CREATE TABLE

Function

Creates a new empty table in the current database.

This table is owned by the user who executes the command. However, if the system administrator creates a table in the schema with the same name as a common user, the owner of the table is the user (not the system administrator).

Precautions

• For details about the data types supported by column-store tables, see Data Types Supported by Column-Store Tables.
• It is recommended that the number of column-store and HDFS partitioned tables do not exceed 1000.
• The primary key constraint and unique constraint in the table must contain a distribution column.
• The data type of the distribution column in an existing table cannot be modified.
• A system column cannot be designated as a primary key in a row-store REPLICATION distributed table.
• If an error occurs during table creation, after it is fixed, the system may fail to delete the empty disk files created before the last automatic clearance. This problem seldom occurs.
• Column-store tables support the PARTIAL CLUSTER KEY and table-level primary key and unique constraints, but do not support table-level foreign key constraints.
• Only the NULL, NOT NULL, and DEFAULT constant values can be used as column-store table column constraints.
• Whether column-store tables support a delta table is specified by the enable_delta parameter. The threshold for storing data into a delta table is specified by the deltarow_threshold parameter. Using column-store tables with delta tables is not recommended. This may cause disk bloat and performance deterioration due to delayed merge.
• Multi-temperature tables support only partitioned column-store tables and depend on available OBS tablespaces.
• Multi-temperature tables support only the default tablespace default_obs_tbs. If you need to add an OBS tablespace, contact technical support.
• To create a column-store table, explicitly set orientation to column. When creating a local table in the storage-compute decoupling edition with all data stored on EVS disks, make sure to explicitly set colversion to 2.0.
• Once a table is created, it cannot be switched from a non-V3 table to a V3 table using the ALTER TABLE syntax. In other words, a non-V3 table with a colversion of 2.0 cannot be converted to a V3 table with a colversion of 3.0.
• Delta tables and column-store level-2 partitions cannot be set for V3 tables (for example, colversion = 3.0, storage and compute separated tables).
• V3 tables cannot be configured as H-Store tables, hot-and-cold-data-separated tables, or time sequence tables.
• Global temporary tables and temporary tables cannot be created for V3 tables. Any created temporary tables are automatically converted to temporary tables with colversion 2.0.

• You are not advised to specify a user-defined tablespace when creating an ordinary table.
• Do not specify the COMPRESS compression attribute when creating a row-store table.
• When creating a hash-distributed table object, ensure that data is evenly distributed. (For a table with more than 10 GB data, the skew rate must be less than 10%.)
• When creating a table object for REPLICATION distribution, ensure that the number of rows in the table is less than 1 million.
• When creating an H-Store table, ensure that the database GUC parameter settings meet the following requirements:
  • autovacuum is set to on.
  • The value of autovacuum_max_workers_hstore is greater than 0.
  • The value of autovacuum_max_workers is greater than that of autovacuum_max_workers_hstore.
• For a large table (with more than 50 million rows of data) that contains the time field, the table must be designed as a partition table and the partition interval must be properly designed based on the query characteristics.
• For a table where a large amount of data needs to be added, deleted, or modified, it is recommended that the number of indexes be less than or equal to three. The maximum number of indexes is five.
• For more information about development and design specifications, see Development and Design Proposal.

Syntax

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12

CREATE [ [ GLOBAL | LOCAL | VOLATILE ] { TEMPORARY | TEMP } | UNLOGGED ] TABLE [ IF NOT EXISTS ] table_name 
    { ({ column_name data_type [ compress_mode ] [ COLLATE collation ] [ column_constraint [ ... ] ]
        | table_constraint
        | LIKE source_table [ like_option [...] ] }
        [, ... ])|
        LIKE source_table [ like_option [...] ] } 
    [ WITH ( {storage_parameter = value} [, ... ] ) ]
    [ ON COMMIT { PRESERVE ROWS | DELETE ROWS } ]
    [ COMPRESS | NOCOMPRESS ]
    [ DISTRIBUTE BY { REPLICATION | ROUNDROBIN | { HASH ( column_name [,...] ) } } ]
    [ TO { GROUP groupname | NODE ( nodename [, ... ] ) } ]
    [ COMMENT [=] 'text' ];

• column_constraint is as follows:

  1 2 3 4 5 6 7 8 9 10 11

  [ CONSTRAINT constraint_name ] { NOT NULL | NULL | CHECK ( expression ) | DEFAULT default_expr | ON UPDATE on_update_expr | COMMENT 'text' | UNIQUE [ NULLS [NOT] DISTINCT | NULLS IGNORE ] index_parameters | PRIMARY KEY index_parameters } REFERENCES reftable [ ( refcolumn ) ] } [ DEFERRABLE | NOT DEFERRABLE | INITIALLY DEFERRED | INITIALLY IMMEDIATE ]

• compress_mode of a column is as follows:

  1	{ DELTA | PREFIX | DICTIONARY | NUMSTR | NOCOMPRESS }

• table_constraint is as follows:

  1 2 3 4 5 6

  [ CONSTRAINT constraint_name ] { CHECK ( expression ) | UNIQUE [ NULLS [NOT] DISTINCT | NULLS IGNORE ] ( column_name [, ... ] ) index_parameters | PRIMARY KEY ( column_name [, ... ] ) index_parameters | PARTIAL CLUSTER KEY ( column_name [, ... ] ) } [ DEFERRABLE | NOT DEFERRABLE | INITIALLY DEFERRED | INITIALLY IMMEDIATE ]

• like_option is as follows:

  1	{ INCLUDING | EXCLUDING } { DEFAULTS | CONSTRAINTS | INDEXES | STORAGE | COMMENTS | PARTITION | RELOPTIONS | DISTRIBUTION | DROPCOLUMNS | ALL }

• index_parameters is as follows:

  1	[ WITH ( {storage_parameter = value} [, ... ] ) ]

Table Design Reference

GaussDB(DWS) is compatible with the PostgreSQL ecosystem. Row storage and its B-tree index are similar to those of PostgreSQL. Column storage and its index are self-developed. When creating a table, it is crucial to choose the right storage method, distribution column, partition key, and index. This ensures efficient data access during SQL execution, reducing I/O consumption. The following figure illustrates the process from SQL statement initiation to data acquisition, helping you understand the function of each technical method for performance optimization.

