Interface ServerSummaries


public interface ServerSummaries

Server Summaries

In Smart GWT, "summarization" refers to applying functions to DataSource fields to obtain summary records calculated from the values of multiple DataSource records.

For example, by applying an "average" function to a DataSourceField representing the dollar value of an order, you can calculate the average dollar amount of a set of orders.

Server Summaries is a feature of the Smart GWT Server allowing similar summaries to be computed server-side, generally by the database engine itself, or client-side for clientOnly DataSources. Client-side calculation of summary records is also supported on ListGrids via the showGridSummary feature.

Server-side calculation is available with Power or better licenses only. See smartclient.com/product for details.

See also the related feature which allows a single field to receive a value summarized from several related records, DataSourceField.includeSummaryFunction.

Summarization can be statically configured directly in a .ds.xml file, or can be dynamically configured when sending a DSRequest, either client-side or server-side. The following examples all assume a DataSource with fields like this:

  <DataSource ID="dsOrders" ...>
       <fields>
           <field name="customerName" type="text" />
           <field name="orderDate" type="date" />
           <field name="deliveryStatus" type="string" />
           ...
       </fields>
  </DataSource>
  

Summaries in DataSource .ds.xml files

You can declare summarization directly on an operationBinding in a .ds.xml file, using OperationBinding.summaryFunctions and OperationBinding.groupBy. For example:

   <DataSource ID="dsOrders" ...>
       ...
       <operationBindings>
           <operationBinding operationType="fetch" operationId="lastOrderDateByCustomer">
               <summaryFunctions>
                   <orderDate>max</orderDate>
               </summaryFunctions>
               <groupBy>customerName</groupBy>
           </operationBinding>
       </operationBindings>
   </DataSource>
  
This would return summary records representing the most recent order per customer. Represented in JSON, the returned records would look like:
    { customerName: "JBar Struts", orderDate:"2012/02/05" },
    { customerName: "KFoo Widgets", orderDate:"2012/03/01" },
    ...
  
This is analogous to the result of a SQL query like:
  SELECT
   max(order.orderDate)
   order.customerName
  FROM
   order
  GROUP BY
   order.customerName
  
Note that, as with SQL, the returned records will only include the fields where summary functions were applied or which were used for grouping - "deliveryDate" and other fields are not included because in general, summary records may represent data from more than one record (there may be more than one record with the "max" value, and consider also "sum" or "avg" functions), so it's ambiguous which record's values should be returned for non-grouped, non-summarized fields.

Dynamically Requested Summaries

DSRequest.summaryFunctions and DSRequest.groupBy allow you to dynamically request a server summary from client-side code. For example:

   DSRequest requestProperties = new DSRequest();
   requestProperties.setGroupBy("customerName");
   requestProperties.setSummaryFunctions(new HashMap() {{
      put("orderDate", SummaryFunctionType.MAX);
   }});
   dsOrders.fetchData(null, callback, request);
  
By default such requests are allowed, but such requests can be disallowed on a per-DataSource or system-wide level if you have concerns - see DataSource.allowClientRequestedSummaries.

You can also dynamically request summaries from server-side code (for example, in a DMI method):

    DSRequest dsRequest = new DSRequest("dsOrders", "fetch");
    dsRequest.setSummaryFunctions(new HashMap() {{
      put("orderDate", SummaryFunctionType.MAX);
    }});
    dsRequest.setGroupBy("customerName");
    DSResponse dsResponse = dsRequest.execute();
  

Sort & Data Paging

Sort directions are supported for queries that involve server summarization although may only target fields that are returned by the query (only those fields included in groupBy or where a summaryFunction was applied).

Data paging is also supported, however, consider that for aggregated queries, when asked for a second page of data, the database is likely to have to repeat all the work of calculating aggregated values. Turning paging off or setting a generous ListGrid.dataPageSize is advised.

