If specified, the table is created as a temporary table. Temporary tables are automatically dropped at the end of a session, or optionally at the end of the current transaction (see ON COMMIT below). The default search_path includes the temporary schema first and so identically named existing permanent tables are not chosen for new plans while the temporary table exists, unless they are referenced with schema-qualified names. Any indexes created on a temporary table are automatically temporary as well.

Unlike for permanent tables, we extend the physical file for a temporary table only when the table does not fit into temp_buffers cache anymore. Thus, the total and on-disk size of a temporary table can differ. Blocks are not allocated in advance, so we may run out of free space on disk when trying to evict a temporary buffer from cache. If the available disk space is smaller than temp_buffers size, an error occurs. This case is quite unusual, though.

The autovacuum daemon cannot access and therefore cannot vacuum or analyze temporary tables. For this reason, appropriate vacuum and analyze operations should be performed via session SQL commands. For example, if a temporary table is going to be used in complex queries, it is wise to run ANALYZE on the temporary table after it is populated.

Optionally, GLOBAL or LOCAL can be written before TEMPORARY or TEMP. This presently makes no difference in Postgres Pro and is deprecated; see Compatibility below.

If specified, the table is created as an unlogged table. Data written to unlogged tables is not written to the write-ahead log (see Chapter 29), which makes them considerably faster than ordinary tables. However, they are not crash-safe: an unlogged table is automatically truncated after a crash or unclean shutdown. The contents of an unlogged table are also not replicated to standby servers. Any indexes created on an unlogged table are automatically unlogged as well.

If this is specified, any sequences created together with the unlogged table (for identity or serial columns) are also created as unlogged.

If specified, the table is created as read-only. No data can be modified or added to constant tables, and they are not processed by autovacuum. Constant tables cannot be changed to read-write mode, so there is no much sense to create them as constant; use CREATE TABLE AS or ALTER TABLE instead.

Do not throw an error if a relation with the same name already exists. A notice is issued in this case. Note that there is no guarantee that the existing relation is anything like the one that would have been created.

The name (optionally schema-qualified) of the table to be created.

Creates a typed table, which takes its structure from the specified composite type (name optionally schema-qualified). A typed table is tied to its type; for example the table will be dropped if the type is dropped (with DROP TYPE ... CASCADE).

When a typed table is created, then the data types of the columns are determined by the underlying composite type and are not specified by the CREATE TABLE command. But the CREATE TABLE command can add defaults and constraints to the table and can specify storage parameters.

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

The data type of the column. This can include array specifiers. For more information on the data types supported by Postgres Pro, refer to Chapter 8.

The COLLATE clause assigns a collation to the column (which must be of a collatable data type). If not specified, the column data type's default collation is used.

This form sets the storage mode for the column. This controls whether this column is held inline or in a secondary TOAST table, and whether the data should be compressed or not. PLAIN must be used for fixed-length values such as integer and is inline, uncompressed. MAIN is for inline, compressible data. EXTERNAL is for external, uncompressed data, and EXTENDED is for external, compressed data. Writing DEFAULT sets the storage mode to the default mode for the column's data type. EXTENDED is the default for most data types that support non-PLAIN storage. Use of EXTERNAL will make substring operations on very large text and bytea values run faster, at the penalty of increased storage space. See Section 73.2 for more information.

The COMPRESSION clause sets the compression method for the column. Compression is supported only for variable-width data types, and is used only when the column's storage mode is main or extended. (See ALTER TABLE for information on column storage modes.) Setting this property for a partitioned table has no direct effect, because such tables have no storage of their own, but the configured value will be inherited by newly-created partitions. The supported compression methods are pglz and lz4. (lz4 is available only if --with-lz4 was used when building Postgres Pro.) In addition, compression_method can be default to explicitly specify the default behavior, which is to consult the default_toast_compression setting at the time of data insertion to determine the method to use.

The optional INHERITS clause specifies a list of tables from which the new table automatically inherits all columns. Parent tables can be plain tables or foreign tables.

Use of INHERITS creates a persistent relationship between the new child table and its parent table(s). Schema modifications to the parent(s) normally propagate to children as well, and by default the data of the child table is included in scans of the parent(s).

If the same column name exists in more than one parent table, an error is reported unless the data types of the columns match in each of the parent tables. If there is no conflict, then the duplicate columns are merged to form a single column in the new table. If the column name list of the new table contains a column name that is also inherited, the data type must likewise match the inherited column(s), and the column definitions are merged into one. If the new table explicitly specifies a default value for the column, this default overrides any defaults from inherited declarations of the column. Otherwise, any parents that specify default values for the column must all specify the same default, or an error will be reported.