Parameters

• UNLOGGED

  If this key word is specified, the created table is not a log table. Data written to unlogged tables is not written to the write-ahead log, which makes them considerably faster than ordinary tables. However, an unlogged table is automatically truncated after a crash or unclean shutdown, incurring data loss risks. The contents of an unlogged table are also not replicated to standby servers. Any indexes created on an unlogged table are not automatically logged as well.

  Usage scenario: Unlogged tables do not ensure safe data. You can back up data before using unlogged tables; for example, you should back up the data before a system upgrade.

  Troubleshooting: If data is missing in the indexes of unlogged tables due to some unexpected operations such as an unclean shutdown, you should re-create the indexes with errors.

  

  1. The UNLOGGED table uses no primary/standby mechanism. In the case of system faults or abnormal breakpoints, data loss may occur. Therefore, the UNLOGGED table cannot be used to store basic data.
  2. Starting from version 9.1.0, UNLOGGED tables are automatically saved in the pg_unlogged tablespace and cannot be moved or assigned to other tablespaces.
  3. After an earlier version is upgraded to 9.1.0, the UNLOGGED table created in the earlier version is still stored in the original tablespace.
  4. If the instance restarts unexpectedly, the UNLOGGED table will be reset, which can impact the instance's recovery time objective (RTO). Version 9.1.0 has a script called switch_unlogged_tablespace.py that can move unlogged tables to optimize the recovery time objective (RTO). This script works together with the GUC parameter enable_unlogged_tablespace_compat.
• GLOBAL | LOCAL | VOLATILE

  Specify the keywords GLOBAL, LOCAL, and VOLATILE before TEMP or TEMPORARY to create temporary tables with different attributes.

  • Currently, the keywords GLOBAL and LOCAL are introduced for standard SQL compatibility. No matter whether GLOBAL or LOCAL is specified, the GaussDB(DWS) creates a LOCAL temporary table.
  • If VOLATILE is specified, a temporary volatile table is created.
  • If default_temptable_type is set to local, temporary tables created without keywords are local temporary tables. If default_temptable_type is set to volatile, temporary tables created without keywords are volatile temporary tables.
• TEMPORARY | TEMP

  If TEMP or TEMPORARY is specified, the created table is a temporary table. Temporary tables are automatically dropped at the end of a session, or optionally at the end of the current transaction. Therefore, apart from CN and other CN errors connected by the current session, you can still create and use temporary table in the current session. Temporary tables are created only in the current session. If a DDL statement involves operations on temporary tables, a DDL error will be generated. Therefore, you are not advised to perform operations on temporary tables in DDL statements. TEMP is equivalent to TEMPORARY.

  

  • Local or volatile temporary tables are visible to the current session through schema of the pg_temp start. Users should not delete schema started with pg_temp, pg_toast_temp.
  • If TEMPORARY or TEMP is not specified when you create a table and the schema of the specified table starts with pg_temp_, the table is created as a temporary table.
  • Similar to common tables, all metadata of local temporary tables is stored in system catalogs. Volatile temporary tables store table structure metadata in memory, except the schema metadata. Compared with local temporary tables, volatile temporary tables have the following constraints:
    • After a CN or DN is restarted, data in its memory will be lost, and accordingly, volatile temporary tables on it will become invalid.
    • Currently, volatile temporary tables do not support table structure modification, such as ALTER and GRANT.
    • Volatile temporary tables and local temporary tables share temporary schemas. Therefore, in the same session, the VOLATILE temporary table and local temporary table cannot have the same name.
    • Volatile temporary table information is not stored in system catalogs. Therefore, Volatile metadata cannot be queried by running DML statements in system catalogs.
    • Volatile temporary tables support only common row-store and column-store tables. Delta tables, time series tables, and cold and hot tables are not supported.
    • Views cannot be created based on volatile temporary tables.
    • A tablespace cannot be specified when a temporary table is created. The default tablespace of a volatile temporary table is pg_volatile.
    • The following constraints cannot be specified when a volatile temporary table is created: CHECK, UNIQUE, PRIMARY KEY, TRIGGER, EXCLUDE, and PARTIAL CLUSTER.
  • Similar to common tables, all metadata in global temporary tables is stored in system catalogs.
    • The difference between a global temporary table and a local temporary table is that in a global temporary table, the metadata is not deleted when a session exits, but the session data is deleted. Data of different sessions is independent but shares the metadata of the same global temporary table.
    • A global temporary table has a schema that is similar to a regular table, unlike a local or volatile temporary table which has a schema starting with pg_temp. This means that a global temporary table can have the same name as a local or volatile temporary table.
    • Global temporary tables support only common row-store and column-store tables. Delta tables, time series tables, and cold and hot tables are not supported.
    • Operations cannot be performed on global temporary tables of other logical clusters.
• IF NOT EXISTS

  If IF NOT EXISTS is specified, a table will be created if there is no table using the specified name. If there is already a table using the specified name, no error will be reported. A message will be displayed indicating that the table already exists, and the database will skip table creation.

• table_name

  Specifies the name of the table to be created.

  The table name can contain a maximum of 63 characters, including letters, digits, underscores (_), dollar signs ($), and number signs (#). It must start with a letter or underscore (_).

  A table name enclosed in double quotation marks can contain spaces and special characters. However, you are not advised to use these characters in a table name because they may make it difficult to reference and use. In addition, they may be processed differently under different database compatibility modes.

• column_name

  Specifies the name of a column to be created in the new table.

