Examples of com.inadco.hbl.client.AggregateQuery

com.inadco.hbl.client.AggregateQuery
Aggregate query interface.
Major capabilities:
- define slice ( {@link #addHalfOpenSlice(String,Object,Object)}, etc. The slice specifications are consistent with mathematical definition of closed and open intervals. Closed boundary means the boundary value is included into the result. Open boundary beans it is not included. Unbounded queries are implicitly supported thru putting closed boundary with maximum value possible for that particular data type. IMPORTANT at this time of this writing, only one range per dimension is supported. If another range is added for the same dimension, it would just simply overwrite the previous one. There's no technical reason not to support multiple slices on the same query except that it is just did not happen to be common case enough to have immediate support for. If multiple slices are needed, they could always be obtained by running a query per slice and then aggregating the results manuallly at this pont.
- define projection ( {@link #addGroupBy(String)}. Note that unlike with RDBMS, there's never a query without projections. If you don't specify any dimension for grouping, it assumes full group projection rather than access to individual facts (i.e. such query will always return only one tuple corresponding to the entire slice). Access to individual facts from original fact stream is in fact not possible, as we never ever save individual facts anywhere. This project will assume one could obtain individual facts from some other system of record.
- define desired measures to be output ( {@link #addMeasure(String)}.
Note that query initialization is relatively expensive. If you use HBL system table to pull cube model definitions from, then this will be executed during AggregateQuery construction. To make sure there's best performance, try to reuse query objects, with help of {@link #reset()}.
This contracts also assume that actual implementation is non-reentrant (not thread safe). @author dmitriy

             * this should be equivalent to select aggr_func(impCnt),
             * aggr_func(click) from ... where dim1<=ids[0] and dim1>=ids[0]
             * group by dim1
             */


            AggregateQuery query = queryClient.createQuery();
            query.setCube(cubeName).addMeasure("impCnt").addMeasure("click");
            query.addClosedSlice("dim1", ids[0], ids[0]).addGroupBy("dim1");
            AggregateResultSet rs = query.execute();
            closeables.addFirst(rs);
            while (rs.hasNext()) {
                rs.next();
                AggregateResult ar = rs.current();
                System.out.printf("%032X sum/cnt: impCnt %.4f/%d, click %.4f/%d\n",

View Full Code Here

             * result in composite key filtering with a restart (most
             * fundamental composite key filtering capability but only one part
             * of the key). This now passes too.
             * 
             */
            AggregateQuery query = queryClient.createQuery();


            query.setCube(cubeName).addMeasure("impCnt").addMeasure("click");
            query.addClosedSlice("dim1", ids[0], ids[1]).addGroupBy("dim1");
            AggregateResultSet rs = query.execute();
            closeables.addFirst(rs);
            while (rs.hasNext()) {
                rs.next();
                AggregateResult ar = rs.current();
                System.out.printf("%032X sum/cnt: impCnt %.4f/%d, click %.4f/%d\n",

View Full Code Here

            /**
             * same as client2 but print the summaries separately (no grouping).
             * This is obviously not terribly useful, the queries have got to
             * have group specification -- -- unless we group up all of it.
             */
            AggregateQuery query = queryClient.createQuery();


            query.setCube(cubeName).addMeasure("impCnt").addMeasure("click");
            query.addClosedSlice("dim1", ids[0], ids[1])/* .addGroupBy("dim1") */;
            AggregateResultSet rs = query.execute();
            closeables.addFirst(rs);
            while (rs.hasNext()) {
                rs.next();
                AggregateResult ar = rs.current();
                System.out.printf("%s sum/cnt: impCnt %.4f/%d, click %.4f/%d\n",

View Full Code Here


    private void testClient4(String cubeName) throws IOException, HblException {
        Deque<Closeable> closeables = new ArrayDeque<Closeable>();
        try {


            AggregateQuery query = queryClient.createQuery();
            query.setCube(cubeName);
            for (int i = 0; i < 5; i++) {


                byte ids[][] = new byte[2][];
                ids[0] = new byte[16];
                ids[1] = new byte[16];
                HblUtil.incrementKey(ids[1], 0, 16);


                /*
                 * will try to also constrain for half-open [1:00am,3:00am)
                 */


                GregorianCalendar startTime = IOUtil.tryClone(START_BASE);
                GregorianCalendar endTime = IOUtil.tryClone(START_BASE);


                /*
                 * this will be in local time, whereas example was generated in
                 * UTC. So for PST we get actually normally hours from 9,10 am.
                 * which would result in impression count of 17 for key 00000,
                 * and 21 ifor key 000001.
                 */
                startTime.add(Calendar.HOUR_OF_DAY, 1);
                endTime.add(Calendar.HOUR_OF_DAY, 3);
                // recalculate the calendars
                startTime.getTimeInMillis();
                endTime.getTimeInMillis();


                /*
                 * same as client2 but print the summaries separately (no
                 * grouping).
                 */


                query.addMeasure("impCnt").addMeasure("click");
                query.addClosedSlice("dim1", ids[0], ids[1]).addGroupBy("dim1");
                query.addHalfOpenSlice("impressionTime", startTime, endTime);


                long ms = System.currentTimeMillis();
                AggregateResultSet rs = query.execute();
                closeables.addFirst(rs);
                while (rs.hasNext()) {
                    rs.next();
                    AggregateResult ar = rs.current();
                    System.out.printf("%032X sum/cnt: impCnt %.4f/%d, click %.4f/%d\n",

View Full Code Here

TOP

Related Classes of com.inadco.hbl.client.AggregateQuery

Example1

All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.