System and method for caching and parameterizing IR

• US 10,339,137 B2
• Filed: 12/07/2015
• Issued: 07/02/2019
• Est. Priority Date: 12/07/2015
• Status: Active Grant

• Current Assignee: Futurewei Technologies Incorporated (Huawei Investment & Holding Co., Ltd.)
• Original Assignee: Futurewei Technologies Incorporated (Huawei Investment & Holding Co., Ltd.)
• Inventors: Rastogi, Kumar Rajeev, Ding, Yonghua, Zhu, Cheng
• Primary Examiner(s): Obisesan, Augustine K.
• Assistant Examiner(s): Ennaji, Lahcen
• Application Number:
  US14/960,894
• Publication Number:
  US 20170161325A1
• Time in Patent Office: 1,303 Days
• CPC Class Codes: G06F 16/24524 Access plan code generation and invalidation; Reuse of access plans
• CPC Class Codes: G06F 16/24542 Plan optimisation

Abstract

A system and method of caching and parameterizing intermediate representation code includes receiving, by a database, a query, parsing, by the database, the query to obtain a plan tree comprising a plurality of plan nodes arranged in hierarchical order descending from a top plan node, generating, by the database, node intermediate representations (IRs) for the plan nodes, executing, by the database, a first query using the node IRs, and reusing, by the database, the node IRs to execute subsequent queries.

Claims

1. A method comprising:

receiving, by a database, a first query;

parsing, by the database, the first query to obtain a plan tree comprising a plurality of plan nodes arranged in hierarchical order descending from a top plan node;

generating, by the database, first node intermediate representations (IRs) for the plan nodes;

parameterizing, by the database, the first node IRs to replace one or more constants or tuple attributes in bodies of the first node IRs each with a respective parameter;

combining, by the database, the parameterized first node IRs to obtain a first module for the first query;

executing, by the database, the first query using the first module;

receiving, by the database, a second query;

generating, by the database, second node IRs for the second query;

parameterizing, by the database, the second node IRs to replace one or more constants or tuple attributes in bodies of the second node IRs each with a respective parameter;

matching, by the database, one or more of the parameterized first node IRs each with a corresponding one of the parameterized second node IRs, the matching parameterized first node IRs differing from the corresponding parameterized second node IRs in the respective parameters for each;

reusing, by the database, the parameterized first node IRs to obtain a second module for the second query, the second module being compiled from the matching parameterized first node IRs and unmatched parameterized second node IRs; and

executing, by the database, the second query using the second module.

2. The method of claim 1, further comprising:

storing the first module with the top plan node of the plan tree in a plan cache; and

producing an executable object for the first module.

3. The method of claim 2, wherein producing the executable object for the first module comprises compiling the first module to obtain the executable object and storing the executable object in the plan cache.

4. The method of claim 2, further comprising:

retrieving the first module from the plan cache;

determining whether the plan cache contains the executable object for the first module;

retrieving the executable object, in response to the plan cache containing the executable object;

remapping a memory address of the executable object and producing a function pointer to the memory address; and

executing the executable object with the function pointer.

5. The method of claim 2, further comprising retrieving the parameterized first node IRs from the plan cache.

6. A method comprising:

receiving, by a database, a first query;

parsing, by the database, the first query to obtain a first plan tree comprising a first plurality of plan nodes;

generating, by the database, first node intermediate representations (IRs) for the first plurality of plan nodes;

parameterizing, by the database, the first node IRs to produce first parameterized IRs, the parameterizing comprising replacing one or more constants or tuple attributes in bodies of the first node IRs with corresponding parameters;

combining, by the database, the first parameterized IRs to produce a first module;

storing, by the database, the first module, the first parameterized IRs, and the first plan tree in a plan cache;

executing, by the database, the first query with the first module;

receiving, by the database, a second query;

parsing the second query to obtain a second plan tree comprising a second plurality of plan nodes, the second plurality of plan nodes being different from the first plurality of plan nodes;

generating, by the database, second node IRs for the second plurality of plan nodes;

parameterizing, by the database, the second node IRs to produce second parameterized IRs;

locating, by the database, matching ones of the first parameterized IRs for the second parameterized IRs, the matching ones of the first parameterized IRs differing from corresponding ones of the second parameterized IRs by the corresponding parameters;

combining, by the database, the matching ones of the first parameterized IRs and unmatched ones of the second parameterized IRs to produce a second module; and

executing, by the database, the second query using the second module.