  The column name can contain a maximum of 63 characters, including letters, digits, underscores (_), dollar signs ($), and number signs (#). It must start with a letter or underscore (_).

• data_type

  Specifies the data type of the column.

  

  In a database compatible with Teradata or MySQL syntax, if the data type of a column is set to DATE, the DATE type is returned. Otherwise, the TIMESTAMP type is returned.

• compress_mode

  Specifies the compress option of the table, only available for row-store table. The option specifies the algorithm preferentially used by table columns.

  Value range: DELTA, PREFIX, DICTIONARY, NUMSTR, NOCOMPRESS

• COLLATE collation

  Assigns a collation to the column (which must be of a collatable data type). If no collation is specified, the default collation is used.

• LIKE source_table [ like_option ... ]

  Specifies a table from which the new table automatically copies all column names, their data types, and their not-null constraints.

  The new table and the source table are decoupled after creation is complete. Changes to the source table will not be applied to the new table, and it is not possible to include data of the new table in scans of the source table.

  Columns and constraints copied by LIKE are not merged with the same name. If the same name is specified explicitly or in another LIKE clause, an error is reported.

  • The default expressions or the ON UPDATE expressions are copied from the source table to the new table only if INCLUDING DEFAULTS is specified. The default behavior is to exclude default expressions, resulting in the copied columns in the new table having default values NULL.
  • The CHECK constraints are copied from the source table to the new table only when INCLUDING CONSTRAINTS is specified. Other types of constraints are never copied to the new table. NOT NULL constraints are always copied to the new table. These rules also apply to column constraints and table constraints.
  • Any indexes on the source table will not be created on the new table, unless the INCLUDING INDEXES clause is specified.
  • STORAGE settings for the copied column definitions are copied only if INCLUDING STORAGE is specified. The default behavior is to exclude STORAGE settings.
  • If INCLUDING COMMENTS is specified, comments for the copied columns, constraints, and indexes are copied. The default behavior is to exclude comments.
  • If INCLUDING PARTITION is specified, the partition definitions of the source table are copied to the new table, and the new table no longer uses the PARTITION BY clause. The default behavior is to exclude partition definition of the source table.
  • If INCLUDING RELOPTIONS is specified, the storage parameter (WITH clause of the source table) of the source table is copied to the new table. The default behavior is to exclude partition definition of the storage parameter of the source table.
    

    PERIOD and TTL in the WITH clause are partition-related parameters. LIKE INCLUDING RELOPTIONS will not be copied to the new table. To copy LIKE INCLUDING RELOPTIONS, use INCLUDING PARTITION.

  • If INCLUDING DISTRIBUTION is specified, the distribution information of the source table is copied to the new table, including distribution type and column, and the new table no longer use the DISTRIBUTE BY clause. The default behavior is to exclude distribution information of the source table.
  • If INCLUDING DROPCOLUMNS is specified, the deleted column information in the source table is copied to the new table. By default, the deleted column information of the source table is not copied.
  • INCLUDING ALL contains the meaning of INCLUDING DEFAULTS, INCLUDING CONSTRAINTS, INCLUDING INDEXES, INCLUDING STORAGE, INCLUDING COMMENTS, INCLUDING PARTITION, INCLUDING RELOPTIONS, INCLUDING DISTRIBUTION, and INCLUDING DROPCOLUMNS.
  • If EXCLUDING is specified, the specified parameters are not included.
  • For an OBS multi-temperature table, all partitions of the new table are local hot partitions after INCLUDING PARTITION is specified.
  

  • If the source table contains a sequence with the SERIAL, BIGSERIAL, or SMALLSERIAL data type, or a column in the source table is a sequence by default and the sequence is created for this table by using CREATE SEQUENCE... OWNED BY, these sequences will not be copied to the new table, and another sequence specific to the new table will be created. This is different from earlier versions. To share a sequence between the source table and new table, create a shared sequence (do not use OWNED BY) and set a column in the source table to this sequence.
  • You are not advised to set a column in the source table to the sequence specific to another table especially when the table is distributed in specific Node Groups, because doing so may result in CREATE TABLE ... LIKE execution failures. In addition, doing so may cause the sequence to become invalid in the source sequence because the sequence will also be deleted from the source table when it is deleted from the table that the sequence is specific to. To share a sequence among multiple tables, you are advised to create a shared sequence for them.
• WITH ( { storage_parameter = value } [, ... ] )

  Specifies an optional storage parameter for a table or an index.

  

  Using Numeric of any precision to define column, specifies precision p and scale s. When precision and scale are not specified, the input will be displayed.

  The description of parameters is as follows:

  • FILLFACTOR

    The fillfactor of a table is a percentage between 10 and 100. 100 (complete packing) is the default value. When a smaller fillfactor is specified, INSERT operations pack table pages only to the indicated percentage. The remaining space on each page is reserved for updating rows on that page. This gives UPDATE a chance to place the updated copy of a row on the same page, which is more efficient than placing it on a different page. For a table whose records are never updated, setting the fillfactor to 100 (complete packing) is the appropriate choice, but in heavily updated tables smaller fillfactors are appropriate. The parameter has no meaning for column–store tables.

    Value range: 10 to 100

  • ORIENTATION

    Specifies the storage mode (row-store, column-store) for table data. This parameter cannot be modified once it is set.

    Valid value:

    • ROW indicates that table data is stored in rows.

      ROW applies to OLTP service, which has many interactive transactions. An interaction involves many columns in the table. Using ROW can improve the efficiency.

    • COLUMN indicates that the data is stored in columns.

      COLUMN applies to the data warehouse service, which has a large amount of aggregation computing, and involves a few column operations.

    Default value: ROW (row-store)

  • COMPRESSION

    Specifies the compression level of the table data. It determines the compression ratio and time. Generally, the higher the level of compression, the higher the ratio, the longer the time, and the lower the level of compression, the lower the ratio, the shorter the time. The actual compression ratio depends on the distribution characteristics of loading table data.

