The subject matter described herein relates to count distinct operations using vertical unions.
Advanced databases are providing functionality similar to conventional business warehouse architectures by handling more complex calculations. For example, databases can calculate the count distinct of attributes such as count distinct on customers for each region. In order to calculate such a query, different aggregation levels and intermediate results must be combined. Such operations are typically done by using a join operation that can unnecessarily consume time, processing resources, and/or memory.
In one aspect, a query statement is received that specifies a count distinct. Thereafter, a data flow graph that comprises a plurality of nodes for executing the query is generated. The nodes provide aggregation operations, sorting of results on join attributes and vertically appending columns of count distinct results with intermediate results from at least one of the aggregation operations. Thereafter, execution of the query is initiated using the data flow graph.
The query can be executed against an in-memory database such as an in-memory columnar database. In addition, columns of count distinct results can be vertically appended using a vertical union operation.
Articles of manufacture are also described that comprise computer executable instructions permanently stored (e.g., non-transitorily stored, etc.) on computer readable media, which, when executed by a computer, causes the computer to perform operations herein. Similarly, computer systems are also described that may include a processor and a memory coupled to the processor. The memory may temporarily or permanently store one or more programs that cause the processor to perform one or more of the operations described herein. In addition, methods can be implemented by one or more data processors either within a single computing system or distributed among two or more computing systems.
The subject matter described herein provides many advantages. For example, the current subject matter enables the calculation of count distinct of attributes that reduces query runtime as compared to conventional techniques while additionally consuming fewer processing resources. In particular, the current subject matter avoids complex joins which can consume significant resources.
The details of one or more variations of the subject matter described herein are set forth in the accompanying drawings and the description below. Other features and advantages of the subject matter described herein will be apparent from the description and drawings, and from the claims.
The current subject matter can be implemented, for example, in connection with a calculation engine environment such as that illustrated in the diagram 200 of
Calculation scenarios can include individual calculation nodes 211-214, which in turn each define operations such as joining various physical or logical indexes and other calculation nodes (e.g., CView 4 is a join of CView 2 and CView 3). That is, the input for a calculation node 211-214 can be one or more physical, join, or OLAP indexes or calculation nodes.
In calculation scenarios, two different representations can be provided. First, a pure calculation scenario in which all possible attributes are given. Second, an instantiated model that contains only the attributes requested in the query (and required for further calculations). Thus, calculation scenarios can be created that can be used for various queries. With such an arrangement, calculation scenarios can be created which can be reused by multiple queries even if such queries do not require every attribute specified by the calculation scenario. Example environments for implementing calculation scenarios can be found, for example, in U.S. patent application Ser. No. 12/914,445, the contents of which are hereby fully incorporated by reference.
Every calculation scenario can be uniquely identifiable by a name (i.e., the calculation scenario can be a database object with a unique identifier, etc.). This means, that the calculation scenario can be queried in a manner similar to a view in a SQL database. Thus, the query is forwarded to the calculation node 211-214 for the calculation scenario that is marked as the corresponding default node. In addition, a query can be executed on a particular calculation node 211-214 (as specified in the query). Furthermore, nested calculation scenarios can be generated in which one calculation scenario is used as source in another calculation scenario (via a calculation node 211-214 in this calculation scenario). Each calculation node 211-214 can have one or more output tables. One output table can be consumed by several calculation nodes 211-214.
For this example, that data source is an OLAP cube “SalesCube” 330, which has the three dimensions Customer, Year, and Product as well as the measure Sales. As stated, this data source 310 can be entered as a special DataSource node in the logical layer 220 in the calculation scenario. The DataSource is now referenced from two calculation nodes. The calculation node TotalsCV 320 on the left side calculates the sales total, by simply summing the sales without any GroupBy attributes. The calculation node SalesCV 330 on the right side calculates the sales according to the GroupBys. To calculate their relationship, the two calculation nodes 320, 330 are joined with each other using a CrossJoin. In the calculation node RatioCV 340 after the join, all the attributes needed for the calculation are available and a new calculated attribute Ratio is provided.
The implementation of
The calculation scenario 515 can be a directed acyclic graph with arrows representing data flows and nodes that represent operations. Each calculation node has a set of inputs and outputs and an operation that transforms the inputs into the outputs. In addition to their primary operation, each calculation node can also have a filter condition for filtering the result set. The inputs and the outputs of the operations can be table valued parameters (i.e., user-defined table types that are passed into a procedure or function and provide an efficient way to pass multiple rows of data to the application server). Inputs can be connected to tables or to the outputs of other calculation nodes. Calculation scenarios 515 can support a variety of node types such as (i) nodes for set operations such as projection, aggregation, join, union, minus, intersection, and (ii) SQL nodes that execute a SQL statement which is an attribute of the node. In addition, to enable parallel execution, a calculation scenario 515 can contain split and merge operations. A split operation can be used to partition input tables for subsequent processing steps based on partitioning criteria. Operations between the split and merge operation can then be executed in parallel for the different partitions. Parallel execution can also be performed without split and merge operation such that all nodes on one level can be executed in parallel until the next synchronization point. Split and merge allows for enhanced/automatically generated parallelization. If a user knows that the operations between the split and merge can work on portioned data without changing the result he or she can use a split. Then, the nodes can be automatically multiplied between split and merge and partition the data.
