WO2024239782A1 - Query plan construction method and apparatus, electronic device and storage medium - Google Patents

Abstract

The present application discloses a query plan construction method and apparatus, an electronic device and a storage medium. The method of embodiments of the present application comprises: acquiring a data query expression, the data query expression comprising a keyword; obtaining basic sub-expressions in the data query expression and an association relationship between the basic sub-expressions according to the keyword; expanding the basic sub-expressions to obtain equivalent sub-expressions of the basic sub-expressions, wherein an equivalent sub-expression and a corresponding basic sub-expression have an equivalent expression meaning; from the basic sub-expressions and the equivalent sub-expressions of the basic sub-expressions, determining sub-expressions to be constructed; and constructing a data query plan by means of the sub-expressions to be constructed and the association relationship. In the present application, the data query plan can be constructed according to the association relationship by a plurality of sub-expressions to be constructed, so that the plurality of sub-expressions to be constructed do not conflict with each other during construction of the data query plan, thereby facilitating the construction of the data query plan, and improving the query performance.

Description

Query plan construction method, device, electronic device and storage medium

This application claims the priority of the Chinese patent application filed with the China Patent Office on May 19, 2023, with application number 202310575987.9 and application name “Query plan construction method, device, electronic device and storage medium”, all contents of which are incorporated by reference in this application.

Technical Field

The present application relates to the field of computers, and in particular to a query plan construction method, device, electronic device and storage medium.

Background Art

A database management system is a software system used to manage and organize data. It can help users create, manage, store and retrieve data. When retrieving data, the database management system can expand data queries through query optimization technology, build query plans based on the expanded data queries, and then query data based on the query plans.

However, the operation steps in the current query plan may conflict with each other, resulting in the data obtained by the query plan constructed being inconsistent with expectations, which affects query efficiency.

Summary of the invention

The embodiments of the present application provide a query plan construction method, device, electronic device and storage medium, which can facilitate the construction of data query plans and improve query efficiency.

The embodiment of the present application provides a query plan construction method, which is executed by an electronic device and includes:

Get data query expressions, which include keywords;

According to the keywords, basic sub-expressions in the data query expression and the association relationship between the basic sub-expressions are obtained;

Expanding the basic sub-expression to obtain an equivalent sub-expression of the basic sub-expression, the basic sub-expression and the equivalent sub-expression of the basic sub-expression have equivalent expression meanings;

Determine the sub-expression to be constructed from the basic sub-expression and the equivalent sub-expression of the basic sub-expression;

Construct a data query plan based on the sub-expressions to be constructed and the associated relationships.

The embodiment of the present application further provides a query plan construction device, which is deployed on an electronic device and includes:

An acquisition unit, used for acquiring a data query expression, where the data query expression includes keywords;

The key unit is used to obtain the basic sub-expressions in the data query expression and the association relationship between the basic sub-expressions according to the keywords;

The equivalent unit is used to expand the basic sub-expression to obtain an equivalent sub-expression of the basic sub-expression. The basic sub-expression and the equivalent sub-expression of the basic sub-expression have equivalent expression meanings;

A determination unit, used for determining a sub-expression to be constructed from a basic sub-expression and an equivalent sub-expression of the basic sub-expression;

A construction unit is used to construct a data query plan through sub-expressions to be constructed and associated relationships.

An embodiment of the present application also provides an electronic device, including a memory storing multiple instructions; the processor loads instructions from the memory to execute the steps in any query plan construction method provided in the embodiment of the present application.

An embodiment of the present application also provides a computer-readable storage medium, which stores a plurality of instructions, and the instructions are suitable for a processor to load to execute the steps in any query plan construction method provided in the embodiment of the present application.

An embodiment of the present application also provides a computer program product, including a computer program/instruction, which, when executed by a processor, implements the steps in any query plan construction method provided in the embodiment of the present application.

The embodiment of the present application can obtain a data query expression, which includes keywords. At the same time, there can be multiple basic sub-expressions in the data query expression, and there is an association relationship between each basic sub-expression, so the basic sub-expressions in the data query expression and the association relationship between the basic sub-expressions can be obtained according to the keywords. Then the basic sub-expression is expanded to obtain the equivalent sub-expression of the basic sub-expression. Since the equivalent sub-expression and the basic sub-expression have equivalent expression meanings, the equivalent sub-expression has the association relationship of the basic sub-expression, and the sub-expression to be constructed can be screened out from the basic sub-expression and the equivalent sub-expression. The sub-expression to be constructed can be a basic sub-expression or an equivalent sub-expression. Then, a data query plan is constructed by the sub-expression to be constructed and the association relationship. Multiple sub-expressions to be constructed can construct a data query plan based on the association relationship, so that multiple sub-expressions to be constructed do not conflict with each other when constructing the data query plan. In this way, it is convenient to construct a data query plan and improve query performance.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

FIG. 1a is a schematic diagram of a scenario of a query plan construction method provided in an embodiment of the present application;

FIG1b is a flow chart of a query plan construction method provided in an embodiment of the present application;

FIG2a is a schematic diagram of the structure of a query system provided in an embodiment of the present application;

FIG2b is a schematic structural diagram of a query optimizer provided in an embodiment of the present application;

FIG2c is a schematic diagram of creating a sub-expression provided in an embodiment of the present application;

FIG2d is a schematic diagram of obtaining metadata information provided by an embodiment of the present application;

FIG. 2e is a schematic diagram of a structure for constructing a data query plan provided in an embodiment of the present application.

FIG2f is a schematic diagram of a flow chart of constructing a data query plan provided in an embodiment of the present application;

FIG3 is a schematic diagram of the structure of a query plan construction device provided in an embodiment of the present application;

FIG. 4 is a schematic diagram of the structure of a server provided in an embodiment of the present application.

DETAILED DESCRIPTION

The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

Embodiments of the present application provide a query plan construction method, device, electronic device and storage medium.

The query plan construction device can be integrated into an electronic device, which can be a terminal, a server, or other devices. The terminal can be a mobile phone, a tablet computer, a smart Bluetooth device, a laptop, or a personal computer (PC), etc. The server can be a single server or a server cluster composed of multiple servers.

In some embodiments, the query plan construction device can also be integrated into multiple electronic devices. For example, the query plan construction device can be integrated into multiple servers, and the query plan construction method of the present application can be implemented by multiple servers.

In some embodiments, the server may also be implemented in the form of a terminal.

For example, referring to Figure 1a, the electronic device can obtain a data query expression, which includes keywords; based on the keywords, obtain basic sub-expressions in the data query expression, and the association relationship between the basic sub-expressions; expand the basic sub-expressions to obtain equivalent sub-expressions of the basic sub-expressions, and the equivalent sub-expressions and the basic sub-expressions have equivalent expression meanings; determine the sub-expressions to be constructed from the basic sub-expressions and the equivalent sub-expressions of the basic sub-expressions; and construct a data query plan through the sub-expressions to be constructed and the association relationships.

Among them, there can be multiple basic sub-expressions in the data query expression, and there is an association relationship between each basic sub-expression. The basic sub-expression can be extended through the equivalent sub-expression. Since the equivalent sub-expression and the basic sub-expression have equivalent expression meanings, the equivalent sub-expression has an association relationship with the basic sub-expression. The sub-expression to be constructed can be screened out from the basic sub-expression and the equivalent sub-expression. The sub-expression to be constructed can be a basic sub-expression or an equivalent sub-expression. Multiple sub-expressions to be constructed can construct a data query plan based on the association relationship, so that multiple sub-expressions to be constructed do not conflict with each other when constructing the data query plan. In this way, it is easy to construct a data query plan and improve the query performance.

It should be noted that the serial numbers of the following embodiments are not intended to limit the preferred order of the embodiments.

In this embodiment, a query plan construction method is provided, as shown in FIG1b , and the specific process of the query plan construction method can be as follows:

110. Obtain a data query expression, where the data query expression includes keywords.

Among them, data query expression is a grammatical structure used to query and filter data from multiple database management systems. For example, the language of data query expression can be simple expression language (Expression Language, EL), query expression language (Query Domain Specific Language, Query DSL), a Java persistence standard query language (Java Persistence API Query Language, JPQL), an expression language in the Spring framework (Spring Expression Language, SpEL), etc.

Keywords are special words or characters used to specify query conditions in data query expressions, where a data query expression can contain multiple query conditions. For example, keywords may include query (SELECT): used to select the data columns to be queried from the database, from (FROM) used to specify the table of data to be queried, condition (WHERE) used to specify the filtering conditions for query results, sort (ORDER BY) used to sort the query results according to the specified column, and so on.

It is understandable that the reason why the data query expression of EL language can interact with multiple database management systems is that it does not directly access the database, but accesses the database through Java Database Connectivity (JDBC), where JDBC is a Java interface used to execute SQL statements in Java applications and interact with multiple database management systems. In this way, through the data query expression of EL language, applications can be easily connected to multiple types and versions of database management systems, thereby obtaining a wider range of functions and interoperability. In addition, the data query expression of EL language also has good portability and cross-platform, and can run on multiple platforms and application servers.

The method for obtaining data query expressions can be:

(1) If the query optimizer runs outside the database management system, multiple database management systems can use one query optimizer. After the user enters a data query command into any database management system, any database management system can convert the data query command into a data query expression and then send the data query expression to the query optimizer.

(2) If the query optimizer is running in a database management system, after the user inputs a data query command into the database management system, the database management system can convert the data query command into a data query expression, and then transmit the data query expression to the query optimizer;

Among them, the query optimizer is a component that provides query optimization technology.

In some embodiments, considering that the types of database management systems may include relational database management systems, non-relational database management systems, in-memory database management systems, distributed database management systems, and data warehouses, etc., in order to enable the query optimizer to be deployed in multiple different database management systems, obtaining a data query expression includes:

Get data query command, the data query command includes keywords;

According to the keywords, the data query command is parsed and processed to obtain the syntax tree of the data query command;

The preset data exchange language is used to perform format conversion on the syntax tree to obtain a data query expression.