    Valid value:

    The valid values for column-store tables are YES/NO and LOW/MIDDLE/HIGH, and the default is LOW. When this parameter is set to YES, the compression level is LOW by default.

    

    • Currently, row-store table compression is not supported.
    • To determine the size of a new GaussDB(DWS) cluster, consider the size of ORC data compressed and migrated to column-store tables in GaussDB(DWS). If the compression level is low, the size of a copy is about 1.5 to 2 times that of ORC. If the compression level is high, the size of a copy is basically the same as that of ORC.
    • The middle compression of column-stores uses dictionary compression. For data not suitable for dictionary compression, the file size after middle compression may be greater than that of after low compression.

    GaussDB(DWS) provides the following compression algorithms:

    Table 2 Compression algorithms for column-based storage

    COMPRESSION	NUMERIC	STRING	INT
    LOW	Delta compression + RLE compression	LZ4 compression	Delta compression (RLE is optional.)
    MIDDLE	Delta compression + RLE compression + LZ4 compression	dict compression or LZ4 compression	Delta compression or LZ4 compression (RLE is optional)
    HIGH	Delta compression + RLE compression + zlib compression	dict compression or zlib compression	Delta compression or zlib compression (RLE is optional)

  • COMPRESSLEVEL

    Specifies the compression level of the table data. It determines the compression ratio and time. This divides a compression level into sublevels, providing you with more choices for compression rate and duration. As the value becomes greater, the compression rate becomes higher and duration longer at the same compression level. The parameter is only valid for column-store tables.

    Value range: 0–3.

    Default value: 0

  • TTL

    Schedules the partition deletion tasks in a partitioned table. By default, no partition deletion task is created.

    Value range: 1 hour–100 years

  • PERIOD

    Schedules the partition creation tasks in a partitioned table. If TTL has been configured, PERIOD cannot be greater than TTL.

    Value range: 1 hour–100 years

    Default value: 1 day

  • MAX_BATCHROW

    Specifies the maximum of a storage unit during data loading process. The parameter is only valid for column-store tables.

    Value range: 10000 to 60000

    Default value: 60,000

  • PARTIAL_CLUSTER_ROWS

    Specifies the number of records to be partial cluster stored during data loading process. The parameter is only valid for column-store tables.

    Value range: 600000 to 2147483647

    Default value: 4,200,000

  • time_format

    You can use auto-increment and decrement partitions for INT4, INT8, VARCHAR, and TEXT columns, which are commonly used for storing time-related data. The time_format option is applicable only when the partition key is INT4, INT8, VARCHAR, or TEXT and a period is specified. This is supported only by clusters of version 9.1.0.200 or later.

    1. The value of time_format must comply with the PostgreSQL specifications, for example, yyyymmdd.
    2. Restrictions on time_format differs by partition key type.

       VARCHAR/TEXT:

       • The precision can be accurate to seconds.
       • The value cannot contain letters, for example, month, am, and pm.
       • The time must be arranged in descending order, for example, year, month, day, hour, minute, and second.

       INT4/INT8:

       • The precision can be accurate to hours.
       • The value can contain only Y, M, D, and HH24.
       • The time must be arranged in descending order, for example, year, month, day, and hour.
    3. ALTER restrictions:
       • The set operation is not supported.
       • When period is reset (indicating that automatic partitioning is disabled and a message is displayed), you can reset this option.
  • enable_delta

    Specifies whether to enable delta tables in column-store tables. The parameter is only valid for column-store tables. If COLVERSION is set to 3.0, enable_delta cannot be turned on because this parameter is not supported by V3 tables.

    Using column-store tables with delta tables is not recommended. This may cause disk bloat and performance deterioration due to delayed merge.

    Default value: off

  • enable_hstore

    Specifies whether an H-Store table will be created (based on column-store tables). The parameter is only valid for column-store tables. This parameter is supported by version 8.2.0.100 or later clusters. If COLVERSION is set to 3.0, enable_delta cannot be turned on because this parameter is not supported by V3 tables.

    Default value: off

    

    If this parameter is enabled, the following GUC parameters must be set to ensure that H-Store tables are cleared.

    Set autovacuum to on, autovacuum_max_workers to 6, and autovacuum_max_workers_hstore to 3.

  • enable_disaster_cstore

    Specifies whether fine-grained DR will be enabled for column-store tables. This parameter only takes effect on column-store tables whose COLVERSION is 2.0 and cannot be set to true if enable_hstore is true. This parameter is supported by version 8.2.0.100 or later clusters.

  • fine_disaster_table_role

    This parameter has been discarded in version 8.2.1 and is reserved for compatibility with earlier versions. This parameter is invalid in the current version.

    Specifies whether the fine-grained DR table will be set as a primary or secondary table. This parameter can be true only when the enable_disaster_cstore parameter has been set to true.

    Valid value:

    • primary: Specifies the primary fine-grained DR table.
    • standby: Specifies the standby fine-grained DR table.
  • DELTAROW_THRESHOLD

    Specifies the upper limit of to-be-imported rows for triggering the data import to a delta table when data is to be imported to a column-store table. This parameter takes effect only if the enable_delta table parameter is set to on. The parameter is only valid for column-store tables.

    Value range: 0 to 60000

    Default value: 6000

  • COLVERSION

    Specifies the version of the column-store format. You can switch between different storage formats.

    Valid value:

    1.0: Each column in a column-store table is stored in a separate file. The file name is relfilenode.C1.0, relfilenode.C2.0, relfilenode.C3.0, or similar.

    2.0: All columns of a column-store table are combined and stored in a file. The file is named relfilenode.C1.0.

    3.0: Each column of a column-store table is stored in a file. The file is stored in the OBS file system and named C1_fileid.0.

    Default value: The default value for the storage-compute coupled version is 2.0, while for the storage-compute decoupling version, it is 3.0.