Criteria

SQL & clientOnly DataSource

Criteria is automatically split into pre and post summarization parts:

  • Fields that are not summarized will have criteria applied to them before summarization. This includes group by fields and fields that are there just for filtering purposes, i.e. might not be fetched at all. For example if a "price" field would be a regular field in a fetch, i.e. not summarized, criteria like "price < 5" will eliminate records where price is less than 5, so such records would not affect the average or total computed.
  • Fields that are summarized will have criteria applied to them after summarization. For example if the "avg" summary function is being applied to a "price" field would, same criteria "price < 5" would eliminate records where summarized value "avg(price)" is less than 5.
This post-summarization server filtering matches client-side filtering, since it is always post-summarization on the client-side. Disabling filter for summarized fields on the client-side is no longer necessary to get the same results, as it was in previous versions.

Previous versions applied criteria on the server for all fields before summarization. You can get that behavior back via OperationBinding.applyCriteriaBeforeAggregation setting. Note that in that case you need to turn off client-side filtering for aggregated fields, because client-side filtering cannot replicate pre-summarization filtering, as client sees only the final computed aggregates. See OperationBinding.applyCriteriaBeforeAggregation docs for more details.

HB and JPA DataSources

Criteria apply to record before summaries are applied. For example, if the "avg" function is being applied to a "price" field, criteria like "price < 5" will eliminate records where price is less than 5 before the average price is calculated. This means that client-side filtering may not work as expected with summarized results: client-side filter criteria are necessarily applied after summary functions have been applied, so may not match the server's behavior. You can set ResultSet.useClientFiltering to disable client-side filtering on a grid via ListGrid.dataProperties. Or individual fields can be marked canFilter:false.

SQL Templating & Aggregation

With the SQL Templating feature you can customize portions of the query without ever having to re-create portions that the framework knows how to generate. This allows to create partially or entirely custom complex aggregation queries to use in a regular "fetch" operation. The SQL Templating feature supports aggregated queries just as regular ones with some additions, see below.

In clause-by-clause substitution there are two additional aggregation specific clauses: groupClause providing "group by" part of aggregated query and groupWhereClause providing "having" part of aggregated query (or outer "where" part if sub-select approach is used, see OperationBinding.useHavingClause for more details). The automatically generated groupClause and groupWhereClause clauses are also available as $defaultGroupClause and $defaultGroupWhereClause SQL templating variables. Note that if OperationBinding.applyCriteriaBeforeAggregation is set to true, groupWhereClause is not generated.

SQLDataSource.getSQLClause() server-side API can generate the entire query, in case you wanted to use an aggregated query as just part of a larger query (perhaps a sub-select), or different parts of a query, including groupClause and groupWhereClause aggregated query clauses.

Also note SQLDataSource.getPartialHaving() and SQLDataSource.getHavingWithout() server-side APIs which generate partial SQL condition expressions to be used as a complete or partial "having" or outer "where" clause. This is also can be accessed via $sql.partialHaving and $sql.havingWithout functions in SQL templates, see $sql variable description in VelocitySupport.

To see an example of wrapping the main query as a sub-select to achieve additional aggregation level and splitting provided criteria into different chunks of condition expressions to apply them to specific parts of a completely customized aggregation query, see the Custom Aggregation Query sample.

Fields with customSelectExpression

Fields with customSelectExpression can be used with server summaries as both groupBy fields or fields with summaryFunction. In case of summaryFunction requested on field with customSelectExpression we will wrap SQL function around the expression, which may or may not be correct.

Summarizing without Grouping

Declaring just <summaryFunctions> without declaring <groupBy> is allowed. This will always give you exactly one summary record in the result, which will represent the summary functions as applied to all records that match the criteria in the DSRequest.

Grouping without Summarizing

Declaring just <groupBy> without <summaryFunctions> is also allowed. This gives results similar to a SQL "select distinct": one record per distinct set of values for the grouped fields. This kind of result can be used in various ways; one common use case is populating a ComboBoxItem with a list of existing values for a field that already appear in DataSource records.

Check out this example of grouping without summarizing being used to determine all unique values of a field.

Custom Aggregation

You can implement your own custom aggregation functions in addition to the built-in ones by providing custom DMI implementation for the "fetch" operation. To see the example of using custom function see Custom Aggregation sample. Note, that in order to access custom summary function on the server-side you need to use DSRequest.getRawSummaryFunctions() API.

See Also: