Big-Data-HPC-Platform/docs/ECLLanguageReference/ECLR_mods/SpecStruc-Transform.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="TRANSFORM_Structure" role="nobrk">
  <title>TRANSFORM Structure<indexterm>
      <primary>TRANSFORM structure</primary>
    </indexterm></title>

  <para><emphasis>resulttype funcname </emphasis><emphasis
  role="bold">(</emphasis><emphasis> parameterlist </emphasis><emphasis
  role="bold">) := TRANSFORM<indexterm>
      <primary>TRANSFORM</primary>
    </indexterm> [, SKIP<indexterm>
      <primary>SKIP</primary>
    </indexterm>( </emphasis><emphasis>condition</emphasis><emphasis
  role="bold"> )]</emphasis><emphasis role="bold"> </emphasis></para>

  <para><emphasis role="bold"> [ </emphasis><emphasis>locals</emphasis>
  <emphasis role="bold">]</emphasis></para>

  <para><emphasis role="bold"></emphasis><emphasis role="bold">
  SELF</emphasis><indexterm>
      <primary>SELF</primary>
    </indexterm>.<emphasis>outfield</emphasis> :=
  <emphasis>transformation</emphasis>;<emphasis role="bold"></emphasis></para>

  <para><emphasis role="bold">END;</emphasis><emphasis
  role="bold"></emphasis></para>

  <para><emphasis role="bold"></emphasis></para>

  <para><emphasis role="bold">TRANSFORM( </emphasis><emphasis>resulttype,
  assignments </emphasis><emphasis role="bold"> )</emphasis><emphasis
  role="bold"></emphasis></para>

  <para><emphasis role="bold"></emphasis></para>

  <para><emphasis role="bold">TRANSFORM( </emphasis><emphasis>datarow
  </emphasis><emphasis role="bold"> )</emphasis></para>

  <informaltable colsep="1" frame="all" rowsep="1">
    <tgroup cols="2">
      <colspec align="left" colwidth="122.40pt" />

      <colspec />

      <tbody>
        <row>
          <entry><emphasis>resulttype</emphasis></entry>

          <entry>The name of a RECORD structure Attribute that specifies the
          output format of the function. You may use TYPEOF here to specify a
          dataset. Any implicit relationality of the input dataset is not
          inherited.</entry>
        </row>

        <row>
          <entry><emphasis>funcname</emphasis></entry>

          <entry>The name of the function the TRANSFORM structure
          defines.</entry>
        </row>

        <row>
          <entry><emphasis>parameterlist</emphasis></entry>

          <entry>A comma separated list of the value types and labels of the
          parameters that will be passed to the TRANSFORM function. These are
          usually the dataset records or COUNTER parameters but are not
          limited to those.</entry>
        </row>

        <row>
          <entry><emphasis role="bold">SKIP</emphasis></entry>

          <entry>Optional. Specifies the <emphasis>condition</emphasis> under
          which the TRANSFORM function operation is skipped.</entry>
        </row>

        <row>
          <entry><emphasis>condition</emphasis></entry>

          <entry>A logical expression defining under what circumstances the
          TRANSFORM operation does not occur. This may use data from the
          <emphasis>parameterlist</emphasis> in the same manner as a
          <emphasis>transformation</emphasis> expression.</entry>
        </row>

        <row>
          <entry><emphasis>locals</emphasis></entry>

          <entry>Optional. Definitions of local Attributes useful within the
          TRANSFORM function. These may be defined to receive parameters and
          may use any parameters passed to the TRANSFORM.</entry>
        </row>

        <row>
          <entry><emphasis role="bold">SELF</emphasis></entry>

          <entry>Specifies the resulting output recordset from the
          TRANSFORM.</entry>
        </row>

        <row>
          <entry><emphasis>outfield</emphasis></entry>

          <entry>The name of a field in the <emphasis>resulttype</emphasis>
          structure.</entry>
        </row>

        <row>
          <entry><emphasis>transformation</emphasis></entry>

          <entry>An expression specifying how to produce the value for the
          <emphasis>outfield</emphasis>. This may include other TRANSFORM
          function operations (nested transforms).</entry>
        </row>