    The value of COLVERSION can only be set to 2.0 for OBS multi-temperature tables.

    

    • For clusters of version 8.1.0, the default value of this parameter is 1.0. For clusters of version 8.1.1 or later, the default value of this parameter is 2.0. If the cluster version is upgraded from 8.1.0 to 8.1.1 or later, the default value of this parameter changes from 1.0 to 2.0.
    • When creating a column-store table, set COLVERSION to 2.0. Compared with the 1.0 storage format, the performance is significantly improved:
      • The time required for creating a column-store wide table is significantly reduced.
      • In the Roach data backup scenario, the backup time is significantly reduced.
      • The build and catch up time is greatly reduced.
      • The occupied disk space decreases significantly.
    • The storage-compute decoupling 3.0 version is compatible with all column-store versions. When creating a table, you need to explicitly specify the value of colversion (1.0, 2.0, or 3.0). If colversion is set to 3.0, a table in decoupled storage and computing mode is created. If colversion is not explicitly specified, a column-store table of version 3.0 is created by default. When creating a table with decoupled storage and compute nodes, set colversion to 3.0 and set orientation to column.
    • V3 storage-compute decoupling tables do not support colversion switching using ALTER TABLE, for example, from 2.0 to 3.0.
  • analyze_mode

    Specifies the mode of table-level auto-analyze.

    Valid value:

    • frozen: disables all ANALYZE operations (dynamic sampling can still be triggered when no statistics are collected).
    • backend: allows only ANALYZE triggered by AUTOVACUUM polling.
    • runtime: allows only runtime ANALYZE triggered by the optimizer.
    • all: Both backend and runtime AUTO-ANALYZE can be triggered.

    Default value: all

  • incremental_analyze

    Specifies whether to enable the incremental analyze mode for partitioned tables. This parameter is valid only for partitioned tables and cannot be set for replicated tables. This is supported only by clusters of version 9.1.0.100 or later.

    The default value is false.

  • SKIP_FPI_HINT

    Indicates whether to skip the hint bits operation when the full-page writes (FPW) log needs to be written during sequential scanning.

    Default value: false

    

    If SKIP_FPI_HINT is set to true and the checkpoint operation is performed on a table, no Xlog will be generated when the table is sequentially scanned. This applies to intermediate tables that are queried less frequently, reducing the size of Xlogs and improving query performance.

  • enable_column_autovacuum_garbage

    Determines whether to enable CU rewriting logic for column-store tables using AUTOVACUUM. This parameter is supported only by clusters of version 8.2.1.100 or later.

    There is a low probability that an error is reported when lightweight UPDATE and AUTOVACUUM are executed concurrently. You can set the table-level parameter to off to avoid this problem.

    Default value: true

  • secondary_part_column

    Specifies the name of a level-2 partition column in a column-store table. Only one column can be specified as the level-2 partition column. This parameter applies only to H-Store column-store tables. This parameter is supported only by clusters of version 8.3.0 or later.

    

    • The column specified as a level-2 partition column cannot be deleted or modified.
    • The level-2 partition column can be specified only when a table is created. After a table is created, the level-2 partition column cannot be modified.
    • You are not advised to specify a distribution column as a level-2 partition column.
    • The level-2 partition column determines how the table is logically split into hash partitions on DNs, which enhances the query performance for that column.
  • secondary_part_num

    Specifies the number of level-2 partitions in a column-store table. This parameter applies only to H-Store column-store tables. This parameter is supported only by clusters of version 8.3.0 or later.

    Value range: 1 to 32

    Default value: 8

    

    • This parameter can be specified only when secondary_part_column is specified.
    • The number of level-2 partitions can be specified only when a table is created and cannot be modified after the table is created.
    • You are not advised to change the default value, which may affect the import and query performance.
  • enable_hstore_opt

    If the enable_hstore_opt table-level parameter is enabled, the enable_hstore table-level parameter is also enabled by default. This parameter is supported only in cluster 8.3.0 and later versions. This parameter supports V2 .

    Default value: false

  • bitmap_columns

    bitmap index is only applicable to the new H-Store (hstore_opt table). To generate the bitmap index mapping, you need to enable the enable_hstore_opt table-level parameter and set bitmap_columns to Specified Columns. This parameter is supported only by clusters of version 8.3.0 or later.

  • enable_turbo_store

    Determines whether to create a turbo table (column-store tables). The parameter is only valid for column-store tables.

    Default value: off

    

    Turbo tables enhance the storage efficiency of numeric and varchar data types using the integers, leading to accelerated processing speeds for these types.

    • For enhanced performance, it is advised to define the numeric type with precision and scale (p, s). The storage allocation is as follows: For p ≤ 9, 4-byte integers are used for storage. For p ≤ 18, 8-byte integers are used. For p ≤ 37, 16-byte integers are used. If p > 37 or remain unspecified, a variable-length format is used, which is less space-efficient and yields suboptimal performance.
    • If the maximum length (n) of the varchar type is less than or equal to 2 bytes, 2-byte integers are used for storage. If the maximum length (n) of the varchar type is less than or equal to 4 bytes, 4-byte integers are used for storage. If the maximum length (n) of the varchar type is less than or equal to 16 bytes, 16-byte integers are used for storage. For varchar type columns involved in GROUP BY or HashJoin operations, it is advisable to limit the maximum byte length n to 16 or fewer to leverage integer storage for strings, thereby enhancing performance.
    • At present, the turbo table does not have support for the timing table, delta table, and lightweight update.
  • cache_policy (This parameter is supported only in 9.0.2 and later versions where the storage and compute nodes are decoupled.)

    Specifies the cache mode of tables or partitioned tables (disks). If one of the following values is specified in the cache policy, hot cache is used. Otherwise, cold cache is used. Hot cache occupies more space than cold cache and uses more complex replacement policies.

