Search in sources :

Example 1 with DistinctPrefixFilter

use of org.apache.phoenix.filter.DistinctPrefixFilter in project phoenix by apache.

the class BaseResultIterators method initializeScan.

private static void initializeScan(QueryPlan plan, Integer perScanLimit, Integer offset, Scan scan) throws SQLException {
    StatementContext context = plan.getContext();
    TableRef tableRef = plan.getTableRef();
    PTable table = tableRef.getTable();
    Map<byte[], NavigableSet<byte[]>> familyMap = scan.getFamilyMap();
    // Hack for PHOENIX-2067 to force raw scan over all KeyValues to fix their row keys
    if (context.getConnection().isDescVarLengthRowKeyUpgrade()) {
        // We project *all* KeyValues across all column families as we make a pass over
        // a physical table and we want to make sure we catch all KeyValues that may be
        // dynamic or part of an updatable view.
        familyMap.clear();
        scan.setMaxVersions();
        // Remove any filter
        scan.setFilter(null);
        // Traverse (and subsequently clone) all KeyValues
        scan.setRaw(true);
        // Pass over PTable so we can re-write rows according to the row key schema
        scan.setAttribute(BaseScannerRegionObserver.UPGRADE_DESC_ROW_KEY, UngroupedAggregateRegionObserver.serialize(table));
    } else {
        FilterableStatement statement = plan.getStatement();
        RowProjector projector = plan.getProjector();
        boolean optimizeProjection = false;
        boolean keyOnlyFilter = familyMap.isEmpty() && context.getWhereConditionColumns().isEmpty();
        if (!projector.projectEverything()) {
            // not match the actual column families of the table (which is bad).
            if (keyOnlyFilter && table.getColumnFamilies().size() == 1) {
                // Project the one column family. We must project a column family since it's possible
                // that there are other non declared column families that we need to ignore.
                scan.addFamily(table.getColumnFamilies().get(0).getName().getBytes());
            } else {
                optimizeProjection = true;
                if (projector.projectEveryRow()) {
                    if (table.getViewType() == ViewType.MAPPED) {
                        // Since we don't have the empty key value in MAPPED tables, 
                        // we must project all CFs in HRS. However, only the
                        // selected column values are returned back to client.
                        context.getWhereConditionColumns().clear();
                        for (PColumnFamily family : table.getColumnFamilies()) {
                            context.addWhereConditionColumn(family.getName().getBytes(), null);
                        }
                    } else {
                        byte[] ecf = SchemaUtil.getEmptyColumnFamily(table);
                        // been projected in its entirety.
                        if (!familyMap.containsKey(ecf) || familyMap.get(ecf) != null) {
                            scan.addColumn(ecf, EncodedColumnsUtil.getEmptyKeyValueInfo(table).getFirst());
                        }
                    }
                }
            }
        }
        // Add FirstKeyOnlyFilter if there are no references to key value columns
        if (keyOnlyFilter) {
            ScanUtil.andFilterAtBeginning(scan, new FirstKeyOnlyFilter());
        }
        if (perScanLimit != null) {
            ScanUtil.andFilterAtEnd(scan, new PageFilter(perScanLimit));
        }
        if (offset != null) {
            ScanUtil.addOffsetAttribute(scan, offset);
        }
        int cols = plan.getGroupBy().getOrderPreservingColumnCount();
        if (cols > 0 && keyOnlyFilter && !plan.getStatement().getHint().hasHint(HintNode.Hint.RANGE_SCAN) && cols < plan.getTableRef().getTable().getRowKeySchema().getFieldCount() && plan.getGroupBy().isOrderPreserving() && (context.getAggregationManager().isEmpty() || plan.getGroupBy().isUngroupedAggregate())) {
            ScanUtil.andFilterAtEnd(scan, new DistinctPrefixFilter(plan.getTableRef().getTable().getRowKeySchema(), cols));
            if (plan.getLimit() != null) {
                // We can push the limit to the server
                ScanUtil.andFilterAtEnd(scan, new PageFilter(plan.getLimit()));
            }
        }
        scan.setAttribute(BaseScannerRegionObserver.QUALIFIER_ENCODING_SCHEME, new byte[] { table.getEncodingScheme().getSerializedMetadataValue() });
        scan.setAttribute(BaseScannerRegionObserver.IMMUTABLE_STORAGE_ENCODING_SCHEME, new byte[] { table.getImmutableStorageScheme().getSerializedMetadataValue() });
        // we use this flag on the server side to determine which value column qualifier to use in the key value we return from server.
        scan.setAttribute(BaseScannerRegionObserver.USE_NEW_VALUE_COLUMN_QUALIFIER, Bytes.toBytes(true));
        // So there is no point setting the range.
        if (!ScanUtil.isAnalyzeTable(scan)) {
            setQualifierRanges(keyOnlyFilter, table, scan, context);
        }
        if (optimizeProjection) {
            optimizeProjection(context, scan, table, statement);
        }
    }
}
Also used : NavigableSet(java.util.NavigableSet) FirstKeyOnlyFilter(org.apache.hadoop.hbase.filter.FirstKeyOnlyFilter) DistinctPrefixFilter(org.apache.phoenix.filter.DistinctPrefixFilter) PColumnFamily(org.apache.phoenix.schema.PColumnFamily) PTable(org.apache.phoenix.schema.PTable) Hint(org.apache.phoenix.parse.HintNode.Hint) StatementContext(org.apache.phoenix.compile.StatementContext) RowProjector(org.apache.phoenix.compile.RowProjector) FilterableStatement(org.apache.phoenix.parse.FilterableStatement) PageFilter(org.apache.hadoop.hbase.filter.PageFilter) TableRef(org.apache.phoenix.schema.TableRef)