        <row>
          <entry><emphasis>assignments</emphasis></entry>

          <entry>A semi-colon delimited list of
          SELF.<emphasis>outfield</emphasis>:=
          <emphasis>transformation</emphasis> definitions.</entry>
        </row>

        <row>
          <entry><emphasis>datarow</emphasis></entry>

          <entry>A single record to transform, typically the keyword
          LEFT.</entry>
        </row>
      </tbody>
    </tgroup>
  </informaltable>

  <para>The <emphasis role="bold">TRANSFORM </emphasis>structure makes
  operations that must be performed on entire datasets (such as a JOIN) and
  any iterative type of record processing (PROJECT, ITERATE, etc.), possible.
  A TRANSFORM defines the specific operations that must occur on a
  record-by-record basis. It defines the function that is called each time the
  operation that uses the TRANSFORM needs to process record(s). One TRANSFORM
  function may be defined in terms of another, and they may be nested.</para>

  <para>The TRANSFORM structure specifies exactly how each field in the output
  record set is to receive its value. That result value may simply be the
  value of a field in an input record set, or it may be the result of some
  complex calculation or conditional expression evaluation.</para>

  <para>The TRANSFORM structure itself is a generic tool; each operation that
  uses a TRANSFORM function defines what its TRANSFORM needs to receive and
  what basic functionality it should provide. Therefore, the real key to
  understanding TRANSFORM structures is in understanding how it is used by the
  calling function -- each function that uses a TRANSFORM documents the type
  of TRANSFORM required to accomplish the goal, although the TRANSFORM itself
  may also provide extra functionality and receive extra parameters beyond
  those required by the operation itself.</para>

  <para>The SKIP option specifies the <emphasis>condition</emphasis> that
  results in no output from that iteration of the TRANSFORM. However, COUNTER
  values are incremented even when SKIP eliminates generating the current
  record.</para>

  <sect2 id="Transformation_Attribute_Definitions">
    <title>Transformation Attribute Definitions</title>

    <para>The attribute definitions inside the TRANSFORM structure are used to
    convert the data passed in as parameters to the output
    <emphasis>resulttype</emphasis> format. Every field in the
    <emphasis>resulttype</emphasis> record layout must be fully defined in the
    TRANSFORM. You can explicitly define each field, using the
    <emphasis>SELF.outfield</emphasis> := <emphasis>transformation;</emphasis>
    expression, or you can use one of these shortcuts:</para>

    <programlisting>SELF := [ ];</programlisting>

    <para>clears all fields in the <emphasis>resulttype</emphasis> output that
    have not previously been defined in the transform function, while this
    form:</para>

    <programlisting>SELF.<emphasis>outfield</emphasis> := [];   //the <emphasis>outfield</emphasis> names a child DATASET in
                       // the <emphasis>resulttype </emphasis>RECORD Structure</programlisting>

    <para>clears only the child fields in the <emphasis>outfield</emphasis>,
    and this form:</para>

    <programlisting>SELF := <emphasis>label</emphasis>; //the <emphasis>label</emphasis> names a RECORD structure<indexterm>
        <primary>RECORD structure</primary>
      </indexterm> parameter
// in the <emphasis>parameterlist</emphasis></programlisting>

    <para>defines the output for each field in the
    <emphasis>resulttype</emphasis> output format that has not previously been
    defined as coming from the <emphasis>label</emphasis> parameter's matching
    named field.</para>

    <para>You may also define <emphasis>local</emphasis> attributes inside the
    TRANSFORM structure to better organize the code. These
    <emphasis>local</emphasis> attributes may receive parameters.</para>
  </sect2>

  <sect2 id="TRANSFORM_Functions">
    <title>TRANSFORM Functions</title>

    <para>This form of TRANSFORM must be terminated by the END keyword. The
    <emphasis>resulttype</emphasis> must be specified, and the function itself
    takes parameters in the <emphasis>parameterlist</emphasis>. These
    parameters are typically RECORD structures, but may be any type of
    parameter depending upon the type of TRANSFORM function the using function
    expects to call. The exact form a TRANSFORM function must take is always
    directly associated with the operation that uses it.</para>