CHECK constraints are merged in essentially the same way as columns: if multiple parent tables and/or the new table definition contain identically-named CHECK constraints, these constraints must all have the same check expression, or an error will be reported. Constraints having the same name and expression will be merged into one copy. A constraint marked NO INHERIT in a parent will not be considered. Notice that an unnamed CHECK constraint in the new table will never be merged, since a unique name will always be chosen for it.

Column STORAGE settings are also copied from parent tables.

If a column in the parent table is an identity column, that property is not inherited. A column in the child table can be declared identity column if desired.

The optional PARTITION BY clause specifies a strategy of partitioning the table. The table thus created is called a partitioned table. The parenthesized list of columns or expressions forms the partition key for the table. When using range or hash partitioning, the partition key can include multiple columns or expressions (up to 32, but this limit can be altered when building Postgres Pro), but for list partitioning, the partition key must consist of a single column or expression.

Range and list partitioning require a btree operator class, while hash partitioning requires a hash operator class. If no operator class is specified explicitly, the default operator class of the appropriate type will be used; if no default operator class exists, an error will be raised. When hash partitioning is used, the operator class used must implement support function 2 (see Section 40.16.3 for details).

A partitioned table is divided into sub-tables (called partitions), which are created using separate CREATE TABLE commands. The partitioned table is itself empty. A data row inserted into the table is routed to a partition based on the value of columns or expressions in the partition key. If no existing partition matches the values in the new row, an error will be reported, unless you use pg_pathman for partitioning as explained below.

Partitioned tables do not support EXCLUDE constraints; however, you can define these constraints on individual partitions.

See Section 5.11 for more discussion on table partitioning.

Optionally, you can choose the backend that performs partitioning by specifying the USING partition_backend clause, where partition_backend can be one of the following:

If you omit this clause, the backend for partitioning is defined by the partition_backend parameter. The default value is internal, which enables declarative partitioning implementation provided by Postgres Pro core.

Do not confuse USING partition_backend with the USING method clause, which cannot be used when creating partitioned tables.

Starting from Postgres Pro 12, using pg_pathman is not recommended.

When using pg_pathman for partitioning, you must provide the pg_pathman_partitioning_clause. Depending on the chosen partitioning strategy, this clause can be one of the following:

Creates the table as a partition of the specified parent table. The table can be created either as a partition for specific values using FOR VALUES or as a default partition using DEFAULT. Any indexes, constraints and user-defined row-level triggers that exist in the parent table are cloned on the new partition. This clause is not supported by pg_pathman.

The partition_bound_spec must correspond to the partitioning method and partition key of the parent table, and must not overlap with any existing partition of that parent. The form with IN is used for list partitioning, the form with FROM and TO is used for range partitioning, and the form with WITH is used for hash partitioning.

partition_bound_expr is any variable-free expression (subqueries, window functions, aggregate functions, and set-returning functions are not allowed). Its data type must match the data type of the corresponding partition key column. The expression is evaluated once at table creation time, so it can even contain volatile expressions such as CURRENT_TIMESTAMP.

When creating a list partition, NULL can be specified to signify that the partition allows the partition key column to be null. However, there cannot be more than one such list partition for a given parent table. NULL cannot be specified for range partitions.

When creating a range partition, the lower bound specified with FROM is an inclusive bound, whereas the upper bound specified with TO is an exclusive bound. That is, the values specified in the FROM list are valid values of the corresponding partition key columns for this partition, whereas those in the TO list are not. Note that this statement must be understood according to the rules of row-wise comparison (Section 9.24.5). For example, given PARTITION BY RANGE (x,y), a partition bound FROM (1, 2) TO (3, 4) allows x=1 with any y>=2, x=2 with any non-null y, and x=3 with any y<4.

The special values MINVALUE and MAXVALUE may be used when creating a range partition to indicate that there is no lower or upper bound on the column's value. For example, a partition defined using FROM (MINVALUE) TO (10) allows any values less than 10, and a partition defined using FROM (10) TO (MAXVALUE) allows any values greater than or equal to 10.

When creating a range partition involving more than one column, it can also make sense to use MAXVALUE as part of the lower bound, and MINVALUE as part of the upper bound. For example, a partition defined using FROM (0, MAXVALUE) TO (10, MAXVALUE) allows any rows where the first partition key column is greater than 0 and less than or equal to 10. Similarly, a partition defined using FROM ('a', MINVALUE) TO ('b', MINVALUE) allows any rows where the first partition key column starts with "a".

Note that if MINVALUE or MAXVALUE is used for one column of a partitioning bound, the same value must be used for all subsequent columns. For example, (10, MINVALUE, 0) is not a valid bound; you should write (10, MINVALUE, MINVALUE).

