8 Understanding Cursor Classes and Concepts

Creating a Cursor

You create a Cursor for a Source by doing the following:

1. Creating a CursorManager by calling one of the createCursorManager methods of the DataProvider and passing it the Source. If you want to alter the behavior of the Cursor, then you can create a CursorInfoSpecification and use the methods of it to specify the behavior. You then create a CursorManager with a method that takes the Source and the CursorInfoSpecification.

2. Creating a Cursor by calling the createCursor method of the CursorManager.

Sources For Which You Cannot Create a Cursor

Some Source objects do not specify data that a Cursor can retrieve from the data store. The following are Source objects for which you cannot create a Cursor that contains values.

• A Source that specifies an operation that is not computationally possible. An example is a Source that specifies an infinite recursion.

• A Source that defines an infinite result set. An example is the fundamental Source that represents the set of all String objects.

• A Source that has no elements or includes another Source that has no elements. Examples are a Source returned by the getEmptySource method of DataProvider and another Source derived from the empty Source. Another example is a derived Source that results from selecting a value from a primary Source that you got from an MdmDimension and the selected value does not exist in the dimension.

If you create a Cursor for such a Source and try to get the values of the Cursor, then an Exception occurs.

Cursor Objects and Transaction Objects

When you create a derived Source or change the state of a Template, you create the Source in the context of the current Transaction. The Source is active in the Transaction in which you create it or in a child Transaction of that Transaction. A Source must be active in the current Transaction for you to be able to create a Cursor for it.

Creating a derived Source occurs in a write Transaction. Creating a Cursor occurs in a read Transaction. After creating a derived Source, and before you can create a Cursor for that Source, you must change the write Transaction into a read Transaction by calling the commitCurrentTransaction methods of the TransactionProvider your application is using. For information on Transaction and TransactionProvider objects, see Chapter 7, "Using a TransactionProvider".

For a Cursor that you create for a query that includes a parameterized Source, you can change the value of the Parameter object and then get the new values of the Cursor without having to commit the Transaction again. For information on parameterized Source objects, see Chapter 5, "Understanding Source Objects".

Structure of a Cursor

The structure of a Cursor mirrors the structure of the Source associated with it. If the Source does not have any outputs, then the Cursor for that Source is a ValueCursor. If the Source has one or more outputs, then the Cursor for that Source is a CompoundCursor. A CompoundCursor has as children a base ValueCursor, which has the values of the base of the Source of the CompoundCursor, and one or more output Cursor objects.

The output of a Source is another Source. An output Source can itself have outputs. The child Cursor for an output of a Source is a ValueCursor if the output Source does not have any outputs and a CompoundCursor if it does.

Example 8-1 creates a query that specifies the prices of selected product items for selected months. In the example, timeHier is a Source for a hierarchy of a dimension of time values, and prodHier is a Source for a hierarchy of a dimension of product values.

If you create a Cursor for prodSel or for timeSel, then either Cursor is a ValueCursor because both prodSel and timeSel have no outputs.

The unitPrice object is a Source for an MdmBaseMeasure that represents values for the price of product units. The MdmBaseMeasure has as inputs the MdmPrimaryDimension objects representing products and times, and the unitPrice Source has as inputs the Source objects for those dimensions.

The example selects elements of the dimension hierarchies and then joins the Source objects for the selections to that of the measure to produce querySource, which has prodSel and timeSel as outputs.

Source timeSel = timeHier.selectValues(new String[] 
                                          {"CALENDAR_YEAR::MONTH::2001.01",
                                           "CALENDAR_YEAR::MONTH::2001.04",
                                           "CALENDAR_YEAR::MONTH::2001.07",
                                           "CALENDAR_YEAR::MONTH::2001.10"});
 
Source prodSel = prodHier.selectValues(new String[]
                                          {"PRODUCT_PRIMARY::ITEM::ENVY ABM",
                                           "PRODUCT_PRIMARY::ITEM::ENVY EXE",
                                           "PRODUCT_PRIMARY::ITEM::ENVY STD"});
 
Source querySource = unitPrice.join(timeSel).join(prodSel);

The result set defined by querySource is the unit price values for the selected products for the selected months. The results are organized by the outputs. Since timeSel is joined to the Source produced by the unitPrice.join(prodSel) operation, timeSel is the slower varying output, which means that the result set specifies the set of selected products for each selected time value. For each time value the result set has three product values so the product values vary faster than the time values. The values of the base ValueCursor of querySource are the fastest varying of all, because there is one price value for each product for each day.