    <para>Example:</para>

    <programlisting>Ages := RECORD
  AgedRecs.id;
  AgedRecs.id1;
  AgedRecs.id2;
END;
SequencedAges := RECORD
  Ages;
  INTEGER4 Sequence := 0;
END;

SequencedAges AddSequence(AgedRecs L, INTEGER C) :=
          TRANSFORM, SKIP(C % 2 = 0) //skip even recs
  INTEGER1 rangex(UNSIGNED4 divisor) := (l.id DIV divisor) % 100;
  SELF.id1 := rangex(10000);
  SELF.id2 := rangex(100);
  SELF.Sequence := C;
  SELF := L;
END;

SequencedAgedRecs := PROJECT(AgedRecs, AddSequence(LEFT,COUNTER<indexterm>
        <primary>COUNTER</primary>
      </indexterm>));
//Example of defining a TRANSFORM function in terms of another
namesIdRecord assignId(namesRecord l, UNSIGNED value) :=  TRANSFORM
  SELF.id := value;
  SELF := l;
END;

assignId1(namesRecord l) := assignId(l, 1);
         //creates an assignId1 TRANSFORM that uses assignId
assignId2(namesRecord l) := assignId(l, 2);
         //creates an assignId2 TRANSFORM that uses assignId</programlisting>
  </sect2>

  <sect2 id="Inline_TRANSFORMs">
    <title>Inline TRANSFORMs<indexterm>
        <primary>Inline TRANSFORMs</primary>
      </indexterm></title>

    <para>This form of TRANSFORM is used in-line within the operation that
    uses it. The <emphasis>resulttype</emphasis> must be specified along with
    all the <emphasis>assignments</emphasis>. This form is mainly for use
    where the transform <emphasis>assignments</emphasis> are trivial (such as
    SELF := LEFT;).</para>

    <para>Example:</para>

    <programlisting>namesIdRecord assignId(namesRecord L) := TRANSFORM
  SELF := L; //more like-named fields across
  SELF := []; //clear all other fields
END;

projected1 := PROJECT(namesTable, assignId(LEFT));
projected2 := PROJECT(namesTable, TRANSFORM(namesIdRecord,
          SELF := LEFT;
          SELF := []));
//projected1 and projected2 do the same thing</programlisting>
  </sect2>

  <sect2 id="Shorthand_Inline_TRANSFORMs">
    <title>Shorthand Inline TRANSFORMs<indexterm>
        <primary>Inline TRANSFORMs</primary>
      </indexterm></title>

    <para>This form of TRANSFORM is a shorthand version of Inline TRANSFORMs.
    In this form,</para>

    <programlisting>TRANSFORM(LEFT)</programlisting>

    <para>is directly equivalent to</para>

    <programlisting>TRANSFORM(RECORDOF(LEFT), SELF := LEFT)</programlisting>

    <para>Example:</para>

    <programlisting>namesIdRecord assignId(namesRecord L) := TRANSFORM
  SELF := L; //move like-named fields across
END;
projected1 := PROJECT(namesTable, assignId(LEFT));
projected2 := PROJECT(namesTable, TRANSFORM(namesIdRecord,
              SELF := LEFT));
projected3 := PROJECT(namesTable, TRANSFORM(LEFT));
//projected1, projected2, and projected3 all do the same thing</programlisting>

    <para>See Also: <link linkend="RECORD_Structure">RECORD Structure</link>,
    <link linkend="RECORDOF">RECORDOF</link>, <link
    linkend="TYPEOF">TYPEOF</link>, <link linkend="JOIN">JOIN</link>, <link
    linkend="PROJECT">PROJECT</link>, <link linkend="ITERATE">ITERATE</link>,
    <link linkend="ROLLUP">ROLLUP</link>, <link
    linkend="NORMALIZE">NORMALIZE</link>, <link
    linkend="DENORMALIZE">DENORMALIZE</link>, <link
    linkend="FETCH">FETCH</link>, <link linkend="PARSE">PARSE</link>, <link
    linkend="ROW">ROW</link></para>
  </sect2>
</sect1>