Skip to content

Latest commit

 

History

History
398 lines (309 loc) · 18.8 KB

09_datatypes.md

File metadata and controls

398 lines (309 loc) · 18.8 KB
Raw

Trusted Language Base types

PostgreSQL provides CREATE TYPE command to register a new base type (scalar type) for use in the current database. A base type allows you to customized how the data is stored internally and how to convert it from/to an external textual representation. The support functions input_function and output_function are required. The input_function converts the type's external textual representation to the internal representation. output_function performs the reverse transformation. Generally they have to be coded in C or another low-level language. Also, you must be a superuser to create a new base type.

pg_tle enables you to build Trusted Language base data types through a set of SQL APIs and use a trusted language to define the support functions. This section of the documentation describes the available APIs and provides examples for how to use them to create your own base data types.

Scope

The pg_tle base data type is scoped to an individual PostgreSQL database (e.g. an object created using CREATE DATABASE).

Functions

pgtle.create_shell_type(typenamespace regnamespace, typename name)

create_shell_type provides a way to create a shell type, which is simply a placeholder for a type to be defined later. This is similar to shell type form of CREATE TYPE.

Role

pgtle_admin

Arguments

  • typenamespace: The namespace where the new shell type will be created.
  • typename: The name of the new type.

Example

SELECT pgtle.create_shell_type('public', 'test_citext');

pgtle.create_shell_type_if_not_exists(typenamespace regnamespace, typename name)

create_shell_type_if_not_exists provides a way to create a shell type, which is simply a placeholder for a type to be defined later. It returns true if the type is created, otherwise it returns false if the type already exists.

Role

pgtle_admin

Arguments

  • typenamespace: The namespace where the new shell type will be created.
  • typename: The name of the new type.

Example

SELECT pgtle.create_shell_type_if_not_exists('public', 'test_citext');

pgtle.create_base_type(typenamespace regnamespace, typename name, infunc regprocedure, outfunc regprocedure, internallength int4, alignment text default 'int4', storage text default 'plain')

create_base_type provides a way to create a new base data type. The type must be a shell type previously defined by create_shell_type. This is similar to base type form of CREATE TYPE. Internally, a base data type created by pg_tle is stored as bytea, it can be cast to bytea explicitly after creation.

Role

pgtle_admin

Arguments

  • typenamespace: The namespace where the base data type will be created.
  • typename: The name of the base data type.
  • infunc: The name of a previously defined function to convert the type's external textual representation to the internal representation (bytea). The function must take one argument of type text and return bytea. The function must also be declared as IMMUTABLE and STRICT. It will not be called with a NULL parameter.
  • outfunc: The name of a previously defined function to convert the type's internal binary representation (bytea) to the external textual representation. The function must take one argument of type bytea and return text. The function must also be declared as IMMUTABLE and STRICT. It will not be called with a NULL parameter.
  • internallength: Total length of the base data type as bytes. Base data types can be fixed-length, in which case internallength is a positive integer, or variable-length, in which case internallength is -1.
  • alignment: The optional alignment parameter specifies the storage alignment requirement of the data type. Allowed values are 'char', 'int2', 'int4', 'double'. The allowed values equate to alignment on 1, 2, 4, or 8 byte boundaries. The default is 'int4', alignment on 4 byte boundaries. A variable-length type (e.g., text) must have a byte-alignment of at least 4 due to the size of its header.
  • storage: The optional storage parameter allows selection of storage strategies for variable-length data types. Allowed values are 'plain', 'external', 'extended', 'main'. Only 'plain' is allowed for fixed-length types. The default is 'plain'. See PostgreSQL doc on CREATE TYPE and TOAST for more details.

Example

-- Create a variable-length data type
SELECT pgtle.create_base_type('public', 'test_citext', 'test_citext_in(text)'::regprocedure, 'test_citext_out(bytea)'::regprocedure, -1);
-- Create a fixed-length (2 bytes) data type
SELECT pgtle.create_base_type('public', 'test_int2', 'test_int2_in(text)'::regprocedure, 'test_int2_out(bytea)'::regprocedure, 2);
-- Create a TOASTable variable-length data type 
SELECT pgtle.create_base_type('public', 'test_citext', 'test_citext_in(text)'::regprocedure, 'test_citext_out(bytea)'::regprocedure, -1, storage => 'extended');

pgtle.create_base_type_if_not_exists(typenamespace regnamespace, typename name, infunc regprocedure, outfunc regprocedure, internallength int4, alignment text default 'int4', storage text default 'plain')

create_base_type_if_not_exists create_base_type provides a way to create a new base data type. It returns true if the type is created, otherwise it returns false if the type already exists. The type must be a shell type previously defined by create_shell_type. This is similar to base type form of CREATE TYPE. Internally, a base data type created by pg_tle is stored as bytea, it can be cast to bytea explicitly after creation.