Example 2 with DistinctPrefixFilter

use of org.apache.phoenix.filter.DistinctPrefixFilter in project phoenix by apache.

the class ScanUtil method setRowKeyOffset.

private static void setRowKeyOffset(Filter filter, int offset) {
    if (filter instanceof BooleanExpressionFilter) {
        BooleanExpressionFilter boolFilter = (BooleanExpressionFilter) filter;
        IndexUtil.setRowKeyExpressionOffset(boolFilter.getExpression(), offset);
    } else if (filter instanceof SkipScanFilter) {
        SkipScanFilter skipScanFilter = (SkipScanFilter) filter;
        skipScanFilter.setOffset(offset);
    } else if (filter instanceof DistinctPrefixFilter) {
        DistinctPrefixFilter prefixFilter = (DistinctPrefixFilter) filter;
        prefixFilter.setOffset(offset);
    }
}
Also used : DistinctPrefixFilter(org.apache.phoenix.filter.DistinctPrefixFilter) BooleanExpressionFilter(org.apache.phoenix.filter.BooleanExpressionFilter) SkipScanFilter(org.apache.phoenix.filter.SkipScanFilter)

Example 3 with DistinctPrefixFilter

use of org.apache.phoenix.filter.DistinctPrefixFilter in project phoenix by apache.

the class ExplainTable method explain.