Also note that some element types, such as timestamp, have a notion of "infinity", which is just another value that can be stored. This is different from MINVALUE and MAXVALUE, which are not real values that can be stored, but rather they are ways of saying that the value is unbounded. MAXVALUE can be thought of as being greater than any other value, including "infinity" and MINVALUE as being less than any other value, including "minus infinity". Thus the range FROM ('infinity') TO (MAXVALUE) is not an empty range; it allows precisely one value to be stored — "infinity".

If DEFAULT is specified, the table will be created as the default partition of the parent table. This option is not available for hash-partitioned tables. A partition key value not fitting into any other partition of the given parent will be routed to the default partition.

When a table has an existing DEFAULT partition and a new partition is added to it, the default partition must be scanned to verify that it does not contain any rows which properly belong in the new partition. If the default partition contains a large number of rows, this may be slow. The scan will be skipped if the default partition is a foreign table or if it has a constraint which proves that it cannot contain rows which should be placed in the new partition.

When creating a hash partition, a modulus and remainder must be specified. The modulus must be a positive integer, and the remainder must be a non-negative integer less than the modulus. Typically, when initially setting up a hash-partitioned table, you should choose a modulus equal to the number of partitions and assign every table the same modulus and a different remainder (see examples, below). However, it is not required that every partition have the same modulus, only that every modulus which occurs among the partitions of a hash-partitioned table is a factor of the next larger modulus. This allows the number of partitions to be increased incrementally without needing to move all the data at once. For example, suppose you have a hash-partitioned table with 8 partitions, each of which has modulus 8, but find it necessary to increase the number of partitions to 16. You can detach one of the modulus-8 partitions, create two new modulus-16 partitions covering the same portion of the key space (one with a remainder equal to the remainder of the detached partition, and the other with a remainder equal to that value plus 8), and repopulate them with data. You can then repeat this -- perhaps at a later time -- for each modulus-8 partition until none remain. While this may still involve a large amount of data movement at each step, it is still better than having to create a whole new table and move all the data at once.

A partition must have the same column names and types as the partitioned table to which it belongs. Modifications to the column names or types of a partitioned table will automatically propagate to all partitions. CHECK constraints will be inherited automatically by every partition, but an individual partition may specify additional CHECK constraints; additional constraints with the same name and condition as in the parent will be merged with the parent constraint. Defaults may be specified separately for each partition. But note that a partition's default value is not applied when inserting a tuple through a partitioned table.

Rows inserted into a partitioned table will be automatically routed to the correct partition. If no suitable partition exists, an error will occur.

Operations such as TRUNCATE which normally affect a table and all of its inheritance children will cascade to all partitions, but may also be performed on an individual partition.

Note that creating a partition using PARTITION OF requires taking an ACCESS EXCLUSIVE lock on the parent partitioned table. Likewise, dropping a partition with DROP TABLE requires taking an ACCESS EXCLUSIVE lock on the parent table. It is possible to use ALTER TABLE ATTACH/DETACH PARTITION to perform these operations with a weaker lock, thus reducing interference with concurrent operations on the partitioned table.

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

Unlike INHERITS, the new table and original table are completely decoupled after creation is complete. Changes to the original 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 original table.

Also unlike INHERITS, columns and constraints copied by LIKE are not merged with similarly named columns and constraints. If the same name is specified explicitly or in another LIKE clause, an error is signaled.

The optional like_option clauses specify which additional properties of the original table to copy. Specifying INCLUDING copies the property, specifying EXCLUDING omits the property. EXCLUDING is the default. If multiple specifications are made for the same kind of object, the last one is used. The available options are:

The LIKE clause can also be used to copy column definitions from views, foreign tables, or composite types. Inapplicable options (e.g., INCLUDING INDEXES from a view) are ignored.

An optional name for a column or table constraint. If the constraint is violated, the constraint name is present in error messages, so constraint names like col must be positive can be used to communicate helpful constraint information to client applications. (Double-quotes are needed to specify constraint names that contain spaces.) If a constraint name is not specified, the system generates a name.

The column is not allowed to contain null values.

The column is allowed to contain null values. This is the default.

This clause is only provided for compatibility with non-standard SQL databases. Its use is discouraged in new applications.

The CHECK clause 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. Should any row of an insert or update operation produce 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 that column's value only, while an expression appearing in a table constraint can reference multiple columns.

Currently, CHECK expressions cannot contain subqueries nor refer to variables other than columns of the current row (see Section 5.4.1). The system column tableoid may be referenced, but not any other system column.

