Crate rupg_catalog

Expand description

The rupg catalog of user objects: schemas, tables, columns, constraints, indexes, roles, privileges.

At M2 the catalog holds the schemas, tables, sequences, indexes, views, column defaults, constraints and text search objects that DDL makes. Each object gets its OID in the order that PostgreSQL 19 gives OIDs, and each object without a name in the statement gets the name that PostgreSQL gives it. Thus after the same statements on a new cluster, the OIDs and the names are the same as in PostgreSQL.

The catalog does not read SQL. The analyzer turns a statement into calls of this crate in the order of DefineRelation, DefineIndex and ATAddForeignKeyConstraint, and it gives each expression as text that only the analyzer reads back. The catalog makes the TOAST tables that PostgreSQL makes, with their indexes, but they hold no data. It also makes the four internal triggers of each foreign key, which have no work to do because the tables have no rows. The catalog also records the rows of pg_depend in the order that PostgreSQL inserts them.

This crate first ships in milestone M2. See spec/22-crate-layout.md section 22.4 and spec/23-milestones.md.

Re-exports§

pub use database::Database;
pub use tablespace::Tablespace;

Modules§

database: The rows of pg_database. The catalog of a database holds the objects of that database, and the rows of pg_database are the same for all databases of the cluster, so the store of the server keeps them.
flag: The reasons that the walk reached an object, the bits DEPFLAG_* of dependency.c.
names: The names that PostgreSQL makes for objects that a statement does not name, as makeObjectName, ChooseIndexNameAddition and ChooseIndexColumnNames of indexcmds.c make them.
size: The pages of a new relation in PostgreSQL. rupg keeps no files, so the catalog gives the pages that the files of PostgreSQL have when the relation is new.
tablespace: The rows of pg_tablespace that CREATE TABLESPACE makes. A tablespace belongs to the cluster, so the store of the server keeps them, as it keeps the rows of pg_database. The rows of pg_default and pg_global are static rows.
toast: The TOAST decision of PostgreSQL. rupg has no TOAST tables, but a PostgreSQL table that has one uses two OIDs for it, so the catalog must know when PostgreSQL makes one.

Structs§

Aggregate: An aggregate function, one row of pg_aggregate.
AmOp: An operator of an operator family, one row of pg_amop.
AmProc: A support function of an operator family, one row of pg_amproc.
AttrDefault: A column default, one row of pg_attrdef.
Cast: A cast, one row of pg_cast.
Catalog: The objects of one cluster, with the OID counter.
Collation: A collation, one row of pg_collation. The catalog holds the collations that pg_import_system_collations makes when the cluster starts.
Column: A column of a relation, one row of pg_attribute with a positive attnum.
Constraint: A constraint, one row of pg_constraint.
DefaultAcl: The default privileges of a role for the new objects of a kind, one row of pg_default_acl.
Depend: One row of pg_depend.
Domain: The facts of a domain that pg_type holds for a type of the kind d.
EnumLabel: A label of an enum type, one row of pg_enum.
Extension: An extension, one row of pg_extension.
ForeignKey: The facts of a foreign key.
Function: A function, one row of pg_proc: a function of the setup of a new cluster, such as a function in SQL of information_schema, or a function that CREATE FUNCTION makes.
FunctionBody: The body of a function in SQL, which the analyzer reads again each time a query calls the function, as it reads the query of a view.
IndexInfo: The facts of an index, one row of pg_index.
Language: A procedural language, one row of pg_language.
MemberLink: The dependency facts of a member of a family, the fields ref_is_hard, ref_is_family and refobjid of OpFamilyMember.
NewDomain: A new domain for Catalog::create_domain.
NewEnum: A new enum type for Catalog::create_enum.
NewForeignKey: A new foreign key before it has an OID.
NewIndex: A new index before it has an OID.
NewIndexColumn: A column of a new index.
NewRange: A new range type for Catalog::create_range.
NewRelation: A new table, a new index or a new sequence before it has an OID.
NewRule: The facts of a new rule for Catalog::create_rule.
NewTrigger: The facts of a new trigger for Catalog::create_trigger.
ObjRef: An object as pg_depend names it: the OID of its catalog, its OID and a column number or 0.
OpClass: An operator class, one row of pg_opclass.
OpFamily: An operator family, one row of pg_opfamily.
Operator: An operator, one row of pg_operator.
PartitionKey: The partition key of a partitioned table, one row of pg_partitioned_table.
PartitionOf: The parent table and the bound of a partition.
RangeInfo: The facts of a range type, one row of pg_range.
Relation: A relation: a table, an index, a sequence or a TOAST table.
Rule: A rule of CREATE RULE, one row of pg_rewrite. The _RETURN rule of a view is part of ViewInfo, not a Rule.
Schema: A schema.
SequenceData: The row of the file of a sequence, FormData_pg_sequence_data.
SequenceInfo: The facts of a sequence, one row of pg_sequence.
StatExt: An extended statistics object, one row of pg_statistic_ext.
Target: An object that a removal takes away, ObjectAddressExtra with its address.
Trigger: A trigger, one row of pg_trigger: a trigger of CREATE TRIGGER, or an internal trigger of a foreign key.
TsConfig: A text search configuration, one row of pg_ts_config.
TsDictionary: A text search dictionary, one row of pg_ts_dict.
TsMapping: One row of pg_ts_config_map: for a token type of a configuration, the dictionary with this place in the list.
TsParser: A text search parser, one row of pg_ts_parser.
TsTemplate: A text search template, one row of pg_ts_template.
Type: A row type, a domain, or the array type of one of them.
ViewInfo: The query of a view and its _RETURN rule.

