Search in sources :

Example 16 with Filter

use of org.apache.parquet.filter2.compat.FilterCompat.Filter in project Gaffer by gchq.

the class JavaPredicateToParquetPredicate method getIsMoreThanOrEqualToFilter.

private FilterPredicate getIsMoreThanOrEqualToFilter(final String colName, final Object[] parquetObjects, final String group, final SchemaUtils schemaUtils) {
    String[] paths = schemaUtils.getPaths(group, colName);
    if (null == paths) {
        paths = new String[1];
        paths[0] = colName;
    }
    FilterPredicate filter = null;
    for (int i = 0; i < paths.length; i++) {
        final String path = paths[i];
        FilterPredicate tempFilter;
        if (parquetObjects[i] instanceof String) {
            tempFilter = gtEq(binaryColumn(path), Binary.fromString((String) parquetObjects[i]));
        } else if (parquetObjects[i] instanceof Double) {
            tempFilter = gtEq(doubleColumn(path), (Double) parquetObjects[i]);
        } else if (parquetObjects[i] instanceof Float) {
            tempFilter = gtEq(floatColumn(path), (Float) parquetObjects[i]);
        } else if (parquetObjects[i] instanceof Integer) {
            tempFilter = gtEq(intColumn(path), (Integer) parquetObjects[i]);
        } else if (parquetObjects[i] instanceof Long) {
            tempFilter = gtEq(longColumn(path), (Long) parquetObjects[i]);
        } else if (parquetObjects[i] instanceof java.util.Date) {
            tempFilter = gtEq(longColumn(path), ((java.util.Date) parquetObjects[i]).getTime());
        } else if (parquetObjects[i] instanceof java.sql.Date) {
            tempFilter = gtEq(longColumn(path), ((java.sql.Date) parquetObjects[i]).getTime());
        } else if (parquetObjects[i] instanceof Short) {
            tempFilter = gtEq(intColumn(path), ((Short) parquetObjects[i]).intValue());
        } else if (parquetObjects[i] instanceof byte[]) {
            tempFilter = gtEq(binaryColumn(path), Binary.fromReusedByteArray((byte[]) parquetObjects[i]));
        } else {
            fullyApplied = false;
            LOGGER.warn(parquetObjects[i].getClass().getCanonicalName() + " is not a natively supported type for the IsMoreThanOrEqualTo filter, therefore execution will take longer to perform this filter.");
            return null;
        }
        if (null == filter) {
            filter = tempFilter;
        } else {
            filter = and(filter, tempFilter);
        }
    }
    return filter;
}
Also used : FilterPredicate(org.apache.parquet.filter2.predicate.FilterPredicate)

Example 17 with Filter

use of org.apache.parquet.filter2.compat.FilterCompat.Filter in project parquet-mr by apache.

the class ParquetLoader method buildFilter.

