document, I will be using throughout the examples is -

<a>
    <b>test-value</b>
    <d>fragment-d</d>
    <c-root>
        <d>fragment-d</d>
        <e>fragment-e</e>
    </c-root>
</a>

The following queries returns the number of documents with value fragment-d for element d -

• Using a cts:search

  fn:count(cts:search(fn:doc(), cts:element-value-query(xs:QName("d"), "fragment-d")))

• Using XPath

  fn:count(fn:doc()[//d="fragment-d"]))

xdmp:estimate(cts:search(fn:doc(), cts:element-value-query(xs:QName("d"), "fragment-d")))

xdmp:estimate can not be used on XPaths unlike fn:count is used in previous example

xdmp:estimate actually gives the number of matching fragments

The XML document to consider in this example -

<a>
    <b>test-value</b>
    <d>fragment-d</d>
    <c-root>
        <d>fragment-d</d>
        <e>fragment-e</e>
    </c-root>
</a>

A fragment root is declared on <c-root>

If this is the only document in the database, xdmp:estimate and fn:count are going to behave differently -

xdmp:estimate(cts:search(fn:doc(), cts:element-value-query(xs:QName("d"), "fragment-d")))

Result of the above query will be 2 (Number of fragments)

fn:count(cts:search(fn:doc(), cts:element-value-query(xs:QName("d"), "fragment-d")))

Result of the above query will be 1 (Number of documents)

In terms of performance xdmp:estimate is much better than fn:count as it takes the advantages of indexes while resolving the search results