A constraint marked with NO INHERIT will not propagate to child tables.

When a table has multiple CHECK constraints, they will be tested for each row in alphabetical order by name, after checking NOT NULL constraints. (PostgreSQL versions before 9.5 did not honor any particular firing order for CHECK constraints.)

The DEFAULT clause assigns a default data value for the column whose column definition it appears within. The value is any variable-free expression (in particular, cross-references to other columns in the current table are not allowed). Subqueries are not allowed either. 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 for a column, then the default is null.

This clause creates the column as a generated column. The column cannot be written to, and when read the result of the specified expression will be returned.

The keyword STORED is required to signify that the column will be computed on write and will be stored on disk.

The generation expression can refer to other columns in the table, but not other generated columns. Any functions and operators used must be immutable. References to other tables are not allowed.

This clause creates the column as an identity column. It will have an implicit sequence attached to it and in newly-inserted rows the column will automatically have values from the sequence assigned to it. Such a column is implicitly NOT NULL.

The clauses ALWAYS and BY DEFAULT determine how explicitly user-specified values are handled in INSERT and UPDATE commands.

In an INSERT command, if ALWAYS is selected, a user-specified value is only accepted if the INSERT statement specifies OVERRIDING SYSTEM VALUE. If BY DEFAULT is selected, then the user-specified value takes precedence. See INSERT for details. (In the COPY command, user-specified values are always used regardless of this setting.)

In an UPDATE command, if ALWAYS is selected, any update of the column to any value other than DEFAULT will be rejected. If BY DEFAULT is selected, the column can be updated normally. (There is no OVERRIDING clause for the UPDATE command.)

The optional sequence_options clause can be used to override the parameters of the sequence. The available options include those shown for CREATE SEQUENCE, plus SEQUENCE NAME name, LOGGED, and UNLOGGED, which allow selection of the name and persistence level of the sequence. Without SEQUENCE NAME, the system chooses an unused name for the sequence. Without LOGGED or UNLOGGED, the sequence will have the same persistence level as the table.

The UNIQUE constraint specifies that a group of one or more columns of a table can contain only unique values. The behavior of a unique table constraint is the same as that of a unique column constraint, with the additional capability to span multiple columns. The constraint therefore enforces that any two rows must differ in at least one of these columns.

For the purpose of a unique constraint, null values are not considered equal, unless NULLS NOT DISTINCT is specified.

Each unique constraint should name a set of columns that is different from the set of columns named by any other unique or primary key constraint defined for the table. (Otherwise, redundant unique constraints will be discarded.)

When establishing a unique constraint for a multi-level partition hierarchy, all the columns in the partition key of the target partitioned table, as well as those of all its descendant partitioned tables, must be included in the constraint definition.

Adding a unique constraint will automatically create a unique btree index on the column or group of columns used in the constraint.

The optional INCLUDE clause adds to that index one or more columns that are simply “payload”: uniqueness is not enforced on them, and the index cannot be searched on the basis of those columns. However they can be retrieved by an index-only scan. Note that although the constraint is not enforced on included columns, it still depends on them. Consequently, some operations on such columns (e.g., DROP COLUMN) can cause cascaded constraint and index deletion.

The PRIMARY KEY constraint specifies that a column or columns of a table can contain only unique (non-duplicate), nonnull values. Only one primary key can be specified for a table, whether as a column constraint or a table constraint.

The primary key constraint should name a set of columns that is different from the set of columns named by any unique constraint defined for the same table. (Otherwise, the unique constraint is redundant and will be discarded.)

PRIMARY KEY enforces the same data constraints as a combination of UNIQUE and NOT NULL. However, identifying a set of columns as the primary key also provides metadata about the design of the schema, since a primary key implies that other tables can rely on this set of columns as a unique identifier for rows.

When placed on a partitioned table, PRIMARY KEY constraints share the restrictions previously described for UNIQUE constraints.

Adding a PRIMARY KEY constraint will automatically create a unique btree index on the column or group of columns used in the constraint.

The optional INCLUDE clause adds to that index one or more columns that are simply “payload”: uniqueness is not enforced on them, and the index cannot be searched on the basis of those columns. However they can be retrieved by an index-only scan. Note that although the constraint is not enforced on included columns, it still depends on them. Consequently, some operations on such columns (e.g., DROP COLUMN) can cause cascaded constraint and index deletion.

The EXCLUDE clause defines an exclusion constraint, which guarantees that if any two rows are compared on the specified column(s) or expression(s) using the specified operator(s), not all of these comparisons will return TRUE. If all of the specified operators test for equality, this is equivalent to a UNIQUE constraint, although an ordinary unique constraint will be faster. However, exclusion constraints can specify constraints that are more general than simple equality. For example, you can specify a constraint that no two rows in the table contain overlapping circles (see Section 8.8) by using the && operator. The operator(s) are required to be commutative.