private FilterPredicate buildFilter(Expression e) {
    OpType op = e.getOpType();
    if (e instanceof BinaryExpression) {
        Expression lhs = ((BinaryExpression) e).getLhs();
        Expression rhs = ((BinaryExpression) e).getRhs();
        switch(op) {
            case OP_AND:
                return and(buildFilter(lhs), buildFilter(rhs));
            case OP_OR:
                return or(buildFilter(lhs), buildFilter(rhs));
            case OP_BETWEEN:
                BetweenExpression between = (BetweenExpression) rhs;
                return and(buildFilter(OpType.OP_GE, (Column) lhs, (Const) between.getLower()), buildFilter(OpType.OP_LE, (Column) lhs, (Const) between.getUpper()));
            case OP_IN:
                FilterPredicate current = null;
                for (Object value : ((InExpression) rhs).getValues()) {
                    FilterPredicate next = buildFilter(OpType.OP_EQ, (Column) lhs, (Const) value);
                    if (current != null) {
                        current = or(current, next);
                    } else {
                        current = next;
                    }
                }
                return current;
        }
        if (lhs instanceof Column && rhs instanceof Const) {
            return buildFilter(op, (Column) lhs, (Const) rhs);
        } else if (lhs instanceof Const && rhs instanceof Column) {
            return buildFilter(op, (Column) rhs, (Const) lhs);
        }
    } else if (e instanceof UnaryExpression && op == OpType.OP_NOT) {
        return LogicalInverseRewriter.rewrite(not(buildFilter(((UnaryExpression) e).getExpression())));
    }
    throw new RuntimeException("Could not build filter for expression: " + e);
}
Also used : BinaryExpression(org.apache.pig.Expression.BinaryExpression) InExpression(org.apache.pig.Expression.InExpression) UnaryExpression(org.apache.pig.Expression.UnaryExpression) BinaryExpression(org.apache.pig.Expression.BinaryExpression) Expression(org.apache.pig.Expression) BetweenExpression(org.apache.pig.Expression.BetweenExpression) Column(org.apache.pig.Expression.Column) BetweenExpression(org.apache.pig.Expression.BetweenExpression) Const(org.apache.pig.Expression.Const) InExpression(org.apache.pig.Expression.InExpression) OpType(org.apache.pig.Expression.OpType) UnaryExpression(org.apache.pig.Expression.UnaryExpression) FilterPredicate(org.apache.parquet.filter2.predicate.FilterPredicate)

Example 18 with Filter

use of org.apache.parquet.filter2.compat.FilterCompat.Filter in project parquet-mr by apache.

the class ParquetLoader method setPushdownPredicate.

@Override
public void setPushdownPredicate(Expression e) throws IOException {
    LOG.info("Pig pushdown expression: {}", e);
    FilterPredicate pred = buildFilter(e);
    LOG.info("Parquet filter predicate expression: {}", pred);
    storeInUDFContext(ParquetInputFormat.FILTER_PREDICATE, pred);
}
Also used : FilterPredicate(org.apache.parquet.filter2.predicate.FilterPredicate)

Example 19 with Filter

use of org.apache.parquet.filter2.compat.FilterCompat.Filter in project parquet-mr by apache.

the class MessageColumnIO method getRecordReader.

public <T> RecordReader<T> getRecordReader(final PageReadStore columns, final RecordMaterializer<T> recordMaterializer, final Filter filter) {
    checkNotNull(columns, "columns");
    checkNotNull(recordMaterializer, "recordMaterializer");
    checkNotNull(filter, "filter");
    if (leaves.isEmpty()) {
        return new EmptyRecordReader<T>(recordMaterializer);
    }
    return filter.accept(new Visitor<RecordReader<T>>() {

        @Override
        public RecordReader<T> visit(FilterPredicateCompat filterPredicateCompat) {
            FilterPredicate predicate = filterPredicateCompat.getFilterPredicate();
            IncrementallyUpdatedFilterPredicateBuilder builder = new IncrementallyUpdatedFilterPredicateBuilder(leaves);
            IncrementallyUpdatedFilterPredicate streamingPredicate = builder.build(predicate);
            RecordMaterializer<T> filteringRecordMaterializer = new FilteringRecordMaterializer<T>(recordMaterializer, leaves, builder.getValueInspectorsByColumn(), streamingPredicate);
            return new RecordReaderImplementation<T>(MessageColumnIO.this, filteringRecordMaterializer, validating, new ColumnReadStoreImpl(columns, filteringRecordMaterializer.getRootConverter(), getType(), createdBy));
        }

        @Override
        public RecordReader<T> visit(UnboundRecordFilterCompat unboundRecordFilterCompat) {
            return new FilteredRecordReader<T>(MessageColumnIO.this, recordMaterializer, validating, new ColumnReadStoreImpl(columns, recordMaterializer.getRootConverter(), getType(), createdBy), unboundRecordFilterCompat.getUnboundRecordFilter(), columns.getRowCount());
        }