A calculation scenario 515 can be defined as part of database metadata and invoked multiple times. A calculation scenario 515 can be created, for example, by a SQL statement “ALTER SYSTEM ADD SCENARIO <xml OR json representing the scenario>”. Once a calculation scenario 515 is created, it can be queried (e.g., “SELECT A, B, C FROM <scenario name>”, etc.). In some cases, databases can have pre-defined calculation scenarios 515 (default, previously defined by users, etc.). The calculation scenarios 515 can be persisted in a repository (coupled to the database server) or in transient scenarios, the calculation scenarios 515 can be kept in-memory.
Calculation scenarios 515 are more powerful than traditional SQL queries or SQL views for many reasons. One reason is the possibility to define parameterized calculation schemas that are specialized when the actual query is issued. Unlike a SQL view, a calculation scenario 515 does not describe the actual query to be executed. Rather, it describes the structure of the calculation. Further information is supplied when the calculation scenario is executed. This further information can include parameters that represent values (for example in filter conditions). To obtain more flexibility, it is also possible to refine the operations when the model is invoked. For example, at definition time, the calculation scenario 515 may contain an aggregation node containing all attributes. Later, the attributes for grouping can be supplied with the query. This allows having a predefined generic aggregation, with the actual aggregation dimensions supplied at invocation time. The calculation engine 520 can use the actual parameters, attribute list, grouping attributes, and the like supplied with the invocation to instantiate a query specific calculation scenario 515. This instantiated calculation scenario 515 is optimized for the actual query and does not contain attributes, nodes or data flows that are not needed for the specific invocation.
When the calculation engine 520 gets a request to execute a calculation scenario 515, it can first optimize the calculation scenario 515 using a rule based model optimizer 522. Examples for optimizations performed by the model optimizer can include “pushing down” filters and projections so that intermediate results 526 are narrowed down earlier, or the combination of multiple aggregation and join operations into one node. The optimized model can then be executed by a calculation engine model executor 524 (a similar or the same model executor can be used by the database directly in some cases). This includes decisions about parallel execution of operations in the calculation scenario 515. The model executor 524 can invoke the required operators (using, for example, a calculation engine operators module 528) and manage intermediate results. Most of the operators are executed directly in the calculation engine 520 (e.g., creating the union of several intermediate results). The remaining nodes of the calculation scenario 515 (not implemented in the calculation engine 520) can be transformed by the model executor 524 into a set of logical database execution plans. Multiple set operation nodes can be combined into one logical database execution plan if possible.
The calculation scenarios 515 of the calculation engine 520 can be exposed as a special type of database views called calculation views. That means a calculation view can be used in SQL queries and calculation views can be combined with tables and standard views using joins and sub queries. When such a query is executed, the database executor inside the SQL processor needs to invoke the calculation engine 520 to execute the calculation scenario 515 behind the calculation view. In some implementations, the calculation engine 520 and the SQL processor are calling each other: on one hand the calculation engine 520 invokes the SQL processor for executing set operations and SQL nodes and, on the other hand, the SQL processor invokes the calculation engine 520 when executing SQL queries with calculation views.
Diagram 900 of
Various implementations of the subject matter described herein may be realized in digital electronic circuitry, integrated circuitry, specially designed ASICs (application specific integrated circuits), computer hardware, firmware, software, and/or combinations thereof. These various implementations may include implementation in one or more computer programs that are executable and/or interpretable on a programmable system including at least one programmable processor, which may be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.
These computer programs (also known as programs, software, software applications or code) include machine instructions for a programmable processor, and may be implemented in a high-level procedural and/or object-oriented programming language, and/or in assembly/machine language. As used herein, the term “machine-readable medium” refers to any computer program product, apparatus and/or device (e.g., magnetic discs, optical disks, memory, Programmable Logic Devices (PLDs)) used to provide machine instructions and/or data to a programmable processor, including a machine-readable medium that receives machine instructions as a machine-readable signal. The term “machine-readable signal” refers to any signal used to provide machine instructions and/or data to a programmable processor.
The subject matter described herein may be implemented in a computing system that includes a back-end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a client computer having a graphical user interface or a Web browser through which a user may interact with an implementation of the subject matter described herein), or any combination of such back-end, middleware, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (“LAN”), a wide area network (“WAN”), and the Internet.
The computing system may include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other.
Although a few variations have been described in detail above, other modifications are possible. For example, the logic flow depicted in the accompanying figure(s) and described herein do not require the particular order shown, or sequential order, to achieve desirable results. Other embodiments may be within the scope of the following claims.