Example 9-1 in Chapter 9, creates a Cursor, queryCursor, for querySource. Since querySource has outputs, queryCursor is a CompoundCursor. The base ValueCursor of queryCursor has values from unitPrice, which is the base Source of the operation that created querySource. The values from unitPrice are those specified by the outputs. The outputs for queryCursor are a ValueCursor that has values from prodSel and a ValueCursor that has values from timeSel.

Figure 8-1 illustrates the structure of queryCursor. The base ValueCursor and the two output ValueCursor objects are the children of queryCursor, which is the parent CompoundCursor.

Figure 8-1 Structure of the queryCursor CompoundCursor

Description of "Figure 8-1 Structure of the queryCursor CompoundCursor"

The following table displays the values from queryCursor in a table. The left column has time values, the middle column has product values, and the right column has the unit price of the product for the month.

For examples of getting the values from a ValueCursor, see Chapter 9.

Specifying the Behavior of a Cursor

CursorSpecification objects specify some aspects of the behavior of their corresponding Cursor objects. You must specify the behavior on a CursorSpecification before creating the corresponding Cursor. To specify the behavior, use the following CursorSpecification methods:

• setDefaultFetchSize

• setExtentCalculationSpecified

• setParentEndCalculationSpecified

• setParentStartCalculationSpecified

• specifyDefaultFetchSizeOnChildren (for a CompoundCursorSpecification only)

A CursorSpecification also has methods that you can use to discover if the behavior is specified. Those methods are the following:

• isExtentCalculationSpecified

• isParentEndCalculationSpecified

• isParentStartCalculationSpecified

If you have used the CursorSpecification methods to set the default fetch size, or to calculate the extent or the starting or ending positions of a value in the parent of the value, then you can successfully use the following Cursor methods:

• getExtent

• getFetchSize

• getParentEnd

• getParentStart

• setFetchSize

For examples of specifying Cursor behavior, see Chapter 9. For information on fetch sizes, see "About Fetch Sizes". For information on the extent of a Cursor, see "What is the Extent of a Cursor?". For information on the starting and ending positions in a parent Cursor of the current value of a Cursor, see "About the Parent Starting and Ending Positions in a Cursor".

Updating the CursorInfoSpecification for a CursorManager

If your application is using OLAP Java API Template objects and the state of a Template changes in a way that alters the structure of the Source produced by the Template, then any CursorInfoSpecification objects for the Source are no longer valid. You need to create new CursorInfoSpecification objects for the changed Source.

After creating a new CursorInfoSpecification, you can create a new CursorManager for the Source. You do not, however, need to create a new CursorManager. You can call the updateSpecification method of the existing CursorManager to replace the previous CursorInfoSpecification with the new CursorInfoSpecification. You can then create a new Cursor from the CursorManager.

Positions of a ValueCursor

The current position of a ValueCursor specifies a value, which you can retrieve. For example, prodSel, a derived Source described in "Structure of a Cursor", is a selection of three products from a primary Source that specifies a dimension of products and their hierarchical groupings. The ValueCursor for prodSel has three elements. The following example gets the position of each element of the ValueCursor, and displays the value at that position.

// prodSelValCursor is the ValueCursor for prodSel
println("ValueCursor Position  Value ");
println("--------------------  ------------------------");
do 
{
  println("          " + prodSelValCursor.getPosition() + 
          "           " + prodSelValCursor.getCurrentValue());
} while(prodSelValCursor.next());

The preceding example displays the following:

ValueCursor Position               Value 
--------------------  -------------------------------
          1           PRODUCT_PRIMARY::ITEM::ENVY ABM
          2           PRODUCT_PRIMARY::ITEM::ENVY EXE
          3           PRODUCT_PRIMARY::ITEM::ENVY STD

The following example sets the current position of prodSelValCursor to 2 and retrieves the value at that position.

prodSelValCursor.setPosition(2);
println(prodSelValCursor.getCurrentString());

The preceding example displays the following:

PRODUCT_PRIMARY::ITEM::ENVY EXE

For more examples of getting the current value of a ValueCursor, see Chapter 9.