        @Override
        public RecordReader<T> visit(NoOpFilter noOpFilter) {
            return new RecordReaderImplementation<T>(MessageColumnIO.this, recordMaterializer, validating, new ColumnReadStoreImpl(columns, recordMaterializer.getRootConverter(), getType(), createdBy));
        }
    });
}
Also used : ColumnReadStoreImpl(org.apache.parquet.column.impl.ColumnReadStoreImpl) NoOpFilter(org.apache.parquet.filter2.compat.FilterCompat.NoOpFilter) FilteringRecordMaterializer(org.apache.parquet.filter2.recordlevel.FilteringRecordMaterializer) RecordMaterializer(org.apache.parquet.io.api.RecordMaterializer) FilterPredicateCompat(org.apache.parquet.filter2.compat.FilterCompat.FilterPredicateCompat) IncrementallyUpdatedFilterPredicateBuilder(org.apache.parquet.filter2.recordlevel.IncrementallyUpdatedFilterPredicateBuilder) FilterPredicate(org.apache.parquet.filter2.predicate.FilterPredicate) IncrementallyUpdatedFilterPredicate(org.apache.parquet.filter2.recordlevel.IncrementallyUpdatedFilterPredicate) UnboundRecordFilterCompat(org.apache.parquet.filter2.compat.FilterCompat.UnboundRecordFilterCompat) IncrementallyUpdatedFilterPredicate(org.apache.parquet.filter2.recordlevel.IncrementallyUpdatedFilterPredicate)

Example 20 with Filter

use of org.apache.parquet.filter2.compat.FilterCompat.Filter in project parquet-mr by apache.

the class FilterCompat method get.

/**
 * Given a FilterPredicate, return a Filter that wraps it.
 * This method also logs the filter being used and rewrites
 * the predicate to not include the not() operator.
 */
public static Filter get(FilterPredicate filterPredicate) {
    checkNotNull(filterPredicate, "filterPredicate");
    LOG.info("Filtering using predicate: {}", filterPredicate);
    // rewrite the predicate to not include the not() operator
    FilterPredicate collapsedPredicate = LogicalInverseRewriter.rewrite(filterPredicate);
    if (!filterPredicate.equals(collapsedPredicate)) {
        LOG.info("Predicate has been collapsed to: {}", collapsedPredicate);
    }
    return new FilterPredicateCompat(collapsedPredicate);
}
Also used : FilterPredicate(org.apache.parquet.filter2.predicate.FilterPredicate)

Aggregations

FilterPredicate (org.apache.parquet.filter2.predicate.FilterPredicate)24 ArrayList (java.util.ArrayList)7 BlockMetaData (org.apache.parquet.hadoop.metadata.BlockMetaData)6 Path (org.apache.hadoop.fs.Path)5 ParquetMetadata (org.apache.parquet.hadoop.metadata.ParquetMetadata)5 MessageType (org.apache.parquet.schema.MessageType)5 HashMap (java.util.HashMap)4 FilterCompat (org.apache.parquet.filter2.compat.FilterCompat)4 HashSet (java.util.HashSet)3 List (java.util.List)3 TypeInfo (org.apache.hadoop.hive.serde2.typeinfo.TypeInfo)3 ParquetFileReader (org.apache.parquet.hadoop.ParquetFileReader)3 Pair (uk.gov.gchq.gaffer.commonutil.pair.Pair)3 FileSystem (org.apache.hadoop.fs.FileSystem)2 SearchArgument (org.apache.hadoop.hive.ql.io.sarg.SearchArgument)2 UnboundRecordFilter (org.apache.parquet.filter.UnboundRecordFilter)2 Filter (org.apache.parquet.filter2.compat.FilterCompat.Filter)2 FilterPredicateCompat (org.apache.parquet.filter2.compat.FilterCompat.FilterPredicateCompat)2 ParquetInputSplit (org.apache.parquet.hadoop.ParquetInputSplit)2 Test (org.junit.Test)2