    Valid value:

    • ALL: Hot cache is used for the entire table.
    • NONE: Cold cache is used for the entire table.
    • HPN: The first N partitions in a partitioned table use hot cache. The rest of the partitions use cold cache.
    • HPL: P1, P2, .... In a partitioned table, the specified partitions use hot cache. The rest of the partitions use cold cache.

    The default value is ALL.

    

    • For foreign tables and non-partitioned tables, only the ALL and NONE cache policies are supported.
    • Only range-partitioned and list-partitioned internal tables support HPN and HPL cache policies.
• ON COMMIT { PRESERVE ROWS | DELETE ROWS }

  ON COMMIT determines what to do when you commit a temporary table creation operation.

  • PRESERVE ROWS (Default): No special action is taken at the ends of transactions. The temporary table and its table data are unchanged.
  • DELETE ROWS: All rows in the temporary table will be deleted at the end of each transaction block.
• COMPRESS | NOCOMPRESS

  If you specify COMPRESS in the CREATE TABLE statement, the compression feature is triggered in the case of a bulk INSERT operation. If this feature is enabled, a scan is performed for all tuple data within the page to generate a dictionary and then the tuple data is compressed and stored. If NOCOMPRESS is specified, the table is not compressed.

  Default value: NOCOMPRESS, tuple data is not compressed before storage.

• DISTRIBUTE BY

  Specifies how the table is distributed or replicated between DNs.

  Valid value:

  • REPLICATION: Each row in the table exists on all DNs, that is, each DN has complete table data.
  • ROUNDROBIN: Each row in the table is sent to each DN in turn. Therefore, data is evenly distributed on each DN. This value is supported only in 8.1.2 or later.
  • HASH (column_name): Each row of the table will be placed into all the DNs based on the hash value of the specified column.
    

    • When DISTRIBUTE BY HASH (column_name) is specified, the primary key and its unique index must contain the column_name column.
    • When DISTRIBUTE BY HASH (column_name) in a referenced table is specified, the foreign key of the reference table must contain the column_name column.
    • If TO GROUP is set to a replication table node group (supported in 8.1.2 or later), DISTRIBUTE BY must be set to REPLICATION. If DISTRIBUTE BY is not specified, the created table is automatically set as a replication table.
    • The hybrid data warehouse (standalone) has only one DN. Therefore, the distribution rule is ignored and cannot be modified.
  Default value: determined by the GUC parameter default_distribution_mode

  • When default_distribution_mode is set to roundrobin, the default value of DISTRIBUTE BY is selected according to the following rules:
    1. If the primary key or unique constraint is included during table creation, hash distribution is selected. The distribution column is the column corresponding to the primary key or unique constraint.
    2. If the primary key or unique constraint is not included during table creation, round-robin distribution is selected.
  • When default_distribution_mode is set to hash, the default value of DISTRIBUTE BY is selected according to the following rules:
    1. If the primary key or unique constraint is included during table creation, hash distribution is selected. The distribution column is the column corresponding to the primary key or unique constraint.
    2. If the primary key or unique constraint is not included during table creation but there are columns whose data types can be used as distribution columns, hash distribution is selected. The distribution column is the first column whose data type can be used as a distribution column.
    3. If the primary key or unique constraint is not included during table creation and no column whose data type can be used as a distribution column exists, round-robin distribution is selected.
  The following data types can be used as distribution columns:

  • Integer types: TINYINT, SMALLINT, INT, BIGINT, and NUMERIC/DECIMAL
  • Character types: CHAR, BPCHAR, VARCHAR, VARCHAR2, NVARCHAR2, and TEXT
  • Date/time types: DATE, TIME, TIMETZ, TIMESTAMP, TIMESTAMPTZ, INTERVAL, and SMALLDATETIME
  

  When you create a table, the choices of distribution keys and partition keys have major impact on SQL query performance. Therefore, choosing proper distribution column and partition key with strategies.

  • Selecting an Appropriate Distribution Column
    In the data distributed table using Hash, an appropriate distributed array should be used to distribute and store data on multiple DNs evenly, preventing data skew (uneven data distribution across several DNs). Determine the proper distribution column based on the following principles:

    1. Determine whether data is skewed.

       Connect to the database and run the following statements to check the number of tuples on each DN: Replace tablename with the actual name of the table to be analyzed.

       SELECT a.count,b.node_name FROM (SELECT count(*) AS count,xc_node_id FROM tablename GROUP BY xc_node_id) a, pgxc_node b WHERE a.xc_node_id=b.node_id ORDER BY a.count DESC;

       If tuple numbers vary greatly (several times or tenfold) in each DN, a data skew occurs. Change the data distribution key based on the following principles:

    2. Run the ALTER TABLE statement to adjust the distribution column. The rules for selecting a distribution column are as follows:

       The column value of the distribution column should be discrete so that data can be evenly distributed on each DN. For example, you are advised to select the primary key of a table as the distribution column, and the ID card number as the distribution column in a personnel information table.

       With the above principles met, you can select join conditions as distribution keys so that join tasks can be pushed down to DNs, reducing the amount of data transferred between the DNs.

    3. If a proper distribution column cannot be found to make data evenly distributed on each DN, you can use the REPLICATION or ROUNDROBIN data distribution mode. The REPLICATION data distribution mode stores complete data on each DN. Therefore, if a table is large and no proper distribution column can be found, the ROUNDROBIN data distribution mode is recommended. The ROUNDROBIN data distribution mode is supported in 8.1.2 or later.
  • Selecting appropriate partition keys

    In range partitioning, the table is partitioned into ranges defined by a key column or set of columns, with no overlap between the ranges of values assigned to different partitions. Each range has a dedicated partition for data storage.

    Modify partition keys to make the query result stored in the same or least partitions (partition pruning). Obtaining consecutive I/O to improve the query performance.

    In actual services, time is used to filter query objects. Therefore, you can use time as a partition key, and change the key value based on the total data volume and single data query volume.