The data query command is a command input by a user when querying data from a database management system. For example, the data query command can be a query field input by a user, a voice command of a user, or a command output when a data query script written by a user is executed, etc.

Keywords can correspond to query conditions in data query commands.

For example, the data query command is to filter out the same data from the ath column of table T1 and the bth column of table T2. In this way, the query conditions in the data query command may include querying the ath column of table T1, querying the bth column of table T2, and filtering out the same data from the ath column of table T1 and the bth column of table T2. The keyword can be used to indicate querying the ath column of table T1, to indicate querying the bth column of table T2, and to indicate filtering out the same data from the ath column of table T1 and the bth column of table T2.

The syntax tree represents the query logic of the data query command, wherein the syntax tree includes multiple nodes representing different operations.

For example, in the syntax tree, the SELECT node represents the operation of querying data, column 1 and column 2 nodes represent the columns to be queried, the FROM node represents which tables to query from, table1 and table2 are the names of these tables, the WHERE node represents the conditions to limit the query results, and so on.

The preset data interaction language can enable the query optimizer to interact with different types of database management systems. For example, the preset data interaction language can be EL language, Query DSL language, JPQL language, etc.

120. According to the keywords, basic sub-expressions in the data query expression and the association relationship between the basic sub-expressions are obtained.

The basic sub-expression is used to implement the query conditions indicated by the keywords in the data query expression. For example, the basic sub-expression can be used to select the data columns of the table to be queried from the database management system, to specify the filtering conditions of the query results, to sort the query results according to the specified columns, and so on.

The association relationship is used to indicate the logical order of execution of the basic sub-expressions. For example, the association relationship can include parallel relationships, progressive relationships, and AND relationships between basic sub-expressions, etc. Among them, the AND relationship is used to limit multiple query conditions to the same query. Only when all conditions are met at the same time can the data that meets the requirements be returned.

For example, if the data query expression is "SELECT T1.a FROM T1,T2 WHERE T1.a＝T2.b ORDER BY T1.a", the data query expression indicates that the table join method is used to connect table T1 and table T2. The join condition is T1.a＝T2.b, which means that only the value of column a in table T1 is equal to the value of column b in table T2 will be returned. That is, the data that meets the query conditions is filtered out through the join condition T1.a＝T2.b, and the data is sorted according to the value of column T1.a.

Through "SELECT T1.a", "FROM T1, T2", "WHERE T1.a＝T2.b", "ORDER BY T1.a" in the data query expression, we can get basic sub-expression 1, basic sub-expression 2 and basic sub-expression 3. Basic sub-expression 1 is used to retrieve the a-th column data of table T1, basic sub-expression 2 is used to retrieve the b-th column data of table T2, and basic sub-expression 3 is used to indicate the inner join between table T1 and table T2 (Inner Join(T1.a＝T2.b)), which is used to filter the query results of basic sub-expression 1 and basic sub-expression 2.

Among them, there is a parallel relationship between basic sub-expression 1 and basic sub-expression 2, and there is an AND relationship between basic sub-expression 1 and basic sub-expression 2 and basic sub-expression 3 respectively, that is, after basic sub-expression 1 and basic sub-expression 2 obtain query results respectively, the query results of basic sub-expression 1 and the query results of basic sub-expression 2 can be filtered through basic sub-expression 3.

130. Expand the basic sub-expression to obtain an equivalent sub-expression of the basic sub-expression, and the equivalent sub-expression and the basic sub-expression have equivalent expression meanings.

The equivalent sub-expression can be used to replace the basic sub-expression, and has the same query result as the basic sub-expression, but the query logic is different, wherein the basic sub-expression corresponds to one or more equivalent sub-expressions.

The equivalent expression meaning is used to limit the equivalent sub-expression and the basic sub-expression to have the same semantics, so that the query result of the equivalent sub-expression is the same as the query result of the basic sub-expression.

For example, if the basic sub-expression is a sub-expression that reads (Get) the data in the a-th column of table T1, then the equivalent sub-expression can be a sub-expression that scans (Scan) the data in the a-th column of table T1, or it can be a sub-expression that first scans the data in table T1 and then redistributes (Redistribute) the data in the a-th column of table T1, and so on. In this way, both the basic sub-expression and the equivalent sub-expression can query the data in the a-th column of table T1.

In some embodiments, in order to facilitate obtaining an equivalent sub-expression, the basic sub-expression is expanded to obtain an equivalent sub-expression of the basic sub-expression, including:

Obtaining a preset expression set, the preset expression set including multiple preset expressions, and the multiple preset expressions have equivalent expression meanings;

Based on the base sub-expression, an equivalent sub-expression is determined from a preset set of expressions.

The preset expression set is a set of multiple preset expressions with equivalent expression meanings.

For example, a preset expression set records multiple preset expressions with different query logics, but each preset expression in the preset expression set has the same expression meaning and can obtain the same query result. The preset expression set makes it easy to determine equivalent sub-expressions of basic sub-expressions.

140. Determine the sub-expression to be constructed from the basic sub-expression and the equivalent sub-expression of the basic sub-expression.

The sub-expression to be constructed is a sub-expression whose query cost satisfies a preset condition among the basic sub-expression and the equivalent sub-expression of the basic sub-expression. For example, the sub-expression to be constructed can be a sub-expression with the smallest query cost among the basic sub-expression and the equivalent sub-expression, or any sub-expression whose query cost satisfies the preset condition, and so on.

In some embodiments, considering that a sub-expression may indicate a table join, which is used to combine data between two or more tables so as to query information across multiple tables, in order to obtain a sub-expression with a small query cost from a basic sub-expression and an equivalent sub-expression, a sub-expression to be constructed is determined from the basic sub-expression and the equivalent sub-expression of the basic sub-expression, including:

Get a first condition number and a second condition number, the first condition number is the condition number of the data connection condition associated with the base sub-expression, and the second condition number is the condition number of the data connection condition associated with the equivalent sub-expression of the base sub-expression;

According to the first conditional quantity and the second conditional quantity, a sub-expression to be constructed is determined from a basic sub-expression and an equivalent sub-expression of the basic sub-expression.

The data connection condition is used to specify the connection relationship between at least two tables so as to combine the data in at least two tables, wherein the data connection condition can be combined using comparison operators (＝, <, >, <＝, >＝, etc.) and logical operators (AND, OR, etc.).

It is understandable that the number of data connection conditions can be multiple. For example, suppose there are two tables, employee data table and department data table, which contain employee and department information respectively. The employee table and the department table are connected using inner join, and two data connection conditions are used: employee department ID (employee.department_id) = department ID (department.id), and employee city (employee.city) = department city (department.city). In this way, it can be ensured that only employee and department records in the same city and in the same department are returned, and the required employee name, department name and address information are obtained.

The first conditional number represents conditional data of a connection relationship between at least two tables as a base sub-expression.

The second condition number is condition data of a connection relationship between at least two tables represented by an equivalent sub-expression.

For example, Inner Join usually refers to an operation to connect two tables based on a comparison operator (such as the equal sign "="), that is, judging whether the corresponding rows need to be connected based on whether the corresponding column values in the two tables are equal. Therefore, in the Inner Join operation, there is usually only one or a few data connection conditions.

Inner Hash Join is based on hash tables to connect tables, and has excellent performance when processing large-scale data. The main steps are to store the data in the two tables to be connected in the hash table in memory respectively, and use the hash function to map them to the corresponding hash buckets. Then, for the data in each hash bucket, the connection is performed by comparing the connection conditions between them.

Since Inner Hash Join needs to build additional data structures such as hash tables, it usually requires more memory and computing resources in actual situations.

To determine the sub-expression to be constructed from the basic sub-expression and the equivalent sub-expression of the basic sub-expression according to the first conditional quantity and the second conditional quantity:

When the amount of data in the table is small, a smaller number of conditions can be screened out from the first number of conditions and the second number of conditions. If the smaller number of conditions is the first number of conditions, the basic sub-expression is used as the sub-expression to be constructed. If the smaller number of conditions is the second number of conditions, the equivalent sub-expression corresponding to the second number of conditions is used as the sub-expression to be constructed.

When a large amount of data is recorded in the table, the condition quantity with a large number can be screened out from the first condition quantity and the second condition quantity. If the condition quantity with a large number is the first condition quantity, the basic sub-expression is used as the sub-expression to be constructed. If the larger number of conditions is the second number of conditions, the equivalent sub-expression corresponding to the second number of conditions is used as the sub-expression to be constructed. In this way, data query can be executed efficiently, so as to provide better performance when processing large-scale data.

In some embodiments, considering that multiple expressions may be used to query the same data set, each expression has a different query logic, and thus each expression has a different query cost, in order to obtain a sub-expression with a small query cost from a basic sub-expression and an equivalent sub-expression of the basic sub-expression, determining a sub-expression to be constructed from the basic sub-expression and the equivalent sub-expression of the basic sub-expression includes:

Obtain metadata information of the data to be queried associated with the basic sub-expression;

Obtain the data volume of the data to be queried from the metadata information;

According to the amount of data, the sub-expression to be constructed is determined from the basic sub-expression and the equivalent sub-expression of the basic sub-expression.

The data to be queried is the data in the table that the basic sub-expression needs to query. For example, the basic sub-expression is used to query table A, and the data to be queried is the data in table A.

Metadata information is data information that describes the data to be queried. For example, metadata information is used to describe the data to be queried, including information such as the format, size, encoding method, and access protocol of the data.

The data volume is the size of the data to be queried that the user needs to query, usually expressed in the form of number of records or file size, etc. For example, the size of the data to be queried can be obtained through metadata information.

For example, the basic sub-expression is Get(T1), and the equivalent sub-expression of the basic sub-expression is Scan(T1). Get(T1) and Scan(T1) are used to query the data to be queried in table T1.