Enums§

ConKind: The kind of a constraint, contype.
DepKind: The kind of a dependency, deptype.
FkSide: The side of a foreign key that a row of pg_constraint serves, as addFkConstraint takes it.
PartStrategy: The strategy of a partition key, partstrat.
PartitionBound: The bound of a partition, relpartbound.
RangeDatum: A datum of the bound of a range partition.
RangePart: The part of a type in a range type.
Refusal: The reason that a removal cannot start.
RelKind: The kind of a relation, relkind.

Constants§

BOOTSTRAP_SUPERUSER: BOOTSTRAP_SUPERUSERID, the owner of the built-in objects and of the constructors of a range type.
BTREE: The OID of the btree access method.
FIRST_NORMAL_OID: The first OID for a user object, FirstNormalObjectId.
FIRST_UNPINNED_OID: FirstUnpinnedObjectId. Most objects with a lower OID are pinned, and no pg_depend row refers to them. See is_pinned.
MAX_COLUMNS: MaxHeapAttributeNumber.
PG_AGGREGATE: The OID of pg_aggregate.
PG_AM: The OID of pg_am.
PG_AMOP: The OID of pg_amop.
PG_AMPROC: The OID of pg_amproc.
PG_ATTRDEF: The OID of pg_attrdef.
PG_CAST: The OID of the catalog pg_cast.
PG_CLASS: The OID of pg_class, the catalog of relations.
PG_COLLATION: The OID of the catalog pg_collation.
PG_CONSTRAINT: The OID of pg_constraint.
PG_DATABASE: The OID of pg_database.
PG_DEFAULT_ACL: The OID of pg_default_acl.
PG_EXTENSION: The OID of the catalog pg_extension.
PG_LANGUAGE: The OID of the catalog pg_language.
PG_LARGEOBJECT: The OID of pg_largeobject.
PG_NAMESPACE: The OID of pg_namespace.
PG_OPCLASS: The OID of the catalog pg_opclass.
PG_OPERATOR: The OID of pg_operator.
PG_OPFAMILY: The OID of pg_opfamily.
PG_PROC: The OID of pg_proc.
PG_REWRITE: The OID of pg_rewrite.
PG_STATISTIC_EXT: The OID of pg_statistic_ext.
PG_TABLESPACE: The OID of pg_tablespace.
PG_TOAST_NAMESPACE: The OID of the schema pg_toast.
PG_TRIGGER: The OID of the catalog pg_trigger.
PG_TS_CONFIG: The OID of pg_ts_config.
PG_TS_DICT: The OID of pg_ts_dict.
PG_TS_PARSER: The OID of pg_ts_parser.
PG_TS_TEMPLATE: The OID of pg_ts_template.
PG_TYPE: The OID of pg_type.
PUBLIC: The OID of the schema public.
TRIGGER_BEFORE: TRIGGER_TYPE_BEFORE. A trigger with neither this bit nor TRIGGER_INSTEAD fires after the event.
TRIGGER_DELETE: TRIGGER_TYPE_DELETE.
TRIGGER_INSERT: TRIGGER_TYPE_INSERT.
TRIGGER_INSTEAD: TRIGGER_TYPE_INSTEAD.
TRIGGER_ROW: TRIGGER_TYPE_ROW.
TRIGGER_TRUNCATE: TRIGGER_TYPE_TRUNCATE.
TRIGGER_UPDATE: TRIGGER_TYPE_UPDATE.

Functions§

acl_new_owner: aclnewowner: the items with the old owner as grantee or grantor get the new owner. When the new owner was in the list before, two items with the same grantee and grantor become one, and an item with no privileges goes away. An item that is not in the text form stays as it is.
collation_conflict: True when a new collation of the encoding new cannot have the name of a collation of the encoding existing in the same schema, as CollationCreate checks it. A collation of any encoding (-1) conflicts with every collation, and a collation of UTF8 conflicts with a new collation of any encoding.
eliminate_duplicates: eliminate_duplicate_dependencies: the objects of refs from the highest OID down, without duplicates. A column takes the place of its whole relation, and each other column of the relation stays.
foreign_key_function: The place of a function of the foreign keys in the order of ri_triggers.c, from 0 for RI_FKey_check_ins to 11 for RI_FKey_noaction_upd, or None for another function. The functions of the referencing table are 0 and 1. For the referenced table, an even place is a function of DELETE and an odd place a function of UPDATE.
is_pinned: IsPinnedObject: true for an object that initdb made, except the schema public, the databases and the large objects.