Semantic Graph Developer's Guide — Chapter 14

Generating an Execution Plan

You can use sem:sparql-plan to generate a query execution plan for a SPARQL query to see how the query will be handled internally.

For example, this SPARQL query produces an execution plan:

sem:sparql-plan("select * { ?s ?p ?o }",(),"optimize=1")

The query outputs the following execution plan:

<plan:plan xmlns:plan="http://marklogic.com/plan">
 <plan:select>
  <plan:project order="1,0,2">
   <plan:variable name="s" column-index="0" static-type="NONE">
   </plan:variable>
   <plan:variable name="p" column-index="1" static-type="NONE">
   </plan:variable>
   <plan:variable name="o" column-index="2" static-type="NONE">
   </plan:variable>
   <plan:triple-index order="1,0,2" permutation="PSO" dedup="true">
    <plan:subject>
      <plan:variable name="s" column-index="0" static-type="NONE">
     </plan:variable>
    </plan:subject>
    <plan:predicate>
     <plan:variable name="p" column-index="1" static-type="NONE">
     </plan:variable>
    </plan:predicate>
    <plan:object>
     <plan:variable name="o" column-index="2" static-type="NONE">
     </plan:variable>
    </plan:object>
   </plan:triple-index>
  </plan:project>
 </plan:select>
</plan:plan>

Parsing an Execution Plan

This section breaks down and describes each portion of the execution plan.

The intro specifies the type of plan, in this case for a SELECT statement:

<plan:plan xmlns:plan="http://marklogic.com/plan">
 <plan:select>

This section identifies the projected order of the elements of the triples...Äìsubject, predicate, object...Äìwith their variable names (s,p,o) and column indexes:

<plan:project order="1,0,2">
   <plan:variable name="s" column-index="0" static-type="NONE">
   </plan:variable>
   <plan:variable name="p" column-index="1" static-type="NONE">
   </plan:variable>
   <plan:variable name="o" column-index="2" static-type="NONE">
   </plan:variable>

At the end is the order of the triple variables in the triple index - predicate, subject, object (p,s,o).

   <plan:triple-index order="1,0,2" permutation="PSO" dedup="true">
    <plan:subject>
      <plan:variable name="s" column-index="0" static-type="NONE">
     </plan:variable>
    </plan:subject>
    <plan:predicate>
     <plan:variable name="p" column-index="1" static-type="NONE">
     </plan:variable>
    </plan:predicate>
    <plan:object>
     <plan:variable name="o" column-index="2" static-type="NONE">
     </plan:variable>
    </plan:object>
   </plan:triple-index>
  </plan:project>
 </plan:select>
</plan:plan>

If you run this query, the variables and the values are projected in three columns (s, p, o):

[{"s":"<http://example.com/ns/directory#jp>","p":"<http://example.com/ns/person#firstName>","o":"\"John-Paul\""}]

Here is an example of a more complicated SPARQL SELECT query, which includes prefixes:

sem:sparql-plan("
PREFIX rdf:    <http://www.w3.org/1999/02/22-rdf-syntax-ns#> 
PREFIX rdfs:   <http://www.w3.org/2000/01/rdf-schema#>
PREFIX prod:   <http://example.com/products/> 
PREFIX ex:     <http://example.com/>

SELECT ?product
FROM <http://marklogic.com/semantics/products/>
WHERE 
  { 
    ?product  rdf:type  ex:Shirt ;
              ex:color  'blue' }")

Here is the output for this query execution plan. The intro specifies the namespace and the type of plan:

<plan:plan xmlns:plan="http://marklogic.com/plan">
 <plan:select>

The first section identifies the order of projected values, in this case just one product.

  <plan:project order="order(0 ASC)">
   <plan:variable name="product" column-index="0" static-type="NONE">
    </plan:variable>

This section describes what sort of hash join order.

   <plan:hash-join order="order(0 ASC)">
    <plan:hash left="0" right="0" operator="=">
     </plan:hash>

Here is the projected order of triple elements...Äìobject, predicate, subject (o, p, s)...Äìfirst for the triple for product of type Shirt:

<plan:triple-index order="order(0 ASC)" permutation="OPS" dedup="true">
  <plan:subject>
    <plan:variable name="product" column-index="0" static-type="NONE">
    </plan:variable>
  </plan:subject>
  <plan:predicate>
    <plan:iri name="http://www.w3.org/1999/02/22-rdf-syntax-ns#type" static-type="NONE">
    </plan:iri>
  </plan:predicate>
  <plan:object>
    <plan:iri name="http://example.com/Shirt" static-type="NONE">
    </plan:iri>
  </plan:object>
</plan:triple-index>

And for the product with the color value blue:

<plan:triple-index order="order(0 ASC)" permutation="OPS" dedup="true">
  <plan:subject>
    <plan:variable name="product" column-index="0" static-type="NONE">
    </plan:variable>
  </plan:subject>
  <plan:predicate>
    <plan:iri name="http://example.com/color" static-type="NONE">
    </plan:iri>
  </plan:predicate>
  <plan:object>
    <plan:value datatype="http://www.w3.org/2001/XMLSchema#string" value="blue">
    </plan:value>
  </plan:object>
</plan:triple-index>

And then the close of the plan:

    </plan:hash-join>
   </plan:project>
  </plan:select>
</plan:plan>

For more about query execution plans, see Execution Plan in the SQL Data Modeling Guide.