Scenarios where the basic sub-expression Get(T1) is used include:

1. You need to get a single record with a specified row key;

2. Only a small amount of data records need to be queried, such as obtaining basic information of a user;

3. You need to query multiple versions of a row.

Scenarios where the equivalent subexpression Scan(T1) is used include:

1. Need to query a large number of data records, such as batch statistical analysis data;

2. Need to filter or sort according to specific conditions, such as time, value, etc.;

3. It is necessary to traverse the entire table or multiple rows of data in a specified range, such as data backup and archiving.

If the amount of data is small, the basic sub-expression Get(T1) can be used as the sub-expression to be constructed. If the amount of data is large, the equivalent sub-expression Scan(T1) can be used as the sub-expression to be constructed.

150. Build a data query plan through the sub-expressions to be built and the associated relationships.

The data query plan is an executable data query solution constructed based on the association relationship of the sub-expressions to be constructed.

For example, if there are three basic sub-expressions in the data query expression, basic sub-expression 1, basic sub-expression 2 and basic sub-expression 3, the association relationship between basic sub-expression 1 and basic sub-expression 2 can be a parallel relationship, and the association relationship between basic sub-expression 1 and basic sub-expression 2 and basic sub-expression 3 respectively can be an AND relationship.

The subexpression to be constructed can be a basic subexpression or an equivalent subexpression of a basic subexpression, that is, there is a subexpression to be constructed 1 determined from basic subexpression 1 and the equivalent subexpression 1 of basic subexpression 1, a subexpression to be constructed 2 determined from basic subexpression 2 and the equivalent subexpression 2 of basic subexpression 2, and a subexpression to be constructed 3 determined from basic subexpression 2 and the equivalent subexpression 2 of basic subexpression 2. The sub-expressions 3 to be constructed determined in the sub-expressions 3 and the equivalent sub-expressions 3 of the basic sub-expressions 3, each sub-expression to be constructed has an association relationship with other sub-expressions to be constructed.

According to the equivalent sub-expression and the basic sub-expression having equivalent expression meanings, the association relationship between equivalent sub-expression 1 and basic sub-expression 2 or equivalent sub-expression 2 is a parallel relationship; the association relationship between equivalent sub-expression 1 and basic sub-expression 3 or equivalent sub-expression 3 can be an AND relationship; the association relationship between equivalent sub-expression 2 and basic sub-expression 3 or equivalent sub-expression 3 can be an AND relationship.

If the association relationship between sub-expression 1 to be constructed and sub-expression 2 to be constructed is a parallel relationship, and the association relationship between sub-expression 1 to be constructed and sub-expression 2 to be constructed and sub-expression 3 to be constructed is an AND relationship respectively, then based on the association relationship between sub-expression 1 to be constructed and sub-expression 2 to be constructed being a parallel relationship, control sub-expression 1 to be constructed and sub-expression 2 to be constructed to construct query methods independently, and then based on the association relationship between sub-expression 1 to be constructed and sub-expression 2 to be constructed and sub-expression 3 to be constructed being an AND relationship respectively, construct a query plan for sub-expression 3 to be constructed to filter the query results of sub-expression 1 to be constructed and sub-expression 2 to be constructed.

In some embodiments, considering that a sub-expression can construct multiple data query plans with the same query function, in order to filter a data query plan with a low query cost from multiple data query plans, a data query plan is constructed by using the sub-expression to be constructed and the association relationship, including:

Obtain the current query optimization request associated with the sub-expression to be constructed from the data query expression;

Construct an initial subquery plan through the sub-expression to be constructed;

Determine the subquery plan to be processed from the initial subquery plan according to the current query optimization request and the association relationship associated with the subexpression to be constructed;

The sub-query plans to be processed are merged to obtain a data query plan.

Among them, the current query optimization request is a query optimization request constructed by the query conditions associated with the sub-expression to be constructed in the data query expression. The query conditions can be the data distribution, index type, etc. associated with the sub-expression to be constructed in the data query expression. The current query optimization request can be used to optimize the data query plan constructed by the sub-expression to be constructed.

For example, the current query optimization request may be a request for quickly acquiring data to be queried, a request for reducing query costs, a request for satisfying query speed and query costs within a certain range, and so on.

The initial subquery plan may be multiple unoptimized subquery plans constructed for the sub-expression to be constructed. For example, the multiple unoptimized subquery plans include subquery plans with high query costs, subquery plans with low query costs, subquery plans with slow query speeds, subquery plans with fast query speeds, and so on.

The subquery plan to be processed may be an initial subquery plan obtained from the initial subquery plan based on the association relationship and satisfying the current query optimization request.

For example, based on the data query expression, sub-expression 1 to be constructed, sub-expression 2 to be constructed and sub-expression 3 to be constructed are determined, sub-expression 1 to be constructed is used to query table T1, sub-expression 2 to be constructed is used to query the data in column b of table T2, sub-expression 3 to be constructed is used to filter data from the data obtained from sub-expression 1 to be constructed and sub-expression 2 to be constructed, and the data query expression also includes sorting according to column a of table T1. The association relationship between sub-expression 1 to be constructed and sub-expression 2 to be constructed is a parallel relationship between the basic sub-expressions, and the association relationship between sub-expression 1 to be constructed and sub-expression 2 to be constructed and sub-expression 3 to be constructed is an AND relationship.

The current query optimization request associated with sub-expression 1 to be constructed can be used to enable sub-expression 1 to be constructed to quickly obtain query table T1, and to sort the data queried from table T1 according to the ath column of the table. The current query optimization request associated with sub-expression 2 to be constructed can be used to enable sub-expression 2 to be constructed to quickly obtain the bth column data in query table T2. The current query optimization request associated with sub-expression 3 to be constructed can be used to merge the sorted (sorted according to the ath column of table T1) filtered data collected by sub-expression 3 to be constructed from all database nodes.

The initial subquery plan constructed by the sub-expression 1 to be constructed may include Get(T1), Scan(T1), Scan(T1)+Sort(T1.a), and the initial subquery plan constructed by the sub-expression 2 to be constructed may include Scan(T2)+Redistribute(T2.b), Get(T2.b).

According to the current query optimization request and the association relationship associated with the sub-expression to be constructed, the sub-query plan to be processed can be determined from the initial sub-query plan:

Due to the parallel relationship, the initial subquery plans constructed by subexpression 1 to be constructed and subexpression 2 to be constructed can be independently filtered. Through the current query optimization request associated with subexpression 1 to be constructed, Scan(T1) can be filtered out from the initial subquery plan Get(T1), Scan(T1), Scan(T1)+Sort(T1.a), and through the current query optimization request associated with subexpression 2 to be constructed, Scan(T2)+Redistribute(T2.b) can be filtered out from the initial subquery plan Scan(T2)+Redistribute(T2.b), Get(T2.b).

According to the AND relationship between sub-expression 1 to be constructed and sub-expression 2 to be constructed and sub-expression 3 to be constructed, since sub-expression 1 to be constructed does not sort the data in table T1 according to the a-th column of the table, the initial sub-plan for constructing sub-expression 3 to be constructed can be to sort the data after filtering by Inner Hash Join (T1.a＝T1.b) according to the a-th column of table T1, so as to realize the data distribution in the data query expression.

That is, the initial subquery plan constructed by the subexpression 3 to be constructed can be Inner Hash Join(T1.a＝T1.b)+Sort(T1.a)+GatherMerge(T1.a), Inner Hash Join(T1.a＝T1.b)+Gather(T1.a)+Sort(T1.a), and so on, where the GatherMerge operator is used to collect sorted data from all segments of the database to the master node, and Gather is used to collect data from all segments to the master node. Through the current query optimization request associated with the subexpression 3 to be constructed, one can be selected from Inner Hash Join(T1.a＝T1.b)+Sort(T1.a)+GatherMerge(T1.a), Inner Hash Join(T1.a＝T1.b)+Gather(T1.a)+Sort(T1.a), etc. as the subquery plan to be processed.

Merge the pending subquery plan Scan(T1) constructed by the pending subexpression 1, the pending subquery plan Scan(T2)+Redistribute(T2.b) constructed by the pending subexpression 2, and the pending subquery plan Inner Hash Join(T1.a＝T1.b)+Sort(T1.a)+GatherMerge(T1.a) constructed by the pending subexpression 3 to obtain the data query plan.

In some embodiments, in order to enable the expression to construct a data query plan, an initial sub-query plan is constructed by using the sub-expression to be constructed, including:

Get the query operator associated with the sub-expression to be constructed;

According to the sub-expressions to be constructed, the connection relationship between the query operators is constructed;

The query operators are controlled to be arranged according to the connection relationship to obtain the initial subquery plan.

The query operator may support obtaining required data from the database, for example, the query operator may be Get, Scan, Sort, GatherMerge, Redistribute, Replicate, and the like.

The connection relationship is used to restrict the order in which query operators are executed. For example, sub-expression 2 to be constructed is used to obtain the data in column b of query table T2, that is, multiple query operators (Get, Scan, and Redistribute) associated with sub-expression 2 to be constructed, and the connection relationship between the query operators constructed by sub-expression 2 to be constructed can be Scan first and then Redistribute, or only Get. In this way, the query operators are arranged according to the connection relationship, and the initial sub-query plan Scan(T2)+Redistribute(T2.b), Get(T2.b), or Get(T2.b) can be obtained.

In some embodiments, in order to calculate the query cost of the data query plan constructed by the expression, according to the current query optimization request and the association relationship associated with the sub-expression to be constructed, the sub-query plan to be processed is determined from the initial sub-query plan, including:

According to the current query optimization request, metadata information of the to-be-queried data associated with the to-be-constructed sub-expression is obtained;

Analyze the amount of computation required when executing the initial subquery plan based on the metadata information of the data to be queried;

According to the computation amount and the association relationship, the subquery plan to be processed is determined from the initial subquery plan.