protected void explain(String prefix, List<String> planSteps) {
    StringBuilder buf = new StringBuilder(prefix);
    ScanRanges scanRanges = context.getScanRanges();
    Scan scan = context.getScan();
    if (scan.getConsistency() != Consistency.STRONG) {
        buf.append("TIMELINE-CONSISTENCY ");
    }
    if (hint.hasHint(Hint.SMALL)) {
        buf.append(Hint.SMALL).append(" ");
    }
    if (OrderBy.REV_ROW_KEY_ORDER_BY.equals(orderBy)) {
        buf.append("REVERSE ");
    }
    if (scanRanges.isEverything()) {
        buf.append("FULL SCAN ");
    } else {
        explainSkipScan(buf);
    }
    buf.append("OVER ").append(tableRef.getTable().getPhysicalName().getString());
    if (!scanRanges.isPointLookup()) {
        appendKeyRanges(buf);
    }
    planSteps.add(buf.toString());
    if (context.getScan() != null && tableRef.getTable().getRowTimestampColPos() != -1) {
        TimeRange range = context.getScan().getTimeRange();
        planSteps.add("    ROW TIMESTAMP FILTER [" + range.getMin() + ", " + range.getMax() + ")");
    }
    PageFilter pageFilter = null;
    FirstKeyOnlyFilter firstKeyOnlyFilter = null;
    BooleanExpressionFilter whereFilter = null;
    DistinctPrefixFilter distinctFilter = null;
    Iterator<Filter> filterIterator = ScanUtil.getFilterIterator(scan);
    if (filterIterator.hasNext()) {
        do {
            Filter filter = filterIterator.next();
            if (filter instanceof FirstKeyOnlyFilter) {
                firstKeyOnlyFilter = (FirstKeyOnlyFilter) filter;
            } else if (filter instanceof PageFilter) {
                pageFilter = (PageFilter) filter;
            } else if (filter instanceof BooleanExpressionFilter) {
                whereFilter = (BooleanExpressionFilter) filter;
            } else if (filter instanceof DistinctPrefixFilter) {
                distinctFilter = (DistinctPrefixFilter) filter;
            }
        } while (filterIterator.hasNext());
    }
    if (whereFilter != null) {
        planSteps.add("    SERVER FILTER BY " + (firstKeyOnlyFilter == null ? "" : "FIRST KEY ONLY AND ") + whereFilter.toString());
    } else if (firstKeyOnlyFilter != null) {
        planSteps.add("    SERVER FILTER BY FIRST KEY ONLY");
    }
    if (distinctFilter != null) {
        planSteps.add("    SERVER DISTINCT PREFIX FILTER OVER " + groupBy.getExpressions().toString());
    }
    if (!orderBy.getOrderByExpressions().isEmpty() && groupBy.isEmpty()) {
        // with GROUP BY, sort happens client-side
        planSteps.add("    SERVER" + (limit == null ? "" : " TOP " + limit + " ROW" + (limit == 1 ? "" : "S")) + " SORTED BY " + orderBy.getOrderByExpressions().toString());
    } else {
        if (offset != null) {
            planSteps.add("    SERVER OFFSET " + offset);
        }
        if (pageFilter != null) {
            planSteps.add("    SERVER " + pageFilter.getPageSize() + " ROW LIMIT");
        }
    }
    Integer groupByLimit = null;
    byte[] groupByLimitBytes = scan.getAttribute(BaseScannerRegionObserver.GROUP_BY_LIMIT);
    if (groupByLimitBytes != null) {
        groupByLimit = (Integer) PInteger.INSTANCE.toObject(groupByLimitBytes);
    }
    groupBy.explain(planSteps, groupByLimit);
    if (scan.getAttribute(BaseScannerRegionObserver.SPECIFIC_ARRAY_INDEX) != null) {
        planSteps.add("    SERVER ARRAY ELEMENT PROJECTION");
    }
}
Also used : PInteger(org.apache.phoenix.schema.types.PInteger) TimeRange(org.apache.hadoop.hbase.io.TimeRange) FirstKeyOnlyFilter(org.apache.hadoop.hbase.filter.FirstKeyOnlyFilter) BooleanExpressionFilter(org.apache.phoenix.filter.BooleanExpressionFilter) FirstKeyOnlyFilter(org.apache.hadoop.hbase.filter.FirstKeyOnlyFilter) DistinctPrefixFilter(org.apache.phoenix.filter.DistinctPrefixFilter) PageFilter(org.apache.hadoop.hbase.filter.PageFilter) Filter(org.apache.hadoop.hbase.filter.Filter) Scan(org.apache.hadoop.hbase.client.Scan) PageFilter(org.apache.hadoop.hbase.filter.PageFilter) DistinctPrefixFilter(org.apache.phoenix.filter.DistinctPrefixFilter) ScanRanges(org.apache.phoenix.compile.ScanRanges) BooleanExpressionFilter(org.apache.phoenix.filter.BooleanExpressionFilter)

Aggregations

DistinctPrefixFilter (org.apache.phoenix.filter.DistinctPrefixFilter)3 FirstKeyOnlyFilter (org.apache.hadoop.hbase.filter.FirstKeyOnlyFilter)2 PageFilter (org.apache.hadoop.hbase.filter.PageFilter)2 BooleanExpressionFilter (org.apache.phoenix.filter.BooleanExpressionFilter)2 NavigableSet (java.util.NavigableSet)1 Scan (org.apache.hadoop.hbase.client.Scan)1 Filter (org.apache.hadoop.hbase.filter.Filter)1 TimeRange (org.apache.hadoop.hbase.io.TimeRange)1 RowProjector (org.apache.phoenix.compile.RowProjector)1 ScanRanges (org.apache.phoenix.compile.ScanRanges)1 StatementContext (org.apache.phoenix.compile.StatementContext)1 SkipScanFilter (org.apache.phoenix.filter.SkipScanFilter)1 FilterableStatement (org.apache.phoenix.parse.FilterableStatement)1 Hint (org.apache.phoenix.parse.HintNode.Hint)1 PColumnFamily (org.apache.phoenix.schema.PColumnFamily)1 PTable (org.apache.phoenix.schema.PTable)1 TableRef (org.apache.phoenix.schema.TableRef)1 PInteger (org.apache.phoenix.schema.types.PInteger)1