Role

pgtle_admin

Arguments

  • typenamespace: The namespace where the base data type will be created.
  • typename: The name of the base data type.
  • infunc: The name of a previously defined function to convert the type's external textual representation to the internal representation (bytea). The function must take one argument of type text and return bytea. The function must also be declared as IMMUTABLE and STRICT. It will not be called with a NULL parameter.
  • outfunc: The name of a previously defined function to convert the type's internal binary representation (bytea) to the external textual representation. The function must take one argument of type bytea and return text. The function must also be declared as IMMUTABLE and STRICT. It will not be called with a NULL parameter.
  • internallength: Total length of the base data type as bytes. Base data types can be fixed-length, in which case internallength is a positive integer, or variable-length, in which case internallength is -1.
  • alignment: The optional alignment parameter specifies the storage alignment requirement of the data type. Allowed values are 'char', 'int2', 'int4', 'double'. The allowed values equate to alignment on 1, 2, 4, or 8 byte boundaries. The default is 'int4', alignment on 4 byte boundaries. A variable-length type (e.g., text) must have a byte-alignment of at least 4 due to the size of its header.
  • storage: The optional storage parameter allows selection of storage strategies for variable-length data types. Allowed values are 'plain', 'external', 'extended', 'main'. Only 'plain' is allowed for fixed-length types. The default is 'plain'. See PostgreSQL doc on CREATE TYPE and TOAST for more details.

Example

-- Create a variable-length data type
SELECT pgtle.create_base_type_if_not_exists('public', 'test_citext', 'test_citext_in(text)'::regprocedure, 'test_citext_out(bytea)'::regprocedure, -1);
-- Create a fixed-length (2 bytes) data type
SELECT pgtle.create_base_type_if_not_exists('public', 'test_int2', 'test_int2_in(text)'::regprocedure, 'test_int2_out(bytea)'::regprocedure, 2);
-- Create a TOASTable variable-length data type 
SELECT pgtle.create_base_type_if_not_exists('public', 'test_citext', 'test_citext_in(text)'::regprocedure, 'test_citext_out(bytea)'::regprocedure, -1, storage => 'extended');

pgtle.create_operator_func(typenamespace regnamespace, typename name, opfunc regprocedure)

create_operator_func provides a way to create an operator function on the base data type previously defined by create_base_type. This function takes an operator function which accepts one or two arguments of type bytea, and creates an overloaded version which accpets the base data type as the arguments instead. This is not required to create an operator function, but it can be helpful while working with certain languages such as plrust.

Role

pgtle_admin

Arguments

  • typenamespace: The namespace where the new operator function will be created. It must be the same namespace where the base data type is.
  • typename: The name of the base data type.
  • opfunc: The name of a previously defined operator function. The function must take exactly one or two arguments of type bytea.

Example

SELECT pgtle.create_operator_func('public', 'test_citext', 'public.test_citext_cmp(bytea, bytea)'::regprocedure);

pgtle.create_operator_func_if_not_exists(typenamespace regnamespace, typename name, opfunc regprocedure)

create_operator_func_if_not_exists provides a way to create an operator function on the base data type previously defined by create_base_type. It returns true if the operator function is created, otherwise it returns false if the operator function already exists. This function is not required to create an operator function, but it can be helpful while working with certain languages such as plrust.

Role

pgtle_admin

Arguments

  • typenamespace: The namespace where the new operator function will be created. It must be the same namespace where the base data type is.
  • typename: The name of the base data type.
  • opfunc: The name of a previously defined operator function. The function must be declared as taking one or two arguments of type bytea.

Example

SELECT pgtle.create_operator_func_if_not_exists('public', 'test_citext', 'public.test_citext_cmp(bytea, bytea)'::regprocedure);

Examples

The following examples demonstrate how to use pg_tle data type API functions to create a base data type. After running this example, a base data type called test_citext (case-insentive text) will be available for use in the current database.

Create shell type

First, use pgtle.create_shell_type to create a shell type in the target namespace (PUBLIC in this example).

SELECT pgtle.create_shell_type('public', 'test_citext');

Create I/O functions

Second, define the input/output function on the data type. Remember the input function takes one argument of type text and returns bytea; while the output function takes one argument of type bytea and returns text.

CREATE FUNCTION public.test_citext_in(input text) RETURNS bytea AS
$$
  SELECT pg_catalog.convert_to(input, 'UTF8');
$$ IMMUTABLE STRICT LANGUAGE sql;

CREATE FUNCTION public.test_citext_out(input bytea) RETURNS text AS
$$
  SELECT pg_catalog.convert_from(input, 'UTF8');
$$ IMMUTABLE STRICT LANGUAGE sql;

Create base type

We can now use pgtle.create_base_type to create the base data type using the I/O functions defined previously. Because we are defining a variable-length type, -1 is used as internallength.