The metadata information is data information describing the data to be queried. For example, the metadata information records various attributes and characteristics of the data, such as data type, data source, data storage method, data access rights, etc. The various attributes of the data include the attributes and characteristics describing the table, such as table name, column name, column data type, column constraints, index information, data size, etc.

The computational amount is the computational operations and resource consumption required to execute the initial subquery plan, which can be understood as the query cost. For example, the computational amount may include the size or time of scanning data files, the size or time of reading indexes, and so on.

The association relationship between sub-expression 1 to be constructed and sub-expression 2 to be constructed is a parallel relationship between sub-expressions, and the association relationship between sub-expression 1 to be constructed and sub-expression 2 to be constructed and sub-expression 3 to be constructed is an AND relationship. According to the calculation amount and the association relationship, the sub-query plan determined from the initial sub-query plan can be:

Since the association relationship between sub-expression 1 to be constructed and sub-expression 2 to be constructed is a parallel relationship, the sub-query plan to be processed can be directly screened from the initial sub-query plan constructed by sub-expression 1 to be constructed and sub-expression 2 to be constructed based on the calculation amount.

Since the association relationship between sub-expression 1 to be constructed and sub-expression 2 to be constructed and sub-expression 3 to be constructed is an AND relationship, when constructing the initial subquery plan by filtering sub-expression 3 to be constructed through calculation amount, it is also necessary to consider sub-query plan 1 to be processed constructed by sub-expression 1 to be constructed and sub-query plan 2 to be processed constructed by sub-expression 2 to be constructed, so as to avoid overlapping plans between sub-query plan 3 to be processed constructed by sub-expression 3 to be constructed and sub-query plan 1 to be processed and sub-query plan 2 to be processed, or to avoid sub-query plan 1 to be processed, sub-query plan 2 to be processed and sub-query plan 3 to be processed from failing to fully execute all query conditions.

In some embodiments, in order to ensure that the data query plan constructed by each sub-expression does not have conflicting or repeated logic, the sub-expression to be constructed includes a first expression and a second expression, the first expression is used to query the data to be queried, and the second expression is used to filter the data to be queried, and according to the current query optimization request and the association relationship associated with the sub-expression to be constructed, the sub-query plan to be processed is determined from the initial sub-query plan, including:

Determine a first subquery plan from a first initial subquery plan according to a current query optimization request associated with the first expression, where the first initial subquery plan is an initial subquery plan constructed by using the first expression;

Based on the current query optimization request satisfied by the first subquery plan, determining unsatisfied requests from the current query optimization request associated with the first expression;

According to the unsatisfied requests and the association relationship, determine a final optimization request from the current query optimization requests associated with the second expression, where the final optimization request includes the unsatisfied request;

According to the final optimization request, a second subquery plan is determined from the second initial subquery plan, where the second initial subquery plan is an initial subquery plan constructed by using the second expression, and the subquery plans to be processed include the first subquery plan and the second subquery plan.

The first expression is used to query the data to be queried. For example, the first expression can be the above-mentioned sub-expression 1 to be constructed and the sub-expression 2 to be constructed, the sub-expression 1 to be constructed is used to query table T1, and the sub-expression 2 to be constructed is used to query the data in column b of table T2.

The current query optimization request associated with the first expression is used to filter the initial subquery plan required for retrieving the data to be queried.

The first initial sub-query plan is an initial sub-query plan constructed by the first expression, and is used to query the data to be queried, wherein the first initial sub-query plan constructed by the first expression may be one or more.

The first sub-query plan is a first initial sub-query that satisfies a condition of a query cost selected according to a current query optimization request associated with a first expression.

The current query optimization request satisfied by the first sub-query plan is a query condition that can be satisfied when the first sub-query plan is executed.

The unsatisfied request is a query optimization request that is not fulfilled by the first subquery plan in the current query optimization request associated with the first expression.

For example, when the first expression is sub-expression 1 to be constructed, it can be seen from the above example that the current query optimization request associated with sub-expression 1 to be constructed can be used to enable sub-expression 1 to be constructed to quickly obtain query table T1, and sort the data queried from table T1 according to the a-th column of the table. If the first sub-query plan constructed by the first expression according to the current query optimization request associated with the first expression is Scan(T1), then the current query optimization request satisfied by the first sub-query plan only includes obtaining query table T1, and the unsatisfied request is to sort the data queried from table T1 according to the a-th column of the table.

The final optimization request is a current query optimization request associated with the second expression obtained based on the unsatisfied request, so that all query conditions in the data query expression can be realized through the query optimization request satisfied by the final optimization request and the first sub-query plan.

For example, as can be seen from the above example, when the second expression is the sub-expression 3 to be constructed, the current query optimization request associated with the sub-expression 3 to be constructed can be used to merge the sorted (sorted according to the ath column of table T1) filtered data collected by the sub-expression 3 to be constructed from all database nodes. Since the first sub-query plan constructed by the sub-expression 1 to be constructed can only obtain the query table T1, the data queried from table T1 is not sorted according to the ath column of the table (the request is not met), thus, the current query optimization request associated with the second expression needs to include the unsatisfied request to realize the query condition in the data query expression.

The second initial subquery plan is an initial subquery plan constructed by using the second expression.

The second sub-query plan is a second initial sub-query that meets the condition based on the query cost selected by the final optimization request.

In some embodiments, in order to shorten the time of constructing a data query plan, the method provided in the embodiment of the present application further includes:

Obtain the historical query optimization request associated with the sub-expression to be constructed, and the historical query plan corresponding to the historical query optimization request;

Determine, from the historical query optimization requests, a pending historical request corresponding to the current query optimization request;

If there are pending historical requests, the historical query plan corresponding to the pending historical request is used as the pending sub-query plan;

If there is no pending historical request, the step of constructing an initial subquery plan through the sub-expression to be constructed is triggered.

The historical query optimization request is a query optimization request that has been used by the sub-expression to be constructed in a historical time period.

The historical query plan is a query plan constructed by the sub-expression to be constructed according to the historical query optimization request.

The pending historical request is a historical query optimization request that is the same as the current query optimization request.

For example, when the same sub-expression to be constructed is constructing a sub-query plan to be processed, it may face multiple different query optimization requests in different query time periods. If the current query optimization request is the same as the historical query optimization request, that is, there is a historical request to be processed, then the historical query plan corresponding to the historical request to be processed is used as the sub-query plan to be processed. If there is no historical request to be processed, the initial sub-query plan is directly constructed through the sub-expression to be constructed, and then the initial sub-query plan is filtered through the current query optimization request to obtain the sub-query plan to be processed.

In some embodiments, in order to facilitate obtaining the historical query optimization request associated with the sub-expression to be constructed, and the historical query plan corresponding to the historical query optimization request, the following steps are included:

Obtain a hash table associated with the sub-expression to be constructed, the hash table including historical query optimization requests, and first tags and plan retrieval links associated with the historical query optimization requests, the first tag being a hash value obtained after the historical query optimization request is calculated by a hash function;

Determine the pending historical request corresponding to the current query optimization request from the historical query optimization requests, including:

Determine a second tag, where the second tag is a hash value obtained after the current query optimization request is calculated by a hash function;

Determine, from the historical query optimization requests, a to-be-processed historical request corresponding to the current query optimization request according to the first tag and the second tag;

If there are pending historical requests, the historical query plan corresponding to the pending historical request is used as the pending sub-query plan, including:

If there are pending historical requests, the plan retrieval link associated with the pending historical request is used to retrieve the historical query plan corresponding to the historical query optimization request as the pending sub-query plan.

Among them, the hash table can map the hash value with the historical query optimization request, and the historical query optimization request carries the historical query plan, so as to quickly filter the pending historical requests that are the same as the current query optimization request from the historical query optimization request, so that the historical query plan corresponding to the pending historical request can be obtained.

The first tag is the hash value obtained after the historical query optimization request is calculated by the hash function.

The plan retrieval link is a link for retrieving the query optimization plan used by the pending historical request.

The second tag is the hash value obtained after the current query optimization request is calculated using the same hash function.

For example, since the first label is the hash value obtained by calculating the historical query optimization request through the hash function, and the second label is the hash value obtained by calculating the current query optimization request through the hash function, it is convenient to quickly find the label with the same hash value as the second label from the first label through the second label, so as to quickly treat the historical query optimization request of the label as the same historical request to be processed as the current query optimization request, avoiding traversing the historical query optimization requests.

In some embodiments, considering that data may be distributed and stored in multiple nodes, in order to quickly obtain query data, after constructing a data query plan through the sub-expressions to be constructed and the association relationships, the following is further included:

Determine the query instruction corresponding to the data query plan, and the copy instruction of the query instruction;

The replica instruction is sent to multiple data storage nodes of the database, so that the data storage nodes return the query data according to the replica instruction.

The query instruction is a representation of a data query plan so that the database can return query data through the query instruction. The copy instruction is an instruction obtained by copying the query instruction.

The data storage node is a node for storing data in the database, and each data storage node returns data according to the received replica instruction.

For example, as can be seen from the above, sub-expression 3 to be constructed is used to filter data from the data obtained from sub-expression 1 to be constructed and sub-expression 2 to be constructed. In this way, after receiving the data returned by each data storage node according to the replica instruction, the returned data is filtered according to sub-expression 3 to be constructed.

It is understandable that the data queried through the data query plan can be applied to fields such as data analysis, such as advertising content query, financial risk control analysis, information recommendation, etc.

In some embodiments, when querying for advertising content, advertisements can be quickly queried through a data query plan, and the advertisements can be displayed in the content played by the application.

For example, it is necessary to display a specific type of advertisement in the content played by an application. Considering that the query optimizer provided by the relevant technology adopts multi-stage query optimization, each optimization stage is optimized independently, which may lead to conflicts in the optimization strategies of each stage, resulting in high uncertainty and error rate in the query process. Therefore, when querying the target type of advertisements through the query optimizer provided by the relevant technology, its query efficiency is low.

