Search in sources :

Example 1 with SimpleBlockWriter

use of com.amazonaws.athena.connector.lambda.data.SimpleBlockWriter in project aws-athena-query-federation by awslabs.

the class MetadataHandler method doGetTableLayout.

/**
 * Used to get the partitions that must be read from the request table in order to satisfy the requested predicate.
 *
 * @param allocator Tool for creating and managing Apache Arrow Blocks.
 * @param request Provides details of the catalog, database, and table being queried as well as any filter predicate.
 * @return A GetTableLayoutResponse which primarily contains:
 * 1. An Apache Arrow Block with 0 or more partitions to read. 0 partitions implies there are 0 rows to read.
 * 2. Set<String> of partition column names which should correspond to columns in your Apache Arrow Block.
 * @note Partitions are opaque to Amazon Athena in that it does not understand their contents, just that it must call
 * doGetSplits(...) for each partition you return in order to determine which reads to perform and if those reads
 * can be parallelized. This means the contents of this response are more for you than they are for Athena.
 * @note Partitions are partially opaque to Amazon Athena in that it only understands your partition columns and
 * how to filter out partitions that do not meet the query's constraints. Any additional columns you add to the
 * partition data are ignored by Athena but passed on to calls on GetSplits.
 */
public GetTableLayoutResponse doGetTableLayout(final BlockAllocator allocator, final GetTableLayoutRequest request) throws Exception {
    SchemaBuilder constraintSchema = new SchemaBuilder().newBuilder();
    SchemaBuilder partitionSchemaBuilder = new SchemaBuilder().newBuilder();
    /**
     * Add our partition columns to the response schema so the engine knows how to interpret the list of
     * partitions we are going to return.
     */
    for (String nextPartCol : request.getPartitionCols()) {
        Field partitionCol = request.getSchema().findField(nextPartCol);
        partitionSchemaBuilder.addField(nextPartCol, partitionCol.getType());
        constraintSchema.addField(nextPartCol, partitionCol.getType());
    }
    enhancePartitionSchema(partitionSchemaBuilder, request);
    Schema partitionSchema = partitionSchemaBuilder.build();
    if (partitionSchema.getFields().isEmpty() && partitionSchema.getCustomMetadata().isEmpty()) {
        // Even though our table doesn't support complex layouts, partitioning or metadata, we need to convey that there is at least
        // 1 partition to read as part of the query or Athena will assume partition pruning found no candidate layouts to read.
        Block partitions = BlockUtils.newBlock(allocator, PARTITION_ID_COL, Types.MinorType.INT.getType(), 1);
        return new GetTableLayoutResponse(request.getCatalogName(), request.getTableName(), partitions);
    }
    /**
     * Now use the constraint that was in the request to do some partition pruning. Here we are just
     * generating some fake values for the partitions but in a real implementation you'd use your metastore
     * or knowledge of the actual table's physical layout to do this.
     */
    try (ConstraintEvaluator constraintEvaluator = new ConstraintEvaluator(allocator, constraintSchema.build(), request.getConstraints());
        QueryStatusChecker queryStatusChecker = new QueryStatusChecker(athena, athenaInvoker, request.getQueryId())) {
        Block partitions = allocator.createBlock(partitionSchemaBuilder.build());
        partitions.constrain(constraintEvaluator);
        SimpleBlockWriter blockWriter = new SimpleBlockWriter(partitions);
        getPartitions(blockWriter, request, queryStatusChecker);
        return new GetTableLayoutResponse(request.getCatalogName(), request.getTableName(), partitions);
    }
}
Also used : Field(org.apache.arrow.vector.types.pojo.Field) GetTableLayoutResponse(com.amazonaws.athena.connector.lambda.metadata.GetTableLayoutResponse) QueryStatusChecker(com.amazonaws.athena.connector.lambda.QueryStatusChecker) Schema(org.apache.arrow.vector.types.pojo.Schema) SchemaBuilder(com.amazonaws.athena.connector.lambda.data.SchemaBuilder) Block(com.amazonaws.athena.connector.lambda.data.Block) ConstraintEvaluator(com.amazonaws.athena.connector.lambda.domain.predicate.ConstraintEvaluator) SimpleBlockWriter(com.amazonaws.athena.connector.lambda.data.SimpleBlockWriter)

Aggregations

QueryStatusChecker (com.amazonaws.athena.connector.lambda.QueryStatusChecker)1 Block (com.amazonaws.athena.connector.lambda.data.Block)1 SchemaBuilder (com.amazonaws.athena.connector.lambda.data.SchemaBuilder)1 SimpleBlockWriter (com.amazonaws.athena.connector.lambda.data.SimpleBlockWriter)1 ConstraintEvaluator (com.amazonaws.athena.connector.lambda.domain.predicate.ConstraintEvaluator)1 GetTableLayoutResponse (com.amazonaws.athena.connector.lambda.metadata.GetTableLayoutResponse)1 Field (org.apache.arrow.vector.types.pojo.Field)1 Schema (org.apache.arrow.vector.types.pojo.Schema)1