/aggregate/groupby

URL: http://GPUDB_IP_ADDRESS:GPUDB_PORT/aggregate/groupby

Calculates unique combinations (groups) of values for the given columns in a given table or view and computes aggregates on each unique combination. This is somewhat analogous to an SQL-style SELECT...GROUP BY.

For aggregation details and examples, see Aggregation. For limitations, see Aggregation Limitations.

Any column(s) can be grouped on, and all column types except unrestricted-length strings may be used for computing applicable aggregates; columns marked as store-only are unable to be used in grouping or aggregation.

The results can be paged via the input parameter offset and input parameter limit parameters. For example, to get 10 groups with the largest counts the inputs would be: limit=10, options={"sort_order":"descending", "sort_by":"value"}.

Input parameter options can be used to customize behavior of this call e.g. filtering or sorting the results.

To group by columns 'x' and 'y' and compute the number of objects within each group, use: column_names=['x','y','count(*)'].

To also compute the sum of 'z' over each group, use: column_names=['x','y','count(*)','sum(z)'].

Available aggregation functions are: count(*), sum, min, max, avg, mean, stddev, stddev_pop, stddev_samp, var, var_pop, var_samp, arg_min, arg_max and count_distinct.

Available grouping functions are Rollup, Cube, and Grouping Sets

This service also provides support for Pivot operations.

Filtering on aggregates is supported via expressions using aggregation functions supplied to having.

The response is returned as a dynamic schema. For details see: dynamic schemas documentation.

If a result_table name is specified in the input parameter options, the results are stored in a new table with that name--no results are returned in the response. Both the table name and resulting column names must adhere to standard naming conventions; column/aggregation expressions will need to be aliased. If the source table's shard key is used as the grouping column(s) and all result records are selected (input parameter offset is 0 and input parameter limit is -9999), the result table will be sharded, in all other cases it will be replicated. Sorting will properly function only if the result table is replicated or if there is only one processing node and should not be relied upon in other cases. Not available when any of the values of input parameter column_names is an unrestricted-length string.

Input Parameter Description

NameTypeDescription
table_namestringName of an existing table or view on which the operation will be performed, in [schema_name.]table_name format, using standard name resolution rules.
column_namesarray of stringsList of one or more column names, expressions, and aggregate expressions.
offsetlongA positive integer indicating the number of initial results to skip (this can be useful for paging through the results). The default value is 0.The minimum allowed value is 0. The maximum allowed value is MAX_INT.
limitlongA positive integer indicating the maximum number of results to be returned, or END_OF_SET (-9999) to indicate that the max number of results should be returned. The number of records returned will never exceed the server's own limit, defined by the max_get_records_size parameter in the server configuration. Use output parameter has_more_records to see if more records exist in the result to be fetched, and input parameter offset & input parameter limit to request subsequent pages of results. The default value is -9999.
encodingstring

Specifies the encoding for returned records. The default value is binary.

Supported ValuesDescription
binaryIndicates that the returned records should be binary encoded.
jsonIndicates that the returned records should be json encoded.
optionsmap of string to strings

Optional parameters. The default value is an empty map ( {} ).

Supported Parameters (keys)Parameter Description
collection_name[DEPRECATED--please specify the containing schema as part of result_table and use /create/schema to create the schema if non-existent] Name of a schema which is to contain the table specified in result_table. If the schema provided is non-existent, it will be automatically created.
expressionFilter expression to apply to the table prior to computing the aggregate group by.
havingFilter expression to apply to the aggregated results.
sort_order

String indicating how the returned values should be sorted - ascending or descending. The default value is ascending.

Supported ValuesDescription
ascendingIndicates that the returned values should be sorted in ascending order.
descendingIndicates that the returned values should be sorted in descending order.
sort_by

String determining how the results are sorted. The default value is value.

Supported ValuesDescription
keyIndicates that the returned values should be sorted by key, which corresponds to the grouping columns. If you have multiple grouping columns (and are sorting by key), it will first sort the first grouping column, then the second grouping column, etc.
valueIndicates that the returned values should be sorted by value, which corresponds to the aggregates. If you have multiple aggregates (and are sorting by value), it will first sort by the first aggregate, then the second aggregate, etc.
result_tableThe name of a table used to store the results, in [schema_name.]table_name format, using standard name resolution rules and meeting table naming criteria. Column names (group-by and aggregate fields) need to be given aliases e.g. ["FChar256 as fchar256", "sum(FDouble) as sfd"]. If present, no results are returned in the response. This option is not available if one of the grouping attributes is an unrestricted string (i.e.; not charN) type.
result_table_persist

If true, then the result table specified in result_table will be persisted and will not expire unless a ttl is specified. If false, then the result table will be an in-memory table and will expire unless a ttl is specified otherwise. The default value is false. The supported values are:

  • true
  • false
result_table_force_replicated

Force the result table to be replicated (ignores any sharding). Must be used in combination with the result_table option. The default value is false. The supported values are:

  • true
  • false
result_table_generate_pk

If true then set a primary key for the result table. Must be used in combination with the result_table option. The default value is false. The supported values are:

  • true
  • false
ttlSets the TTL of the table specified in result_table.
chunk_sizeIndicates the number of records per chunk to be used for the result table. Must be used in combination with the result_table option.
create_indexesComma-separated list of columns on which to create indexes on the result table. Must be used in combination with the result_table option.
view_idID of view of which the result table will be a member. The default value is ''.
materialize_on_gpu

No longer used. See Resource Management Concepts for information about how resources are managed, Tier Strategy Concepts for how resources are targeted for VRAM, and Tier Strategy Usage for how to specify a table's priority in VRAM. The default value is false. The supported values are:

  • true
  • false
pivotpivot column
pivot_valuesThe value list provided will become the column headers in the output. Should be the values from the pivot_column.
grouping_setsCustomize the grouping attribute sets to compute the aggregates. These sets can include ROLLUP or CUBE operartors. The attribute sets should be enclosed in paranthesis and can include composite attributes. All attributes specified in the grouping sets must present in the groupby attributes.
rollupThis option is used to specify the multilevel aggregates.
cubeThis option is used to specify the multidimensional aggregates.

Output Parameter Description

The GPUdb server embeds the endpoint response inside a standard response structure which contains status information and the actual response to the query. Here is a description of the various fields of the wrapper:

NameTypeDescription
statusString'OK' or 'ERROR'
messageStringEmpty if success or an error message
data_typeString'aggregate_group_by_request' or 'none' in case of an error
dataStringEmpty string
data_strJSON or String

This embedded JSON represents the result of the /aggregate/groupby endpoint:

NameTypeDescription
response_schema_strstringAvro schema of output parameter binary_encoded_response or output parameter json_encoded_response.
binary_encoded_responsebytesAvro binary encoded response.
json_encoded_responsestringAvro JSON encoded response.
total_number_of_recordslongTotal/Filtered number of records.
has_more_recordsbooleanToo many records. Returned a partial set.
infomap of string to strings

Additional information. The default value is an empty map ( {} ).

Possible Parameters (keys)Parameter Description
qualified_result_table_nameThe fully qualified name of the table (i.e. including the schema) used to store the results.

Empty string in case of an error.