In order to improve the efficiency of querying specific types of advertisements, the current sub-expression in the data query expression can be expanded through an equivalent sub-expression, and the sub-expressions to be constructed with high query efficiency can be filtered out from the equivalent sub-expressions and the current sub-expression, and multiple sub-expressions to be constructed can be used to construct a data query plan based on the association relationship between the current sub-expressions, thereby avoiding conflicting or overlapping steps in the data query plan, further speeding up the query efficiency of specific types of advertisements. In this way, specific types of advertisements can be quickly displayed in the content played by the application.

In some embodiments, data query plan construction can also be applied to the field of financial risk control analysis.

For example, when a user analyzes a financial product, financial risk control analysis can quickly query data related to the financial product through the data query plan constructed by this application, and recommend the data related to the financial product to the user.

In some embodiments, data query plan construction can also be applied to the field of information recommendation.

For example, when a user searches for information, the data query plan constructed by this application can quickly search for information and recommend the searched information to the user.

As can be seen from the above, the embodiment of the present application can obtain a data query expression, which includes keywords. At the same time, the data query expression can have multiple basic sub-expressions, and there is an association relationship between each basic sub-expression. Therefore, the basic sub-expressions in the data query expression and the association relationship between the basic sub-expressions can be obtained according to the keywords. Then the basic sub-expression is expanded to obtain the equivalent sub-expression of the basic sub-expression. Since the equivalent sub-expression and the basic sub-expression have equivalent expression meanings, the equivalent sub-expression has an association relationship with the basic sub-expression. The sub-expression to be constructed can be screened out from the basic sub-expression and the equivalent sub-expression. The sub-expression to be constructed can be It is a basic sub-expression or an equivalent sub-expression. Then, a data query plan is constructed through the sub-expression to be constructed and the associated relationship. Multiple sub-expressions to be constructed can construct a data query plan based on the associated relationship, so that multiple sub-expressions to be constructed do not conflict with each other when constructing a data query plan. In this way, it is easy to construct a data query plan and improve query performance.

The method described in the above embodiment will be further described in detail below.

In this embodiment, the query system is taken as an example to describe the method of the embodiment of the present application in detail.

As shown in FIG. 2a , a query system includes a database management system and a query optimizer. The query optimizer is located outside the database management system. The query optimizer is used to construct a data query plan. The database management system specifically includes:

A parser, used for performing syntax analysis on the received data query command to obtain a syntax tree of the data query command;

A query converter is used to receive the syntax tree sent by the parser, and use a preset data exchange language to perform format conversion processing on the syntax tree to obtain a data query expression, so that the query optimizer can construct a data query plan based on the data query expression;

The metadata provider is used to receive a request for obtaining metadata information sent by the query optimizer when the query optimizer constructs a data query plan;

The metadata provider is an interface for exchanging metadata information between the query optimizer and the database management system. The metadata provider component is a plug-in specific to the database management system and is used to retrieve metadata information from the database management system.

The directory is used to receive the acquisition request sent by the metadata provider, and send the acquisition request to the metadata exchanger through the metadata provider, so that the acquisition request can obtain metadata information from the directory through the metadata exchanger, and send the metadata information to the parser;

The directory is a database in a database management system that stores metadata information, and the metadata information is data information that describes the data in the database.

Metadata exchanger: Use the preset data exchange language (EL language) to exchange metadata information. EL language is a data exchange language in XML format, which contains metadata information in the database management system, such as table definition, operator definition, etc. Through the acquisition request of EL language, it can interact with different back-end database management systems and obtain the required metadata information. At the same time, the metadata exchanger can also extract metadata information from the database system and convert it into EL language to achieve decoupling of the query optimizer and the database management system;

EL language: refers to a framework for exchanging information between the query optimizer and the database management system. Separating the query optimizer from the database management system requires building a communication mechanism for processing queries. The framework uses an XML-based language to encode the necessary information required for communication, such as input queries, output plans, and metadata. A simple communication protocol is used on the data exchange language framework to obtain metadata information and send data query plans;

A plan converter, used to convert a data query plan into a query instruction;

The executor is used to receive the query instructions sent by the plan converter and return the query data according to the query instructions.

The executor performs: dispatching a copy of the final query instruction (replica instruction) to each segment (data storage node). During distributed query execution, the assigned executor on each segment acts as both the sender and receiver of data. Receiver. For example, the Redistribute(T2.b) instance running on segment S sends tuples on S to other segments based on the hash value of T2.b, and also receives tuples from other Redistribute(T2.b) instances on other segments.

Here, Redistribute(T2.b) indicates the b-th column of table T2.

The input of the query optimizer in Figure 2a is a data query expression (EL query). The output of the query optimizer is a data query plan (EL plan). When the query optimizer builds a data query plan, it can query the database management system to obtain metadata information (such as table definitions). The query optimizer abstracts metadata access details by allowing the database management system to register a metadata provider, which is responsible for serializing the metadata information into an EL query and sending it to the query optimizer. Metadata information can also be queried from a file containing metadata objects serialized in the EL language format. The database management system needs to include a converter that consumes/emit data in EL format. The query converter converts the syntax tree into a data query expression, while the plan converter converts the data query plan into an executable plan. The implementation of this plan converter is completed entirely outside the optimizer, which allows multiple database management systems to use the query optimizer by providing appropriate plan converters.

By locating the query optimizer outside the database management system, the query optimizer architecture can be highly extensible, and all components in the query optimizer and the database management system can be replaced and configured separately.

As shown in Figure 2b, the query optimizer specifically includes:

Memo, used to encode the data query plan constructed by the query optimizer in a compact memory data structure and store it in the memo;

The structure of the memo consists of a set of containers called expression groups, where each expression group contains logically equivalent expressions. The memo expression group captures different subquery expressions of a data query expression (such as a filter on a table or a table join of two tables), and the group members of the expression group (base subexpressions and equivalent subexpressions of the base subexpressions) implement the same data query in different logical ways (such as different join orders). The base subexpressions or equivalent subexpressions of the base subexpressions within each expression group are constructed through query operators. The recursive structure of the memo allows compact encoding of huge data to build a data query plan space.

A searcher, used for using a search mechanism to construct an initial subquery plan through the sub-expressions to be constructed in the expression group, and estimating a subquery plan to be processed with the minimum query cost from the initial subquery plan;

A task scheduler is used to start the search mechanism, through which parallel work units are created to perform the three main steps of query optimization: exploration (generating equivalent logical expressions), realization (building initial subquery plans), and optimization (optimizing initial subquery plans), where the required physical properties (such as sort order) are enforced;

Optimization tools are used to provide search mechanisms for search engines. The optimization tools include:

A transformation component, used for determining an equivalent sub-expression, wherein the equivalent sub-expression and the basic sub-expression have equivalent expression meanings;

For example, plan alternatives are generated by applying transformation rules, which can produce equivalent logical expressions (e.g. Inner Join(T1, T2) → Inner Join(T2, T1)) or physical realizations of existing sub-expressions (e.g. Join(T1, T2) → Hash Join(T2, T1)). The results of applying transformation rules are copied to the memo, which may result in the creation of new expression groups and/or the addition of new expressions to existing expression groups. Each transformation rule is an independent component that can be explicitly activated and deactivated in the system configuration.

Among them, Join is used to combine the rows in two or more tables according to specific conditions to generate a larger new table that contains all the information in the source table. Hash Join is to join the two tables through hash. The (Hash) function is mapped to several buckets respectively, and then a Join operation is performed on each bucket, and finally the Join results are merged.

Enforcement attributes, used to insert the physical attributes required for enforcement into the initial subquery plan;

It can be understood that the query optimizer includes an extensible framework for describing query requirements and plan characteristics based on formal attribute specifications. Attributes are of different types, including logical attributes (such as output columns), physical attributes (such as sort order and data distribution), and scalar attributes (such as columns used in input conditions). When the query optimizer builds a data query plan, each query operator may request specific attributes from the sub-expression to be built. The optimized initial sub-query plan (sub-query plan) may automatically meet the current query optimization request associated with the sub-expression to be built (the required attributes (such as the index scan (IndexScan) plan provides sorted data)), but many times it is necessary to insert some attributes into the initial sub-query plan through enforcement attributes (for example, the initial sub-query plan contains a Sort sorting operation, and the Sort operator needs to be inserted into the plan). The framework allows each query operator to control the settings of the enforcer based on the attributes of the sub-query plan and the local behavior of the query operator, so as to ensure that the query data meets specific requirements, such as the order of output columns, the distribution of data, etc.

Metadata cache, used to cache metadata information of data to be queried;

The query cost estimation is used to estimate the query cost of the initial sub-query plan constructed by the searcher through the metadata information in the metadata cache, and obtain the query cost of the initial sub-query plan.

For example, since metadata (such as table definitions) rarely changes, sending metadata information for each query will incur a lot of overhead. Therefore, the query system caches metadata information on the query optimizer side and only queries metadata information from the directory when the metadata information is not found in the cache or has changed.

For example, the query language input by the user is converted into Structured Query Language (SQL) (i.e., data query command), where SQL is a standard computer language for processing relational databases. The data query command may be (SELECT T1.a FROM T1, T2 WHERE T1.a = T2.b ORDER BY T1.a), where the distribution of table T1 is grouped according to the hash value of the a-th column in table T1, which can be expressed as Hashed(T1.a), and the distribution of table T2 is grouped according to the hash value of the a-th column in table T1, which can be expressed as Hashed(T2.a). The following example illustrates how the query optimizer constructs a data query plan:

(i) Using the preset data exchange language, the data query command is converted into a data query expression in EL format.