SELECT pgtle.create_base_type('public', 'test_citext', 'test_citext_in(text)'::regprocedure, 'test_citext_out(bytea)'::regprocedure, -1);

After this step, we should be able to use the newly created data type test_citext.

CREATE TABLE public.test_dt(c1 test_citext PRIMARY KEY);
INSERT INTO test_dt VALUES ('SELECT'), ('INSERT'), ('UPDATE'), ('DELETE');
INSERT INTO test_dt VALUES ('select');

'SELECT' and 'select' are considered different values at this moment because we haven't defined related operators and operator class for the type.

Create operator functions

After the base data type is created, we can define operators and operator class if needed.

The following commands define a set of operator functions on type test_citext. We use an explicit cast from test_citext to bytea so that we can utilize existing binary functions in Postgresql.

CREATE FUNCTION public.test_citext_cmp(l test_citext, r test_citext)
RETURNS int AS
$$
BEGIN
  RETURN pg_catalog.bttextcmp(pg_catalog.lower(pg_catalog.convert_from(l::bytea, 'UTF8')), pg_catalog.lower(pg_catalog.convert_from(r::bytea, 'UTF8')));
END;
$$ IMMUTABLE STRICT LANGUAGE plpgsql;

CREATE FUNCTION public.test_citext_eq(l test_citext, r test_citext) 
RETURNS boolean AS
$$
BEGIN
  RETURN public.test_citext_cmp(l, r) == 0;
END;
$$ IMMUTABLE STRICT LANGUAGE plpgsql;

CREATE FUNCTION public.test_citext_ne(l test_citext, r test_citext) 
RETURNS boolean AS
$$
BEGIN
  RETURN public.test_citext_cmp(l, r) != 0;
END;
$$ IMMUTABLE STRICT LANGUAGE plpgsql;

CREATE FUNCTION public.test_citext_lt(l test_citext, r test_citext) 
RETURNS boolean AS
$$
BEGIN
  RETURN public.test_citext_cmp(l, r) < 0;
END;
$$ IMMUTABLE STRICT LANGUAGE plpgsql;

CREATE FUNCTION public.test_citext_le(l test_citext, r test_citext) 
RETURNS boolean AS
$$
BEGIN
  RETURN public.test_citext_cmp(l, r) <= 0;
END;
$$ IMMUTABLE STRICT LANGUAGE plpgsql;

CREATE FUNCTION public.test_citext_gt(l test_citext, r test_citext) 
RETURNS boolean AS
$$
BEGIN
  RETURN public.test_citext_cmp(l, r) > 0;
END;
$$ IMMUTABLE STRICT LANGUAGE plpgsql;

CREATE FUNCTION public.test_citext_ge(l test_citext, r test_citext) 
RETURNS boolean AS
$$
BEGIN
  RETURN public.test_citext_cmp(l, r) >= 0;
END;
$$ IMMUTABLE STRICT LANGUAGE plpgsql;

Alternative way to operator functions

Alternatively, we can use pgtle.create_operator_func to create operator functions without explicit cast. It can be really helpful in certains languages such as plrust where the newly created type is not available.

First, create the operator functions that take bytea as argument type:

CREATE FUNCTION public.test_citext_cmp(l bytea, r bytea) 
RETURNS int AS
$$
  SELECT pg_catalog.bttextcmp(pg_catalog.lower(pg_catalog.convert_from(l, 'UTF8')), pg_catalog.lower(pg_catalog.convert_from(r, 'UTF8')));
$$ IMMUTABLE STRICT LANGUAGE sql;

CREATE FUNCTION public.test_citext_lt(l bytea, r bytea) 
RETURNS boolean AS
$$
    SELECT public.test_citext_cmp(l, r) < 0;
$$ IMMUTABLE STRICT LANGUAGE sql;

CREATE FUNCTION public.test_citext_eq(l bytea, r bytea) 
RETURNS boolean AS
$$
    SELECT public.test_citext_cmp(l, r) = 0;
$$ IMMUTABLE STRICT LANGUAGE sql;

CREATE FUNCTION public.test_citext_le(l bytea, r bytea) 
RETURNS boolean AS
$$
    SELECT public.test_citext_cmp(l, r) <= 0;
$$ IMMUTABLE STRICT LANGUAGE sql;

CREATE FUNCTION public.test_citext_gt(l bytea, r bytea) 
RETURNS boolean AS
$$
    SELECT public.test_citext_cmp(l, r) > 0;
$$ IMMUTABLE STRICT LANGUAGE sql;

