Big-Data-HPC-Platform/docs/ECLLanguageReference/ECLR_mods/Recrd-ImplicitDatasetRel.xml

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE sect1 PUBLIC "-//OASIS//DTD DocBook XML V4.5//EN"
"http://www.oasis-open.org/docbook/xml/4.5/docbookx.dtd">
<sect1 id="Implicit_Dataset_Relationality">
  <title>Implicit Dataset<indexterm>
      <primary>Implicit Dataset</primary>
    </indexterm> Relationality<indexterm>
      <primary>Relationality</primary>
    </indexterm><indexterm>
      <primary>Implicit Dataset Relationality</primary>
    </indexterm></title>

  <para>Nested child datasets<indexterm>
      <primary>Nested child datasets</primary>
    </indexterm> in a Data Refinery (Thor) or Rapid Data Delivery Engine
  (Roxie) cluster are inherently relational, since all the parent-child data
  is contained within a single physical record. The following rules apply to
  all inherent relationships.</para>

  <para>The scope level of a particular query is defined by the primary
  dataset for the query. During the query, the assumption is that you are
  working with a single record from that primary dataset.</para>

  <para>Assuming that you have the following relational structure in your
  database:</para>

  <programlisting>     Household           Parent 
        Person           Child of Household 
          Accounts       Child of Person, Grandchild of Household</programlisting>

  <para>This means that, at the primary scope level:</para>

  <para>a) All fields from any file that has a 1:M relationship with the
  primary file are available. That is, all fields in any parent (or
  grandparent, etc.) record are available to the child. For example, if the
  Person dataset is the primary scope, then all the fields in the Household
  dataset are available.</para>

  <para>b) All child datasets (or grandchildren, etc.) can be used in
  sub-queries to filter the parent, as long as the sub-query uses an aggregate
  function or operates at the level of the existence of a set of child records
  that meet the filter criteria (see EXISTS).You can use specific fields from
  within a child record at the scope level of the parent record by the use of
  EVALUATE or subscripting ([]) to a specific child record. For example, if
  the Person dataset is the primary scope, then you may filter the set of
  related Accounts records and check to see if you've filtered out all the
  related Accounts records.</para>

  <para>c) If a dataset is used in a scope where it is not a child of the
  primary dataset, it is evaluated in the enclosing scope. For example, the
  expression:</para>

  <para><programlisting>Household(Person(personage &gt; AVE(Person,personage))</programlisting></para>

  <para>means “households containing people whose age is above the average age
  for the household.” It does <emphasis role="underline">not</emphasis> mean
  “households containing people whose age is above the average for all the
  households.” This is because the primary dataset (Household) encloses the
  child dataset (Person), making the evaluation of the AVE function operate at
  the level of the persons within the household.</para>

  <para>d) An attribute defined with the STORED() workflow service is
  evaluated at the global level. It is an error if it cannot be evaluated
  independently of other datasets. This can lead to some slightly strange
  behaviour:</para>

  <programlisting>AveAge := AVE(Person,personage);
MyHouses := Household(Person(personage &gt; aveAge));</programlisting>

  <para>means “households containing people whose age is above the average age
  for the household.” However,</para>

  <programlisting>AveAge := AVE(Person,personage) : STORED('AveAge');
MyHouses := Household(Person(personage &gt; aveAge));</programlisting>

  <para>Means “households containing people whose age is above the average for
  all the households.” This is because the AveAge attribute is now evaluated
  outside the enclosing Household scope.</para>
</sect1>