For example, the data query command is converted into EL format, which contains the required output columns, sorting columns, data distribution, and query logic in XML format. Metadata information (such as table and operator definitions) is decorated as a metadata identifier (ID) to allow the query optimizer to request more information when building a data query plan. The metadata ID is a unique identifier consisting of a database system identifier, an object identifier, and a version number. The data query expression in EL format is sent to the query optimizer, parsed and converted into an in-memory logical expression tree, and then copied to the memo.

(ii) According to the keywords, basic sub-expressions in the data query expression and the association relationship between the basic sub-expressions are obtained.

For example, as shown in Figure 2c, through the data query command (SELECT T1.a FROM T1,T2 WHERE T1.a＝T2.b ORDER BY T1.a), the keyword retrieval table T1, retrieval table T2 and inner join (T1.a＝T2.b)) in the data query expression can be obtained, and three basic sub-expressions are created through the retrieval table T1, retrieval table T2 and Inner Join (T1.a＝T2.b).

For the sake of brevity, the join conditions are omitted. Expression group 0 (Group0) is called the root group because it corresponds to the root of the logical plan. The dependencies between operators in the logical plan are captured as references between expression groups, that is, the basic sub-expressions corresponding to Inner Join belong to the root group, and the basic sub-expressions corresponding to tables T1 and T2 belong to sub-groups. For example, Inner Join [T1, T2] means that expression group 1 (Group1) and expression group 2 (Group2) are connected as sub-groups. Query optimization is performed in the following steps:

(1) Exploration: triggering the transformation rules that generate other equivalent logical plan expressions (i.e., determining equivalent sub-expressions, which have equivalent expression meanings to the corresponding basic sub-expressions).

For example, Inner Join[T1,T2] is generated as Inner Join[T2,T1]. The newly generated equivalent expression is added to the existing root group and may create a new expression group. The memo structure has a built-in duplicate detection mechanism based on the expression topology to detect and eliminate any duplicate expressions created by different transformations.

(2) Derived Statistics: At the end of the exploration, the memo maintains the complete logical space of a given query. The derived statistics mechanism of the query optimizer is then triggered to calculate the statistical objects of the memo group. The statistical objects are mainly used to calculate the column histograms of cardinality and data skewness (i.e., metadata information of the data to be queried). Derived statistics are performed on a compact memo structure to avoid expanding the search space. The system selects the expression with the highest reliable statistics to calculate the derived statistical data for the group of sub-expressions to be constructed, and the statistical calculation is based on the sub-expressions to be constructed.

For example, an Inner Join expression (base subexpression) with a few join conditions (first number of conditions) has a smaller query cost than another equivalent Inner Join expression (equivalent subexpression) with more join conditions (second number of conditions) (this may occur when generating multiple join orders). Because the greater the number of join conditions, the higher the propagation and amplification errors may be. Calculating the confidence score of the cardinality is very challenging because the confidence scores need to be aggregated on all nodes of a given subexpression to be constructed. After selecting the subexpression to be constructed with the smallest query cost in the root group, the system recursively triggers derived statistics on the subgroups of the subexpression to be constructed. Finally, by combining the statistical objects of the subgroups, the statistical object of the root group (metadata information of the data to be queried) is constructed.

As shown in Figure 2d, the derived statistics mechanism of the running example is shown. First, a top-down traversal is performed, where the sub-expression of the root group requests statistics (metadata information of column a in table T1 (T1.a) and metadata information of column b in table T2 (T2.b)) from the sub-expressions of its sub-groups (Group1 and Group2). For example, Inner Join(T1.a＝T2.b) requests the histograms of T1.a and T2.b. The requested histograms are loaded on demand from the catalog through the metadata provider, parsed into EL queries (data query expressions) and stored in the metadata cache to serve future requests. Next, a bottom-up traversal is performed to combine the sub-statistics objects into the statistics object of the root group (metadata information of the data to be queried). Because the join condition may affect the column histogram, this combines the (possibly modified) histograms of T1.a and T2.b. The constructed statistics objects are attached to a single group and can be incrementally updated during optimization (for example, by adding new histograms). This is crucial to keep the cost of derived statistics manageable.

(3) Implementation: Triggering transformation rules that create the physical implementation of a logical expression. For example, triggering a scan rule to generate a logical (Get) into a physical table scan (Scan).

(4) Optimization: In this step, attribute enforcement and query cost estimation schemes are performed. Optimization begins by submitting a query optimization request to the root expression group of the memo, specifying query requirements such as result distribution and sort order. Submitting a request to the root expression group is equivalent to requesting the minimum query cost plan that satisfies the request among the physical operators in the root expression group.

For each query optimization request, the sub-expressions in the root expression group pass the corresponding request to the sub-expressions in the sub-group according to the incoming query optimization request and the local requirements of the operator. Multiple identical query optimization requests may be submitted by the same group during optimization. The system caches query optimization requests in a hash table. Incoming requests are only calculated if there is no query optimization request in the hash table (to build a sub-query plan through the query optimization request). In addition, each sub-expression maintains a local hash table to map query optimization requests to corresponding historical query requests. The local hash table provides a plan call link used when extracting historical query plans from the memo.

As shown in Figure 2e, the query optimization request in the running example is shown. The current query optimization request associated with the first expression specifies that the query results need to be collected on the master node based on the order of column a of table T1. Figure 2e also includes the group hash table corresponding to the best group expression, as well as the Sort, Redistribute, and Replicate operators inserted in the memo. The Gather operator collects data from all data storage nodes to the master node. The GatherMerge operator collects sorted data from all data storage nodes to the master node and maintains the sorted order. The Redistribute operator distributes tuples to segments based on the hash value of the given parameter.

FIG2e includes the query operators in Group0 (Nested Loop Join [Group1, Group2] (Inner Nest Loop Join [Group1, Group2]), Inner Nest Loop Join [Group2, Group1], Inner Hash Join [Group1, Group2] and Inner Hash Join [Group2, Group1, (Sort (T1.a), Gather, GatherMerge (T1.a)), the query operators in Group1 (Scan (T1), Sort (T1.a), Replicate), the query operators in Group2 (Scan (T2), Replicate, Redistribute (T2.b)), and according to the query optimization request, the data query plan on the right is constructed by the query operators in Group1, Group2 and Group0. As shown in FIG2f, the optimization of Inner Hash Join [T1, T2] is shown. For this query optimization request, one of the alternatives is to align the distribution of child nodes according to the join condition, so that the nodes to be connected The data can be co-located. This is achieved by requesting a Hashed(T1.a) distribution from the first expression 1 and a Hashed(T2.b) distribution from the first expression 2. Both first expressions require any type of sort order to be provided. After finding the best plan for the subgroups (Group1 and Group2) (first subquery plan), the Inner Hash Join combines the sub-attributes to determine the provided distribution and sort order (final optimization request). Among them, the best plan for the first expression 2 requires hash distribution of T2 on T2.b, because T2 was originally hashed on T2.a. , while the best plan for the first expression 1 is a simple scan because T1 is already hash distributed on T1.a. When it is determined that the provided attributes do not satisfy the original requirements (unsatisfied request), the unsatisfied attributes must be enforced (final optimization request). The attribute enforcement in the system adopts a flexible framework that allows each operator, based on subplans and operator local behavior, to define the behavior of enforcing the required attributes. For example, the order-preserving Nested Loops Join (NL Join) operator may not need to enforce a sort order on the join if the outer child nodes already provide a sort order.

Figure 2f shows two possible initial subquery plans of Group0 that satisfy the query optimization request through attribute execution. The data query plan on the left sorts the query data on the segment, and then collects and merges the sorted results on the main node. The data query plan on the right collects the query data from the segment to the main node and then sorts it. These different alternatives are encoded in the memo, and the cost model calculates the query cost (computational amount) of the initial subquery plan based on the metadata information of the data to be queried. Finally, based on the query structure given by the query optimizer, the data query plan is extracted from the memo. Figure 2f illustrates the plan extraction of the running example. The relevant expressions have corresponding local hash tables. Each local hash table maps the incoming optimization request to the corresponding sub-optimization request. First, in the root expression group to be constructed The sub-expression to find the best group expression is the GatherMerge operator. In the local hash table of GatherMerge, the best group expression to find is Sort. Therefore, the GatherMerge operator is linked to Sort. In the local hash table of Sort, the corresponding best group expression is Inner Hash Join [T1, T2]. Therefore, Sort is linked to Inner Hash Join. The same process is followed to complete the plan extraction, and the final data query plan is shown in Figure 2f. The extracted best data query plan is serialized in EL format and sent to the database management system for execution.

This application is mainly aimed at important business scenarios such as data analysis, such as large-scale real-time data and offline data query systems. The Internet generates a huge amount of data every day, which brings great challenges to the big data query system. The data management and query systems in the business have made great progress in scalability, availability and processing performance, so that large data sets of hundreds of trillions of bytes (TeraByte, TB) or even petabytes (PetaByte, PB) can be analyzed and queried more quickly through structured query language (Structured Query Language, SQL) or SQL-like interfaces. Internet data is usually stored in a distributed computing framework (Hadoop), and the query engine compiles the initial SQL query into Spark or MapReduce jobs, where Spark is a distributed computing framework that uses memory computing, and MapReduce is a distributed computing framework that uses disk read and write IO. Previous query optimizers usually use multi-stage query optimization, and the performance of compiled Spark or MapReduce jobs is poor. For this reason, this application designs a query plan construction method based on the waterfall tree model to accelerate the encoding of complex queries. After actual data testing, it can bring more than several times the query acceleration compared to previous query optimizers.

For SQL queries on big data systems, a common solution in the industry is to use Hive to convert queries into MapReduce tasks, where Hive is a data warehouse tool based on Hadoop, and Hadoop is a distributed computing framework, but this method may lead to poor interactive analysis performance. To solve this problem, the industry has developed a number of professional query engines. However, these methods are often only applicable to specific host systems and are optimized, but cannot support the query requirements of multiple distributed systems at the same time. Since the query optimizer is the main influencing factor in analyzing query processing performance, and there are a large number of complex query processing requirements for big data in the business, this application designs a new type of query optimizer for a distributed query architecture for big data, and the business can quickly develop new optimization technologies and advanced query functions based on this optimizer. This application accelerates queries by designing metadata cache components, memos, mandatory attribute optimization and other technologies, and uses an efficient multi-core scheduler to increase the optimization speed.