CREATE FUNCTION public.test_citext_ge(l bytea, r bytea) 
RETURNS boolean AS
$$
    SELECT public.test_citext_cmp(l, r) >= 0;
$$ IMMUTABLE STRICT LANGUAGE sql;

CREATE FUNCTION public.test_citext_ne(l bytea, r bytea) 
RETURNS boolean AS
$$
    SELECT public.test_citext_cmp(l, r) != 0;
$$ IMMUTABLE STRICT LANGUAGE sql;

then use pgtle.create_operator_func to convert the functions to operate on type test_citext:

SELECT pgtle.create_operator_func('public', 'test_citext', 'public.test_citext_cmp(bytea, bytea)'::regprocedure);
SELECT pgtle.create_operator_func('public', 'test_citext', 'public.test_citext_lt(bytea, bytea)'::regprocedure);
SELECT pgtle.create_operator_func('public', 'test_citext', 'public.test_citext_le(bytea, bytea)'::regprocedure);
SELECT pgtle.create_operator_func('public', 'test_citext', 'public.test_citext_eq(bytea, bytea)'::regprocedure);
SELECT pgtle.create_operator_func('public', 'test_citext', 'public.test_citext_ne(bytea, bytea)'::regprocedure);
SELECT pgtle.create_operator_func('public', 'test_citext', 'public.test_citext_gt(bytea, bytea)'::regprocedure);
SELECT pgtle.create_operator_func('public', 'test_citext', 'public.test_citext_ge(bytea, bytea)'::regprocedure);

Create operators / operator class

After defining operator functions, run the following commands to define operators on type test_citext.

CREATE OPERATOR < (
    LEFTARG = public.test_citext,
    RIGHTARG = public.test_citext,
    COMMUTATOR = >,
    NEGATOR = >=,
    RESTRICT = scalarltsel,
    JOIN = scalarltjoinsel,
    PROCEDURE = public.test_citext_lt
);

CREATE OPERATOR <= (
    LEFTARG = public.test_citext,
    RIGHTARG = public.test_citext,
    COMMUTATOR = >=,
    NEGATOR = >,
    RESTRICT = scalarltsel,
    JOIN = scalarltjoinsel,
    PROCEDURE = public.test_citext_le
);

CREATE OPERATOR = (
    LEFTARG = public.test_citext,
    RIGHTARG = public.test_citext,
    COMMUTATOR = =,
    NEGATOR = <>,
    RESTRICT = eqsel,
    JOIN = eqjoinsel,
    HASHES,
    MERGES,
    PROCEDURE = public.test_citext_eq
);

CREATE OPERATOR <> (
    LEFTARG = public.test_citext,
    RIGHTARG = public.test_citext,
    COMMUTATOR = <>,
    NEGATOR = =,
    RESTRICT = neqsel,
    JOIN = neqjoinsel,
    PROCEDURE = public.test_citext_ne
);

CREATE OPERATOR > (
    LEFTARG = public.test_citext,
    RIGHTARG = public.test_citext,
    COMMUTATOR = <,
    NEGATOR = <=,
    RESTRICT = scalargtsel,
    JOIN = scalargtjoinsel,
    PROCEDURE = public.test_citext_gt
);

CREATE OPERATOR >= (
    LEFTARG = public.test_citext,
    RIGHTARG = public.test_citext,
    COMMUTATOR = <=,
    NEGATOR = <,
    RESTRICT = scalargtsel,
    JOIN = scalargtjoinsel,
    PROCEDURE = public.test_citext_ge
);

Run following command to create operator class. Note superuser permission is required here. If you are using Amazon RDS, superuser permission is not required.

CREATE OPERATOR CLASS public.test_citext_ops
    DEFAULT FOR TYPE public.test_citext USING btree AS
        OPERATOR        1       < ,
        OPERATOR        2       <= ,
        OPERATOR        3       = ,
        OPERATOR        4       > ,
        OPERATOR        5       >= ,
        FUNCTION        1       public.test_citext_cmp(public.test_citext, public.test_citext);

Now we can run some simply queries to verify the operators are working as expected.

DROP TABLE IF EXISTS public.test_dt;
CREATE TABLE public.test_dt(c1 test_citext PRIMARY KEY);
INSERT INTO test_dt VALUES ('SELECT'), ('INSERT'), ('UPDATE'), ('DELETE');
INSERT INTO test_dt VALUES ('select');
ERROR:  duplicate key value violates unique constraint "test_dt_pkey"

'SELECT' and 'select' are considered equal according to the = operator and violates the unique constraint of PRIMARY KEY.