At least two methods may be used to accelerate the execution of queries with xmlview() function. The first method assumes that a path statement following xmlview() function should contain maximum conditions to reduce the result set. For example, the query

select xquery_eval('<w>
  {
    for $prod in xmlview("product")/product[@ProductID="1"]
    return <q>{$prod}</q>
  }</w>',
  xtree_doc('<dummy_tag/>'));

will be executed faster than

select xquery_eval('<w>
  {
    for $prod in xmlview("product")/product where $prod[@ProductID="1"]
    return <q>{$prod}</q>
  }</w>',
  xtree_doc('<dummy_tag/>'));

due to the SQL query produced from the path expression 'product[@ProductID="1"]' reduces the result set in comparison with the SQL query produced from the path expression 'product' in the second query.

If we execute a join of two (or more) XML views (or XML document and XML view), i.e. the query consists of the nested loops, the second method proposes to carry out a piece of query of the nested loop which is independent of the outer loop outside the outer loop and uses LET clause for it. For example, the query

select xquery_eval('<document>
  {
    let $prod_set:=(for $prod in xmlview("product")/product return $prod)
    for $cat in xmlview("cat")/category
    return (<category description={$cat/@description}$gt;
      {distinct(for $prod in $prod_set
                where $cat/@CategoryID=$prod/category/@CategoryID
                return $prod/supplier)}</category>)}</document>',
   xtree_doc('<dummy_tag/>'));

is equivalent to the query in the example 1, but it is about 5 times faster than original one. This method is especially useful for full joins. In this case we do not have a full join and this query may be optimized without the use of temporary result sets if the 'where' clause is replaced with proper filter:

select xquery_eval('<document>
  {
    for $cat in xmlview("cat")/category
    let $catID := $cat/@CategoryID
    return (<category description={$cat/@description}$gt;
      {distinct(
        for $supp in xmlview("product")/product[category/@CategoryID=$catID]/supplier
        return $supp)
      }</category>)
  }
  </document<',
  xtree_doc('<dummy_tag/>'));

This variant requires no memory for storing $prod_set and it never fetches redundant fields from "Demo"."demo"."Products" table but it heavily needs index for "Demo"."demo"."Products" on "CategoryID" field. If such index is built the last variant is about 10 times faster than the query in example 1. Similarly, the query in example 2 may be optimized as follows:

select xquery_eval('
  <supplier_product>
   {
     for $supp in document("suppliers.xml")/supplier
     let $supp_id:=string($supp/@supplier_id)
     return (
      <supplier>{$supp/supplier_name}
        <product_name>
          {
            for $prod in xmlview("product")/product[supplier/@SupplierID=$supp_id]
            return string($prod/@product_name)
          }
        </product_name>
      </supplier>)
   }
  </supplier_product>',
  xtree_doc('<dummy_tag/>')
  );

and it speeds up the operation by more than 15 times.

Virtuoso does not support certain kinds of XPATH expressions applied to the xmlview() function.

1. A path expression must not contain any functions, because it is impossible to translate most of the functions to SQL queries.

2. A path expression must not contain numeric XQUERY variables in the arithmetic expressions. Let a document named products.xml contains product elements; each product element has numeric attribute ProductPrice. A run time error occurs if the following query would be used

select xquery_eval('
  <document>
   {
     for $prod_doc in document("products.xml")/products/product/@ProductPrice
       for $prod_view in xmlview("product")/product[@price>$prod_doc+1]
     return <q>{$prod_doc, $prod_view/@product_name}</q>
   }
   </document>',
   xtree_doc('<dummy_tag/>'));

because the type of $prod_doc is considered as string. As it is mentioned in the previous restriction the using a function in a path expression (e.g. xmlview("product")/product[@price>number($prod_doc)+1]) is not allowed. The correct query is as follows:

select xquery_eval('
  <document>
    {
      for $prod_doc in document("products.xml")/products/product/@ProductPrice
        let $prod_doc2:=number($prod_doc)
        for $prod_view in xmlview("product")/product[@price>$prod_doc2+1]
      return <q>{$prod_doc, $prod_view/@product_name}</q>
    }
  </document>',
  xtree_doc('<dummy_tag/>'));

3. A path expression must not contain XQUERY variables with the following paths. The following query will not be executed

select xquery_eval('
  <document>
    {
      for $cat in xmlview("cat")/category,
          $prod in xmlview("product")/product[@CategoryID=$cat/@CategoryID]
      return <q>{$prod/supplier}</q>
    }
  </document>',
  xtree_doc('<dummy_tag/>'));

The correct query may be given as

select xquery_eval('
  <document>
    {
      for $cat in xmlview("cat")/category
      let $cat_id:=$cat/@CategoryID
      for $prod in xmlview("product")/product[@CategoryID=$cat_id]
      return <q>{$prod/supplier}</q>
    }
  </document>',
  xtree_doc('<dummy_tag/>'));

4. Virtuoso does not support a selection of n-th element in a path expression. The following query will not be executed

select xquery_eval('
  <document>
    {
      for $cat in xmlview("cat")//product[1] return <q>{$cat}</q>
    }
  </document>',
  xtree_doc('<dummy_tag/>'));

5. Virtuoso does not support a dereference (=>) in a path expression.

6. It is not recommended to use the long varchar, long varbinary and long nvarchar data types with the logical and boolean operations in a filter of the path expression. For example, the execution of the following query

select xquery_eval('
  <document>
    {
      for $prod in
        xmlview("product")/product[@SupplierID<5]/category[@Description like "Sw%"]
      return <q>{$prod}</q>
    }
  </document>',
  xtree_doc('<dummy_tag/>'));

may return an error, because the field "Description" has LONG VARCHAR type in the table "demo"."Categories".