• TO { GROUP groupname | NODE ( nodename [, ... ] ) }

  TO GROUP specifies the Node Group in which the table is created. Currently, it cannot be used for HDFS tables. TO NODE is used for internal scale-out tools.

  In logical cluster mode, if TO GROUP is not specified, the table is created in the node group associated with the logical cluster user by default. If the user, such as the administrator or a common user, does not manage the logical cluster, by default the table is created in the first logical cluster, which is the logical cluster with the smallest OID in pgxc_group.

  If the node group specified by TO GROUP is a replication table node group, the table is created on all CNs and DNs, but the replication table data is distributed only on the DNs in the replication table node group.

  The Storage-compute decoupling 3.0 supports read-only logical clusters. If a user is not bound to any read-only logical clusters but sets TO GROUP to a logical cluster in a table creation statement, an error will be reported during table creation. If a user bound to a read-only logical cluster creates a table, the table will be created in the logical cluster specified by the GUC parameter default_storage_nodegroup. If default_storage_nodegroup is set to installation, tables will be created in the first logical cluster.

• COMMENT [=] 'text'

  The COMMENT clause can specify table comments during table creation.

• CONSTRAINT constraint_name

  Specifies a name for a column or table constraint. The optional constraint clauses specify constraints that new or updated rows must satisfy for an insert or update operation to succeed.

  There are two ways to define constraints:

  • A column constraint is defined as part of a column definition, and it is bound to a particular column.
  • A table constraint is not bound to any particular columns but can apply to more than one column.
• NOT NULL

  Indicates that the column is not allowed to contain NULL values.

• NULL

  The column is allowed to contain NULL values. This is the default setting.

  This clause is only provided for compatibility with non-standard SQL databases. You are advised not to use this clause.

• CHECK ( expression )

  Specifies an expression producing a Boolean result which new or updated rows must satisfy for an insert or update operation to succeed. Expressions evaluating to TRUE or UNKNOWN succeed. If any row of an insert or update operation produces a FALSE result, an error exception is raised and the insert or update does not alter the database.

  A check constraint specified as a column constraint should reference only the column's values, while an expression appearing in a table constraint can reference multiple columns.

  

  <>NULL and !=NULL are invalid in an expression. Change them to IS NOT NULL.

• DEFAULT default_expr

  Assigns a default data value for a column. The value can be any variable-free expressions (Subqueries and cross-references to other columns in the current table are not allowed). The data type of the default expression must match the data type of the column.

  The default expression will be used in any insert operation that does not specify a value for the column. If there is no default value for a column, then the default value is NULL.

• ON UPDATE on_update_expr

  The ON UPDATE clause specifies a timestamp function for a column. Ensure that the data type of the column for which the ON UPDATE clause specifies a timestamp function is timestamp or timestamptz.

  When an SQL statement containing the UPDATE operation is executed, this column is automatically updated to the time specified by the timestamp function.

  

  The on_update_expr function supports only CURRENT_TIMESTAMP, CURRENT_TIME, CURRENT_DATE, LOCALTIME, LOCALTIMESTAMP.

• COMMENT 'text'

  The COMMENT clause can specify a comment for a column.

• UNIQUE [ NULLS [ NOT ] DISTINCT | NULLS IGNORE ] index_parameters

  UNIQUE [ NULLS [ NOT ] DISTINCT | NULLS IGNORE ] ( column_name [, ... ] ) index_parameters

  Specifies that a group of one or more columns of a table can contain only unique values.

  The [ NULLS [ NOT ] DISTINCT | NULLS IGNORE ] field is used to specify how to process null values in the index column of the Unique index.

  Default value: This parameter is left empty by default. NULL values can be inserted repeatedly.

  When the inserted data is compared with the original data in the table, the NULL value can be processed in any of the following ways:

  • NULLS DISTINCT: NULL values are unequal and can be inserted repeatedly.
  • NULLS NOT DISTINCT: NULL values are equal. If all index columns are NULL, NULL values cannot be inserted repeatedly. If some index columns are NULL, data can be inserted only when non-null values are different.
  • NULLS IGNORE: NULL values are skipped during the equivalent comparison. If all index columns are NULL, NULL values can be inserted repeatedly. If some index columns are NULL, data can be inserted only when non-null values are different.

  The following table lists the behaviors of the three processing modes.

  Table 3 Processing of NULL values in index columns in unique indexes

  Constraint	All Index Columns Are NULL	Some Index Columns Are NULL.
  NULLS DISTINCT	Can be inserted repeatedly.	Can be inserted repeatedly.
  NULLS NOT DISTINCT	Cannot be inserted repeatedly.	Cannot be inserted if the non-null values are equal. Can be inserted if the non-null values are not equal.
  NULLS IGNORE	Can be inserted repeatedly.	Cannot be inserted if the non-null values are equal. Can be inserted if the non-null values are not equal.

  

  If DISTRIBUTE BY REPLICATION is not specified, the column table that contains only unique values must contain distribution columns.

• PRIMARY KEY index_parameters

  PRIMARY KEY ( column_name [, ... ] ) index_parameters

  Specifies the primary key constraint specifies that a column or columns of a table can contain only unique (non-duplicate) and non-null values.

  Only one primary key can be specified for a table.

  

  If DISTRIBUTE BY REPLICATION is not specified, the column set with a primary key constraint must contain distributed columns.

• REFERENCES reftable [ ( refcolumn ) ]

  The foreign key constraint requires that the group consisting of one or more columns in the new table should contain and match only the referenced column values in the referenced table. The referenced column should be the only column or primary key in the referenced table.

  

  GaussDB(DWS) does not check foreign key constraints. When using foreign key constraints, you need to use the check_foreign_key_constraint function to check whether the data in the foreign key table meets the foreign key constraints.