The present application designs a new query optimizer. A query optimizer based on a waterfall model optimization framework. The query optimizer of the previous data query system was tightly coupled with the entire database management system, but a unique feature of the present application is that the designed optimizer can be run as a separate component outside the database system. This capability makes it possible for products with different computing architectures to use the same optimizer together. In addition, the optimizer can be deployed and run as an independent product, and can be tested and optimized separately in detail without being coupled to the structure of the database system. This reduces the difficulty of deployment and testing.

From the above, it can be seen that this application facilitates the construction of data query plans and improves query efficiency.

In order to better implement the above method, the embodiment of the present application also provides a query plan construction device, which can be integrated in an electronic device, and the electronic device can be a terminal, a server, etc. Among them, the terminal can be a mobile phone, a tablet computer, a smart Bluetooth device, a laptop, a personal computer, etc.; the server can be a single server or a server cluster composed of multiple servers.

For example, in this embodiment, the method of the embodiment of the present application will be described in detail by taking the query plan construction device specifically integrated in an electronic device as an example.

For example, as shown in FIG. 3 , the query plan construction device may include an acquisition unit 310, a key unit 320, an equivalent unit 330, a determination unit 340, and a construction unit 350, as follows:

(i) Acquisition unit 310.

The acquisition unit 310 is used to acquire a data query expression, where the data query expression includes keywords.

In some embodiments, obtaining a data query expression includes:

Get data query command, the data query command includes keywords;

According to the keywords, the data query command is parsed and processed to obtain the syntax tree of the data query command;

The preset data exchange language is used to perform format conversion on the syntax tree to obtain a data query expression.

(ii) Key unit 320.

The key unit 320 is used to obtain the basic sub-expressions in the data query expression and the association relationship between the basic sub-expressions according to the keywords.

(iii) Equivalent unit 330.

The equivalent unit 330 is used to expand the basic sub-expression to obtain an equivalent sub-expression of the basic sub-expression. The basic sub-expression and the equivalent sub-expression of the basic sub-expression have equivalent expression meanings.

(iv) determining unit 340.

The determining unit 340 is used to determine the sub-expression to be constructed from the basic sub-expression and the equivalent sub-expression of the basic sub-expression.

In some embodiments, determining a sub-expression to be constructed from a basic sub-expression and an equivalent sub-expression of the basic sub-expression includes:

Get a first condition number and a second condition number, the first condition number is the condition number of the data connection condition associated with the base sub-expression, and the second condition number is the condition number of the data connection condition associated with the equivalent sub-expression of the base sub-expression;

According to the first conditional quantity and the second conditional quantity, a sub-expression to be constructed is determined from a basic sub-expression and an equivalent sub-expression of the basic sub-expression.

In some embodiments, determining a sub-expression to be constructed from a basic sub-expression and an equivalent sub-expression of the basic sub-expression includes:

Obtain metadata information of the data to be queried associated with the basic sub-expression;

Obtain the data volume of the data to be queried from the metadata information;

According to the amount of data, the sub-expression to be constructed is determined from the basic sub-expression and the equivalent sub-expression of the basic sub-expression.

(V) construct unit 350.

The construction unit 350 is used to construct a data query plan through the sub-expressions to be constructed and the association relationships.

In some embodiments, constructing a data query plan using sub-expressions to be constructed and associated relationships includes:

Obtain the current query optimization request associated with the sub-expression to be constructed from the data query expression;

Construct an initial subquery plan through the sub-expression to be constructed;

Determine the subquery plan to be processed from the initial subquery plan according to the current query optimization request and the association relationship associated with the subexpression to be constructed;

The sub-query plans to be processed are merged to obtain a data query plan.

In some embodiments, constructing an initial subquery plan using the sub-expression to be constructed includes:

Get the query operator associated with the sub-expression to be constructed;

According to the sub-expressions to be constructed, the connection relationship between the query operators is constructed;

The query operators are controlled to be arranged according to the connection relationship to obtain the initial subquery plan.

In some embodiments, according to the current query optimization request and the association relationship, determining a to-be-processed sub-query plan from an initial sub-query plan includes:

According to the current query optimization request associated with the sub-expression to be constructed, metadata information of the to-be-constructed data associated with the to-be-constructed sub-expression is obtained;

Analyze the amount of computation required when executing the initial subquery plan based on the metadata information of the data to be queried;

According to the computation amount and the association relationship, the subquery plan to be processed is determined from the initial subquery plan.

In some embodiments, the sub-expression to be constructed includes a first expression and a second expression, the first expression is used to query the data to be queried, and the second expression is used to filter the data to be queried, and according to the current query optimization request and the association relationship, the sub-query plan to be processed is determined from the initial sub-query plan, including:

Determine a first subquery plan from a first initial subquery plan according to a current query optimization request associated with the first expression, where the first initial subquery plan is an initial subquery plan constructed by using the first expression;

Based on the current query optimization request satisfied by the first subquery plan, determining unsatisfied requests from the current query optimization request associated with the first expression;

According to the unsatisfied requests and the association relationship, determine a final optimization request from the current query optimization requests associated with the second expression, where the final optimization request includes the unsatisfied request;

According to the final optimization request, a second subquery plan is determined from the second initial subquery plan, where the second initial subquery plan is an initial subquery plan constructed by using the second expression, and the subquery plans to be processed include the first subquery plan and the second subquery plan.

In some embodiments, it also includes:

Obtain the historical query optimization request associated with the sub-expression to be constructed, and the historical query plan corresponding to the historical query optimization request;

Determine, from the historical query optimization requests, a pending historical request corresponding to the current query optimization request;

If there are pending historical requests, the historical query plan corresponding to the pending historical request is used as the pending sub-query plan;

If there is no pending historical request, the step of constructing an initial subquery plan through the sub-expression to be constructed is triggered.

In some embodiments, obtaining a historical query optimization request associated with the sub-expression to be constructed and a historical query plan corresponding to the historical query optimization request includes:

Obtain a hash table associated with the sub-expression to be constructed, the hash table including historical query optimization requests, and first tags and plan retrieval links associated with the historical query optimization requests, the first tag being a hash value obtained after the historical query optimization request is calculated by a hash function;

Determine the pending historical request corresponding to the current query optimization request from the historical query optimization requests, including:

Determine a second tag, where the second tag is a hash value obtained after the current query optimization request is calculated by a hash function;

Determine, from the historical query optimization requests, a to-be-processed historical request corresponding to the current query optimization request according to the first tag and the second tag;

If there are pending historical requests, the historical query plan corresponding to the pending historical request is used as the pending sub-query plan, including:

If there are pending historical requests, the plan retrieval link associated with the pending historical request is used to retrieve the historical query plan corresponding to the historical query optimization request as the pending sub-query plan.

In some embodiments, after constructing a data query plan using the sub-expression to be constructed, the method further includes:

Determine the query instruction corresponding to the data query plan, and the copy instruction of the query instruction;

The replica instruction is sent to multiple data storage nodes of the database, so that the data storage nodes return the query data according to the replica instruction.

In specific implementation, the above units can be implemented as independent entities, or can be arbitrarily combined to be implemented as the same or several entities. The specific implementation of the above units can refer to the previous method embodiments, which will not be repeated here.

As can be seen from the above, the query plan construction device of this embodiment obtains a data query expression by an acquisition unit, and the data query expression includes keywords; the key unit obtains the current sub-expression in the data query expression and the association relationship between the current sub-expressions according to the keywords; the equivalent unit expands the basic sub-expression to obtain the equivalent sub-expression of the basic sub-expression, and the basic sub-expression and the equivalent sub-expression of the basic sub-expression have equivalent expression meanings; the determination unit determines the sub-expression to be constructed from the basic sub-expression and the equivalent sub-expression of the basic sub-expression; the construction unit constructs a data query plan through the sub-expression to be constructed and the association relationship. Therefore, the embodiment of the present application can facilitate the construction of a data query plan and improve query efficiency.

The embodiment of the present application also provides an electronic device, which can be a terminal, a server, etc. The terminal can be a mobile phone, a tablet computer, a smart Bluetooth device, a laptop, a personal computer, etc. The server can be a single server or a server cluster composed of multiple servers, etc.

In some embodiments, the query plan construction device can also be integrated into multiple electronic devices. For example, the query plan construction device can be integrated into multiple servers, and the query plan construction method of the present application can be implemented by multiple servers.

In this embodiment, the electronic device of this embodiment is a server as an example for detailed description. For example, as shown in FIG. 4 , it shows a schematic diagram of the structure of the server involved in the embodiment of the present application. Specifically:

The server may include one or more processing core processors 410, one or more computer-readable storage media memories 420, a power supply 430, an input module 440, and a communication module 450. Those skilled in the art will appreciate that the server structure shown in FIG. 4 does not limit the server, and may include more or fewer components than shown, or combine certain components, or arrange the components differently.

The processor 410 is the control center of the server, and uses various interfaces and lines to connect various parts of the entire server. It executes various functions of the server and processes data by running or executing software programs and/or modules stored in the memory 420, and calling data stored in the memory 420. In some embodiments, the processor 410 may include one or more processing cores; in some embodiments, the processor 410 may integrate an application processor and a modem processor, wherein: The application processor mainly processes the operating system, user interface, and application programs, and the modem processor mainly processes wireless communications. It is understandable that the modem processor may not be integrated into the processor 410 .

The memory 420 can be used to store software programs and modules. The processor 410 executes various functional applications and data processing by running the software programs and modules stored in the memory 420. The memory 420 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area may store data created according to the use of the server, etc. In addition, the memory 420 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage devices. Accordingly, the memory 420 may also include a memory controller to provide the processor 410 with access to the memory 420.