7. The method of claim 6, wherein parameterizing the first node IRs to produce parameterized IRs comprises:

determining whether any bodies of the first node IRs contain expressions with a constant; and

replacing the constant with an input parameter.

8. The method of claim 7, wherein the constant comprises a literal value.

9. The method of claim 7, wherein the constant comprises a bind value.

10. The method of claim 6, wherein parameterizing the first node IRs to produce parameterized IRs comprises:

determining whether any bodies of the first node IRs contain expressions with a tuple attribute; and

replacing the tuple attribute with an attribute position parameter.

11. The method of claim 10, wherein the tuple attribute comprises a column of a table.

12. The method of claim 6, wherein storing the first module and the first plan tree in the plan cache comprises:

compiling the first parameterized IRs in the first module to produce an executable object; and

storing the executable object in the plan cache with a top plan node of the first plurality of plan nodes.

13. The method of claim 12, wherein a quantity of modules in the plan cache is less than a quantity of plan trees in the plan cache.

14. A device comprising:

a processor; and

a computer-readable storage medium storing a program to be executed by the processor, the program including instructions for;

receiving a first query;

parsing the first query to obtain a plan tree comprising a plurality of plan nodes arranged in hierarchical order descending from a top plan node;

generating first node intermediate representations (IRs) for the plan nodes;

parameterizing the first node IRs to replace one or more constants or tuple attributes in bodies of the first node IRs each with a respective parameter;

combining the parameterized first node IRs to obtain a first module for the first query;

executing the first query using the first module;

receiving a second query;

generating second node IRs for the second query;

parameterizing the second node IRs to replace one or more constants or tuple attributes in bodies of the second node IRs each with a respective parameter;

matching one or more of the parameterized first node IRs each with a corresponding one of the parameterized second node IRs, the matching parameterized first node IRs differing from the corresponding parameterized second node IRs in the respective parameters for each;

reusing the parameterized first node IRs to obtain a second module for the second query, the second module being compiled from the matching parameterized first node IRs and unmatched parameterized second node IRs; and

executing the second query using the second module.

15. The device of claim 14, where the program further includes instructions for:

storing the first module and the top plan node of the plan tree alongside one another in a plan cache;

compiling the first module to obtain an executable object; and

storing the executable object in the plan cache.

16. A device comprising:

a processor; and

a computer-readable storage medium storing a program to be executed by the processor, the program including instructions for;

receiving a first query;

parsing the first query to obtain a first plan tree comprising a first plurality of plan nodes;

generating first node intermediate representations (IRs) for the first plurality of plan nodes;

parameterizing the first node IRs to produce first parameterized IRs, the parameterizing comprising replacing one or more constants or tuple attributes in bodies of the first node IRs with corresponding parameters;

combining the first parameterized IRs to produce a first module;

storing the first module, the first parameterized IRs, and the first plan tree in a plan cache;

executing the first query with the first module;

receiving a second query;

parsing the second query to obtain a second plan tree comprising a second plurality of plan nodes, the second plurality of plan nodes being different from the first plurality of plan nodes;

generating second node IRs for the second plurality of plan nodes;

parameterizing the second node IRs to produce second parameterized IRs;

locating matching ones of the first parameterized IRs for the second parameterized IRs, the matching ones of the first parameterized IRs differing from corresponding ones of the second parameterized IRs by the corresponding parameters;

combining the matching ones of the first parameterized IRs and unmatched ones of the second parameterized IRs to produce a second module; and

executing the second query with the second module.

17. The device of claim 16, wherein the instruction for parameterizing the first node IRs to produce the first parameterized IRs comprises:

determining whether any bodies of the first node IRs contain expressions with constants; and

replacing the constants with corresponding input parameters.

18. The device of claim 16, wherein the instruction for parameterizing the first node IRs to produce the first parameterized IRs comprises:

determining whether any bodies of the first node IRs contain expressions with a tuple attribute; and

replacing the tuple attribute with an attribute position parameter.

19. The device of claim 16, wherein the instruction for storing the first module and the first plan tree in the plan cache comprises:

compiling the first parameterized IRs in the first module to produce an executable object; and

storing the executable object in the plan cache with a top plan node of the first plurality of plan nodes.