• DEFERRABLE | NOT DEFERRABLE

  Controls whether the constraint can be deferred. A constraint that is not deferrable will be checked immediately after every command. Checking of constraints that are deferrable can be postponed until the end of the transaction using the SET CONSTRAINTS command. NOT DEFERRABLE is the default value. Currently, only UNIQUE and PRIMARY KEY constraints of row-store tables accept this clause. All the other constraints are not deferrable.

• PARTIAL CLUSTER KEY

  Specifies a partial cluster key for storage. When importing data to a column-store table, you can perform local data sorting by specified columns (single or multiple).

• INITIALLY IMMEDIATE | INITIALLY DEFERRED

  If a constraint is deferrable, this clause specifies the default time to check the constraint.

  • If the constraint is INITIALLY IMMEDIATE (default value), it is checked after each statement.
  • If the constraint is INITIALLY DEFERRED, it is checked only at the end of the transaction.

  The constraint check time can be altered using the SET CONSTRAINTS command.

Examples

Create a V3 table with storage and compute decoupled (supported only in the storage-compute decoupling 3.0 version).

1
2
3
4
5
6
7
8

CREATE TABLE  public.t1 
( 
id integer not null,  
data integer, 
age integer 
)  
WITH (ORIENTATION =COLUMN, COLVERSION =3.0) 
DISTRIBUTE BY ROUNDROBIN;

Specify the cache policy when creating a table (supported only in clusters of the storage-compute decoupling 3.0 version).

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17

CREATE TABLE Sports
(
    N_NATIONKEY  INT NOT NULL
  , N_NAME       CHAR(25) NOT NULL
  , N_REGIONKEY  INT NOT NULL
  , N_COMMENT    VARCHAR(152)
) WITH (orientation = column, colversion = 3.0, cache_policy = 'HPL: Balls, Basketball')
tablespace cu_obs_tbs
DISTRIBUTE BY ROUNDROBIN
partition by list(N_NAME)
(
  partition Balls values ('Basketball', 'football', 'badminton'),
  partition Athletics values ('High jump', 'long jump', 'javelin'),
  partition Water_Sports values ('Surfing', 'diving', 'swimming'),
  partition Shooting values ('air guns', 'Rifles', 'archery'),
  partition rest values (DEFAULT)
);

Define a unique column constraint for the table:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10

CREATE TABLE CUSTOMER
(    
    C_CUSTKEY     BIGINT NOT NULL CONSTRAINT C_CUSTKEY_pk PRIMARY KEY  , 
    C_NAME        VARCHAR(25)  , 
    C_ADDRESS     VARCHAR(40)  , 
    C_NATIONKEY   INT          , 
    C_PHONE       CHAR(15)     , 
    C_ACCTBAL     DECIMAL(15,2)  
)
DISTRIBUTE BY HASH(C_CUSTKEY);

Define a primary key table constraint for the table. You can define a primary key table constraint on one or more columns of a table:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11

CREATE TABLE CUSTOMER
(    
    C_CUSTKEY     BIGINT       , 
    C_NAME        VARCHAR(25)  , 
    C_ADDRESS     VARCHAR(40)  , 
    C_NATIONKEY   INT          , 
    C_PHONE       CHAR(15)     , 
    C_ACCTBAL     DECIMAL(15,2)   , 
    CONSTRAINT C_CUSTKEY_KEY PRIMARY KEY(C_CUSTKEY,C_NAME)
)
DISTRIBUTE BY HASH(C_CUSTKEY,C_NAME);

Define the CHECK column constraint:

1
2
3
4
5
6
7
8

CREATE TABLE CUSTOMER
(    
    C_CUSTKEY     BIGINT NOT NULL CONSTRAINT C_CUSTKEY_pk PRIMARY KEY  , 
    C_NAME        VARCHAR(25)  , 
    C_ADDRESS     VARCHAR(40)  , 
    C_NATIONKEY   INT NOT NULL  CHECK (C_NATIONKEY > 0)  
)
DISTRIBUTE BY HASH(C_CUSTKEY);

Define the CHECK table constraint:

CREATE TABLE CUSTOMER
(    
    C_CUSTKEY     BIGINT NOT NULL CONSTRAINT C_CUSTKEY_pk PRIMARY KEY  , 
    C_NAME        VARCHAR(25)      , 
    C_ADDRESS     VARCHAR(40)      , 
    C_NATIONKEY   INT              , 
    CONSTRAINT C_CUSTKEY_KEY2 CHECK(C_CUSTKEY > 0 AND C_NAME <> '')
)
DISTRIBUTE BY HASH(C_CUSTKEY);

Create a column-store table and specify the storage format and compression mode:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11

CREATE TABLE customer_address
(
    ca_address_sk       INTEGER                  NOT NULL   ,
    ca_address_id       CHARACTER(16)            NOT NULL   ,
    ca_street_number    CHARACTER(10)                       ,
    ca_street_name      CHARACTER varying(60)               ,
    ca_street_type      CHARACTER(15)                       ,
    ca_suite_number     CHARACTER(10)                    
)
WITH (ORIENTATION = COLUMN, COMPRESSION=HIGH,COLVERSION=2.0)
DISTRIBUTE BY HASH (ca_address_sk);

Use DEFAULT to declare a default value for column W_STATE:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10

CREATE TABLE warehouse_t
(
    W_WAREHOUSE_SK            INTEGER                NOT NULL,
    W_WAREHOUSE_ID            CHAR(16)               NOT NULL,
    W_WAREHOUSE_NAME          VARCHAR(20)   UNIQUE DEFERRABLE,
    W_WAREHOUSE_SQ_FT         INTEGER                        ,
    W_COUNTY                  VARCHAR(30)                    ,
    W_STATE                   CHAR(2)            DEFAULT 'GA',
    W_ZIP                     CHAR(10)                       
);

Create the CUSTOMER_bk table in LIKE mode:

1

CREATE TABLE CUSTOMER_bk (LIKE CUSTOMER INCLUDING ALL);

Helpful Links

ALTER TABLE, 12.101-RENAME TABLE, and DROP TABLE