Positions of a CompoundCursor

A CompoundCursor has one position for each set of the elements of the descendent ValueCursor objects. The current position of the CompoundCursor specifies one of those sets.

For example, querySource, the Source created in Example 8-1, has values from a measure, unitPrice. The values are the prices of product units at different times. The outputs of querySource are Source objects that represent selections of four month values from a time dimension and three product values from a product dimension.

The result set for querySource has one measure value for each tuple (each set of output values), so the total number of values is twelve (one value for each of the three products for each of the four months). Therefore, the queryCursor CompoundCursor created for querySource has twelve positions.

Each position of queryCursor specifies one set of positions of the outputs and the base ValueCursor. For example, position 1 of queryCursor defines the following set of positions for the outputs and the base ValueCursor:

• Position 1 of output 1 (the ValueCursor for timeSel)

• Position 1 of output 2 (the ValueCursor for prodSel)

• Position 1 of the base ValueCursor for queryCursor (This position has the value from the unitPrice measure that is specified by the values of the outputs.)

Figure 8-2 illustrates the positions of queryCursor CompoundCursor, the base ValueCursor, and the outputs.

Figure 8-2 Cursor Positions in queryCursor

Description of "Figure 8-2 Cursor Positions in queryCursor"

The ValueCursor for queryCursor has only one position because only one value of unitPrice is specified by any one set of values of the outputs. For a query such as querySource, the ValueCursor of the Cursor has only one value, and therefore only one position, at a time for any one position of the root CompoundCursor.

Figure 8-3 illustrates one possible display of the data from queryCursor. It is a crosstab view with four columns and five rows. In the left column are the month values. In the top row are the product values. In each of the intersecting cells of the crosstab is the price of the product for the month.

Figure 8-3 Crosstab Display of queryCursor

Description of "Figure 8-3 Crosstab Display of queryCursor"

A CompoundCursor coordinates the positions of the ValueCursor objects relative to each other. The current position of the CompoundCursor specifies the current positions of the descendent ValueCursor objects. Example 8-2 sets the position of queryCursor and then gets the current values and the positions of the child Cursor objects.

CompoundCursor rootCursor = (CompoundCursor) queryCursor;
ValueCursor baseValueCursor = rootCursor.getValueCursor();
List outputs = rootCursor.getOutputs();
ValueCursor output1 = (ValueCursor) outputs.get(0);
ValueCursor output2 = (ValueCursor) outputs.get(1);
int pos = 5;
rootCursor.setPosition(pos);
println("CompoundCursor position set to " + pos + ".");
println("The current position of the CompoundCursor is = " +
        rootCursor.getPosition() + ".");
println("Output 1 position = " + output1.getPosition() +
        ", value = " + output1.getCurrentValue());
println("Output 2 position = " + output2.getPosition() +
        ", value = " + output2.getCurrentValue());
println("VC position = " + baseValueCursor.getPosition() +
        ", value = " + baseValueCursor.getCurrentValue());

CompoundCursor position set to 5.
The current position of the CompoundCursor is 5.
Output 1 position = 2, value = CALENDAR_YEAR::MONTH::2001.04
Output 2 position = 2, value = PRODUCT_PRIMARY::ITEM::ENVY EXE
VC position = 1, value = 3107.65

// Create the query
prodSel.join(chanSel).join(custSel).join(timeSel).select(units.gt(1));
 
// Commit the current Transaction.
try
{  // The DataProvider is dp.
  (dp.getTransactionProvider()).commitCurrentTransaction();
}
catch(Exception e)
{
  output.println("Cannot commit current Transaction " + e);
}

// Create the CursorManager and the Cursor.
CursorManager cursorManager = dp.createCursorManager(querySource2);
Cursor queryCursor2 = cursorManager.createCursor();

CompoundCursor rootCursor = (CompoundCursor) queryCursor2;
ValueCursor baseValueCursor = rootCursor.getValueCursor();
List outputs = rootCursor.getOutputs();
ValueCursor output1 = (ValueCursor) outputs.get(0);

