CREATE_OPERATOR_CLASS(7)
๐ CREATE OPERATOR CLASS(7) ๐
okay bestie, welcome to CREATE_OPERATOR_CLASS โ yeah it's from the PostgreSQL 18.6 Documentation, and yeah it's a lot, but we're gonna get through it together, hand in hand, no cap. ๐ซถ
โจ NAME โจ
CREATE_OPERATOR_CLASS โ define a new operator class ๐
๐ธ SYNOPSIS ๐ธ
CREATE OPERATOR CLASS name [ DEFAULT ] FOR TYPE data_type
USING index_method [ FAMILY family_name ] AS
{ OPERATOR strategy_number operator_name [ ( op_type, op_type ) ] [ FOR SEARCH | FOR ORDER BY sort_family_name ]
| FUNCTION support_number [ ( op_type [ , op_type ] ) ] function_name ( argument_type [, ...] )
| STORAGE storage_type
} [, ... ]
๐ฆ DESCRIPTION ๐ฆ
so like. CREATE OPERATOR CLASS creates a new operator class. and what even is an operator class, you ask? it's the lil blueprint that defines how a particular data type can be used with an index. iconic behavior honestly. โจ
the operator class specifies that certain operators will fill particular roles or โstrategiesโ for this data type and this index method. it also specifies the support functions the index method should use when this operator class gets picked for an index column. and listen โ all the operators and functions used by an operator class must be defined before the operator class can be created. no putting the cart before the horse, bestie. ๐ด
if you give a schema name, the operator class gets created in that schema. if you don't, it's created in the current schema. and here's the tea: two operator classes in the same schema can have the same name only if they're for different index methods. that's the rule, periodt. ๐
whoever defines an operator class becomes its owner. and rn โ presently, the creating user must be a superuser. ๐ญ why? because a messed-up operator class definition could confuse or even crash the server. like, genuinely. so they gatekeep it and honestly? valid.
also, CREATE OPERATOR CLASS does not presently check whether the operator class definition includes all the operators and functions required by the index method, nor whether the operators and functions form a self-consistent set. it's literally on you, the user, to define a valid operator class. no vibe check from postgres here. ๐
related operator classes can be grouped into operator families. to add a new operator class to an existing family, specify the FAMILY option in CREATE OPERATOR CLASS. without that option, the new class gets placed into a family named the same as the new class (creating that family if it doesn't already exist โ it's giving self-made ๐
).
refer to Section 36.16 for further information. ๐
๐ PARAMETERS ๐
nameโ the name of the operator class to be created. it can be schema-qualified, so flex that namespace if you need to.DEFAULTโ if present, the operator class becomes the default operator class for its data type. at most one operator class can be the default for a specific data type and index method. so like, one main character per category, okay? ๐data_typeโ the column data type this operator class is for.index_methodโ the name of the index method this operator class is for.family_nameโ the name of the existing operator family to add this operator class to. if not specified, a family named the same as the operator class is used (creating it, if it doesn't already exist).strategy_numberโ the index method's strategy number for an operator associated with the operator class.operator_nameโ the name (optionally schema-qualified) of an operator associated with the operator class.op_typeโ okay this one's got layers, pay attention bestie:- in an
OPERATORclause, it's the operand data type(s) of the operator, orNONEto signify a prefix operator. the operand data types can be omitted in the normal case where they're the same as the operator class's data type. - in a
FUNCTIONclause, it's the operand data type(s) the function is intended to support, if different from the input data type(s) of the function (for B-tree comparison functions and hash functions) or the class's data type (for B-tree sort support functions, B-tree equal image functions, and all functions in GiST, SP-GiST, GIN and BRIN operator classes). these defaults are correct, and soop_typeneed not be specified inFUNCTIONclauses โ except for the case of a B-tree sort support function that's meant to support cross-data-type comparisons. that one special lil guy needs it. ๐ง