The server also includes a power supply 430 for supplying power to various components. In some embodiments, the power supply 430 can be logically connected to the processor 410 through a power management system, so that the power management system can manage charging, discharging, and power consumption. The power supply 430 can also include any components such as one or more DC or AC power supplies, recharging systems, power failure detection circuits, power converters or inverters, and power status indicators.

The server may further include an input module 440, which may be used to receive input digital or character information and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function controls.

The server may further include a communication module 450. In some embodiments, the communication module 450 may include a wireless module. The server may perform short-range wireless transmission through the wireless module of the communication module 450, thereby providing wireless broadband Internet access to the user. For example, the communication module 450 may be used to help the user send and receive emails, browse web pages, and access streaming media.

Although not shown, the server may further include a display unit, etc., which will not be described in detail herein. Specifically in this embodiment, the processor 410 in the server will load the executable files corresponding to the processes of one or more applications into the memory 420 according to the following instructions, and the processor 410 will run the application stored in the memory 420, thereby implementing any of the methods provided in the aforementioned embodiments.

The specific implementation of the above methods can be found in the previous embodiments, which will not be described in detail here.

From the above, it can be seen that multiple sub-expressions to be constructed can construct a query plan based on the association relationship, so that there is no conflict between the multiple sub-expressions to be constructed when constructing the data query plan. This makes it easier to construct the data query plan and improves the query performance.

A person of ordinary skill in the art will appreciate that all or part of the steps in the various methods of the above embodiments may be completed by instructions, or by controlling related hardware through instructions. The instructions may be stored in a computer-readable storage medium and loaded and executed by a processor.

To this end, an embodiment of the present application provides a computer-readable storage medium, in which a plurality of instructions are stored, and the instructions can be loaded by a processor to execute the steps in any query plan construction method provided in the embodiment of the present application. For example, the instructions can execute any method provided in the aforementioned embodiment.

Among them, the storage medium may include: read-only memory (ROM), random access memory (RAM), disk or CD, etc.

According to one aspect of the present application, a computer program product is provided, the computer program product comprising a computer program/instructions stored in a computer readable storage medium. The machine-readable storage medium reads the computer program/instructions, and the processor executes the computer program/instructions, so that the computer device executes the methods provided in various optional implementations of the query plan construction provided in the above embodiments.

Since the instructions stored in the storage medium can execute the steps in any query plan construction method provided in the embodiments of the present application, the beneficial effects that can be achieved by any query plan construction method provided in the embodiments of the present application can be achieved. Please refer to the previous embodiments for details and will not be repeated here.

The above is a detailed introduction to a query plan construction method, device, electronic device and storage medium provided in the embodiments of the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for technical personnel in this field, according to the ideas of the present application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims (15)

A query plan construction method, the method being executed by an electronic device, comprising: Obtaining a data query expression, wherein the data query expression includes keywords; According to the keywords, basic sub-expressions in the data query expression and the association relationship between the basic sub-expressions are obtained; Expanding the basic sub-expression to obtain an equivalent sub-expression of the basic sub-expression, wherein the basic sub-expression and the equivalent sub-expression of the basic sub-expression have equivalent expression meanings; Determine a sub-expression to be constructed from the basic sub-expression and the equivalent sub-expression of the basic sub-expression; A data query plan is constructed using the sub-expressions to be constructed and the association relationships. The query plan construction method according to claim 1, wherein determining the sub-expression to be constructed from the basic sub-expression and the equivalent sub-expression of the basic sub-expression comprises: Acquire a first condition quantity and a second condition quantity, wherein the first condition quantity is the condition quantity of the data connection condition associated with the basic sub-expression, and the second condition quantity is the condition quantity of the data connection condition associated with the equivalent sub-expression of the basic sub-expression; The sub-expression to be constructed is determined from the basic sub-expression and the equivalent sub-expression of the basic sub-expression according to the first conditional quantity and the second conditional quantity. The query plan construction method according to claim 1, wherein determining the sub-expression to be constructed from the basic sub-expression and the equivalent sub-expression of the basic sub-expression comprises: Obtaining metadata information of the to-be-queried data associated with the basic sub-expression; Acquire the data volume of the to-be-queried data from the metadata information; According to the data amount, a sub-expression to be constructed is determined from the basic sub-expression and the equivalent sub-expression of the basic sub-expression. The query plan construction method according to claim 1, wherein the step of constructing a data query plan using the sub-expressions to be constructed and the association relationship comprises: Acquire a current query optimization request associated with the sub-expression to be constructed from the data query expression; Constructing an initial subquery plan through the sub-expression to be constructed; Determining a to-be-processed sub-query plan from the initial sub-query plan according to the current query optimization request and the association relationship; The sub-query plans to be processed are merged to obtain the data query plan. The query plan construction method according to claim 4, wherein the step of constructing an initial subquery plan using the subexpression to be constructed comprises: Obtaining the query operator associated with the sub-expression to be constructed; Constructing a connection relationship between the query operators according to the sub-expressions to be constructed; The query operators are controlled to be arranged according to the connection relationship to obtain the initial sub-query plan. The query plan construction method according to claim 4, wherein determining the sub-query plan to be processed from the initial sub-query plan according to the current query optimization request and the association relationship comprises: According to the current query optimization request, metadata information of the to-be-queried data associated with the to-be-constructed sub-expression is obtained; Analyzing the amount of computation required when executing the initial subquery plan based on the metadata information of the data to be queried; The to-be-processed sub-query plan is determined from the initial sub-query plan according to the calculation amount and the association relationship. The query plan construction method according to claim 4, wherein the sub-expression to be constructed includes a first expression and a second expression, the first expression is used to query the data to be queried, and the second expression is used to filter the data to be queried, and the determining the sub-query plan to be processed from the initial sub-query plan according to the current query optimization request and the association relationship comprises: Determine, according to the current query optimization request associated with the first expression, a first sub-query plan from a first initial sub-query plan, wherein the first initial sub-query plan is the initial sub-query plan constructed by using the first expression; Based on the current query optimization request satisfied by the first subquery plan, determining unsatisfied requests from the current query optimization request associated with the first expression; Determine, according to the unsatisfied request and the association relationship, a final optimization request from the current query optimization requests associated with the second expression, wherein the final optimization request includes the unsatisfied request; According to the final optimization request, a second sub-query plan is determined from a second initial sub-query plan, wherein the second initial sub-query plan is the initial sub-query plan constructed by the second expression, and the sub-query plan to be processed includes the first sub-query plan and the second sub-query plan. The query plan construction method according to claim 4, further comprising: Obtaining a historical query optimization request associated with the sub-expression to be constructed, and a historical query plan corresponding to the historical query optimization request; Determine, from the historical query optimization requests, a to-be-processed historical request corresponding to the current query optimization request; If the pending historical request exists, the historical query plan corresponding to the pending historical request is used as the pending sub-query plan; If the to-be-processed historical request does not exist, the step of constructing the initial subquery plan through the to-be-constructed sub-expression is triggered to execute. The query plan construction method according to claim 8, wherein the step of obtaining the historical query optimization request associated with the sub-expression to be constructed and the historical query plan corresponding to the historical query optimization request comprises: Obtaining a hash table associated with the sub-expression to be constructed, wherein the hash table includes a historical query optimization request, and a first label and a plan retrieval link associated with the historical query optimization request, wherein the first label is a hash value obtained after the historical query optimization request is calculated by a hash function; The determining, from the historical query optimization requests, the to-be-processed historical request corresponding to the current query optimization request includes: Determine a second tag, where the second tag is a hash value of the current query optimization request obtained after the hash function is used to calculate the current query optimization request; Determine, from the historical query optimization requests, a to-be-processed historical request corresponding to the current query optimization request according to the first tag and the second tag; If the pending historical request exists, taking the historical query plan corresponding to the pending historical request as the pending sub-query plan includes: If the pending historical request exists, the plan retrieval link associated with the pending historical request is used to retrieve the historical query plan corresponding to the historical query optimization request as the pending sub-query plan. The query plan construction method according to any one of claims 1 to 9, wherein obtaining the data query expression comprises: Acquire a data query command, wherein the data query command includes the keyword; According to the keywords, the data query command is subjected to syntax analysis to obtain a syntax tree of the data query command; The syntax tree is format converted using a preset data exchange language to obtain the data query expression. The query plan construction method according to any one of claims 1 to 9, after constructing the data query plan by using the sub-expression to be constructed and the association relationship, further comprising: Determine a query instruction corresponding to the data query plan, and a copy instruction of the query instruction; The replica instruction is sent to multiple data storage nodes of the database, so that the data storage nodes return the query data according to the replica instruction. A query plan construction device, the device being deployed on an electronic device, comprising: An acquisition unit, used to acquire a data query expression, wherein the data query expression includes a keyword; A key unit, used to obtain basic sub-expressions in the data query expression and the association relationship between the basic sub-expressions according to the keywords; An equivalent unit, used for expanding the basic sub-expression to obtain an equivalent sub-expression of the basic sub-expression, wherein the basic sub-expression and the equivalent sub-expression of the basic sub-expression have equivalent expression meanings; A determination unit, configured to determine a sub-expression to be constructed from the basic sub-expression and an equivalent sub-expression of the basic sub-expression; A construction unit is used to construct a data query plan through the sub-expression to be constructed and the association relationship. An electronic device comprises a processor and a memory, wherein the memory stores a plurality of instructions; the processor loads instructions from the memory to execute the steps in the query plan construction method as described in any one of claims 1 to 11. A computer-readable storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded by a processor to execute the steps in the query plan construction method according to any one of claims 1 to 11. A computer program product comprises a computer program/instruction, wherein when the computer program/instruction is executed by a processor, the steps in the query plan construction method according to any one of claims 1 to 11 are implemented.