// Get the positions and values and display them.
println("CompoundCursor  Output ValueCursor        ValueCursor");
println("  position      position  |  value    position  |  value");
do
{
  println(sp6 + rootCursor.getPosition() +  // sp6 is 6 spaces
          sp13 + output1.getPosition() +    // sp13 is 13 spaces
          sp7 + getLocalValue(output1.getCurrentString()) + //sp7 is 7 spaces
          sp7 + baseValueCursor.getPosition() +
          sp7 + getLocalValue(baseValueCursor.getCurrentString()));
}
while(queryCursor2.next());

CompoundCursor  Output ValueCursor        ValueCursor
  position      position |   value    position |   value
      1             1       2001.01       1       ENVY ABM
      2             1       2001.01       2       ENVY EXE
      3             1       2001.01       3       ENVY STD
      4             2       2001.04       1       ENVY ABM
      5             3       2001.07       1       ENVY ABM
      6             3       2001.07       2       ENVY EXE
      7             4       2001.10       1       ENVY EXE
      8             4       2001.10       2       ENVY STD

Because not every combination of product and time selections has unit sales greater than 1 for the specified channel and customer selections, the number of elements of the ValueCursor for the values derived from prodSel is not the same for each value of the output ValueCursor. For time value 2001.01, all three products have sales greater than one, but for time value 2001.04, only one of the products does. The other two time values, 2001.07 and 2001.10, have two products that meet the criteria. Therefore, the ValueCursor for the CompoundCursor has three positions for time 2001.01, only one position for time 2001.04, and two positions for times 2001.07 and 2001.10.

About the Parent Starting and Ending Positions in a Cursor

To effectively manage the display of the data that you get from a CompoundCursor, you sometimes need to know how many faster varying values exist for the current slower varying value. For example, suppose that you are displaying in a crosstab one row of values from an edge of a cube, then you might want to know how many columns to draw in the display for the row.

To determine how many faster varying values exist for the current value of a child Cursor, you find the starting and ending positions of that current value in the parent Cursor. Subtract the starting position from the ending position and then add 1, as in the following.

long span = (cursor.getParentEnd() - cursor.getParentStart()) + 1;

The result is the span of the current value of the child Cursor in the parent Cursor, which tells you how many values of the fastest varying child Cursor exist for the current value. Calculating the starting and ending positions is costly in time and computing resources, so you should only specify that you want those calculations performed when your application needs the information.

An Oracle OLAP Java API Cursor enables your application to have only the data that it is currently displaying actually present on the client computer. For information on specifying the amount of data for a Cursor, see "About Fetch Sizes".

From the data on the client computer, however, you cannot determine at what position of the parent Cursor the current value of a child Cursor begins or ends. To get that information, you use the getParentStart and getParentEnd methods of a Cursor.

To specify that you want Oracle OLAP to calculate the starting and ending positions of a value of a child Cursor in the parent Cursor, call the setParentStartCalculationSpecified and setParentEndCalculationSpecified methods of the CursorSpecification corresponding to the Cursor. You can determine whether calculating the starting or ending positions is specified by calling the isParentStartCalculationSpecified or isParentEndCalculationSpecified methods of the CursorSpecification. For an example of specifying these calculations, see Chapter 9.

What is the Extent of a Cursor?

The extent of a Cursor is the total number of elements it contains relative to any slower varying outputs.

The extent is information that you can use, for example, to display the correct number of columns or correctly-sized scroll bars. The extent, however, can be expensive to calculate. For example, a Source that represents a cube might have four outputs. Each output might have hundreds of values. If all null values and zero values of the measure for the sets of outputs are eliminated from the result set, then to calculate the extent of the CompoundCursor for the Source, Oracle OLAP must traverse the entire result space before it creates the CompoundCursor. If you do not specify that you wants the extent calculated, then Oracle OLAP only needs to traverse the sets of elements defined by the outputs of the cube as specified by the fetch size of the Cursor and as needed by your application.

To specify that you want Oracle OLAP to calculate the extent for a Cursor, call the setExtentCalculationSpecified method of the CursorSpecification corresponding to the Cursor. You can determine whether calculating the extent is specified by calling the isExtentCalculationSpecified method of the CursorSpecification. For an example of specifying the calculation of the extent of a Cursor, see Chapter 9.