- in an
sort_family_nameโ the name (optionally schema-qualified) of an existing btree operator family that describes the sort ordering associated with an ordering operator.- if neither
FOR SEARCHnorFOR ORDER BYis specified,FOR SEARCHis the default. ๐
- if neither
support_numberโ the index method's support function number for a function associated with the operator class.function_nameโ the name (optionally schema-qualified) of a function that's an index method support function for the operator class.argument_typeโ the parameter data type(s) of the function.storage_typeโ the data type actually stored in the index. normally this is the same as the column data type, but some index methods (currently GiST, GIN, SP-GiST and BRIN) allow it to be different. theSTORAGEclause must be omitted unless the index method allows a different type to be used. and if the columndata_typeis specified asanyarray, thestorage_typecan be declared asanyelementto indicate that the index entries are members of the element type belonging to the actual array type that each particular index is created for. yeah, read that one twice. ๐
the OPERATOR, FUNCTION, and STORAGE clauses can appear in any order. no choreography required, they're flexible like that. ๐
๐ NOTES ๐
because the index machinery does not check access permissions on functions before using them, including a function or operator in an operator class is tantamount to granting public execute permission on it. so like, be aware. usually not an issue for the sorts of functions that are useful in an operator class, but still โ eyes open. ๐
and the big one: the operators should not be defined by SQL functions. an SQL function is likely to be inlined into the calling query, which will prevent the optimizer from recognizing that the query matches an index. so don't do it. it's giving "why is my index not being used" energy and we hate that for you. ๐ญ
โญ EXAMPLES โญ
the following example command defines a GiST index operator class for the data type _int4 (array of int4). see the intarray module for the complete example. ๐
CREATE OPERATOR CLASS gist__int_ops
DEFAULT FOR TYPE _int4 USING gist AS
OPERATOR 3 &&,
OPERATOR 6 = (anyarray, anyarray),
OPERATOR 7 @>,
OPERATOR 8 <@,
OPERATOR 20 @@ (_int4, query_int),
FUNCTION 1 g_int_consistent (internal, _int4, smallint, oid, internal),
FUNCTION 2 g_int_union (internal, internal),
FUNCTION 3 g_int_compress (internal),
FUNCTION 4 g_int_decompress (internal),
FUNCTION 5 g_int_penalty (internal, internal, internal),
FUNCTION 6 g_int_picksplit (internal, internal),
FUNCTION 7 g_int_same (_int4, _int4, internal);
ate. no crumbs. ๐ฅ
๐ COMPATIBILITY ๐
CREATE OPERATOR CLASS is a PostgreSQL extension. there is no CREATE OPERATOR CLASS statement in the SQL standard. so like, this is our thing, nobody else has it. ๐
โจ
๐ SEE ALSO ๐
ALTER OPERATOR CLASS(ALTER_OPERATOR_CLASS(7))DROP OPERATOR CLASS(DROP_OPERATOR_CLASS(7))CREATE OPERATOR FAMILY(CREATE_OPERATOR_FAMILY(7))ALTER OPERATOR FAMILY(ALTER_OPERATOR_FAMILY(7))
okay that's the whole thing, you made it, so proud of you fr fr ๐ซถ
PostgreSQL 18.6 ยท 2026 ยท CREATE OPERATOR CLASS(7) โ mwah, go define some operator classes bestie, you got this ๐๐๐ฆ
๐ฆ where this page comes from
- alpine 3.23: postgresql18-doc 18.6-r0 ยท PostgreSQL
- alpine 3.24: postgresql18-doc 18.6-r0 ยท PostgreSQL
- arch rolling: postgresql 18.6-4 ยท PostgreSQL
- fedora 43: postgresql 18.6-1.fc43 ยท PostgreSQL
- fedora 43: postgresql 18.6-1.fc43 ยท PostgreSQL
- fedora 44: postgresql 18.6-1.fc44 ยท PostgreSQL
- fedora 44: postgresql 18.6-1.fc44 ยท PostgreSQL
- nixos 26.05: postgresql 18.6
- ubuntu resolute: postgresql-client-18 18.6-0ubuntu0.26.04.1 ยท PostgreSQL
- ubuntu stonking: postgresql-client-18 18.6-3build2 ยท PostgreSQL
- ๐ girlified by deepseek-ai/DeepSeek-V4.1-Flash on : a language model's rewrite of the original, under its terms