Exclusion constraints are implemented using an index, so each specified operator must be associated with an appropriate operator class (see Section 11.10) for the index access method index_method. Each exclude_element defines a column of the index, so it can optionally specify a collation, an operator class, operator class parameters, and/or ordering options; these are described fully under CREATE INDEX.

The access method must support amgettuple (see Chapter 64); at present this means GIN cannot be used. Although it's allowed, there is little point in using B-tree or hash indexes with an exclusion constraint, because this does nothing that an ordinary unique constraint doesn't do better. So in practice the access method will always be GiST or SP-GiST.

The predicate allows you to specify an exclusion constraint on a subset of the table; internally this creates a partial index. Note that parentheses are required around the predicate.

These clauses specify a foreign key constraint, which requires that a group of one or more columns of the new table must only contain values that match values in the referenced column(s) of some row of the referenced table. If the refcolumn list is omitted, the primary key of the reftable is used. Otherwise, the refcolumn list must refer to the columns of a non-deferrable unique or primary key constraint or be the columns of a non-partial unique index. The user must have REFERENCES permission on the referenced table (either the whole table, or the specific referenced columns). The addition of a foreign key constraint requires a SHARE ROW EXCLUSIVE lock on the referenced table. Note that foreign key constraints cannot be defined between temporary tables and permanent tables.

A value inserted into the referencing column(s) is matched against the values of the referenced table and referenced columns using the given match type. There are three match types: MATCH FULL, MATCH PARTIAL, and MATCH SIMPLE (which is the default). MATCH FULL will not allow one column of a multicolumn foreign key to be null unless all foreign key columns are null; if they are all null, the row is not required to have a match in the referenced table. MATCH SIMPLE allows any of the foreign key columns to be null; if any of them are null, the row is not required to have a match in the referenced table. MATCH PARTIAL is not yet implemented. (Of course, NOT NULL constraints can be applied to the referencing column(s) to prevent these cases from arising.)

In addition, when the data in the referenced columns is changed, certain actions are performed on the data in this table's columns. The ON DELETE clause specifies the action to perform when a referenced row in the referenced table is being deleted. Likewise, the ON UPDATE clause specifies the action to perform when a referenced column in the referenced table is being updated to a new value. If the row is updated, but the referenced column is not actually changed, no action is done. Referential actions other than the NO ACTION check cannot be deferred, even if the constraint is declared deferrable. There are the following possible actions for each clause:

If the referenced column(s) are changed frequently, it might be wise to add an index to the referencing column(s) so that referential actions associated with the foreign key constraint can be performed more efficiently.

This 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. Currently, only UNIQUE, PRIMARY KEY, EXCLUDE, and REFERENCES (foreign key) constraints accept this clause. NOT NULL and CHECK constraints are not deferrable. Note that deferrable constraints cannot be used as conflict arbitrators in an INSERT statement that includes an ON CONFLICT DO UPDATE clause.

If a constraint is deferrable, this clause specifies the default time to check the constraint. If the constraint is INITIALLY IMMEDIATE, it is checked after each statement. This is the default. If the constraint is INITIALLY DEFERRED, it is checked only at the end of the transaction. The constraint check time can be altered with the SET CONSTRAINTS command.

This optional clause specifies the table access method to use to store the contents for the new table; the method needs be an access method of type TABLE. See Chapter 63 for more information. If this option is not specified, the default table access method is chosen for the new table. See default_table_access_method for more information.

This clause specifies optional storage parameters for a table or index; see Storage Parameters below for more information. For backward-compatibility the WITH clause for a table can also include OIDS=FALSE to specify that rows of the new table should not contain OIDs (object identifiers), OIDS=TRUE is not supported anymore.

This is backward-compatible syntax for declaring a table WITHOUT OIDS, creating a table WITH OIDS is not supported anymore.

The behavior of temporary tables at the end of a transaction block can be controlled using ON COMMIT. The three options are:

The tablespace_name is the name of the tablespace in which the new table is to be created. If not specified, default_tablespace is consulted, or temp_tablespaces if the table is temporary. For partitioned tables, since no storage is required for the table itself, the tablespace specified overrides default_tablespace as the default tablespace to use for any newly created partitions when no other tablespace is explicitly specified.

This clause allows selection of the tablespace in which the index associated with a UNIQUE, PRIMARY KEY, or EXCLUDE constraint will be created. If not specified, default_tablespace is consulted, or temp_tablespaces if the table is temporary.